The Best of Hands On Stronger Software, The Next Step in Worker FOR THE

The aspirational delight of Fewer Bugs The case for Productivity EEG Brain TECHNOLOGY
INSIDER
our top DIY projects functional programming Scans
P.18 P.40 P.46 DECEMBER 2022

The
Transistor
at 75
The Past, Present,
and Future of
the World’s Most
Important Device
exp.section(uid="control"):
.play(pulse=pi_half, signal="
.delay(time= sweep ="charge_l
Qubit Control
.play(Just
pulse=pi_half,s_line")
Got Easier.
That Simple.
exp.section(uid="readout"):
.acquire(signal="measure_line

LabOne Q is the new Zurich Instruments software

to control quantum computers. Start now, accelerate
your progess, and enjoy.

∏Intuitive coding
∏High duty cycle
∏Scalable performance
Watch now

Zurich
Instruments
 VOLUME 59 / ISSUE 12 DECEMBER 2022

A New Way to
Squash Bugs
40
Functional programming can make
complex software less fragile.
By Charles Scalfani

Are You Ready for

Workplace Brain
46
Scanning?
The tech can boost productivity,
but will it make us happier?
By Evan Ackerman
& Eliza Strickland

EDITOR’S NOTE 2

NEWS 6
Scientific Machine Learning
5G and Aircraft
Meet the Posits
Sulfuric Acid Test
Geothermal Energy Storage

CAREERS 16
Inflation Hits U.S.
Engineering Salaries

HANDS ON 18
IEEE Spectrum’s Top
DIY Projects

NUMBERS DON’T LIE 21

Batteries Still Fall

The 22 The State of

the Transistor
30 Short of Liquid Fuels

PAST FORWARD 76
TOP: TRANSISTOR MUSEUM/JONATHAN HOPPE COLLECTION; BOTTOM: LIZ PINTO

Transistor In 75 years, it’s become

tiny, mighty, ubiquitous,
The Music of the Transistor

at 75 and just plain weird.

By Samuel K. Moore
THE INSTITUTE
69
& David Schneider Keeping Shelves
The First Transistor
and How It Worked
24 Stocked

The point-contact transistor

Taking Moore’s Law
to New Heights
32 AI platform spots low inventory.

revolutionized electronics even

When transistors can’t
as its inventors struggled to
get any smaller, the only
understand it.
direction is up.
By Glenn Zorpette
By Marko Radosavljevic &
Jack Kavalieros
The Ultimate
Transistor Timeline
29
The transistor has been
The Transistor
of 2047
38
reinvented over and over again.
By Stephen Cass Experts predict what
transistors will be like on
ON THE COVER: their 100th birthday.
Illustration by Lisa Sheehan By Samuel K. Moore

DECEMBER 2022  SPECTRUM.IEEE.ORG  1
 EDITOR’S NOTE BY HARRY GOLDSTEIN
The Device
That Changed
Everything
Transistors are civilization’s
invisible infrastructure
I
was roaming around the IEEE Spectrum office
a couple of months ago, looking at the display
cases the IEEE History Center has installed in
the corridor that runs along the conference
rooms at 3 Park. They feature photos of illustrious
engineers, plaques for IEEE milestones, and a
handful of vintage electronics and memorabilia
including an original Sony Walkman, an Edison
Mazda lightbulb, and an RCA Radiotron vacuum
tube. And, to my utter surprise and delight, a
replica of the first point-contact transistor
invented by John Bardeen, Walter Brattain, and
William Shockley 75 years ago this month.
I dashed over to our photography director,
Randi Klett, and startled her with my excitement,
which, when she saw my discovery, she under-
stood: We needed a picture of that replica, which
she expertly shot and now accompanies this
column.
What amazed me most besides the fact that
the very thing this issue is devoted to was here
with us? I’d passed by it countless times and
never noticed it, even though it is tens of billions
times the size of one of today’s transistors. In fact,
each of us is surrounded by billions, if not trillions
of transistors, none of which are visible to the
naked eye. It is a testament to imagination and
ingenuity of three generations of electronics engi-
neers who took the (by today’s standards) mam-
moth point-contact transistor and shrunk it
down to the point where transistors are so ubiq-
uitous that civilization as we know it would not
exist without them.
2  SPECTRUM.IEEE.ORG  DECEMBER 2022
This replica of the original point-contact
transistor is on display outside IEEE Spectrum’s
conference rooms.
Of course, this wouldn’t be a Spectrum special
issue if we didn’t tell you how the original
point-contact transistor worked, something that
The best even the inventors seemed a little fuzzy on.
explanation According to our editorial director for content
development, Glenn Zorpette, the best explana-
of the point- tion of the point-contact transistor is in Bardeen’s
contact 1956 Nobel Prize lecture, but even that left out
transistor is important details, which Zorpette explores in clas-
in Bardeen’s sic Spectrum style in “The First Transistor and
1956 Nobel How It Worked,” on page 24.
Prize lecture, And while we’re celebrating this historic
but even accomplishment, Senior Editor Samuel K. Moore,
that left out who covers semiconductors for Spectrum and
curated this special issue, looks at what the tran-
important
sistor might be like when it turns 100. For “The
details. Transistor of 2047” [p. 38], Moore talked to the
leading lights of semiconductor engineering, many
of them IEEE Fellows, to get a glimpse of a future
where transistors are stacked on top of each other
and are made of increasingly exotic 2D materials,
even as the OG of transistor materials, germanium,
is poised for a comeback.
When I was talking to Moore a few weeks ago
PORTRAIT BY SERGIO ALBIAC; RANDI KLETT
about this issue, he mentioned that he’s attending
his favorite conference just as this issue comes
out, the 68th edition of IEEE’s International Elec-
tron Devices Meeting, in San Francisco. The
mind-bending advances that emerge from that
conference always get him excited about the engi-
neering feats occurring in today’s labs and on
tomorrow’s production lines. This year he’s most
excited about new devices that combine comput-
ing capability with memory to speed machine
learning. Who knows, maybe the transistor of 2047
will make its debut there, too.
 CONTRIBUTORS
 DAN GARISTO
Garisto is a freelance science
journalist who reports in this issue
on the latest machine-learning
techniques in particle physics and
math [p. 6]. In the process, he
effectively wound up reporting
on how people working on these
problems “think about how
machines think,” he says.
 ALLISON MARSH
Marsh, a historian of technology at
the University of South Carolina,
writes this month about a
transistorized music box owned by
John Bardeen [p. 76]. By the 1990s,
the box had succumbed to “inherent
vice,” a museum term for the
degradation of materials over time.
“‘Inherent vice’ is one of my favorite
phrases,” Marsh says. “It conjures up
images of objects behaving badly.”
 MARKO RADOSAVLJEVIC
Radosavljevic is a principal engineer
in the components research group
at Intel. He and Jack Kavalieros, a
Fellow and vice president of device
integration in the group, describe
a likely next step in the evolution of
Moore’s Law on page 32. “We are
very optimistic that the industry will
continue to find ways in which to
increase density in logic,” he says.
 CHARLES SCALFANI
Scalfani is the chief technology
officer of Panoramic Software,
based in Dana Point, Calif. He has
experienced firsthand the tendency
of software systems to become
increasingly complex and fragile. “It
just gets out of control,” he says.
“It’s like running downhill.” In an effort
to manage that complexity better,
Scalfani was drawn to a paradigm
called functional programming, which
he describes on page 40.
 LISA SHEEHAN
Sheehan, an illustrator based in
Bedfordshire, England, took a
hands-on approach for this month’s
special report. Her 3D-graphics
software couldn’t get the internal
texture of the cake right, so she
needed something real. “I paid my
14-year-old daughter, Isla, to bake a
cake,” she says. She then matched
the illustration’s lighting,
photographed it, and inserted the
image into the illustration. The
mouth-watering result is on page 22.
4  SPECTRUM.IEEE.ORG  DECEMBER 2022
ACTING EDITOR IN CHIEF Harry Goldstein, h.goldstein@ieee.org
EXECUTIVE EDITOR Jean Kumagai, j.kumagai@ieee.org
MANAGING EDITOR Elizabeth A. Bretz, e.bretz@ieee.org
CREATIVE DIRECTOR
Mark Montgomery, m.montgomery@ieee.org
PRODUCT MANAGER, DIGITAL Erico Guizzo, e.guizzo@ieee.org
EDITORIAL DIRECTOR, CONTENT DEVELOPMENT
Glenn Zorpette, g.zorpette@ieee.org
SENIOR EDITORS
Evan Ackerman (Digital), ackerman.e@ieee.org
Stephen Cass (Special Projects), cass.s@ieee.org
Samuel K. Moore, s.k.moore@ieee.org
Tekla S. Perry, t.perry@ieee.org
Philip E. Ross, p.ross@ieee.org
David Schneider, d.a.schneider@ieee.org
Eliza Strickland, e.strickland@ieee.org
ART & PRODUCTION
DEPUTY ART DIRECTOR Brandon Palacio, b.palacio@ieee.org
PHOTOGRAPHY DIRECTOR Randi Klett, randi.klett@ieee.org
ONLINE ART DIRECTOR Erik Vrielink, e.vrielink@ieee.org
PRINT PRODUCTION SPECIALIST
Sylvana Meneses, s.meneses@ieee.org
MULTIMEDIA PRODUCTION SPECIALIST
Michael Spector, m.spector@ieee.org
WEB PRODUCTION Michael Novakovic, m.novakovic@ieee.org
NEWS MANAGER Margo Anderson, m.k.anderson@ieee.org
ASSOCIATE EDITORS
Willie D. Jones (Digital), w.jones@ieee.org
Michael Koziol, m.koziol@ieee.org
SENIOR COPY EDITOR Joseph N. Levine, j.levine@ieee.org
COPY EDITOR Michele Kogon, m.kogon@ieee.org
EDITORIAL RESEARCHER Alan Gardner, a.gardner@ieee.org
EDITORIAL INTERN Dina Genkina, d.genkina@ieee.org
CONTRACT SPECIALIST Ramona L. Foster, r.foster@ieee.org
CONTRIBUTING EDITORS Robert N. Charette, Steven ­Cherry,
Charles Q. Choi, Peter Fairley, Edd Gent, W. Wayt Gibbs, Mark Harris,
Allison Marsh, Prachi Patel, Julianne Pepitone, Lawrence Ulrich,
Emily Waltz
THE INSTITUTE
EDITOR IN CHIEF Kathy Pretz, k.pretz@ieee.org
ASSISTANT EDITOR Joanna Goodrich, j.goodrich@ieee.org
DIRECTOR, PERIODICALS PRODUCTION SERVICES Peter Tuohy
ADVERTISING PRODUCTION MANAGER
Felicia Spagnoli, f.spagnoli@ieee.org
SENIOR ADVERTISING PRODUCTION COORDINATOR
Nicole Evans Gyimah, n.gyimah@ieee.org
ADVERTISING PRODUCTION +1 732 562 6334
EDITOR IN CHIEF EMERITUS Susan Hassler, s.hassler@ieee.org
EDITORIAL ADVISORY BOARD, IEEE SPECTRUM
Susan Hassler, Chair; Ella M. Atkins, Robert N. Charette, Francis J.
Doyle III, Matthew Eisler, Shahin Farshchi, Alissa Fitzgerald, Jonathan
Garibaldi, Benjamin Gross, Lawrence O. Hall, Jason K. Hui, Leah H.
Jamieson, Mary Lou Jepsen, M ­ ichel M. Maharbiz, Somdeb Majumdar,
Lisa May, Carmen S. Menoni, Ramune Nagisetty, Paul Nielsen, Sofia
Olhede, Christopher Stiller, Wen Tong, Boon-Lock Yeo
EDITORIAL ADVISORY BOARD, THE INSTITUTE
Kathy Pretz, Chair; Qusi Alqarqaz, Stamatis Dragoumanos,
Jonathan Garibaldi, Madeleine Glick, Lawrence O. Hall,
Susan Hassler, Francesca Iacopi, Cecilia Metra, Shashi Raj Pandey,
John Purvis, Chenyang Xu
MANAGING DIRECTOR, PUBLICATIONS Steven Heffner
DIRECTOR, BUSINESS DEVELOPMENT,
MEDIA & ADVERTISING Mark David, m.david@ieee.org
EDITORIAL CORRESPONDENCE
IEEE Spectrum, 3 Park Ave., 17th Floor,
New York, NY 10016-5997 TEL: +1 212 419 7555
BUREAU Palo Alto, Calif.; Tekla S. Perry +1 650 752 6661
ADVERTISING INQUIRIES Naylor Association Solutions,
Erik Albin +1 352 333 3371, ealbin@naylor.com
REPRINT SALES +1 212 221 9595, ext. 319
REPRINT PERMISSION / LIBRARIES Articles may be
photocopied for private use of patrons. A per-copy fee must
be paid to the Copyright Clearance Center, 29 Congress St.,
Salem, MA 01970. For other copying or republication, contact
Managing Editor, IEEE Spectrum.
COPYRIGHTS AND TRADEMARKS
IEEE Spectrum is a registered trademark owned by The Institute of
Electrical and Electronics Engineers Inc. Responsibility for the substance
of articles rests upon the authors, not IEEE, its organizational units, or its
members. Articles do not represent official positions of IEEE. Readers
may post comments online; comments may be excerpted for publication.
IEEE reserves the right to reject any advertising.
IEEE BOARD OF DIRECTORS
PRESIDENT & CEO K.J. Ray Liu, president@ieee.org
+1 732 562 3928 Fax: +1 732 981 9515
PRESIDENT-ELECT Saifur Rahman
TREASURER Mary Ellen Randall
SECRETARY John W. Walz
PAST PRESIDENT Susan K. “Kathy” Land
VICE PRESIDENTS
Stephen M. Phillips, Educational Activities; Lawrence O. Hall,
Publication Services & Products; David A. Koehler, Member &
Geographic Activities; Bruno Meyer, Technical Activities;
James E. Matthews, President, Standards Association;
Deborah M. Cooper, President, IEEE-USA
DIVISION DIRECTORS
Franco Maloberti (I); Ruth A. Dyer (II); Khaled Ben Letaief (III);
­Manfred “Fred” J. Schindler (IV); Cecilia Metra (V); Paul M.
­Cunningham (VI); Claudio Cañizares (VII); Christina M. Schober
(VIII); Ali H. Sayed (IX); Dalma Novak (X)
REGION DIRECTORS
Greg T. Gdowski (1); Barry C. Tilton (2); Theresa A. Brunasso (3);
Johnson A. Asumadu (4); Bob G. Becnel (5);  Timothy T. Lee (6);
Robert L. Anderson (7); Antonio Luque (8); Enrique A. Tejera (9);
Deepak Mathur (10)
DIRECTOR EMERITUS Theodore W. Hissey
IEEE STAFF
EXECUTIVE DIRECTOR & COO Stephen Welby
+1 732 562 5400, s.p.welby@ieee.org
CHIEF INFORMATION OFFICER Cherif Amirat
+1 732 562 6017, c.amirat@ieee.org
CHIEF MARKETING OFFICER Karen L. Hawkins
+1 732 562 3964, k.hawkins@ieee.org
PUBLICATIONS Steven Heffner
+1 212 705 8958, s.heffner@ieee.org
CORPORATE ACTIVITIES Donna Hourican
+1 732 562 6330, d.hourican@ieee.org
MEMBER & GEOGRAPHIC ACTIVITIES Cecelia Jankowski
+1 732 562 5504, c.jankowski@ieee.org
STANDARDS ACTIVITIES Konstantinos Karachalios
+1 732 562 3820, constantin@ieee.org
EDUCATIONAL ACTIVITIES Jamie Moesch
+1 732 562 5514, j.moesch@ieee.org
GENERAL COUNSEL & CHIEF COMPLIANCE OFFICER
Sophia A. Muirhead +1 212 705 8950, s.muirhead@ieee.org
CHIEF FINANCIAL OFFICER Thomas R. Siegert
+1 732 562 6843, t.siegert@ieee.org
TECHNICAL ACTIVITIES Mary Ward-Callan
+1 732 562 3850, m.ward-callan@ieee.org
ACTING MANAGING DIRECTOR, IEEE-USA Russell T. Harrison
+1 202 530 8326, r.t.harrison@ieee.org
IEEE PUBLICATION SERVICES & PRODUCTS BOARD
Lawrence O. Hall, Chair; Stefano Galli, Nazanin Bassiri Gharb,
James Irvine, Clem Karl, Hulya Kirkici, Yong Lian, Fabrizio Lombardi,
Peter Luh, Anna Scaglione, Gaurav Sharma, Isabel Trancoso,
Peter Winzer, Bin Zhao, Weihua Zhuang
IEEE OPERATIONS CENTER
445 Hoes Lane, Box 1331
Piscataway, NJ 08854-1331 U.S.A.
Tel: +1 732 981 0060 Fax: +1 732 981 1721
IEEE SPECTRUM (ISSN 0018-9235) is published monthly by
The Institute of Electrical and Electronics Engineers, Inc. All rights
reserved. © 2022 by The Institute of Electrical and Electronics
Engineers, Inc., 3 Park Avenue, New York, NY 10016-5997, U.S.A.
Volume No. 59, Issue No. 12. The editorial content of IEEE Spectrum
magazine does not represent official positions of the IEEE or its
organizational units. Canadian Post International Publications Mail
(Canadian Distribution) Sales Agreement No. 40013087. Return
undeliverable Canadian addresses to: Circulation Department,
IEEE Spectrum, Box 1051, Fort Erie, ON L2A 6C7. Cable address:
ITRIPLEE. Fax: +1 212 419 7570. INTERNET: spectrum@ieee.org.
ANNUAL SUBSCRIPTIONS: IEEE Members: $21.40 included in
dues. Libraries/institutions: $399. POSTMASTER: Please send
address changes to IEEE Spectrum, c/o Coding Department, IEEE
Service Center, 445 Hoes Lane, Box 1331, Piscataway, NJ 08855.
Periodicals postage paid at New York, NY, and additional mailing
offices. Canadian GST #125634188. Printed at 120 Donnelley Dr.,
Glasgow, KY 42141-1060, U.S.A. IEEE Spectrum circulation is audited
by BPA Worldwide. IEEE Spectrum is a member of the Association
of Business Information & Media Companies, the Association of
Magazine Media, and Association Media & Publishing. IEEE prohibits
discrimination, harassment, and bullying. For more information, visit
https://www.ieee.org/about/corporate/governance/p9-26.html.
Simulate real-world
designs, devices, Innovate
and processes faster.
with COMSOL Test more design iterations
before prototyping.
Multiphysics®
comsol.com/feature/multiphysics-innovation
Innovate
smarter.
Analyze virtual prototypes and
develop a physical prototype
only from the best design.

Innovate with
multiphysics
simulation.
Base your design decisions
on accurate results with
software that lets you study
unlimited multiple physical
effects on one model.

Nonisothermal PEM Fuel Cell

 THE LATEST DEVELOPMENTS IN TECHNOLOGY, ENGINEERING, AND SCIENCE 
M
ARTIFICIAL INTELLIGENCE
Machine Learning
Rethinks Scientific
Thinking Neural nets
find new approaches to
physics and math
BY DAN GARISTO
6  SPECTRUM.IEEE.ORG  DECEMBER 2022
ath and physics are old friends. Over
time, they’ve witnessed the rise and fall
of technological advances, including
slide rules, calculators, and aids like
Wolfram Alpha. Now, as the latest developments in
machine learning (ML) are being applied to prob-
lems in math and physics, these advances are raising
fundamental questions about what it means to
teach algorithms to think like us.
X Plus Why?
“When we say computers are very good at math,
they’re very good at things that are quite specific,”
says Guy Gur-Ari, a machine-learning expert at
Google Research. Computers are good at arithme-
tic—plugging numbers in and calculating is com-
paratively straightforward stuff. But outside of
formal structures, computers struggle.
Illustration by Chad Hagen

Solving mathematical word prob-
lems, or “quantitative reasoning,” is
deceptively tricky because it requires a
robustness and rigor that many other
problems don’t. Although ML models
trained on more data make fewer mis-
takes, scaling up only goes so far with
quantitative reasoning. Researchers have
begun to realize that the mistakes ML
language-generation models make in
solving word problems require a more
targeted approach.
Last year, two different teams, at
the University of California, Berkeley,
and at OpenAI, released two data sets,
MATH and GSM8K, respectively, which
contain thousands of math problems
across geometry, algebra, precalculus,
and more. “We basically wanted to see
if it was a problem with data sets,” says
Steven Basart, a researcher at the Center
for AI Safety who worked on MATH.
Could machine learning’s quantitative
reasoning errors be fixed by training on
better formatted, bigger data sets? The
MATH group found just how challenging
quantitative reasoning is for top-of-the-
line ML language models: They scored
less than 7 percent. (A human grad stu-
dent scored 40 percent, while an Inter-
national Mathematical Olympiad champ
scored 90 percent.)
Models attacking GSM8K ques-
tions, which had easier grade-school-
level problems, reached about 20
percent accuracy. To do even that well,
OpenAI researchers used two tech-
niques: fine-tuning and verification.
In fine-tuning, researchers take a pre-
trained language model that includes
irrelevant information and then show
the model only the relevant information
(math problems). Verification allows the
models to review their mistakes.
At the time, OpenAI predicted a model
would need to be trained on 100 times
as much data to reach 80 percent accu-
racy on GSM8K. But in June, Google’s
­Minerva announced 78 percent accu-
racy with minimal upward scaling. “It’s
ahead of any of the trends that we were
expecting,” says Karl Cobbe, a machine-­
“We’d like to have a machine learn
to think more like a physicist. We
also just need to learn how to think
a little bit more like a machine.”
—JESSE THALER, MIT
learning expert at OpenAI. Basart agrees.
“That’s shocking. I thought it would take
longer,” he says.
Minerva uses Google’s own Pathways
Language Model (PaLM), fine-tuned on
scientific papers from arXiv and other
sources with formatted math. Two other
strategies helped Minerva. In “chain
of thought prompting,” Minerva was
required to break down larger problems
into more palatable chunks. The model
also used majority voting—instead of
being asked for one answer, it was asked
to solve the problem 100 times. Of
those results, Minerva picked the most
common answer.
The gains from these new strategies
were enormous. Minerva shot up to 50
percent accuracy on MATH and nearly
80 percent accuracy on GSM8K, as well
as the MMLU, a more general set of
STEM questions that includes chemistry
and biology. When Minerva was asked to
redo a random sample of slightly tweaked
questions, it performed just as well, sug-
gesting that its capabilities were not from
mere memorization.
What Minerva knows—or doesn’t
know—about math is fuzzier. Unlike
“proof assistants” used by mathemati-
cians, which come with built-in struc-
ture, Minerva and other language models
have no formal structure. They can have
strange, messy reasoning and still arrive
at the right answer. As numbers grow
larger, the language models’ accuracy fal-
ters, something that would never happen
on a trusty, old TI-84 Plus calculator.
“Just how smart is it? Or isn’t it?”
asks Cobbe. Though models like Min-
erva might arrive at the same answer as
DECEMBER 2022
a human, the actual process they’re fol-
lowing could be wildly different. On the
other hand, “chain of thought prompt-
ing” is familiar to any human student
who’s been asked to “show your work.”
“I think there’s this notion humans
doing math have some rigid reasoning
system that there’s a sharp distinction
between knowing something and not
knowing something,” says Ethan Dyer,
a machine-learning expert at Google
Research. But humans give inconsistent
answers, make errors, and fail to apply
core concepts too. The borders, at this
frontier of machine learning, are blurred.
Think Like a Physicist
Particle physics data is unusual. While
convolutional neural nets (CNNs) have
proven extremely effective at classifying
images of everyday objects like trees,
cats, and food, they’re not great for par-
ticle collisions. The problem, according
to Javier Duarte, a particle physicist at
the University of California, San Diego,
is that collision data such as that from
the Large Hadron Collider, doesn’t nat-
urally work as an image.
Flashy depictions of collisions at the
LHC can misleadingly fill up the entire
detector. In reality, only a few out of mil-
lions of inputs are registering a signal,
like a white screen with a few black
pixels. This makes for a poor image in a
CNN, but it can work well in a different,
newer framework called graph neural
networks (GNNs).
Beyond its strange format, there’s the
sheer amount of data—about one peta-
byte per second, of which only a small
high-quality amount is saved. To better
DECEMBER 2022  SPECTRUM.IEEE.ORG  7
 NEWS
sift through the data, researchers want to train a sharp-eyed
algorithm. To be effective, such an algorithm would need to
be incredibly speedy, executing in microseconds, Duarte says.
Machine-learning techniques like pruning and quantization,
could allow algorithms to get there.
Machine learning is also allowing particle physicists to see
data in a different light. Instead of focusing on a single event—
say, a Higgs boson decaying to two photons—they are learning
to think about the dozens of other events that happen during a
collision. Although there’s no causal relationship between any
two events, researchers are now embracing a more holistic view
of the data, not just the piecemeal point of view that comes
from analyzing individual events.
More dramatically, machine learning has also forced physi-
cists to reconceive basic concepts. “I was imprecise in my own
thinking about what a symmetry was,” says Jesse Thaler, a
theoretical particle physicist at the Massachusetts Institute of
Technology. “Forcing myself to teach a computer what a sym-
metry was, helped me understand what a symmetry actually
is.” Symmetries require a reference frame—in other words, is
the image of a distorted sphere in a mirror actually symmetri-
cal? There’s no way of knowing without knowing if the mirror
itself is distorted.
These are still early days for machine learning in particle
physics, and researchers are effectively treating the technique
like a proverbial kitchen sink. “It may not be the right fit for
every single problem in particle physics,” admits Duarte.
As some particle physicists delve into ML, an uncomfortable
question rears its head: Are they doing physics or computer
science? Stigma against coding—often not considered “real
physics”—already exists; similar concerns swirl around ML.
Researchers who worry that ML will obscure the analysis of
very complicated collisions are building algorithms to provide
feedback in language humans can understand. But algorithms
may not be the only ones with responsibilities to communicate.
“On the one hand, we’d like to have a machine learn to think
more like a physicist. We also just need to learn how to think a
little bit more like a machine,” Thaler says. “We need to learn
to speak each other’s language.”
Graph neural networks (GNNs) help particle physicists
recreate and reimagine the complex and manifold colli-
sions at the Large Hadron Collider—pictured here in a
cross section of the heavy-ion ALICE detector.
8  SPECTRUM.IEEE.ORG  DECEMBER 2022
TELECOMMUNICATIONS
How 5G’s Rollout Rattled
Hundreds of Pilots Did
C-band signals foul up
planes’ altimeters?
BY MARK HARRIS
I
n January this year, at least three flights above the
U.S. state of Tennessee simultaneously experi-
enced altimeter errors that made it “impossible
to maintain assigned altitude,” according to one
of the pilots. One jet lost its autopilot completely, and
reportedly had fire trucks waiting for it on landing.
In February, a passenger plane on approach to
the Louis Armstrong International Airport in New
Orleans experienced erratic low-altitude warnings
as it flew below 1,000 feet. “This sort of erroneous
warning indications would be extremely distracting
in a more challenging environment such as low vis-
ibility, icing conditions, etc,” the pilot wrote later.
In March, a commercial jet landing on autopilot
at Los Angeles International Airport suddenly went
into an aggressive descent just 100 feet above the
ground. “I took control of the aircraft and raised
the nose and landed,” its pilot reported. “It was a
very alarming pushover by the autopilot. In [other]
conditions, it could have caused a crash.”
All three incidents—and many more this year—
were linked by pilots to problems with the aircrafts’
radio (radar) altimeters, which pilots rely on during
takeoff and landing, and to help avoid crashing into
mountains. These altimeters also feed into critical
autopilot, autothrottle, and instrument landing sys-
tems. According to an IEEE Spectrum analysis of
reports made to NASA’s Aviation Safety Reporting
System (ASRS), complaints of malfunctioning and
failing altimeters soared after the rollout earlier this
year of high-speed 5G wireless networks, which use
the same frequency band as the planes’ radar systems.
ASRS is a public database maintained by NASA
to encourage U.S.-based air and ground crew, and
air traffic controllers, to anonymously share safety
incidents and concerns. Between January and May,
there were 93 reports of faulty or failing radar altim-
eters, where a normal year might see only a handful.
January alone saw almost twice as many complaints
of malfunctioning altimeters as the previous five
years combined. In most, including the ­Tennessee
and Los Angeles incidents above, the reporter
referred to 5G interference.
The U.S. Federal Communications Commission
initially played down concerns that the new cell
CERN
towers and devices might interfere with commercial
 jet radar altimeters, which operate a few
hundred megahertz higher in the radio
spectrum. In early 2020, the agency wrote:
“The technical rules on power and emis-
sion limits we set for the 3.7 [gigahertz]
service and the spectral separation of
220 megahertz offers significant protec-
tion of services in the 4.2–4.4 GHz band.”
The U.S. Federal Aviation Administra-
tion was not so certain. It commissioned
an empirical study by RTCA, a nonprofit
that studies aviation electronics, to assess
the operational risks of 5G interference.
That report concluded that C-band 5G
systems would cause harmful interfer-
ence to radar altimeters on all types of civil
aircraft, and that “this risk is widespread
and has the potential for broad impacts to
aviation operations in the United States,
including the possibility of catastrophic
failures leading to multiple fatalities.”
It found the risk would originate from
intentional and spurious signals from
terrestrial cell towers, as well as 5G cell-
phones inadvertently operated onboard
aircraft, exacerbated by altimeter receiv-
ers with poor spectral selectivity. In what
now seems a prescient observation, the
RTCA report said: “[The] possibility
of harmful interference…is particularly
dangerous given that up to the present
time, radar altimeter failures…have been
extremely uncommon.”
Nevertheless, the spectrum auction
went ahead, with wireless providers led by
Verizon and AT&T paying over US $80 bil-
lion for the frequencies. 5G services were
due to go live in 46 markets around the
United States on 5 January this year. But as
the deployment neared, the FAA issued an
airworthiness directive that would have
vastly restricted air travel into those areas.
Ultimately, wireless carriers agreed to
delay the 5G rollout for two weeks, and
the FAA banned certain aircraft—those
with susceptible or untested altime-
ters—from low-visibility landings at
airports with 5G wireless networks. The
networks also agreed to set up “buffer
zones” around 50 airports that would
JETLINER IMAGES/GETTY IMAGES
reduce 5G C-band signal levels within a
mile of takeoff and landing runways to
one-tenth of what was originally called
for in the networks’ designs. The miti-
gations would remain in place for six
months, while the FAA attempted to clear
more aircraft for operation.
On 19 January, the 5G systems were
switched on.
Almost immediately, complaints began
rolling into NASA. On their first flight to
San Francisco after 5G was switched on,
one pilot was horrified when the plane’s
speed brakes unexpectedly activated
before touchdown. “With over 18,000
hours as Captain of Boeing airliners…I’ve
never had the auto speedbrakes deploy
uncommanded before ground contact,”
the pilot wrote. “While I operate in the 5G
environment, I have no intention on being
the first to make a 5G landing.”
“I was predisposed to think the
implementation would be a non-event
and was surprised to experience actual
interference events,” reported another
pilot, who suspected a malfunctioning
altimeter had caused their plane’s auto-
throttle to disengage.
ASRS contains at least 40 reports of
possible 5G interference in the vicinity of
the 50 airports with buffer zones. With
six reports, Phoenix Sky Harbor Inter­
national Airport had the most complaints.
“I’ve been flying an aircraft with a
radar altimeter for years and never once
have I seen a malfunction such as this
until the 5G turn on,” wrote another pilot
in Florida, whose altimeter fluctuated
between 90 and 400 feet prior to takeoff.
“Coincidence? Probably not.”
But could it be? Chris Rudell is an
associate professor in the department of
electrical and computer engineering at
the University of Washington. “I’d sleep
like a baby [on a plane] that flew over
a 5G base station at full power output,”
he told Spectrum. “Probably something
happens that’s unusual and the pilots
attribute it to 5G but maybe it’s not
attributable to 5G. After everything in
the news, they’re now submitting what
they actually see, whereas before they
perhaps weren’t motivated to do that.”
It is true that the FAA does now have
a specific online radio-altimeter anom-
aly reporting form that it asks pilots and
other aviation personnel to complete. The
FAA would not share the full details of
these reports, but it did provide high-level
data that indicate it has received around
550 such submissions since January.
The agency has reviewed over half of
the reports and was unable to rule out
5G interference in around 80 incidents.
The agency claims that, contrary to the
ASRS reports, none of the interference
events identified by the FAA had any
safety-­related system impacts, affecting
direct aircraft control inputs such as
autothrottle or speed brakes. The FAA
has concluded that the mitigations it had
agreed to undertake with wireless pro-
viders “are working.” In June, it agreed
with the carriers to extend those mitiga-
tions for another year.
Although altimeter complaints on
ASRS seem to be declining, many pilots
still want tighter controls on 5G tech-
nology. The Aircraft Owners and Pilots
Association (AOPA), the world’s largest
organization representing general avia-
tion pilots and aircraft owners, told Spec-
trum: “We call for continued discussions
between the public and private sectors to
further mitigate the proven safety risks
that 5G technology poses to radar altime-
ters. Further cooperation and coordination
will bring about solutions that work for all.”
A few pilots were less diplomatic when
sharing their views anonymously. “Shut-
down [sic] 5G until carefully vetted,” sug-
gested one. “Delay implementation of 5G
service until all issues with radar altime-
ters and transport category aircraft have
been dealt with,” wrote another. A third
simply pleaded: “Turn off 5G cell service
near airports.”
DECEMBER 2022  SPECTRUM.IEEE.ORG  9
 NEWS
ARTIFICIAL INTELLIGENCE
A New Type of Number
Improves the Math of AI To
speed training, posits maintain
accuracy with fewer bits
BY DINA GENKINA
T
raining the large neural net-
works behind many modern
AI tools requires real compu-
tational might: For example,
OpenAI’s most advanced language
model, GPT-3, required an astounding
100,000 billion billion operations to
train, and cost about US $5 million in
compute time. Engineers think they
have figured out a way to ease the
burden by using a different way of rep-
resenting numbers.
Back in 2017, John Gustafson, then
jointly appointed at A*STAR Computa-
tional Resource Centre and the National
University of Singapore, and Isaac Yone­
moto, then at Interplanetary Robot and
Electric Brain Co., developed a new
way of representing numbers. These
numbers, called posits, were proposed
10  SPECTRUM.IEEE.ORG  DECEMBER 2022
as an improvement over the standard
floating-point arithmetic processors in
use today.
Now, a team of researchers at the
Complutense University of Madrid have
developed the first processor core imple-
menting the posit standard in hardware
and showed that, bit-for-bit, the accuracy
of a basic computational task increased
by up to four orders of magnitude, com-
pared with computing using standard
floating-point numbers. They presented
their results in September at the IEEE
Symposium on Computer Arithmetic.
“Nowadays it seems that Moore’s law
is starting to fade,” says David Mallasén
Quintana, a graduate researcher in the
ArTeCS group at Complutense. “So we
need to find some other ways of get-
ting more performance out of the same
machines. One of the ways of doing that
is changing how we encode the real num-
bers, how we represent them.”
The Complutense team isn’t alone in
pushing the envelope with number repre-
sentation. Back in September, Arm, Intel,
and Nvidia agreed on a specification
for using 8-bit floating-point numbers
instead of the usual 32-bit or 16-bit ver-
sions for machine-learning applications.
Using the smaller, less-precise format
improves efficiency and memory usage,
at the cost of computational accuracy.
Real numbers can’t be perfectly rep-
resented in hardware, simply because
there are an infinite number of them. To
fit into a designated number of bits, many
real numbers have to be rounded. The
advantage of posits comes from the way
the numbers they represent precisely are
distributed along the number line. In the
middle of the number line, around 1 and
-1, there are more posit representations
than with floating point. And at the
wings, going out to large negative and
positive numbers, posit accuracy falls off
more gracefully than does floating point.
“It’s a better match for the natural
distribution of numbers in a calculation,”
says Gustafson. “It’s the right dynamic
range, and it’s the right accuracy where
you need more accuracy. There’s an awful
lot of bit patterns in floating-point arith-
metic no one ever uses. And that’s waste.”
Posits accomplish this improved
accuracy around 1 and -1 thanks to an
extra component in their representation.
Floats are made up of three parts: a sign
bit (0 for positive, 1 for negative), sev-
eral “mantissa” (fraction) bits denoting
what comes after the binary version of
a decimal point, and the remaining bits
defining the exponent (2exp ).
Posits keep all the components of a
float but add an extra “regime” section,
an exponent of an exponent. The beauty
of the regime is that it can vary in bit
length. For small numbers, it can take
as few as two bits, leaving more preci-
sion for the mantissa. This allows for the
HIROSHI WATANABE/GETTY IMAGES
higher accuracy of posits in their sweet
spot around 1 and -1.
Deep neural networks usually work
with normalized parameters called
weights, making them the perfect can-
didate to benefit from posits’ strengths.
Much of neural-net computation is com-
posed of multiply-accumulate opera-
tions. Every time such a computation is
Posits showed an astounding four-
order-of-magnitude improvement
over their floating-point
counterpart in the accuracy of
matrix multiplication.
performed, each sum has to be truncated
anew, leading to accuracy loss. With
posits, a special register called a quire
can efficiently do the accumulation step
to reduce the accuracy loss. But today’s
hardware implements floats, and so far,
computational gains from using posits
in software have been largely overshad-
owed by losses from converting between
the formats.
With their new hardware imple-
mentation, which they synthesized in a
field-programmable gate array (FPGA),
JOURNAL WATCH
Obstetric Self-
Monitoring App
Although a pregnancy is an exciting
time for expectant parents, it can
involve many trips to a doctor’s
office to monitor the health of the
fetus. Currently, the most common
method for monitoring the heartbeat
of a fetus is through an ultrasound
test—which requires not only a visit
to a doctor’s office but also the
expertise of a trained technician.
In the search for a simpler
approach, researchers have
developed a novel system that
includes a wearable device, a
smartphone, and an algorithm that
could offer at-home monitoring
of a fetus’s cardiac health. The
new algorithm, dubbed Lullaby, is
described in a study published on
19 August in IEEE Sensors Letters.
The wearable device involves a
patch that’s placed on the abdomen
of a pregnant user in addition to
electrodes that monitor the electrical
activity of the fetus’s heart. The
device also includes a microcontroller
that processes the signals and sends
them via Bluetooth to a smartphone
or watch, where the algorithm creates
the Complutense team was able to
compare computations done using
32-bit floats and 32-bit posits side by
side. The team assessed the accuracy
of the results by comparing them with
the much more accurate but compu-
tationally costly 64-bit floating-point
format. Posits showed an astounding
four-­order-of-magnitude improvement
over their floating-point counterpart in
the accuracy of matrix multiplication, a
series of multiply-accumulates inherent
in neural-network training. The team
an electrocardiogram (ECG). Users
can then view the data on an app.
However, a major challenge with
this type of tech is that continuous
ECG monitoring involves a lot of
data to process, which has made
real-time monitoring with wearable
devices challenging. The new Lullaby
algorithm addresses this issue.
“Lullaby was made to push the
boundaries of the field by creating an
algorithm that could process high-
resolution ECG in real time and on
a wearable device,” explains Daniel
Jilani, an undergraduate researcher
at the University of California, Irvine,
who co-led the development of this
technology.
In their study, the researchers
used a data set of abdominal ECG
recordings to test the Lullaby
approach against other existing
ECG-processing algorithms, finding it
to be between 7 times and 1,000 times
as fast as these existing options.
“In terms of power, we believe that
the algorithm is efficient enough to
run continuously [on a smartphone]
for days or weeks,” says Jilani. “In
terms of RAM memory, the Lullaby
algorithm uses memory on the scale
of kilobytes, meaning it can run on
memory-limited devices such as
microcontrollers and smartwatches.”
also found that the improved accuracy
didn’t come at the cost of computation
time, only a slight increase in chip area
used and energy consumed.
Although the numerical accuracy
gains are undeniable, how, exactly, this
would affect the training of large AIs like
GPT-3 remains to be seen.
“It’s possible that posits will speed
up training because you’re not losing
as much information on the way,” says
­Mallasén “But these are things that we
don’t know. Some people have tried it out
in software, but we also want to try that
in hardware now that we have it.”
Other teams are working on their own
hardware implementations to advance
posit usage. “It’s doing exactly what I
hoped it would do; it’s getting adopted
like crazy,” Gustafson says. “The posit
number format caught fire, and there are
dozens of groups, both companies and
universities, that are using it.”
The researchers do not yet know
how accurate the Lullaby approach
is compared with a traditional
ultrasound test, but they note
that ultrasound devices are less
accessible, more expensive, and more
difficult to operate than an ECG.
What’s more, ultrasound tests are
done during a visit to the doctor’s
office, whereas a wearable ECG
device can provide more continuous
monitoring as a pregnant user goes
about a normal day. Significantly,
this technology could have
particular impact by making fetal
heart monitoring more accessible
to low-income and disadvantaged
communities, Jilani notes.
Since this initial study, says
Jilani, the team has already greatly
improved the algorithm’s accuracy
and speed and is now working toward
implementing it in a full system. This
includes work on a mobile app that
can be downloaded to smartphones
to support fetal heart monitoring.
The group has a provisional joint
patent on the Lullaby algorithm with
the University of California, Irvine,
and has teamed up with a sponsoring
company, Sensoriis, to produce
a novel fetal cardiac-monitoring
system that uses it.
—Michelle Hampson
DECEMBER 2022  SPECTRUM.IEEE.ORG  11
 NEWS
California Sulphur Co. churns out 2,000 tonnes of pelletized sulfur per day.
The company says it has the capacity to store over 90,000 tonnes at its
Wilmington, Calif., facility.
ENERGY
Where Will We Get Sulfur in a
Petroleum-Free World? Sulfur,
essential to EV batteries, comes mainly
from fossil fuels
BY RAHUL RAO
L
ook at the periodic table and
think of the elements needed for
a prosperous planet powered by
renewable energy. Sulfur likely
won’t be the first to come to mind.
In a net-zero future, a future where
petroleum and natural-gas production
enter terminal decline, sulfur production
will fall away, too.
Therein lies the problem. Sulfur—
easily turned into sulfuric acid—is a
necessary tool for creating fertilizer and
extracting heavy metals from their ores
before they can go into batteries, wind
turbines, and electric-vehicle compo-
nents. Even as sulfur production is set
to fall, sulfuric-acid demand is set to rise.
That’s a recipe for an impending crisis,
researchers warn in a paper published
21 August in The Geographical Journal.
Today, the world uses 246 million
tonnes of sulfuric acid in a year. The
researchers project that number might
increase to 400 million tonnes by 2040.
That’s because, when it comes to
renewable energy, sulfuric acid has a
very critical use. Extracting heavy metals,
such as nickel, cobalt, and rare earths,
relies on chemical processes that use
12  SPECTRUM.IEEE.ORG  DECEMBER 2022
sulfuric acid to separate the metals from
their ores. Those heavy metals are key
elements in lithium-ion batteries, electric
motors, and other technologies crucial
for the renewable transition.
Where does that sulfur come from,
then? Decades ago, most of it came
from mining. In 1894, a German chemist
named Herman Frasch plotted a process
of tapping a sulfur-containing mineral
deposit by pumping it with superheated
water. The sulfur will melt and bubble up
to the surface. Sulfur miners have used
the same process ever since.
The Frasch process, however, can be
tremendously damaging to the surround-
ing environment and mine workers’
health. The minerals that contain sulfur
often also contain toxic metals like mer-
cury, arsenic, and thallium. The process
also spews out wastewater that contains
hydrogen sulfide, which causes poison-
ing symptoms similar to those of carbon
monoxide. (Although hydrogen sulfide
is in fact some 10 times as toxic as CO.)
By the 1950s, the world had exhausted
the most easily available sulfur deposits.
Analysts warned of an impending “sul-
phur famine.”
Then, in swept the fossil-fuel indus-
try’s waste to save the day. By coinci-
dence, sulfur needs to be removed from
petroleum as a precursor to distilling it
into diesel, jet fuel, and other assorted
products. The refineries filter out the
sulfur—thereby making it available to
the rest of the world. Today, more than
80 percent of the world’s sulfur comes
from these acts of waste management.
Dirty as it sounds, it’s actually cleaner
than mining.
However, if carbon-emitting fossil
fuels go, refineries will likely go too. In
a scenario where the world reaches net-
zero carbon dioxide emissions by 2050,
it’s estimated that sulfur production
would drop by one-half or even more.
“It’s an eye-opener,” says Jean-Michel
Lavoie, a chemist at the Laboratoire des
Technologies de la Biomasse, on the
campus of the University of Sherbrooke,
in Quebec, who was not an author of the
paper. “We never really focus on trying
to solve these problems, but I think we’re
going to start looking into solutions.”
The world could swing back to sourc-
ing sulfur directly from mines.
“Research is urgently needed to
develop low-cost, low environmen-
tal impact methods of extracting large
quantities of elemental sulfur,” said
Mark Maslin, a geographer at University
College London and one of the paper’s
authors, in a press release.
Maslin and his colleagues believe,
however, that there won’t be a single
solution to a sulfur shortage. In some
fields, recycling sulfuric acid is quite
common. A few scientists have begun tin-
kering with sulfuric-acid substitutes—
nitric acid, for example—though that’s
still quite speculative.
Reducing the future need for sulfuric
acid certainly wouldn’t hurt. Already,
there are batteries (such as lithium iron
phosphate batteries) that have rela-
tively low energy-capacity-to-weight
ratios but require less nickel, cobalt,
BING GUAN/BLOOMBERG/GETTY IMAGES
and heavy metals to produce, and thus
need less sulfuric acid. Future research
could yield batteries that deliver the best
of both worlds.
What is clear is that weaning off fossil
fuels may have quite a number of unin-
tended side effects—bad or good—and a
sulfur shortage is just one of them.
“A lot of people, they take sulfur for
granted,” says Lavoie.

Enhanced geothermal systems (EGS) are artificial reservoirs that will mimic
natural geologic formations like this one. The EGS plants under development
will be carved into hot, dry rock formations by drilling into or melting the
rock and then injecting water into the fissures.
ENERGY
Hot Rocks Best Batteries for Energy
Storage Artificial geothermal reservoirs
could boost renewables’ uptake
BY PRACHI PATEL
G
eothermal systems carry
warmth from Earth’s interior
up to the surface for heating
or electricity. But geothermal
power plants are expensive to build and
will get even less economically viable as
wind and solar power get cheaper and
more plentiful. Even as wind and solar
grow, however, so does the need to store
electricity from these intermittent
sources.
A new proposal could solve those
issues and bolster all three renewable
technologies. The idea is simple—use
advanced geothermal reservoirs to store
excess wind and solar power in the form
of hot water or steam and bring up that
heat when wind and solar aren’t avail-
able, to turn turbines for electricity.
“It would allow next-generation geo-
thermal plants to break from the tradi-
tional baseload operating paradigm and
earn much greater value as suppliers
SERGIO PITAMITZ/AP
of wind and solar,” says Wilson Ricks,
a graduate student in mechanical and
aerospace engineering at Princeton
University.
Ricks, his Ph.D. advisor Jesse Jenkins,
and Jack Norbeck, cofounder and chief
technology officer of Houston-based
advanced geothermal developer Fervo
Energy, ran extensive simulations of
such geothermal-reservoir energy stor-
age to see if the technical components
of the system as well as the economics
actually work out. They found that the
systems could indeed store electricity
over a range of time scales, from a few
hours up to many days, as efficiently as
lithium-ion batteries do. Plus, says Ricks,
“the storage capacity effectively comes
free of charge with construction of a geo-
thermal reservoir.”
Their results apply only to enhanced
geothermal plants, like the ones Fervo
and other companies such as Cam-
bridge, Mass.–based Quaise Energy
and Seattle-based AltaRock Energy are
developing.
Conventional geothermal systems
drill wells into naturally occurring hydro-
thermal reservoirs. But these pockets
of hot water deep underground do not
exist everywhere. In the United States,
for instance, they are mostly located in
the West.
Enhanced geothermal systems (EGS)
get around this geological limitation by
making artificial reservoirs. Developers
create fractures in hot, dry rock forma-
tions by drilling into or melting the rock,
and then injecting water into the fissures.
Production wells bring the heated water
up for generating electricity. “For scales
necessary to contribute to national or
global electricity decarbonization, we
need to be able to extract geothermal
heat outside of conventional forma-
tions,” Ricks says.
Fervo Energy raised US $138 million
in venture capital funding in August to
advance its technology. The company
uses innovations from the oil and gas
industries, such as horizontal drilling and
distributed fiber-optic sensing, to create
underground reservoirs. The company
plans to use the new funds to complete
two pilot projects, including one with
Google in Nevada.
Once these EGS systems are in place,
they would be ideal for storing energy as
well as producing electricity. Excess wind
or solar energy could be used to inject
water into the artificial reservoirs, where
it would accumulate and build up pres-
sure. The production wells could then
be opened up when electricity is needed.
“EGS reservoirs are created in rock
formations that are naturally imper-
meable; everything outside the artifi-
cial reservoir is sealed off,” says Ricks.
“It’s very similar to a hydropower res-
ervoir, where you choose when to have
water go through the dam and generate
electricity.”
Depending on the geology and traits
of the rocks, Ricks and his colleagues
found through their simulations that
the systems could store energy with up
to 90 percent efficiency over one cycle.
That’s comparable with lithium-ion and
pumped hydro storage, he says. The cost,
meanwhile, would be minimal compared
with other energy-storage technologies.
It would require larger facilities on the
surface, but the storage space would
be effectively free, because the EGS
reservoirs are being built for electricity
anyway.
In January, the team received $4.5
million in funding from the Advanced
Research Projects Agency–Energy
(ARPA-E) to demonstrate a full-scale test
of geothermal reservoir energy storage in
the field. The group published detailed
findings of their modeling study recently
in the journal Applied Energy.
DECEMBER 2022  SPECTRUM.IEEE.ORG  13
 THE BIG PICTURE

Shot of
Nuclear Fusion
By Willie D. Jones

An old saw regarding

the multitude of dashed
hopes about fusion
energy’s promise goes
“Fusion is 30 years
away—and it always
will be.” After decades
of predictions that
fusion was just around
the corner, a team
at the Joint European
Torus (JET) plasma
physics experiment
did something that
suggests scientists
are homing in on
exactly which corner
that is. In February
2022, the U.K-based JET
experimenters induced
the single greatest
sustained energy
pulse ever created by
humans. It had twice
the energy of the
previous record-setting
blast, triggered
a quarter century
earlier. A doubling
every 25 years is far
behind the pace of the
microchip improvements
described by Moore’s
Law. But that hasn’t
dampened enthusiasm
over an alternative
energy source that
could make fossil fuels
and their effect on
the environment relics
of a bygone era. In
the foreground of the
picture is a trainee
learning how to use
the systems involved in
accomplishing the feat.
PHOTOGRAPH BY
LEON NEAL/GETTY IMAGES

14  SPECTRUM.IEEE.ORG  DECEMBER 2022
DECEMBER 2022  SPECTRUM.IEEE.ORG  15
 SHARING THE EXPERIENCES OF WORKING ENGINEERS

Inflation-
INCOME BY GENDER Median primary income, for those
working full time in their primary area of technical
competence, in thousands of U.S. dollars

Adjusted Income
for U.S. Engineers 180

Drops Insights MEN

WOMEN

from IEEE-USA’s MEDIAN PRIMARY INCOME: $160,000 163

annual salary 160

156

survey 150

BY TEKLA S. PERRY
140

H
ow much does a tech professional in the 130
United States earn? In 2021, the median 128 128
income of U.S. engineers and other tech
professionals who were IEEE members
hit US $160,097, up from $154,443 in 2020. That
120
bump in pay is revealed in the IEEE-USA 2022
Salary & Benefits Survey Report. 112
This increase turns into a nearly $3,500 dip,
however, when adjusted for inflation. It’s the first
significant dip in median tech salary in terms of
spending power recorded by IEEE-USA since 2013.
These numbers—and 65 more pages of detailed 100
2021 salary and job-satisfaction statistics—give
readers a good sense of the United States’ tech 93
employment landscape. The analysis is based on
3,057 responses from professionals working full
time in their areas of technical competence; they
reported their income, excluding overtime pay,
80
bonuses, profit sharing, and side hustles. (When
2018 2019 2020 2021
those are considered, the 2021 median income for
these tech professionals was $167,988, according
to the report.)
The survey report chronicles bad news for
women in engineering, as their incomes fell fur- No. 1 job 10 percent, where it’s been for the past 10 years).
ther behind men’s in 2021. The gap in salaries The salary gap between Caucasian and African
specialty:
between genders grew $5,900 (not adjusted for American engineers decreased by $11,000 to
inflation) to $33,900. The gap is tricky to measure,
solid-state $13,000 in 2021, while the disparity between Cau-
given that men responding to the survey had more circuitry casian and Hispanic engineers’ incomes fell by
years of ex­perience as a group than the women, nearly $6,000 to $12,278.
and more women entering the engineering work- 2021 was a good time to be an engineer working
force could skew the median salary downward. with solid-state circuitry: Salaries continued a
However, the proportion of female engineers in steep climb and claimed the No. 1 spot on the salaries-
the workforce remained flat (on a plateau at under by-specialty list. Last year’s No. 1 on that chart,

16  SPECTRUM.IEEE.ORG  DECEMBER 2022
 DECEMBER 2022

193 INCOME BY SELECTED U.S. CENSUS DIVISIONS SATISFACTION WITH TECH JOBS Average of
Median primary income, for those working responses according to a 5-point scale from
full time in their primary area of technical +2 (very satisfied) to -2 (very dissatisfied)
competence, in thousands of U.S. dollars

1.2
2018
2019
1.1
2020
166
2021
1.0

152
0.9

142 0.8

135
0.7

0.6

0.5

0.4

0.3

—— Technical challenges of the job

0.2 —— Overall satisfaction
—— Compensation
—— Advancement opportunities
0.1
PACIFIC WEST SOUTH MID- EAST WEST 2012 2014 2016 2018 2020
CENTRAL ATLANTIC SOUTH NORTH
CENTRAL CENTRAL

consumer electronics, saw a decline in average
salary. Engineers working with other circuits and
devices, machine learning, image and video pro-
cessing, and engineering in medicine and biology
recorded big gains.
Median salaries for engineers in the Pacific
region increased dramatically compared with those
in the rest of the United States, climbing faster than
hypothetically booming regions like the West South
Central area, which includes Texas. These num­bers
were not adjusted for regional costs of living.
The IEEE-USA survey also looked beyond
dollar figures to consider engineers’ general
­attitudes toward their work—and found that
­overall job satisfaction fell in 2021. The biggest
­­drop-offs involved compensation and advance-
ment opportunities. But passion for the work
itself still runs strong: Satisfaction with the tech-
SOURCE: IEEE-USA nical ­challenge of engineering jobs jumped up
SALARY & BENEFITS
SURVEY REPORT 2022 significantly last year.

DECEMBER 2022  SPECTRUM.IEEE.ORG  17
 TECH TO TINKER WITH

Hi-fi, Radio, Retro,

T
his month we’re celebrating
the launch of our second PDF

and More The DIY

collection of Hands On articles,
which members can download
from IEEE Spectrum’s website and share
projects Spectrum with friends. So we thought we’d take a
look at the relative popularity of Hands
readers love On articles published over the last five
years and share the top 15 projects our
website visitors found most interesting.
Just to give a little peek behind our
analytics curtain, the measure of popu-
larity Spectrum’s editors use is “total
engaged minutes,” (TEM), which com-
BY STEPHEN CASS bines page views of articles with how long

18  SPECTRUM.IEEE.ORG  DECEMBER 2022 Illustrations by James Provost


visitors spend reading them. We use TEM
because we’re not terribly interested in
grabbing folks with a clickbait headline,
only for them to bounce out before they’ve
finished reading the first paragraph.
The first thing that jumps out is that
Spectrum readers love good quality audio,
but unlike some audiophiles, they don’t
see an exorbitant price tag or voodoo
components as a badge of honor. Far and
away our most popular article in the last
five years has been “DIY Pro Audio”
(November 2018). And a follow-up to that
article, “A Next-Gen DIY Audio Amp”
(March 2022), comes in at No. 9.
DECEMBER 2022
Here are some of your favorite Hands On projects: An inexpensive but
high-quality DIY audio amplifier (A) and its Web-enabled successor (B);
an Arduino-powered replica of the groundbreaking Altair 8800 (C); a
Raspberry Pi–powered color mechanical television (D); and (E), a home
computer built with just five digital chips that uses an old hack to
create an analog video signal.
Radio is another popular subject, The final cluster falls under the
although not radio as, say, old-school umbrella of retrotech. Sometimes it’s
hams might know it. A third of the top about a functionally identical replica of
15 articles relate to wireless and radio a legendary computer as in “The Altair,
tech, with one concerning a home-brew Reincarnated” (March 2018), but more
radio telescope (October 2019). The often about remixing new technology
other four are about exchanging data using the principles of the old to better
from distances ranging from a few tens understand the latter. “8 Bits, 6 Chips”
of meters to hundreds of kilometers. (April 2020) revisited the clever hack that
DECEMBER 2022  SPECTRUM.IEEE.ORG  19
 HANDS ON

THE 15 BEST-PERFORMING HANDS ON ARTICLES allowed early home computers with only
1 DIY Pro Audio
digital circuits to display color graphics
on analog televisions. “EZ Mechanical
2 Ham Radio Does Distant Data Networking TV” (June 2022) went even further back
in broadcast history to reveal the surpris-
3 Hacking Ham Radio for Texting ing quality that electromechanical tele-
visions were capable of.
4 Build Your Own RISC-V CPU
The remaining articles are a potpourri
5 8 Bits, 6 Chips of topics, but I think they all share that
magic element of unexpected delight that
6 Build Your Own Radio Telescope engineers are always hoping for. With
just a bit of know-how applied in the right
7 Pedaling Out of the Dark way, the world becomes a bit more inter-
esting. “Build Your Own RISC-V CPU”
8 Passive Radar With the KrakenSDR
(June 2021) showed that it was possible
9 A Next-Gen DIY Audio Amp to design modern computer architec-
tures at home without needing exotic
10 Hack This $50 Watch tools or a semiconductor fab, while
­“Pedaling Out of the Dark” (November
11 The Lego Microscope 2020) spoke to engineers’ natural skep-
ticism of marketing promises regarding
12 The Printable Motor
reliability by showing how you could use
13 Build a Two-Way Pager With LoRa a conventional bicycle as a backup to
your backup power supply.
14 EZ Mechanical TV Are there trends in DIY projects you
From 60 Hands On
projects, these are think we’re missing? Drop me a line at
15 The Altair, Reincarnated the top 25 percent. cass.s@ieee.org!

IEEE Spectrum’s ROBOTS site

features more than 200 robots
from around the world.

• Spin, swipe and tap to make robots move.

• Read up-to-date robotics news.
• Rate robots and check their ranking. Check out
Robots.ieee.org
• View photography, videos and technical specs.
STEPHEN CASS

on your desktop,
• Play Faceoff, an interactive question game. tablet, or phone now!

20  SPECTRUM.IEEE.ORG  DECEMBER 2022
 OPINION, INSIGHT, AND ANALYSIS BY VACLAV SMIL
Numbers Don’t Lie
Waiting for
Superbatteries
They still can’t match the energy
density of liquid fuel
I
f hay must be dragged to market on an oxcart,
how far can it go before the oxen eat up all the
cargo? This, in brief, is the problem faced by any
transportation system in which the vehicle must
carry its own fuel. The key value is the density of
energy, expressed with respect to either mass or
volume.
The era of large steam-powered ocean liners
began during the latter half of the 19th century, when
wood was still the world’s dominant fuel. But no
liners fired their boilers with wood: There would
have been too little space left for passengers and
cargo. Soft wood, such as spruce or pine, packs less
than 10 megajoules per liter, whereas bituminous
coal has 2.5 times as much energy by volume and at
least twice as much by mass. By comparison, gaso-
line has 34 MJ/L and diesel about 38 MJ/L.
But in a world that aspires to leave behind all fuels
(except hydrogen or maybe ammonia) and to electrify
everything, the preferred measure of stored energy
density is watt-hours per liter. By this metric, air-dried
wood contains about 3,500 watt-hours per liter, good
steam coal around 6,500, gasoline 9,600, aviation
kerosene 10,300, and natural gas (methane) merely
9.7—less than 1/1,000 the density of kerosene.
How do batteries compare with the fuels they
are to displace? The first practical battery, Gaston
Planté’s lead-acid cell, introduced in 1859, has grad-
Lead-acid cell battery
DENSITY AS COMPARED TO
NCA lithium-ion battery
SOURCE: VACLAV SMIL
Panasonic 2170 lithium-ion battery
Amprius lithium-ion battery
0 1 2 3 4 5
Thousands of watt-hours per liter
ually improved from less than 60 Wh/L to about
90 Wh/L. The nickel-cadmium battery, invented by
Waldemar Jungner in 1899, now frequently stores
more than 150 Wh/L, and today’s best mass-man-
ufactured performers are lithium-ion batteries, the
first commercial versions of which came out in 1991.
The best energy density now commercially available
in very large quantities for lithium-ion batteries is
at 750 Wh/L, which is widely seen in electric cars.
In 2020 Panasonic promised it would reach about
850 Wh/L by 2025 (and do so without the expensive
Over the cobalt). Eventually, the company aims to reach a
past 50 1,000-Wh/L product.
Claims of new energy-density records for lithi-
years, the um-ion batteries appear regularly. In March 2021,
highest Sion Power announced an 810-Wh/L pouch cell;
energy three months later NanoGraf announced a cylindrical
density cell with 800 Wh/L. Earlier claims spoke of even loft-
of mass- ier energy densities—QuantumScape mentioned a
produced 1,000-Wh/L cell in a December 2020 claim, and Sion
batteries Power referred to a 1,400-Wh/L cell as far back as
has roughly 2018. But Sion’s cells came from a pilot production
line, not from a routine mass-scale operation, and
quintupled.
QuantumScape’s claim was based on laboratory tests
of single-layer cells, not on any commercially avail-
able multilayer products.
The real-world leader seems to be Amprius Tech-
nologies of Fremont, Calif.: In February 2022, the
company announced the first delivery of batteries
rated as high as 1,150 Wh/L, to a maker of a new
generation of high-altitude uncrewed aircraft, to be
used to relay signals. This is obviously a niche
market, orders of magnitude smaller than the poten-
tial market for electric vehicles, but it is a welcome
confirmation of continuous density gains.
There is a long way to go before batteries rival the
energy density of liquid fuels. Over the past 50 years,
the highest energy density of mass-produced batteries
has roughly quintupled, from less than 150 to more
than 700 Wh/L.
BATTERIES’ ENERGY But even if that trend continues for the next 50
years, we would still see top densities of about 3,500
OTHER ENERGY SOURCES Wh/L, no more than a third that of kerosene. The wait
for superbatteries ready to power intercontinental
flight may not be over by even 2070.
Steam coal
Gasoline
Aviation kerosene
6 7 8 9 10
DECEMBER 2022  SPECTRUM.IEEE.ORG  21
The past, present, and future
of the modern world’s most
important invention
The Transistor at 75
SEVENTY-FIVE YEARS is a long time. It’s so long that
most of us don’t remember a time before the transistor,
and long enough for many engineers to have devoted
entire careers to its use and development. In honor of
this most important of technological achievements, this
issue’s package of articles explores the transistor’s his-
torical journey and potential future.
In “The First Transistor and How it Worked” [p. 24],
Glenn Zorpette dives deep into how the point-contact
transistor came to be. Then, in “The Ultimate Transistor
Timeline” [p. 29], Stephen Cass lays out the device’s evo-
lution, from the flurry of successors to the p
­ oint-contact
transistor to the complex devices in today’s laboratories
that might one day go commercial. The transistor would
never have become so useful and so ubiquitous if the
semiconductor industry hadn’t succeeded in making it
Illustration by Lisa Sheehan
small and cheap. We try to give you a sense of that scale
in “The State of the Transistor” [p. 30].
So what’s next in transistor technology? In less than
10 years’ time, transistors could take to the third dimen-
sion, stacked atop each other, write Marko Radosavljevic
and Jack Kavalieros in “Taking Moore’s Law to New
Heights” [p. 32]. And we asked experts what the transis-
tor will be like on the 100th anniversary of its invention
in “The Transistor of 2047” [p. 38].
Meanwhile, IEEE’s celebration of the transistor’s 75th
anniversary continues. The Electron Devices Society has
been at it all year, writes Joanna Goodrich in The Institute
[p. 64], and has events planned into 2023 that you can
get involved in. So go out and celebrate the device that
made the modern world possible.
—Samuel K. Moore
DECEMBER 2022  SPECTRUM.IEEE.ORG  23
 T H E T R A N S I S T O R AT 7 5
THE FIRST
TRANSISTOR
AND HOW IT
WORKED
It emerged from
a combustible mix
of brilliance and
ambition
by Glenn Zorpette
The vacuum-tube triode wasn’t
quite 20 years old when physicists began
trying to create its successor, and the
stakes were huge. Not only had the triode
made long-distance telephony and
movie sound possible, it was driving the
entire enterprise of commercial radio,
an industry worth more than a billion
dollars in 1929. But vacuum tubes were
power-hungry and fragile. If a more
rugged, reliable, and efficient alternative
to the triode could be found, the rewards
would be immense.
The goal was a three-terminal device
made out of semiconductors that would
accept a low-power signal into an input
terminal and use it to control a large cur-
rent flowing between two other termi-
nals, thereby amplifying the original
signal. The underlying principle of such
a device would be something called the
field effect—the ability of electric fields
to modulate the electrical conductivity
of semiconductor materials. The field
effect was already well known in those
days, thanks to diodes and related
research on semiconductors.
24  SPECTRUM.IEEE.ORG  DECEMBER 2022
trating the semiconductor and modulat-
ing the flow of current. No modulation,
no signal amplification.
Sometime late in 1947 they hit on a
solution. It featured two pieces of barely
separated gold foil gently pushed by that
squiggly spring into the surface of a small
slab of germanium.
Textbooks and popular accounts alike
tend to ignore the mechanism of the
point-contact transistor in favor of
explaining how its more recent descen-
dants operate. Indeed, the current edition
of that bible of undergraduate EEs, The
Art of Electronics by Horowitz and Hill,
makes no mention of the point-contact
transistor at all, glossing over its exis-
tence by erroneously stating that the
But building such a device had
proved an insurmountable challenge to
junction transistor was a “Nobel
Prize-winning invention in 1947.” But the
some of the world’s top physicists for transistor that was invented in 1947 was
more than two decades. Patents for
­transistor-like devices had been filed
the point-contact; the junction transistor
was invented by Shockley in 1948.
starting in 1925, but the first recorded So it seems appropriate somehow
instance of a working transistor was the that the most comprehensive explana-
legendary point-contact device built at tion of the point-contact transistor is
AT&T Bell Telephone Laboratories late contained within John Bardeen’s lecture
in 1947. for that Nobel Prize, in 1956. Even so,
Though the point-contact transistor reading it gives you the sense that a few
was the most important invention of the fine details probably eluded even the
20th century, there exists, surprisingly, inventors themselves. “A lot of people
no clear, complete, and authoritative were confused by the point-contact tran-
account of how the thing actually worked. sistor,” says Thomas Misa, former direc-
Modern, more robust junction and planar tor of the Charles Babbage Institute for
transistors rely on the physics in the bulk
of a semiconductor, rather than the sur-
face effects exploited in the first transis-
tor. And relatively little attention has
been paid to this gap in scholarship.
It was an ungainly looking assem-
blage of germanium, plastic, and gold
foil, all topped by a squiggly spring. Its
inventors were a soft-spoken Midwest-
ern theoretician, John Bardeen, and a
voluble and “somewhat volatile” exper-
imentalist, Walter Brattain. Both were
working under William Shockley, a rela-
tionship that would later prove conten-
tious. In November 1947, Bardeen and
Brattain were stymied by a simple prob-
lem. In the germanium semiconductor
they were using, a surface layer of elec-
trons seemed to be blocking an applied
electric field, preventing it from pene-

the History of Science and Technology,
at the University of Minnesota.
In fact, a year after Bardeen’s lecture,
R.D. Middlebrook, a professor of electri-
cal engineering at Caltech who would go
on to do pioneering work in power elec-
AT&T ARCHIVES AND HISTORY CENTER
tronics, wrote: “Because of the three-di-
mensional nature of the device,
theoretical analysis is difficult and the
internal operation is, in fact, not yet com-
pletely understood.”
Nevertheless, and with the benefit of
75 years of semiconductor theory, here
we go. The point-contact transistor was
built around a thumb-size slab of n-type
germanium, which has an excess of neg-
atively charged electrons. This slab was
treated to produce a very thin surface
layer that was p-type, meaning it had an
excess of positive charges. These positive
charges are known as holes. They are
actually localized deficiencies of elec-
trons that move among the atoms of the
semiconductor very much as a real par-
ticle would. An electrically grounded
electrode was attached to the bottom of
this slab, creating the base of the transis-
tor. The two strips of gold foil touching
the surface formed two more electrodes,
known as the emitter and the collector.
That’s the setup. In operation, a small
positive voltage—just a fraction of a
A 1955 AT&T publicity photo shows [in
palm, from left] a photo­transistor, a
junction transistor, and a point-contact
transistor. In the cutaway photo
[opposite page] of a point-contact, two
thin conductors are visible; these
connect to the points that make
contact with a tiny slab of germanium.
One of these points is the emitter and
the other is the collector. A third
contact, the base, is attached to the
reverse side of the germanium.
volt—is applied to the emitter, while a
much larger negative voltage—4 to 40
volts—is applied to the collector, all with
reference to the grounded base. The
interface between the p-type layer and
the n-type slab created a junction just like
the one found in a diode: Essentially, the
junction is a barrier that allows current
to flow easily in only one direction,
toward lower voltage. So current could
flow from the positive emitter across the
barrier, while no current could flow
across that barrier into the collector.
Now, let’s look at what happens down
among the atoms. First, we’ll disconnect
the collector and see what happens
DECEMBER 2022  SPECTRUM.IEEE.ORG  25
 T H E T R A N S I S T O R AT 7 5
around the emitter without it. The emit-
ter injects positive charges—holes—into
the p-type layer, and they begin moving
toward the base. But they don’t make a
beeline toward it. The thin layer forces
them to spread out laterally for some dis-
tance before passing through the barrier
into the n-type slab. Think about slowly
pouring a small amount of fine powder
onto the surface of water. The powder
eventually sinks, but first it spreads out
in a rough circle.
Now we connect the collector. Even
though it can’t draw current by itself
through the barrier of the p-n junction,
its large negative voltage and pointed
shape do result in a concentrated electric
field that penetrates the germanium.
Because the collector is so close to the
emitter, and is also negatively charged,
it begins sucking up many of the holes
that are spreading out from the emitter.
This charge flow results in a concentra-
The Western Electric Type-2
point-contact transistor was the first
transistor to be manufactured in
large quantities, in 1951, at Western
Electric’s plant in Allentown, Pa. By
1960, when this photo was taken, the
plant had switched to producing
junction transistors.
26  SPECTRUM.IEEE.ORG  DECEMBER 2022
tion of holes near the p-n barrier under-
neath the collector. This concentration
effectively lowers the “height” of the
barrier that would otherwise prevent
current from flowing between the col-
lector and the base. With the barrier
lowered, current starts flowing from the
base into the collector—much more cur-
rent than what the emitter is putting into
the transistor.
The amount of current depends on
the height of the barrier. Small decreases
or increases in the emitter’s voltage cause
the barrier to fluctuate up and down,
respectively. Thus very small changes in
the emitter current control very large
changes at the collector, so voilà! Ampli-
fication. (EEs will notice that the func-
tions of base and emitter are reversed
compared with those in later transistors,
where the base, not the emitter, controls
the response of the transistor.)
Ungainly and fragile though it was, it
was a semiconductor amplifier, and its
progeny would change the world. And
its inventors knew it. The fateful day was
16 December 1947, when Brattain hit on
the idea of using a plastic triangle belted
by a strip of gold foil, with that tiny slit
separating the emitter and collector
contacts. This configuration gave reli-
able power gain, and the duo knew then
that they had succeeded. In his carpool
home that night, Brattain told his com-
panions he’d just done “the most
important experiment that I’d ever do
in my life” and swore them to secrecy.
The taciturn Bardeen, too, couldn’t
resist sharing the news. As his wife,
Jane, prepared dinner that night, he
reportedly said, simply, “We discovered
something today.” With their children
scampering around the kitchen, she
responded, “That’s nice, dear.”
It was a transistor, at last, but it was
pretty rickety. The inventors later hit on
the idea of electrically forming the col-
Textbooks and popular accounts alike
tend to ignore the mechanism of the
point-contact transistor in favor of
explaining how its more recent
descendants operate.
lector by passing large currents through
it during the transistor’s manufacturing.
This technique enabled them to get
somewhat larger current flows that
weren’t so tightly confined within the
surface layer. The electrical forming was
a bit hit-or-miss, though. “They would
just throw out the ones that didn’t work,”
Misa notes.
Nevertheless, point-contact transis-
tors went into production at many com-
panies, under license to AT&T, and, in
1951, at AT&T’s own manufacturing arm,
Western Electric. They were used in
hearing aids, oscillators, telephone-rout-
ing gear, an experimental TV receiver
built at RCA, and in the Tradic, the first
airborne digital computer, among other
systems. In fact, point-contact transis-
tors remained in production until 1966,
in part due to their superior speed com-
pared with the alternatives.
The Bell Labs group wasn’t alone in
its successful pursuit of a transistor. In
Aulnay-sous-Bois, a suburb northeast of
Paris, two German physicists, Herbert
Mataré and Heinrich Welker, were also
trying to build a three-terminal semicon-
ductor amplifier. Working for a French
subsidiary of Westinghouse, they were
following up on very intriguing observa-
tions Mataré had made while developing
germanium and silicon rectifiers for the
German military in 1944. The two suc-
ceeded in creating a reliable point-con-
tact transistor in June 1948.
They were astounded, a week or so
later, when Bell Labs finally revealed the
news of its own transistor, at a press con-
ference on 30 June 1948. Though they
were developed completely inde-
pendently, and in secret, the two devices
were more or less identical.
Here the story of the transistor takes
a weird turn, breathtaking in its brilliance
and also disturbing in its details. Bar-
deen’s and Brattain’s boss, William
AT&T ARCHIVES AND HISTORY CENTER
 RCA was one of the first licensees of the Bell Labs
transistors. In 1953, the company used point-contact
transistors to build what the company called the first
“all-transistor” television (yes, it had a cathode-ray
tube). The approximately 22 transistors used in the
experimental set included the TA153 [color photo: top
row, second from left], the TA165 [top, far right], the
TA156 [bottom row, middle] and the TA172 [bottom,
right]. The man in the black-and-white photo is RCA
engineer Gerald Herzog, whose team designed and built
the television.

Shockley, was furious that his name was
not included with Bardeen’s and Brat-
tain’s on the original patent application
for the transistor. He was convinced that
Bardeen and Brattain had merely spun
his theories about using fields in semi-
conductors into their working device,
and had failed to give him sufficient
credit. Yet in 1945, Shockley had built a
LEFT: TRANSISTOR MUSEUM JERRY HERZOG ORAL HISTORY;
RIGHT: TRANSISTOR MUSEUM JONATHAN HOPPE COLLECTION
point-contact transistor and reign as the
dominant transistor until the late 1970s.
The BJT was based on Shockley’s
conviction that charge could, and should,
flow through the bulk semiconductors
rather than through a thin layer on their
surface. The device consisted of three
semiconductor layers, like a sandwich:
an emitter, a base in the middle, and a
electrons flowing in combines with these
holes and are removed from circulation,
while the vast majority (more than 97
percent) of electrons keep flowing
through the thin base and into the col-
lector, setting up a strong flow of current
[see illustration, next page].
But those few electrons that do com-
bine with holes must be drained from the

transistor based on those very theories,
and it hadn’t worked.
At the end of December, barely two
weeks after the initial success of the
point-contact transistor, Shockley trav-
eled to Chicago for the annual meeting
of the American Physical Society. On
New Year’s Eve, holed up in his hotel
room and fueled by a potent mix of jeal-
ousy and indignation, he began designing
a transistor of his own. In three days he
scribbled some 30 pages of notes. By the
end of the month, he had the basic design
for what would become known as the
bipolar junction transistor, or BJT, which
would eventually supersede the
collector. They were alternately doped,
so there were two versions: n-type/p-
type/n-type, called “NPN,” and p-type/n-
type/p-type, called “PNP.”
The BJT relies on essentially the same
principles as the point-contact, but it
uses two p-n junctions instead of one.
When used as an amplifier, a positive
voltage applied to the base allows a small
current to flow between it and the emit-
ter, which in turn controls a large current
between the collector and emitter.
Consider an NPN device. The base is
p-type, so it has excess holes. But it is very
thin and lightly doped, so there are rela-
tively few holes. A tiny fraction of the
base in order to maintain its p-type
nature and the strong flow of current
through it. That removal of the “trapped”
electrons is accomplished by a relatively
small flow of current out of the base. That
trickle of current enables the much stron-
ger flow of current into the collector. So,
in effect, the small base current is con-
trolling the higher-power collector
circuit.
Electric fields come into play, but they
do not modulate the current flow, which
the early theoreticians thought would
have to happen for such a device to func-
tion. Here’s the gist: Both of the p-n junc-
tions in a BJT are straddled by depletion

DECEMBER 2022  SPECTRUM.IEEE.ORG  27
 T H E T R A N S I S T O R AT 7 5

Point-Contact Transistor
Depletion regions
Point-Contact Transistor The BJT was more rugged and reliable
Depletion regions
Emitter Collector
than the point-contact transistor, and
p-type those features primed it for greatness.
Holes
surface Emitter Collector But it took a while for that to become
layer Electrons
Emitter p-type Holes obvious. The BJT was the technology
n-type
surface
current used to make integrated circuits, from
layer Electrons
Emitter n-type the first ones in the early 1960s all the
current
Emitter Collector way until the late 1970s, when
current current ­metal-oxide-semiconductor field-effect
Emitter Collector Collector transistors (MOSFETs) took over. In fact,
current current current
it was these field-effect transistors, first
Collector the junction field-effect transistor and
current
Base
then MOSFETs, that finally realized the
Base
decades-old dream of a three-terminal
semiconductor device whose operation
was based on the field effect—Shockley’s
original idea.
Such a glorious future could scarcely
Bipolar Junction Transistor be imagined in the early 1950s, when
Depletion regions AT&T and others were struggling to come
Bipolar Junction Transistor up with practical and efficient ways to
Depletion regions manufacture the new BJTs. Shockley
n-type p-type n-type himself went on to literally put the silicon
into Silicon Valley. He moved to Palo Alto
n-type p-type n-type
Electrons and in 1956 founded a company that led
Emitter Collector
the switch from germanium to silicon as
Electrons Recombination the electronic semiconductor of choice.
Emitter Collector Employees from his company would go
Recombination on to found Fairchild Semiconductor,
and then Intel.
Holes Later in his life, after losing his com-
Emitter Base Collector
pany because of his terrible management,
current Holes current he became a professor at Stanford and
Base current
Emitter Base Collector began promulgating ungrounded and
current current
Base current unhinged theories about race, genetics,
and intelligence. In 1951 Bardeen left Bell
Labs to become a professor at the Uni-
The very first transistors were a type accomplishes amplification using much versity of Illinois at Urbana-Champaign,
known as point contact, because they the same principles but with two where he won a second Nobel Prize for
relied on metal contacts touching the semiconductor interfaces, or junctions, physics, for a theory of superconductiv-
surface of a semiconductor. They ramped rather than one. As with the point-contact ity. (He is the only person to have won
up output current—labeled “Collector transistor, an applied voltage overcomes
current” in the top diagram—by using an a barrier and enables current flow that is
two Nobel Prizes in physics.) Brattain
applied voltage to overcome a barrier to modulated by a smaller input current. In stayed at Bell Labs until 1967, when he
charge flow. Small changes to the input, particular, the semiconductor junctions joined the faculty at Whitman College, in
SOURCE: THE SEMICONDUCTOR MUSEUM, WIKIPEDIA

or “emitter,” current modulate this barrier,
thus controlling the output current.
The bipolar junction transistor
regions, in which electrons and holes
combine and there are relatively few
mobile charge carriers. Voltage applied
across the junctions sets up electric fields
at each, which push charges across those
regions. These fields enable electrons to
flow all the way from the emitter, across
the base, and into the collector.
are straddled by depletion regions, across
which the charge carriers diffuse under
the influence of an electric field.
In the BJT, “the applied electric
fields affect the carrier density, but
because that effect is exponential, it
only takes a little bit to create a lot of
diffusion current,” explains Ioannis
“John” Kymissis, chair of the depart-
ment of electrical engineering at
Columbia University.
Walla Walla, Wash.
Shockley died a largely friendless
pariah in 1989. But his transistor would
change the world, though it was still not
clear as late as 1953 that the BJT would
be the future. In an interview that year,
Donald G. Fink, who would go on to help
establish the IEEE a decade later, mused,
“Is it a pimpled adolescent, now awk-
ward, but promising future vigor? Or has
it arrived at maturity, full of languor, sur-
rounded by disappointments?”
It was the former, and all of our lives
are so much the better because of it.

28  SPECTRUM.IEEE.ORG  DECEMBER 2022 Illustrations by Chris Philpot

You just can’t reinvent the transistor enough times.

Point-contact transistor

Grown-junction transistor

Tetrode point-contact transistor

Alloy-junction transistor

Junction field-effect transistor (JFET)

Surface-barrier transistor

Drift-field transistor

Modern BJT (diffused junction bipolar junction transistor)

Heterojunction bipolar transistor (HBT)

MOSFET (metal-oxide-semiconductor field-effect transistor)

Thin-film transistor

Schottky transistor

MESFET (metal-semiconductor field-effect transistor)

Floating-gate MOSFET

IGBT (insulated-gate bipolar transistor)

HEMT (high-electron-mobility transistor)

Flash memory cell

FinFET

Carbon nanotube transistor

Single-atom transistor

Tunnel FET

Vertical 3D NAND (modern flash cell)

Graphene transistor
Invented Commercialized Nanosheet (gate-all-around transistor)

Complementary FET

1950 1960 1970 1980 1990 2000 2010 2020

THE ULTIMATE
TRANSISTOR TIMELINE
The transistor’s How it Worked,” p. 24). Since then, engi-
neers have reinvented the transistor over
amazing evolution and over again, raiding condensed-matter
from point contacts physics for anything that might offer even
the possibility of turning a small signal into
to quantum tunnels a larger one.
by Stephen Cass But physics is one thing; mass pro-
duction is another. This timeline shows
the time elapsed between the invention
Even as the initial sales receipts for of a number of transistor types and the
the first transistors to hit the market were year they became commercially available.
being tallied up in 1948, the next genera- To be honest, finding the latter set of
tion of transistors had already been dates was often a murky business, and
invented (see “The First Transistor and we welcome corrections. But it’s clear
that the initial breakneck pace of inno-
vation seems to have slowed from 1970
to 2000, likely because these were the
golden years for Moore’s Law, when scal-
ing down the dimensions of the existing
metal-oxide-semiconductor field-effect
transistors (MOSFETs) led to computers
that doubled in speed every couple of
years for the same money. Then, when
the inevitable end of this exponential
improvement loomed on the horizon, a
renaissance in transistor invention
seems to have begun and continues to
this day.

DECEMBER 2022  SPECTRUM.IEEE.ORG  29
 T H E T R A N S I S T O R AT 7 5

The State of the Transistor The most obvious change in transistor technology in
the last 75 years has been just how many we can make.

10 trillion 10 billion 1 billion

US $ per trillion transistors (log scale)

Metal pitch, nanometers

Trillions of transistors (log scale)

Price Transistors sold

Contacted

Density (log scale)

gate pitch,
nanometers

Transistor density,
devices per square
millimeter
100 100
1955 2022 1970

HUGE VOLUME, SMALL PRICE TRANSISTOR DENSITY

Price per trillion transistors, US $; transistors sold per year, trillions
In 1947, there was only one transistor.
According to TechInsight’s forecast, the
semiconductor industry is on track to
produce almost 2 billion trillion (1021) devices
this year. That’s more transistors than were
cumulatively made in all the years prior to
2017. Behind that barely conceivable number
Maximum devices per mm2,
millions; critical dimensions,
nanometers
is the continued reduction in the price of a
transistor, as engineers have learned to
integrate more and more of them into the Scaling down transistors in the 2D space of
same area of silicon. the plane of the silicon has been a smashing
success: Transistor density in logic circuits
has increased more than 600,000-fold
since 1971. Reducing transistor size requires

WHAT TRANSISTORS HAVE BECOME  Besides making them tiny and numerous, engineers have devoted their

EPHEMERAL: FAST: The first FLAT: Today’s FLEXIBLE: The

Researchers in
Illinois developed
circuits that
dissolve in the
body using a combination of
ultrathin silicon membranes,
magnesium conductors, and
magnesium oxide insulators.
Five minutes in water was
enough to turn the first
generation to mush. But
recently researchers used a
more durable version to make
temporary cardiac pacemakers
that release an anti-inflamma-
tory drug as they disappear.
transistors were
made for radio
frequencies, but
there are now
devices that operate about a
billion times faster. Engineers in
South Korea and Japan
reported the invention of an
indium gallium arsenide
high-electron-mobility
transistor, or HEMT, that
reached a maximum frequency
of 738 gigahertz. Seeking raw
speed, engineers at Northrop
Grumman made a HEMT that
passed 1 terahertz.
(and yesterday’s) world is not flat,
transistors and neither are
depend on the the places
semiconducting transistors need
properties of bulk (3D) to operate. Using indium
materials. Tomorrow’s devices gallium arsenide, engineers in
might rely on 2D semiconduc- South Korea recently made
tors, such as molybdenum high-performance logic
disulfide and tungsten transistors on plastic that
disulfide. These transistors hardly suffered when bent
might be built in the intercon- around a radius of just 4
nect layers above a processor’s millimeters. And engineers in
silicon, researchers say. So 2D Illinois and England have made
semiconductors could help lead microcontrollers that are both
to 3D processors. affordable and bendable.

30  SPECTRUM.IEEE.ORG  DECEMBER 2022 Illustrations by Chris Philpot

 In 75 years, it’s become tiny, mighty, ubiquitous, and just plain weird

Reducing the size of the device has been a titanic effort and a the only feature engineers have been improving.
fantastically successful one, as these charts show. But size isn’t —Samuel K. Moore & David Schneider

10,000 1 trillion

2020
Apple M1: The company’s first
homegrown laptop CPU.

1990

Transistors in CPUs (log scale)

POWER1:
IBM’s
Nanometers (log scale)

multichip CPU
Extreme- 1971 was the first to
ultraviolet Intel 4004: implement the
lithography The first POWER
introduced. microprocessor. architecture.

1979
Intel 8088: The
First x86 processor 1993 Pentium: Intel’s fifth-generation x86 processor
multitier behind the and the first with a “superscalar” architecture, one
CMOS original IBM PC. that executes at least two instructions per clock cycle.
chips?

193-nanometer
immersion
lithography
1985 ARM 1: The original implementation of the ARM architecture.
introduced. 10 1,000
2004 2018 2031 1970 2025

TRANSISTORS PER PROCESSOR

using shorter wavelengths of light, such as Number of transistors in CPUs
extreme ultraviolet, and other lithography
tricks to shrink the space between transistor
gates and between metal interconnects. Perhaps the crowning achievement of all SOURCES: TECHINSIGHTS [LEFT]; IEEE
INTERNATIONAL ROADMAP FOR DEVICES AND
Going forward, it’s the third dimension, this effort is the ability to integrate millions, SYSTEMS, STANFORD NANOELECTRONICS LAB;
where transistors will be built atop one even billions, of transistors into some of the H.-S. P. WONG, ET AL., DATA ACCESSED
17 OCTOBER 2022 [CENTER]; WIKIPEDIA
another, that counts. This trend is more than most complex systems on the planet: CPUs. CONTRIBUTORS, “TRANSISTOR COUNT,”
a decade old in flash memory, but it’s still in Here’s a look at some of the high points WIKIPEDIA, THE FREE ENCYCLOPEDIA,
ACCESSED 28 SEPTEMBER 2022 [RIGHT]
the future for logic (see “Taking Moore’s along the way.
Law to New Heights,” p. 32).

efforts to enhancing the device’s other qualities. Here is a small sampling of what transistors have become in the last 75 years.

INVISIBLE: When MNEMONIC: TALENTED: In TOUGH: Some

you need to hide
your computing in
plain sight, turn to
transparent
transistors. Researchers in
Fuzhou, China, recently made a
see-through analogue of flash
memory using organic
semiconductor thin-film
transistors. And researchers in
Japan and Malaysia produced
transparent diamond devices
capable of handling more than
1,000 volts.
NAND flash 2018, engineers in
memory cells can Canada used an
store multiple bits algorithm to
in a single device. generate all the
Those on the market today possible unique and functional
store either 3 or 4 bits each. elementary circuits that can be
Researchers at Kioxia Corp. made using just two
built a modified NAND flash cell ­metal-oxide field-effect
and dunked it in 77 kelvin liquid transistors. The number of
nitrogen. A single superchilled circuits totaled an astounding
transistor could store up to 582. Increasing the scope to
7 bits of data, or 128 different three transistors netted 56,280
values. circuits, including several
amplifiers previously unknown
to engineering.
transistors can
take otherworldly
punishment.
NASA Glenn
Research Center built
200-transistor silicon carbide
ICs and operated them for 60
days in a chamber that
simulates the environment on
the surface of Venus—460 °C
heat, a planetary-probe-crush-
ing 9.3 megapascals of
pressure, and the hellish
planet’s corrosive atmosphere.

DECEMBER 2022  SPECTRUM.IEEE.ORG  31
 TAKING
T H E T R A N S I S T O R AT 7 5

MOORE’S
LAW TO NEW
HEIGHTS
When transistors can’t
get any smaller, the only
direction is up
by Marko Radosavljevic
& Jack Kavalieros

32  SPECTRUM.IEEE.ORG  DECEMBER 2022
 Illustrations by
Emily Cooper

DECEMBER 2022  SPECTRUM.IEEE.ORG  33
 T H E T R A N S I S T O R AT 7 5
Perhaps the most far-reaching technological
achievement over the last 50 years has been the steady
march toward ever smaller transistors, fitting them more
tightly together, and reducing their power consumption.
And yet, ever since the two of us started our careers at
Intel more than 20 years ago, we’ve been hearing the
alarms that the descent into the infinitesimal was about
to end. Yet year after year, brilliant new innovations
continue to propel the semiconductor industry further.
Along this journey, we engineers had to change the
transistor’s architecture as we continued to scale down
area and power consumption while boosting
performance. The “planar” transistor designs that took
us through the last half of the 20th century gave way to
3D fin-shaped devices by the first half of the 2010s. Now,
these too have an end date in sight, with a new gate-all-
around (GAA) structure rolling into production soon.
But we have to look even further ahead because our
ability to scale down even this new transistor
architecture, which we call RibbonFET, has its limits.
So where will we turn for future scaling? We will
continue to look to the third dimension. We’ve created
experimental devices that stack atop each other,
delivering logic that is 30 to 50 percent smaller. Crucially,
the top and bottom devices are of the two
complementary types, NMOS and PMOS, that are the
foundation of all the logic circuits of the last several
decades. We believe this 3D-stacked complementary
metal-oxide semiconductor (CMOS), or CFET
(complementary field-effect transistor), will be the key to
extending Moore’s Law into the next decade.
TRANSISTOR
EVOLUTION
The shift from a planar
transistor architecture
[left] to the FinFET
[right] provided greater
control of the channel
[covered by blue block],
resulting in a reduction
in power consumption
of 50 percent and an
increase in perfor-
mance of 37 percent.
34  SPECTRUM.IEEE.ORG  DECEMBER 2022
THE EVOLUTION
OF THE TRANSISTOR
Continuous innovation is an
essential underpinning of Moore’s Law,
but each improvement comes with trade-
offs. To understand these trade-offs and
how they’re leading us inevitably toward
3D-stacked CMOS, you need a bit of
background on transistor operation.
Every metal-oxide-semiconductor
field-effect transistor, or MOSFET, has the
same set of basic parts: the gate stack, the
channel region, the source, and the drain.
The source and drain are chemically
doped to make them both either rich in
mobile electrons (n-type) or deficient in
them (p-type). The channel region has the
opposite doping to the source and drain.
In the planar version in use in
advanced microprocessors up to 2011,
the MOSFET’s gate stack is situated just
above the channel region and is designed
to project an electric field into the chan-
nel region. Applying a large enough volt-
age to the gate (relative to the source)
creates a layer of mobile charge carriers
in the channel region that allows current
to flow between the source and drain.
As we scaled down the classic planar
transistors, what device physicists call
short-channel effects took center stage.
Basically, the distance between the
source and drain became so small that
current would leak across the channel
when it wasn’t supposed to, because the
gate electrode struggled to deplete the
channel of charge carriers. To address
this, the industry moved to an entirely
different transistor architecture called a
FinFET. It wrapped the gate around the
channel on three sides to provide better
electrostatic control.
SOURCE: INTEL
 In the RibbonFET, the gate wraps around the transistor
channel region to enhance control of charge carriers. The
new structure also enables better performance and more
refined optimization.
Intel introduced its FinFETs in 2011,
at the 22-nanometer node, with the
third-generation Core processor, and the
device architecture has been the work-
horse of Moore’s Law ever since. With
FinFETs, we could operate at a lower
voltage and still have less leakage, reduc-
ing power consumption by some 50 per-
cent at the same performance level as the
previous-generation planar architecture.
FinFETs also switched faster, boosting
performance by 37 percent. And because
conduction occurs on both vertical sides
of the “fin,” the device can drive more
current through a given area of silicon
than can a planar device, which only con-
ducts along one surface.
However, we did lose something in
moving to FinFETs. In planar devices, the
width of a transistor was defined by
lithography, and therefore it is a highly
flexible parameter. But in FinFETs, the
transistor width comes in the form of
discrete increments—adding one fin at a
time–a characteristic often referred to as
fin quantization. As flexible as the
FinFET may be, fin quantization remains
a significant design constraint. The
design rules around it and the desire to
add more fins to boost performance
increase the overall area of logic cells and
complicate the stack of interconnects
SOURCE: INTEL
that turn individual transistors into com-
plete logic circuits. It also increases the
transistor’s capacitance, thereby sapping
some of its switching speed. So, while the
3D-stacked CMOS puts a PMOS device on top of an NMOS
device in the same footprint a single RibbonFET would
occupy. The NMOS and PMOS gates use different metals.
FinFET has served us well as the indus- providing even tighter control of charge
try’s workhorse, a new, more refined carriers within channels that are now
approach is needed. And it’s that formed by nanometer-scale ribbons of
approach that led us to the 3D transistors silicon. With these nanoribbons (also
we’re introducing soon. called nanosheets), we can again vary
This advance, the RibbonFET, is our the width of a transistor as needed using
first new transistor architecture since lithography.
the FinFET’s debut 11 years ago. In it, With the quantization constraint
the gate fully surrounds the channel, removed, we can produce the appropri-
ately sized width for the application. That
lets us balance power, performance, and
cost. What’s more, with the ribbons
stacked and operating in parallel, the
Taking device can drive more current, boosting
performance without increasing the area
advantage of of the device.
the potential We see RibbonFETs as the best option
for higher performance at reasonable
benefits of 3D power, and we will be introducing them
in 2024 along with other innovations,
stacking means such as PowerVia, our version of back-
solving a side power delivery, with the Intel 20A
fabrication process.
number of STACKED CMOS
process One commonality of planar, FinFET,
integration and RibbonFET transistors is that they
all use CMOS technology, which, as men-
challenges, tioned, consists of n-type (NMOS) and
p-type (PMOS) transistors. CMOS logic
some of which became mainstream in the 1980s because
will stretch the it draws significantly less current than do
the alternative technologies, notably
limits of CMOS NMOS-only circuits. Less current also
led to greater operating frequencies and
fabrication. higher transistor densities.
DECEMBER 2022  SPECTRUM.IEEE.ORG  35
 T H E T R A N S I S T O R AT 7 5
To date, all CMOS technologies place
the standard NMOS and PMOS transis-
tor pair side by side. But in a keynote at
the IEEE International Electron Devices
Meeting (IEDM) in 2019, we introduced
the concept of a 3D-stacked transistor
that places the NMOS transistor on top
of the PMOS transistor. The following
year, at IEDM 2020, we presented the
design for the first logic circuit using this
3D technique, an inverter. Combined
with appropriate interconnects, the
3D-stacked CMOS approach effectively
cuts the inverter footprint in half, dou-
bling the area density and further push-
ing the limits of Moore’s Law.
Taking advantage of the potential
benefits of 3D stacking means solving a
number of process integration chal-
lenges, some of which will stretch the
limits of CMOS fabrication.
We built the 3D-stacked CMOS
inverter using what is known as a self-
aligned process, in which both transistors
are built in one manufacturing step. This
means constructing both n-type and
p-type sources and drains by epitaxy—
crystal deposition—and adding different
metal gates for the two transistors. By
combining the source-drain and dual-­
metal-gate processes, we are able to create
different conductive types of silicon
nanoribbons (p-type and n-type) to make
up the stacked CMOS transistor pairs. It
also allows us to adjust the device’s
threshold voltage—the voltage at which
a transistor begins to switch—separately
for the top and bottom nanoribbons.
In CMOS logic, NMOS and PMOS devices usually
sit side by side on chips. An early prototype has
NMOS devices stacked on top of PMOS devices,
compressing circuit sizes.
36  SPECTRUM.IEEE.ORG  DECEMBER 2022
2D CMOS NR Inverter
By stacking NMOS on top of PMOS transistors, 3D stacking effectively doubles
CMOS transistor density per square millimeter, though the real density depends
on the complexity of the logic cell involved. The inverter cells are shown from
above indicating source and drain interconnects [red], gate interconnects [blue],
and vertical connections [green].
How do we do all that? The self-
aligned 3D CMOS fabrication begins
with a silicon wafer. On this wafer, we
deposit repeating layers of silicon and
silicon germanium, a structure called a
superlattice. We then use lithographic
patterning to cut away parts of the
superlattice and leave a finlike struc-
ture. The superlattice crystal provides
a strong support structure for what
comes later.
Next, we deposit a block of “dummy”
polycrystalline silicon atop the part of the
superlattice where the
device gates will go, pro-
tecting them from the next
step in the procedure. That
step, called the vertically
stacked dual source/drain
process, grows phospho-
rous-doped silicon on both
ends of the top nanorib-
bons (the future NMOS
device) while also selec-
tively growing boron-
doped silicon germanium
on the bottom nanorib-
bons (the future PMOS
device). After this, we
deposit dielectric around
the sources and drains to
electrically isolate them
from one another. The
3D-Stacked CMOS NR Inverter
latter step requires that we then polish the
wafer down to perfect flatness.
Finally, we construct the gate. First,
we remove that dummy gate we’d put in
place earlier, exposing the silicon
nanoribbons. We next etch away only the
silicon germanium, releasing a stack of
parallel silicon nanoribbons, which will
be the channel regions of the transistors.
We then coat the nanoribbons on all
sides with a vanishingly thin layer of an
insulator that has a high dielectric con-
stant. The nanoribbon channels are so
small and positioned in such a way that
we can’t effectively dope them chemically
as we would with a planar transistor.
Instead, we use a property of the metal
gates called the work function to impart
the same effect. We surround the bottom
nanoribbons with one metal to make a
p-doped channel and the top ones with
another to form an n-doped channel.
TOP: SOURCE: INTEL; BOTTOM: INTEL
Thus, the gate stacks are finished off and
the two transistors are complete.
The process might seem complex, but
it’s better than the alternative—a technol-
ogy called sequential 3D-stacked CMOS.
With that method, the NMOS devices and
the PMOS devices are built on separate
wafers, the two are bonded, and the
PMOS layer is transferred to the NMOS
wafer. In comparison, the self-aligned 3D
process takes fewer manufacturing steps
 and keeps a tighter rein on manufacturing cost, something we
demonstrated in research and reported at IEDM 2019.
Importantly, the self-aligned method also circumvents the
problem of misalignment that can occur when bonding two
wafers. Still, sequential 3D stacking is being explored to facilitate
integration of silicon with nonsilicon channel materials, such as
germanium and III-V semiconductor materials. These approaches
and materials may become relevant as we look to tightly integrate
optoelectronics and other functions on a single chip.
The new self-aligned CMOS process, and the 3D-stacked
CMOS it creates, work well and appear to have substantial room
for further miniaturization. At this early stage, that’s highly
encouraging. Devices having a gate length of 75 nm demon-
strated both the low leakage that comes with excellent device
scalability and a high on-state current. Another promising sign:
We’ve made wafers where the smallest distance between two
sets of stacked devices is only 55 nm. While the device perfor-
mance results we achieved are not records in and of themselves,
they do compare well with individual nonstacked control
devices built on the same wafer with the same processing.
In parallel with the process integration and experimental
work, we have many ongoing theoretical, simulation, and design
studies underway looking to provide insight into how best to
use 3D CMOS. Through these, we’ve found some of the key
considerations in the design of our transistors. Notably, we now
know that we need to optimize the vertical spacing between
the NMOS and PMOS—if it’s too short it will increase parasitic An edge-on view of the 3D stacked inverter shows how
capacitance, and if it’s too long it will increase the resistance complicated its connections are.
of the interconnects between the two devices. Either extreme
results in slower circuits that consume more power.
Many design studies, such as one by TEL Research Center
America presented at IEDM 2021, focus on providing all the
necessary interconnects in the 3D CMOS’s limited space and
doing so without significantly increasing the area of the logic
cells they make up. The TEL research showed that there are
many opportunities for innovation in finding the best inter-
connect options. That research also highlights that 3D-stacked
CMOS will need to have interconnects both above and below
the devices. This scheme, called buried power rails, takes the
interconnects that provide power to logic cells but don’t carry
data and removes them to the silicon below the transistors.
Intel’s PowerVIA technology, which does just that and is sched-
uled for introduction in 2024, will therefore play a key role in
making 3D-stacked CMOS a commercial reality.

THE FUTURE OF MOORE’S LAW

With RibbonFETs and 3D CMOS, we have a clear path to
extend Moore’s Law beyond 2024. In a 2005 interview in which
he was asked to reflect on what became of his law, Gordon
Moore admitted to being “periodically amazed at how we’re
able to make progress. Several times along the way, I thought
we reached the end of the line, things taper off, and our creative
engineers come up with ways around them.”
Making all the needed connections to 3D-stacked CMOS is a
With the move to FinFETs, the ensuing optimizations, and challenge. Power connections will need to be made from
SOURCE: INTEL

now the development of RibbonFETs and eventually below the device stack. In this design, the NMOS device [top]
3D-stacked CMOS, supported by the myriad packaging and PMOS device [bottom] have separate source/drain
enhancements around them, we’d like to think Mr. Moore will contacts, but both devices have a gate in common.
be amazed yet again.

DECEMBER 2022  SPECTRUM.IEEE.ORG  37
 T H E T R A N S I S T O R AT 7 5
THE
TRANSISTOR
OF 2047
What will the device
be like on its 100th
anniversary?
by Samuel K. Moore
The 100th anniversary of the
invention of the transistor will happen in
2047. What will transistors be like then?
Will they even be the critical computing
element they are today? IEEE Spectrum
asked experts for their predictions.
WHAT WILL TRANSISTORS BE LIKE
IN 2047?
Expect transistors to be even more
varied than they are now, says one
researcher. Just as processors have
evolved from CPUs to include GPUs,
network processors, AI accelerators,
and other specialized computing chips,
transistors will evolve to fit a variety of
purposes. “Device technology will
become application domain–specific in
the same way that computing architec-
ture has become application domain–
specific,” says H.-S. Philip Wong, an
IEEE Fellow, professor of electrical
engineering at Stanford University, and
former vice president of corporate
research at TSMC.
Despite the variety, the fundamental
operating principle—the field effect
that switches transistors on and off—
will likely remain the same, suggests
Suman Datta, an IEEE Fellow, professor
of electrical and computer engineering
at Georgia Tech, and director of the
38  SPECTRUM.IEEE.ORG  DECEMBER 2022
multi-university nanotech research
center ASCENT. This device will likely
have minimum critical dimensions of
1 nanometer or less, enabling device
densities of 10 trillion per square cen-
timeter, says Tsu-Jae King Liu, an IEEE
Fellow, dean of the college of engineer-
ing at the University of California,
Berkeley, and a member of Intel’s board
of directors.
Experts seem to agree that the transis-
tor of 2047 will need new materials and
probably a stacked or 3D architecture,
expanding on the planned complemen-
tary field-effect transistor (CFET, or
3D-stacked CMOS). [For more on the
CFET, see “Taking Moore’s Law to New
Heights,” in this issue.] And the transistor
channel, which now runs parallel to the
plane of the silicon, may need to become
vertical in order to continue to increase in
density, says Datta.
AMD senior fellow Richard Schultz,
suggests that the main aim in developing
these new devices will be power. “The
focus will be on reducing power and the
need for advanced cooling solutions,” he
says. “Significant focus on devices that
work at lower voltages is required.”
WILL TRANSISTORS STILL BE THE
HEART OF MOST COMPUTING?
It’s hard to imagine a world where com-
puting is not done with transistors, but,
of course, vacuum tubes were once the
digital switch of choice. Startup funding
for quantum computing, which does not
directly rely on transistors, reached
US  $1.4  billion in 2021, according to
Mc­Kinsey & Co.
But advances in quantum computing
won’t happen fast enough to challenge
the transistor by 2047, experts in electron
devices say. “Transistors will remain the
most important computing element,”
says Sayeef Salahuddin, an IEEE Fellow
and professor of electrical engineering
and computer science at the University
of California, Berkeley. “Currently, even
with an ideal quantum computer, the
potential areas of application seem to be
rather limited compared to classical
computers.”
Sri Samavedam, senior vice president
of CMOS technologies at the European
chip R&D center Imec, agrees. “Transis-
tors will still be very important comput-
ing elements for a majority of the
general-purpose compute applications,”
he says. “One cannot ignore the efficien-
cies realized from decades of continuous
optimization of transistors.”
HAS THE TRANSISTOR OF 2047
ALREADY BEEN INVENTED?
Twenty-five years is a long time, but in the
world of semiconductor R&D, it’s not that
long. [See “The Ultimate Transistor Time-
line,” in this issue.] “In this industry, it
usually takes about 20 years from [demon-
The luminaries who dared predict the future
of the transistor for IEEE Spectrum are
[clockwise from left] Gabriel Loh, Sri
Samavedam, Sayeef Salahuddin, Richard
Schultz, Suman Datta, Tsu-Jae King Liu, and
H.-S. Philip Wong.
strating a concept] to introduction into
manufacturing,” says Samavedam. “It is
safe to assume that the transistor or switch
architectures of 2047 have already been
demonstrated on a lab scale” even if the
materials involved won’t be exactly the
same. King Liu, who demonstrated the
modern FinFET about 25 years ago with
colleagues at Berkeley, agrees.
But the idea that the transistor of 2047
is already sitting in a lab somewhere isn’t
universally shared. Salahuddin, for one,
doesn’t think it’s been invented yet. “But
just like the FinFET in the 1990s, it is
possible to make a reasonable prediction
for the geometric structure” of future
transistors, he says.
AMD’s Schultz says you can glimpse
this structure in proposed 3D-stacked
devices made of 2D semiconductors or
carbon-based semiconductors. “Device
materials that have not yet been invented
could also be in scope in this time frame,”
he adds.
Photo-illustration by Gluekit
WILL SILICON STILL BE THE ACTIVE
PART OF MOST TRANSISTORS IN
2047?
Experts say that the heart of most devices,
the transistor channel region, will still be
silicon, or possibly silicon-germanium—
which is already making inroads—or ger-
manium. But in 2047 many chips may use
semiconductors that are considered
exotic today. These could include oxide
semiconductors like indium gallium zinc
oxide; 2D semiconductors, such as the
metal dichalcogenide tungsten disulfide;
and one-dimensional semiconductors,
such as carbon nanotubes. Or even
“others yet to be invented,” says Imec’s
Samavedam.
Silicon-based chips may be integrated
in the same package with chips that rely
on newer materials, just as processor
makers are today integrating chips using
different silicon manufacturing technol-
ogies into the same package, notes IEEE
Fellow Gabriel Loh, a senior fellow at
AMD.
Which semiconductor material is at
the heart of the device may not even be
the central issue in 2047. “The choice
of channel material will essentially be
dictated by which material is the most
compatible with many other materials
that form other parts of the device,”
says Salahuddin. And we know a lot
about integrating materials with
silicon.
WHERE WILL TRANSISTORS BE
COMMON WHERE THEY ARE NOT
FOUND TODAY?
Everywhere. No, seriously. Experts really
do expect some amount of intelligence
and sensing to creep into every aspect of
our lives. That means devices will be
attached to our bodies and implanted
inside them; embedded in all kinds of
infrastructure, including roads, walls,
and houses; woven into our clothing;
stuck to our food; swaying in the breeze
in grain fields; watching just about every
step in every supply chain; and doing
many other things in places nobody has
thought of yet.
Transistors will be “everywhere that
needs computation, command and con-
trol, communications, data collection,
storage and analysis, intelligence, sensing
and actuation, interaction with humans,
or an entrance portal to the virtual and
mixed reality world,” sums up Stanford’s
Wong.
DECEMBER 2022  SPECTRUM.IEEE.ORG  39
 A N ew
Way to
Squ ash
Bugs
Functional
programming is hard
to learn but yields
fewer nasty surprises
BY CHARLES SCALFANI

YOU’D EXPECT THE LONGEST AND MOST COSTLY PHASE in the life cycle of a
software product to be the initial development of the system, when all those great
features are first imagined and then created. In fact, the hardest part comes l­ater,
during the maintenance phase. That’s when programmers pay the price for the
shortcuts they took during development. • So why did they take shortcuts? Maybe
they didn’t realize that they were cutting any corners. Only when their code was
­deployed and exercised by a lot of users did its hidden flaws come to light. And
­maybe the developers were rushed. Time-to-market pressures would almost guar-
antee that their software will contain more bugs than it would otherwise.

Illustrations by Shira Inbar DECEMBER 2022  SPECTRUM.IEEE.ORG  41


The struggle that most companies have
maintaining code causes a second problem:
fragility. Every new feature that gets added
to the code increases its complexity, which
then increases the chance that something
will break. It’s common for software to grow
so complex that the developers avoid chang-
ing it more than is absolutely necessary for
fear of breaking something. In many com-
panies, whole teams of developers are
employed not to develop anything new but
just to keep existing systems going. You
might say that they run a software version
of the Red Queen’s race, running as fast as
they can just to stay in the same place.
It’s a sorry situation. Yet the current tra-
jectory of the software industry is toward
increasing complexity, longer product-­
development times, and greater fragility of
production systems. To address such issues,
companies usually just throw more people
at the problem: more developers, more tes-
ters, and more technicians who intervene
when systems fail.
Surely there must be a better way. I’m
part of a growing group of developers who
think the answer could be functional pro-
gramming. Here I describe what functional
programming is, why using it helps, and why
I’m so enthusiastic about it.
A good way to understand
the rationale for functional
A programming is by consid-
ering something that hap-
pened more than a half
century ago. In the late 1960s,
a programming paradigm emerged that
aimed to improve the quality of code while
reducing the development time needed. It was
called structured programming.
Various languages emerged to foster
structured programming, and some existing
languages were modified to better support
it. One of the most notable features of these
structured-­programming languages was not
Imagine if every time you ran your microwave, your
dishwasher’s cycle setting changed from Normal to Pots
and Pans. In software, this kind of thing goes on all the time.
42  SPECTRUM.IEEE.ORG  DECEMBER 2022
a feature at all: It was the absence of something that had been around a
long time—the GOTO statement.
The GOTO statement is used to redirect program execution. Instead
of carrying out the next statement in sequence, the flow of the program
is redirected to some other statement, the one specified in the GOTO line,
typically when some condition is met.
The elimination of the GOTO was based on what programmers had
learned from using it—that it made the program very hard to understand.
Programs with GOTOs were often referred to as spaghetti code because
the sequence of instructions that got executed could be as hard to follow
as a single strand in a bowl of spaghetti.
The inability of these developers to understand how their code worked,
or why it sometimes didn’t work, was a complexity problem. Software
experts of that era believed that those GOTO statements were creating
unnecessary complexity and that the GOTO had to, well, go.
Back then, this was a radical idea, and many programmers resisted the
loss of a statement that they had grown to rely on. The debate went on for
more than a decade, but in the end, the GOTO went extinct, and no one
today would argue for its return. That’s because its elimination from
higher-­level programming languages greatly reduced complexity and
boosted the reliability of the software being produced. It did this by limiting
what programmers could do, which ended up making it easier for them
to reason about the code they were writing.
Although the software industry has eliminated GOTO from modern
higher-level languages, software nevertheless continues to grow in complex-
ity and fragility. Looking for how else such programming languages could
be modified to avoid some common pitfalls, software designers can find
inspiration, curiously enough, from their counterparts on the hardware side.
In designing hardware for a computer, you can’t have a
resistor shared by, say, both the keyboard and the moni-
I tor’s circuitry. But programmers do this kind of sharing
all the time in their software. It’s called shared global state:
Variables are owned by no one process but can be changed
by any number of processes, even simultaneously.
Now, imagine that every time you ran your microwave, your dish­
washer’s cycle setting changed from Normal to Pots and Pans. That, of
course, doesn’t happen in the real world, but in software, this kind of thing
goes on all the time. Programmers write code that calls a function, expect-
ing it to perform a single task. But many functions have side effects that
change the shared global state, giving rise to unexpected consequences.
In hardware, that doesn’t happen because the laws of physics curtail
what’s possible. Of course, hardware engineers can mess up, but not like
you can with software, where just too many things are possible, for better
or worse.
Another complexity monster lurking in the software quagmire is called
a null reference, meaning that a reference to a place in memory points to
nothing at all. If you try to use this reference, an error ensues. So program-
mers have to remember to check whether something is null before trying
to read or change what it references.
 Nearly every popular language
today has this flaw. The pioneering
computer scientist Tony Hoare intro- Avoiding Null-Reference Surprises
duced the null reference in the A comparison of JavaScript and PureScript shows how
ALGOL language back in 1965, and it the latter can help programmers avoid bugs.
was later incorporated into numer-
ous other languages. Hoare explained JAVASCRIPT CODE
that he did this “simply because it
was so easy to implement,” but today Set up an array of three numbers
const list = [0, 1, 2] (0, 1, and 2).
he considers it to be a “billion-dollar
const num = list[99] Reference the 99th number
mistake.” That’s because it has in that array, which doesn’t exist.
caused countless bugs when a refer- console.log("Number is: ", num + 1) Add 1 to that number and
ence that the programmer expects to display the result.
be valid is really a null reference.
OUTPUT
Software developers need to be
The result is undefined (not a number,
extremely disciplined to avoid such > Number is: NaN or NaN), an unexpected result that
pitfalls, and sometimes they don’t could interfere with further processing.
take adequate precautions. The archi-
tects of structured programming Programming the equivalent in PureScript avoids such pitfalls:
knew this to be true for GOTO state-
ments and left developers no escape PURESCRIPT CODE
hatch. To guarantee the improve-
do
ments in clarity that GOTO-free code
let list = [ 0, 1, 2 ] Set up an array of three numbers
promised, they knew that they’d have (0, 1, and 2).
to eliminate it entirely from their Reference the 99th number
structured-programming languages. num = list !! 99 in that array, which doesn’t exist.
History is proof that removing a Add 1 to that number and
log $ "Number is: " <> show (num + 1)
dangerous feature can greatly display the result.
improve the quality of code. Today,
The PureScript compiler would
we have a slew of dangerous practices catch that the programmer hadn’t
that compromise the robustness and > Compiler error! anticipated the possibility that the
maintainability of software. Nearly referenced number was not defined.
all modern programming languages
have some form of null references, To get the code to compile, the programmer would have to amend
shared global state, and functions one line, perhaps like this:
with side effects—things that are far
worse than the GOTO ever was. do
Set up an array of three numbers
How can those flaws be elimi- let list = [ 0, 1, 2 ] (0, 1, and 2).
nated? It turns out that the answer
Default to 0 if the referenced number
has been around for decades: purely let num = (list !! 99) # fromMaybe 0 doesn’t exist.
functional programming languages. Add 1 to that number and
log $ "Number is: " <> show (num + 1)
The first purely functional lan- display the result.
guage to become popular, called
­Haskell, was created in 1990. So by The referenced number isn’t defined
> Number is: 1 and defaults to 0. The value of 0 + 1
the mid-1990s, the world of software is shown.
development really had the solution
to the vexing problems it still faces.
Sadly, the hardware of the time often
wasn’t powerful enough to make use
of the solution. But today’s processors can easily manage the demands of With this one restriction, you increase stabil-
Haskell and other purely functional languages. ity, open the door to compiler optimizations,
Indeed, software based on pure functions is particularly well suited to and end up with code that’s far easier to
modern multicore CPUs. That’s because pure functions operate only on reason about.
their input parameters, making it impossible to have any interactions But what if a function needs to know or
between different functions. This allows the compiler to be optimized to needs to manipulate the state of the system?
produce code that runs on multiple cores efficiently and easily. In that case, the state is passed through a
As the name suggests, with purely functional programming, the developer long chain of what are called composed
can write only pure functions, which, by definition, cannot have side effects. functions—functions that pass their out-

DECEMBER 2022  SPECTRUM.IEEE.ORG  43
 puts to the inputs of the next function in the
chain. By passing the state from function to
function, each function has access to it and
there’s no chance of another concurrent
programming thread modifying that state—
another common and costly fragility found
in far too many programs.
Functional programming also has a solu-
tion to Hoare’s “billion-dollar mistake,” null
references. It addresses that problem by dis-
allowing nulls. Instead, there is a construct
usually called Maybe (or Option in some lan-
guages). A Maybe can be Nothing or Just some
value. Working with Maybes forces develop-
ers to always consider both cases. They have
no choice in the matter. They must handle
the Nothing case every single time they
encounter a Maybe. Doing so eliminates the
many bugs that null references can spawn.
Functional programming also requires
that data be immutable, meaning that once
you set a variable to some value, it is forever
that value. Variables are more like variables
in math.
For example, to compute the formula y =
x2 + 2x – 11, you pick a value for x and at no
time during the computation of y does x take
on a different value. So, the same value for x
is used when computing x2 as is used when
computing 2x. In most programming lan-
guages, there is no such restriction. You can
compute x2 with one value, then change the
value of x before computing 2x. By disallow- take the value of x, add one to it, and put it back into a variable called x.
ing developers from changing (mutating) In functional programming, there are no statements, only expressions.
values, they can use the same reasoning they Mathematical thinking that we learned in middle school can now be
did in middle-school algebra class. employed when writing code in a functional language.
Unlike most languages, functional pro- Thanks to functional purity, you can reason about code using algebraic
gramming languages are deeply rooted substitution to help reduce code complexity in the same way you reduced
in mathematics. It’s this lineage in the the complexity of equations back in algebra class. In non-functional lan-
highly disciplined field of mathematics that guages (imperative languages), there is no equivalent mechanism for
gives functional languages their biggest reasoning about how the code works.
advantages.
Why is that? It’s because people have Pure functional programming solves many of our indus-
been working on mathematics for thousands try’s biggest problems by removing dangerous features
of years. It’s pretty solid. Most programming
paradigms, such as object-oriented pro-
P from the language, making it harder for developers to shoot
themselves in the foot. At first, these limitations may seem
gramming, have at most half a dozen decades drastic, as I’m sure the 1960s developers felt regarding the
of work behind them. They are crude and removal of GOTO. But the fact of the matter is that it’s both
immature by comparison. liberating and empowering to work in these languages—so much so that
Let me share an example of how pro- nearly all of today’s most popular languages have incorporated functional
gramming is sloppy compared with mathe- features, although they remain fundamentally imperative languages.
matics. We typically teach new programmers The biggest problem with this hybrid approach is that it still allows devel-
to forget what they learned in math class opers to ignore the functional aspects of the language. Had we left GOTO
when they first encounter the statement x = as an option 50 years ago, we might still be struggling with spaghetti code.
x + 1. In math, this equation has zero solu- To reap the full benefits of pure functional programming languages,
tions. But in most of today’s programming you can’t compromise. You need to use languages that were designed with
languages, x = x + 1 is not an equation. It is a these principles from the start. Only by adopting them will you get the
statement that commands the computer to many benefits that I’ve outlined here.

44  SPECTRUM.IEEE.ORG  DECEMBER 2022
To reap the full benefits of pure functional programming
languages, you can’t compromise. You need to use languages
that were designed with these principles from the start.
But functional programming isn’t a bed of roses. It comes at a cost. But there are still challenges when work-
Learning to program according to this functional paradigm is almost like ing with a language that relatively few others
learning to program again from the beginning. In many cases, developers use—in particular, the lack of online help,
must familiarize themselves with math that they didn’t learn in school. The documentation, and example code. And it’s
required math isn’t difficult—it’s just new and, to the math phobic, scary. hard to hire developers with experience in
More important, developers need to learn a new way of thinking. At these languages. Because of that, my com-
first this will be a burden, because they are not used to it. But with time, pany uses recruiters who specialize in find-
this new way of thinking becomes second nature and ends up reducing ing functional programmers. And when we
cognitive overhead compared with the old ways of thinking. The result is hire someone with no background in func-
a massive gain in efficiency. tional programming, we put them through
But making the transition to functional programming can be difficult. a training process for the first few months
My own journey doing so a few years back is illustrative. to bring them up to speed.
I decided to learn Haskell—and needed to do that on a business time-
line. This was the most difficult learning experience of my 40-year career, My company is small. It
in large part because there was no definitive source for helping developers delivers software to govern-
make the transition to functional programming. Indeed, no one had writ-
ten anything very comprehensive about functional programming in the
M mental agencies to enable
them to help veterans
prior three decades. receive benefits from the U.S.
I was left to pick up bits and pieces from here, there, and everywhere. Department of Veteran’s
And I can attest to the gross inefficiencies of that process. It took me three Affairs. It’s extremely rewarding work, but it’s
months of days, nights, and weekends living and breathing Haskell. But not a lucrative field. With razor-slim margins,
finally, I got to the point that I could write better code with it than with we must use every tool available to us to do
anything else. more with fewer developers. And for that,
When I decided that our company should switch to using functional functional programming is just the ticket.
languages, I didn’t want to put my developers through the same nightmare. It’s very common for unglamorous busi-
So, I started building a curriculum for them to use, which became the basis nesses like ours to have difficulty attracting
for a book intended to help developers transition into functional program- developers. But we are now able to hire top-
mers. In my book, I provide guidance for obtaining proficiency in a func- tier people because they want to work on a
tional language called PureScript, which stole all the great aspects of functional codebase. Being ahead of the
Haskell and improved on many of its shortcomings. In addition, it’s able curve on this trend, we can get talent
to operate in both the browser and in a back-end server, making it a great that most companies our size could only
solution for many of today’s software demands. dream of.
While such learning resources can only help, for this transition to take I anticipate that the adoption of pure
place broadly, software-based businesses must invest more in their biggest functional languages will improve the qual-
asset: their developers. At my company, Panoramic Software, where I’m ity and robustness of the whole software
the chief technical officer, we’ve made this investment, and all new work industry while greatly reducing time
is being done in either PureScript or Haskell. wasted on bugs that are simply impossible
We started down the road of adopting functional languages three years to generate with functional programming.
ago, beginning with another pure functional language called Elm because It’s not magic, but sometimes it feels like
it is a simpler language. (Little did we know we would eventually outgrow that, and I’m reminded of how good I have
it.) It took us about a year to start reaping the benefits. But since we got it every time I’m forced to work with a
over the hump, it’s been wonderful. We have had no production runtime non-functional codebase.
bugs, which were so common in what we were formerly using, JavaScript One sign that the software industry as
on the front end and Java on the back. This improvement allowed the team a whole is preparing for a paradigm shift is
to spend far more time adding new features to the system. Now, we spend that functional features are showing up in
almost no time debugging production issues. more and more of the mainstream pro-
gramming languages. It will take much
more work for the industry to make the
transition fully, but the benefits of doing so
are clear, and that is no doubt where things
are headed.

DECEMBER 2022  SPECTRUM.IEEE.ORG  45
 Leveraging brain data will make workers happier
and more productive, backers say
B y evan ac k e r m an & e l i z a st r ic k l a n d
ARE
YOU
WORKPLACE
BRAIN SCANNING?
G
ET READY: Neurotechnology is coming
to the workplace. Neural sensors are now
reliable and affordable enough to support
commercial pilot projects that extract
productivity-enhancing data from work-
ers’ brains. These projects aren’t confined
to specialized workplaces; they’re also
happening in offices, factories, farms, and
airports. The companies and people
behind these neurotech devices are certain that they will
improve our lives. But there are serious questions about
whether work should be organized around certain functions
of the brain, rather than the person as a whole.
To be clear, the kind of neurotech that’s currently available
is nowhere close to reading minds. Sensors detect electrical
activity across different areas of the brain, and the patterns in
Illustration by Nadia Radic
FOR
READY
that activity can be broadly correlated with different feelings
or physiological responses, such as stress, focus, or a reaction
to external stimuli. These data can be exploited to make work-
ers more efficient—and, proponents of the technology say, to
make them happier. Two of the most interesting innovators in
this field are the Israel-based startup InnerEye, which aims to
give workers superhuman abilities, and Emotiv, a Silicon Valley
neurotech company that’s bringing a brain-tracking wearable
to office workers, including those working remotely.
The fundamental technology that these companies rely on
is not new: Electroencephalography (EEG) has been around for
about a century, and it’s commonly used today in both medicine
and neuroscience research. For those applications, the subject
may have up to 256 electrodes attached to their scalp with con-
ductive gel to record electrical signals from neurons in different
parts of the brain. More electrodes, or “channels,” mean that
DECEMBER 2022  SPECTRUM.IEEE.ORG  47
 doctors and scientists can get better spatial resolution
in their readouts—they can better tell which neurons
are associated with which electrical signals.
What is new is that EEG has recently broken out
of clinics and labs and has entered the consumer mar-
ketplace. This move has been driven by a new class
of “dry” electrodes that can operate without conduc-
tive gel, a substantial reduction in the number of elec-
trodes necessary to collect useful data, and advances
in artificial intelligence that make it far easier to inter-
pret the data. Some EEG headsets are even available
directly to consumers for a few hundred dollars.
While the public may not yet be embracing EEG,
experts say this neurotechnology is mature and
ready for commercial applications. “This is not sci-
fi,” says James Giordano, chief of neuroethics stud�-
ies at Georgetown University Medical Center. “This
is quite real.”
I
n an office in Herzliya, Israel, Sergey
Vaisman sits in front of a computer. He’s
relaxed but focused, silent and unmoving,
and not at all distracted by the seven-chan-
nel EEG headset he’s wearing. On the computer
screen, images rapidly appear and disappear, one
after another. At a rate of three images per second,
it’s just possible to tell that they come from an air-
port X-ray scanner. It’s essentially impossible to
focus on anything beyond fleeting impressions of
ghostly bags and their contents.
“Our brain is an amazing machine,” Vaisman
tells us as the stream of images ends. The screen
now shows an album of selected X-ray images that
InnerEye uses commercially available EEG sensors as part of
its target-detection system.
48  SPECTRUM.IEEE.ORG  DECEMBER 2022
were just flagged by Vaisman’s brain, most of which
are now revealed to have hidden firearms. No one
can knowingly identify and flag firearms among the
jumbled contents of bags when three images are
flitting by every second, but Vaisman’s brain has no
problem doing so behind the scenes, with no action
required on his part.
The brain processes visual imagery very quickly.
According to Vaisman, the decision-making process
to determine whether there’s a gun in complex
images like these takes just 300 milliseconds. What
takes much more time are the cognitive and motor
processes that occur after the decision making—
planning a response (such as saying something or
pushing a button) and then executing that response.
If you can skip these planning and execution phases
and instead use EEG to directly access the output
of the brain’s visual-processing and decision-mak-
ing systems, you can perform image-recognition
tasks far faster. The user no longer has to actively
think: For an expert, just that fleeting first impres-
sion is enough for their brain to make an accurate
determination of what’s in the image.
Vaisman is the vice president of R&D of I­ nnerEye,
an Israel-based startup that recently came out of
stealth mode. InnerEye uses deep learning to clas-
sify EEG signals into responses that indicate “tar-
gets” and “nontargets.” Targets can be anything that
a trained human brain can recognize. In addition to
developing the security-screening application,
­InnerEye has worked with doctors to detect tumors
in medical images, with farmers to identify diseased
plants, and with manufacturing experts to spot
product defects. In simple cases, InnerEye has found
that our brains can handle image recognition at rates
of up to 10 images per second. And, Vaisman says,
the company’s system produces results just as accu-
rate as a human would when recognizing and tag-
ging images manually—InnerEye is merely using
EEG as a shortcut to that person’s brain to drasti-
cally speed up the process.
While using the InnerEye technology doesn’t
require active decision making, it does require train-
ing and focus. Users must be experts at the task, well
trained in identifying a given type of target, whether
that’s firearms or tumors. They must also pay close
attention to what they’re seeing—they can’t just
INNEREYE; X-RAY: ISTOCK
zone out and let images flash past. InnerEye’s system
measures focus very accurately, and if the user blinks
or stops concentrating momentarily, the system
detects it and shows the missed images again.
Having a human brain in the loop is especially
important for classifying data that may be open to
interpretation. For example, a well-trained image
 Nontargets Targets

F3

F4
Signal strength

200 300 450 550 700 C3 200 300 450 550 700

Time (milliseconds) Time (milliseconds)

C4

Pz

Signal
Signal

P3

Time P4 Time

Time

InnerEye’s image-classification system operates at high speed by providing a shortcut to the brain of an expert human.
As an expert focuses on a continuous stream of images (from three to 10 images per second, depending on complexity),
a commercial EEG system combined with InnerEye’s software can distinguish the characteristic response the expert’s
brain produces when it recognizes a target. In this example, the target is a weapon in an X-ray image of a suitcase,
representing an airport-security application.
InnerEye uses multiple channels of EEG data collected from different areas of the brain [center left and right]. The
raw EEG signals detected by the headset [center] are processed and analyzed by InnerEye’s system to generate target and
nontarget responses [bottom left and right]. For safety-critical applications, maintaining a human in the loop offers a
significant advantage over AI classifiers, but InnerEye uses its system to train more-nuanced AI at the same time.

classifier may be able to determine with reasonable
accuracy whether an X-ray image of a suitcase
shows a gun, but if you want to determine whether
that X-ray image shows something else that’s
vaguely suspicious, you need human experience.
People are capable of recognizing something
unusual even if they don’t know quite what it is.
“We can see that uncertainty in the brain waves,”
says InnerEye founder and chief technology officer
Amir Geva. “We know when they aren’t sure.”
Humans have a unique ability to recognize and con-
textualize novelty, a substantial advantage that
­InnerEye’s system has over AI image classifiers.
InnerEye then feeds that nuance back into its AI
models. “When a human isn’t sure, we can teach AI
systems to be not sure, which is better training than
teaching the AI system just one or zero,” says Geva.
“There is a need to combine human expertise with
AI.” InnerEye’s system enables this combination, as
every image can be classified by both computer
vision and a human brain.
Using InnerEye’s system is a positive experience
for its users, the company claims. “When we start
working with new users, the first experience is a bit
overwhelming,” Vaisman says. “But in one or two
sessions, people get used to it, and they start to like
it.” Geva says some users do find it challenging to
maintain constant focus throughout a session,

Illustration by Chris Philpot DECEMBER 2022  SPECTRUM.IEEE.ORG  49

 which lasts up to 20 minutes, but once they get used
to working at three images per second, even two
images per second feels “too slow.”
In a security-screening application, three images
per second is approximately an order of magnitude
faster than an expert can manually achieve. I­ nnerEye
says their system allows far fewer humans to handle
far more data, with just two human experts redun-
dantly overseeing 15 security scanners at once, sup-
ported by an AI image-recognition system that is
being trained at the same time, using the output
from the humans’ brains.
InnerEye is currently partnering with a handful
of airports around the world on pilot projects. And
it’s not the only company working to bring neuro-
tech into the workplace.
W
hen it comes to neural monitoring for
productivity and well-being in the work-
place, the San Francisco–based company
Emotiv is leading the charge. Since its
founding 11 years ago, Emotiv has released three
models of lightweight brain-scanning headsets.
Until now the company had mainly sold its hardware
to neuroscientists, with a sideline business aimed
at developers of brain-controlled apps or games.
Emotiv started advertising its technology as an
enterprise solution only this year, when it released
its fourth model, the MN8 system, which tucks
brain-scanning sensors into a pair of discreet Blue-
tooth earbuds.
Tan Le, Emotiv’s CEO and cofounder, sees neuro�-
tech as the next trend in wearables, a way for people
Gamma: 25–100+ Hz
Heightened perception,
cognitive processing
Beta: 12–25 Hz
Alert consciousness,
thinking, excitement
Alpha: 8–12 Hz
Physical and
mental relaxation
Theta: 4–8 Hz
Deep relaxation
and meditation
Delta: .1–4 Hz
Deep, dreamless sleep
Brain waves are repetitive patterns of neural
activity. Waves of different frequencies are
associated with different mental states.
50  SPECTRUM.IEEE.ORG  DECEMBER 2022
to get objective “brain metrics” of mental states,
enabling them to track and understand their cognitive
and mental well-being. “I think it’s reasonable to imag-
ine that five years from now this [brain tracking] will
be quite ubiquitous,” she says. When a company uses
the MN8 system, workers get insight into their indi-
vidual levels of focus and stress, and managers get
aggregated and anonymized data about their teams.
Emotiv launched its enterprise technology into
a world that is fiercely debating the future of the
workplace. Workers are feuding with their employ-
ers about return-to-office plans following the pan-
demic, and companies are increasingly using
“bossware” to keep tabs on employees—whether
staffers or gig workers, working in the office or
remotely. Le says Emotiv is aware of these trends
and is carefully considering which companies to
work with as it debuts its new gear. “The dystopian
potential of this technology is not lost on us,” she
says. “So we are very cognizant of choosing partners
that want to introduce this technology in a respon-
sible way—they have to have a genuine desire to help
and empower employees.”
Lee Daniels, a consultant who works for the
global real estate services company JLL, has spoken
with a lot of C-suite executives lately. “They’re wor-
ried,” says Daniels. “There aren’t as many people
coming back to the office as originally anticipated—
the hybrid model is here to stay, and it’s highly com-
plex.” Executives come to Daniels asking how to
manage a hybrid workforce. “This is where the
­neuroscience comes in,” he says.
Emotiv has partnered with JLL, which has begun
to use the MN8 earbuds to help its clients collect
“true scientific data,” Daniels says, about workers’
attention, distraction, and stress, and how those
factors influence both productivity and well-being.
Daniels says JLL is currently helping its clients run
short-term experiments using the MN8 system to
track workers’ responses to new collaboration tools
and various work settings; for example, employers
could compare the focus of in-office and remote
workers.
Emotiv CTO Geoff Mackellar believes the new
MN8 system will succeed because of its convenient
and comfortable form factor: The multipurpose
earbuds also let the user listen to music and answer
phone calls. The downside of earbuds is that they
provide only two channels of brain data. When the
company first considered this project, Mackellar
says, his engineering team looked at the rich data
set they’d collected from Emotiv’s other headsets
over the past decade. The company boasts that aca-
demics have conducted more than 4,000 studies
 High attention  Low attention
Low cognitive stress High cognitive stress

RAW EEG DATA  RAW EEG DATA

Left ear Left ear

Right ear Right ear

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

Time (seconds) Time (seconds)

BRAIN-WAVE FREQUENCY BANDS BRAIN-WAVE FREQUENCY BANDS

1.0 1.0
Band power
(decibels)

Band power
(decibels)
.8 .8

.6 .6

.4 .4

.2 .2

.0 .0

Theta Alpha Low beta High beta Gamma Theta Alpha Low beta High beta Gamma
(4–8 Hz) (8–12 Hz) (12–16 Hz)  (16–25 Hz)  (25–45 Hz) (4–8 Hz) (8–12 Hz) (12–16 Hz)  (16–25 Hz)  (25–45 Hz)

PERFORMANCE METRICS PERFORMANCE METRICS

Attention

Attention

74

31
Cognitive

Cognitive

75
stress

stress

12

0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180

Time (seconds) Time (seconds)

Emotiv’s MN8 system uses earbuds to capture two channels of EEG data, from which the company’s proprietary algorithms
derive performance metrics for attention and cognitive stress. It’s very difficult to draw conclusions from raw EEG
signals [top], especially with only two channels of data. The MN8 system relies on machine-learning models that Emotiv
developed using a decade’s worth of data from its earlier headsets, which have more electrodes.
To determine a worker’s level of attention and cognitive stress, the MN8 system uses a variety of analyses. One
shown here [middle, bar graphs] reveals increased activity in the low-frequency ranges (theta and alpha) when a
worker’s attention is high and cognitive stress is low; when the worker has low attention and high stress, there’s
more activity in the higher-frequency ranges (beta and gamma). This analysis and many others feed into the models that
present simplified metrics of attention and cognitive stress [bottom] to the worker.

using Emotiv tech. From that trove of data—from data from the two systems in unison, the engineers
headsets with 5, 14, or 32 channels—Emotiv isolated trained a machine-learning algorithm to identify the
the data from the two channels the earbuds could signatures of attention and cognitive stress from the
pick up. “Obviously, there’s less information in the relatively sparse MN8 data. The brain signals asso-
two sensors, but we were able to extract quite a lot ciated with attention and stress have been well stud-
of things that were very relevant,” Mackellar says. ied, Mackellar says, and are relatively easy to track.
Once the Emotiv engineers had a hardware pro- Although everyday activities such as talking and
totype, they had volunteers wear the earbuds and a moving around also register on an EEG, the Emotiv
14-channel headset at the same time. By recording software filters out those artifacts.

Illustration by Chris Philpot DECEMBER 2022  SPECTRUM.IEEE.ORG  51


Emotiv’s MN8 earbuds collect two channels of EEG brain data.
The earbuds can also be used for phone calls and music.
The app that’s paired with the MN8 earbuds
doesn’t display raw EEG data. Instead, it processes
that data and shows workers two simple metrics
relating to their individual performance. One squig-
gly line shows the rise and fall of workers’ attention
to their tasks—the degree of focus and the dips that
come when they switch tasks or get distracted—
while another line represents their cognitive stress.
Although short periods of stress can be motivating,
too much for too long can erode productivity and
well-being. The MN8 system will therefore some-
times suggest that the worker take a break. Workers
can run their own experiments to see what kind of
break activity best restores their mood and focus—
maybe taking a walk, or getting a cup of coffee, or
chatting with a colleague.
W
hile MN8 users can easily access data
from their own brains, employers don’t see
individual workers’ brain data. Instead,
they receive aggregated data to get a sense
of a team or department’s attention and stress levels.
With that data, companies can see, for example, on
which days and at which times of day their workers
are most productive, or how a big announcement
affects the overall level of worker stress.
Emotiv emphasizes the importance of anonymiz-
ing the data to protect individual privacy and prevent
people from being promoted or fired based on their
brain metrics. “The data belongs to you,” says Emo-
tiv’s Le. “You have to explicitly allow a copy of it to
52  SPECTRUM.IEEE.ORG  DECEMBER 2022
be shared anonymously with your employer.” If a
group is too small for real anonymity, Le says, the
system will not share that data with employers. She
also predicts that the device will be used only if work-
ers opt in, perhaps as part of an employee wellness
program that offers discounts on medical insurance
in return for using the MN8 system regularly.
However, workers may still be worried that
employers will somehow use the data against them.
Karen Rommelfanger, founder of the Institute of
Neuroethics, shares that concern. “I think there is
significant interest from employers” in using such
technologies, she says. “I don’t know if there’s sig-
nificant interest from employees.”
Both she and Georgetown’s Giordano doubt that
such tools will become commonplace anytime soon.
“I think there will be pushback” from employees on
issues such as privacy and worker rights, says Gior-
dano. Even if the technology providers and the com-
panies that deploy the technology take a responsible
approach, he expects questions to be raised about
who owns the brain data and how it’s used. “Per-
ceived threats must be addressed early and explic-
itly,” he says.
Giordano says he expects workers in the United
States and other western countries to object to rou-
Brain data tine brain scanning. In China, he says, workers have
reportedly been more receptive to experiments with
can be such technologies. He also believes that brain-mon-
exploited itoring devices will really take off first in industrial
to make settings, where a momentary lack of attention can
workers lead to accidents that injure workers and hurt a com-
more pany’s bottom line. “It will probably work very well
efficient— under some rubric of occupational safety,” Giordano
and, says. It’s easy to imagine such devices being used by
proponents companies involved in trucking, construction, ware� -
house operations, and the like. Indeed, at least one
of the such product, an EEG headband that measures
technology fatigue, is already on the market for truck drivers
say, to
and miners.
make them Giordano says that using brain-tracking devices
happier. for safety and wellness programs could be a slippery
“The slope in any workplace setting. Even if a company
dystopian focuses initially on workers’ well-being, it may soon
potential find other uses for the metrics of productivity and
of this performance that devices like the MN8 provide.
“Metrics are meaningless unless those metrics are
technology standardized, and then they very quickly become
is not lost on comparative,” he says.
us.” —Tan Le, Rommelfanger adds that no one can foresee how
Emotiv CEO workplace neurotech will play out. “I think most
companies creating neurotechnology aren’t pre-
EMOTIV
pared for the society that they’re creating,” she says.
“They don’t know the possibilities yet.”
 NEWS OF THE IEEE  
VOLUME 46 / ISSUE 4

Executive Director Thomas Coughlin Is

2023 IEEE President-
Stephen Welby Reflects Elect P. 60
BRAD TRENT

on His Time at IEEE

P. 56
Arti Agrawal Champions
Inclusivity in STEM P. 62

DECEMBER 2022  THE INSTITUTE  53

 PRESIDENT’S COLUMN
Looking to 2050
and Beyond
WHAT WILL THE future of the world
look like? Everything in the world
evolves. Therefore, IEEE also must
evolve, not only to survive but to thrive.
How will people build communities
and engage with one another and
with IEEE in the future? How will
knowledge be acquired? How will
content be curated, shared, and
accessed? What issues will influence
the development of technical
standards? How should IEEE be
organized to be most impactful?
While no one has a crystal ball,
predictions can be made based on
evidence and trends. To start the
conversation around these questions,
I appointed the 2022 IEEE Ad Hoc
Committee on IEEE in 2050. The
committee, chaired by IEEE Fellow
Roger Fujii, is designed to envision
scenarios looking out to the year
2050 and beyond to gain a global
perspective of what the world may
look like and what potential futures
might mean for IEEE.
54  THE INSTITUTE  DECEMBER 2022
The committee explored plausible
scenarios across IEEE’s range of
interests and scanned for drivers of
change within existing and emerging
technology fields. It also analyzed
the role IEEE should take based on
the identified potential futures and
discussed next steps in IEEE’s major
areas of focus, including conferences,
education, publications, standards,
membership, sustainability, and
governance.
For example, imagine that in
2050, your “cognitive digital twin” is
constantly surfing the large volumes
of research papers and data stored
across open-access repositories to find
information directly relevant to your
interests. It will also use its imaginative
and creative logic to suggest new
concepts and solutions for you. This
FOLLOW ME ON SOCIAL MEDIA FOR UPDATES AND INFORMATION FROM ACROSS IEEE
Facebook: @ieeepresident Instagram: @ieeepresident
Twitter: @IeeePresident LinkedIn: linkedin.com/showcase/ieeepresident
platform will be driven by artificial
intelligence, augmented reality, and
virtual reality developed to help make
you more productive and creative in
your career. What is IEEE’s role in this
new environment?
Molding the IEEE of the future
As a global organization, a
considerable challenge of IEEE is that
it supports a broad community. This
also presents a great opportunity
to learn and pilot various models,
services, products, and solutions to
meet the diverse set of members’
needs predicted for 2050 and beyond.
The technology generation of 2050
will likely be interested in solving
mission-based issues such as climate
change, universal access to health
care, sustainable food sources, and
ubiquitous energy generation and
transmission. Thus, IEEE’s mission—
to advance technology for humanity—
will still be relevant in the future.
However, the way IEEE achieves its
mission must and will change.
The future is multinodular and
digital. IEEE will benefit from its status
and reputation as a trusted, neutral
provider of content and information.
As a knowledge provider, IEEE has
an opportunity to curate and deliver
information to assist its constituents
and the public in understanding the
benefits and risks associated with
several technology areas.
Most importantly, IEEE can help
with the deep integration of artificial
intelligence and virtual reality into a
wide variety of everyday applications.
Adapting to an environment of
constant chaos and change is essential
moving forward. The ebb and flow
of geopolitical tensions are likely to
continue to increase—which will
impact global organizations like
ours. IEEE must become exceedingly
nimble to address rapid changes in
technologies and interdisciplinary
needs, and attract a broader audience.
IEEE will also need to rapidly respond
to selected strategic changes and
IEEE

allocate funding for new approaches.
IEEE’s governance structure will need
to be streamlined to meet the needs of
many future scenarios that will require the
organization to empower local entities to
make decisions within their area.
Trust in IEEE must remain high if the
organization is to maintain relevance and
remain a credible source of information
in the future. Now is the time for the
organization to be thoughtful and bold,
and take risks. IEEE cannot be afraid to
break silos. Some activities will need to be
terminated to make space for new ones.
Products and initiatives should be evaluated
regularly, and decisions must be made on a
continuous basis.
Sound scary? Compounded by global
warming, uneven demographic growth, and
geopolitical challenges, the future likely is
more uncertain than we realize. But often a
crisis can be transformed into opportunity
if we honestly face the unexpected and
become prepared for whatever lies ahead.
Adhering to IEEE’s core principles—trust,
growth and nurturing, global community
building, partnership, service to humanity,
and integrity in action—will serve the
organization well into the future.
I sincerely thank Roger Fujii and the ad
hoc committee members for their efforts.
Their work will aid IEEE in devising
long-term strategies to prepare for the
future, to adapt, and to convert uncertainty
into opportunity.
As I have shared, in an ever-changing and
uncertain world, IEEE—your professional
home—is always here for you, our members,
as well as for humanity, and for our shared
future. After all, serving our members well is
our raison d’être.
By addressing the challenges and
opportunities that lie ahead of us, IEEE
can remain a vibrant organization with
relevance both now and well into 2050. If
IEEE remains true to its central values—
fostering technological innovation and
excellence for the benefit of humanity—I’m
certain that the organization’s future will be
very bright indeed.
It has been my honor and privilege to
work with and for you this year as IEEE
president and CEO.
—K.J. RAY LIU
IEEE president and CEO
Please share your thoughts with me at
president@ieee.org.
EDITOR’S NOTE
Members on a Mission
They are leading diversity initiatives and expanding
STEM education
MANY PEOPLE JOIN IEEE not
for its benefits or access to its
publications but because they
support its mission and because
their affiliation with IEEE can
help them advance their causes.
That’s what drew Stephen
Welby to the job of IEEE
executive director and chief
operating officer in 2018.
The position offered him an
opportunity to influence and
engage a global community.
He did that and more
during the past five years,
leading the organization
through a global pandemic,
a changing publishing
landscape, and soaring
geopolitical tensions. The
IEEE Fellow plans to leave at
the end of the year to spend
more time with his family
while he explores his career
options. On page 56, read
about his accomplishments.
And on page 59, learn about
Welby’s replacement, Sophia
Muirhead, IEEE’s current
general counsel and chief
compliance officer, a position
she has held since 2019. In the
past three years, she improved
the organization’s standards,
finance, and publishing
protocols; oversaw the
evolution of the contracting
process; and deftly navigated
export controls, sanctions, and
other compliance regulations.
Fighting for diversity, equity,
and inclusion in science,
technology, engineering,
and math has become Arti
Agrawal’s purpose in life
[page 62]. The IEEE senior
member has spearheaded
DEI campaigns for the IEEE
Photonics Society and the
universities she worked for
in London and Sydney. This
DECEMBER 2022  THE INSTITUTE  55
year she started a consulting
firm to help businesses around
the world improve their DEI
programs.
“If you want to make a
change in a very difficult and
unchanging environment,”
Agrawal says, “you must be
disruptive.”
IEEE Life Senior Member
John A. Zulaski is on a quest
to make STEM kits available
in public libraries to teach
youngsters what engineers
do. On page 68, learn about
the Science Kits for Public
Libraries program he launched
in IEEE’s Region 4. The kits
can be found at more than
150 public libraries in the
central United States, and he
wants to see them in every
library around the world.
“The kits get kids interested
in pursuing a technical career—
which, by the way, helps
increase IEEE membership
down the road,” Zulaski says.
Jack Dongarra, recipient of
last year’s A.M. Turing Award,
wrote numerical algorithms
and libraries for linear
algebra that many describe
as world-changing [page 66].
Dongarra’s software libraries
contributed to the growth of
high-performance computing
and are used in just about
every computer.
“I’m very happy and proud
to be a member of IEEE,” says
Dongarra, an IEEE Life Fellow.
“I think it provides a valuable
service to the community.”
—KATHY PRETZ
Editor in chief, The Institute
For updates about IEEE and its
members, visit us at
spectrum.ieee.org/the-institute
 PROFILE public about the critical role that

Stephen
technical professionals play in building
a better world.”

A commitment to diversity
Welby lists his most important

Welby:
accomplishments as helping to
strengthen IEEE’s commitment to
diversity, inclusion, and professional
ethics; updating the organization’s

A Man on
publication model to increase its
embrace of open access; and shoring
up its financial footing.
“IEEE is very considerate in trying

a Mission
The departing executive director
steered IEEE through a pandemic
and open access publishing
BY KATHY PRETZ
IN HIS FIVE years as IEEE’s executive is stronger than when I arrived,”
director and chief operating Welby says.
officer, Stephen Welby has led IEEE’s mission of advancing
the organization through a global technology for the benefit of humanity
pandemic, a changing publishing drew him to the organization, he says,
landscape, and soaring geopolitical after he served a term in the Obama
tensions. Welby, an IEEE Fellow, is administration as U.S. assistant
leaving at the end of the year to spend secretary of defense for research
more time with his family while he and engineering.
explores his career options. The IEEE The IEEE position “offered an
Board of Directors has named Sophia opportunity to influence and engage a
Muirhead, IEEE’s current general global community,” Welby says. “The
counsel and chief compliance officer, work that IEEE is doing is important,
as his replacement. and the ideas embedded in our
Welby directs the daily operations mission statement are big drivers for
of IEEE and its approximately 1,000 ensuring that we leave a better world
employees. While the IEEE Board to future generations.
of Directors sets the organization’s “IEEE has an ambitious agenda:
policies and strategic direction, the supporting our members around
executive director’s job the world; deepening our
is to implement them technical engagement
Employer IEEE
and provide input about Title Executive
in emerging areas;
issues affecting the director and building communities and
organization’s future. chief operating disseminating technical
“I feel comfortable with officer information; meeting
stepping out of my role Member grade a growing demand for
Fellow
at this time because I feel Alma mater
technical education;
that, in almost every way, The Cooper Union, exploring new frontiers of
thanks to the hard work of New York City technical standards; and
volunteers and staff, IEEE engaging with the broader
to make sure that it is a voice for all
of our members,” he says. “We’ve
been taking steps to be as inclusive as
possible and to find ways that people
can participate and contribute—which
ties back to our mission. We’ve got
to find ways to be open, welcoming,
and maybe more than encouraging to
ensure that everybody can contribute
to the best of their abilities.”
During his tenure, the IEEE
Board revised its policy on diversity
to ensure that members have a
safe, inclusive place for collegial
discourse. Changes also were made
to the IEEE Code of Ethics to focus
members on key elements of the
code, a commitment not to engage
in harassment, and to protect the
privacy of others.
When Welby began working
for IEEE, he says, the open access
publishing movement was perceived
as a threat, but today he sees it as an
opportunity. The movement calls for
making research publications available
without fees for the reader.
“This requires a shift to new
business models to cover expenses
that were previously funded with
subscription fees,” he says. “IEEE has
been responsive to this change in a
careful, deliberate, and responsible
way. We are offering a wide variety
of opportunities and tools for people
to engage with. The evolution of our
publication activity has been about
responding to community demands for
new and diverse offerings.”
Twenty open access journals
have been launched in the past
five years, for a total of 29 in the
portfolio. Today they make up
more than 20 percent of IEEE’s

56  THE INSTITUTE  DECEMBER 2022

 Welby visits IEEE’s server
room in its offices in
Piscataway, N.J.
journal publishing output, and that
percentage is growing, Welby says.
He says he is proud that IEEE is
much stronger financially than when
he started. In 2018, its total net assets
were US $391 million, and by the end of
last year it had more than $851 million,
he says. He adds that he anticipates
strong results for this year as well.
BRAD TRENT
“To support our mission in
perpetuity, IEEE needs strong
financial reserves—ideally reserves
that are large enough to serve as
reliable and stabilizing sources of
revenue,” he says. “Doubling our
reserves has been a significant
accomplishment and represents an
investment in our future and our long-
term commitment to our mission.”
Other achievements, he says,
include the IEEE Standards
Association’s facilitation of
discussions about the responsible use
of artificial intelligence.
Another is the work that IEEE-
USA has done in helping to shape
public policy concerning technology.
He points to major U.S. legislation
passed this year: the Inflation
Reduction Act, and the Creating
Helpful Incentives for Producing
Semiconductors for America and
Foundries Act. Welby predicts that
other countries will make similar
investments, and he says he hopes
IEEE will support those as well.
DECEMBER 2022  THE INSTITUTE  57
 “IEEE is in an interesting spot
relative to public policy in the United
States, Europe, and elsewhere
because we talk to the concerns of
technologists, and about technology
and how it impacts regulations and
legislation,” he says. “We’re not
arguing for a particular stakeholder;
we’re arguing for investments in
advancing technology. We’re engaged
around the world in helping to ensure
that research is adequately funded
and that technical education gets the
support it needs.”
The impact of the pandemic
Leading a global organization that
depends on in-person meetings and
events has proven especially challenging
in recent years. After the coronavirus
pandemic began spreading in 2020,
the organization suddenly shifted
its business model from building
communities through in-person
engagement to relying almost exclusively
on digital delivery, Welby says.
“IEEE has a reputation for being
a conservative organization, for
being slow to change, and sometimes
for being overly rule-bound and
bureaucratic,” he says. “But in the face
of a global crisis, we adapted quickly.
Staff and volunteers worked closely
together to redesign and redeploy IEEE
activities in a manner that allowed our
communities to remain connected even
while physically distancing themselves
for safety. There was great trust, great
collaboration, and great success.
“The ability to respond, to change,
to evolve, and to adapt is one of IEEE’s
great strengths,” he adds. “That makes
me feel it will be ready to take on
whatever the future brings and deliver
on our mission.”
Geopolitical tensions in recent
years have also impacted IEEE, he
notes. Conflict among countries,
trade restrictions, and proliferating
international sanctions among nations
have created new challenges for IEEE
activities, he says.
“We have worked hard to stay
focused on IEEE’s role in supporting an
international community of engineers
and technical professionals that spans
national borders and supports a
global commitment to expanding the
58  THE INSTITUTE DECEMBER 2022
scientific and technical knowledge
of humankind,” he says. “But IEEE
also must operate in compliance with
applicable laws and regulations while
seeking clarification of those that
appear misapplied in the IEEE context.
At the same time, we are sympathetic to
the hardships that international tension
can place on our members.”
Coordinating the enormous scope
and scale of IEEE activities is complex.
IEEE operates in 160 countries;
has 46 societies and councils; has
343 sections in 10 geographic regions;
maintains a portfolio of more than
1,075 technical standards; publishes
200 transactions, journals, and
magazines; and holds more than
1,900 conferences and events.
“Trying to maintain coordination
and synchronization across this
enormous portfolio, managing
risk, and prioritizing resources is a
continuous challenge,” Welby says.
He says his one regret is that
more progress has not been made in
evolving IEEE’s membership model.
The model has remained relatively
unchanged since the organization was
established in 1963.
He points out that overall
membership growth has largely
flattened, despite the increased role that
technology plays in society. Across IEEE
there has been a decline in the share
of members working in industry, while
at the same time there has been recent
strong growth in student membership.
“We continue to rely on face-
to-face interaction for many of our
core membership activities, despite
the growth of online, virtual, and
digital communities,” Welby says.
“Prospective volunteers are working
longer hours [and] have family
commitments. Plus, there are other
activities competing for their time. The
time and energy commitments that
IEEE has traditionally demanded of its
volunteers may not be a viable strategy
in coming decades. It may be time to
fundamentally rethink the structure of
IEEE membership and explore different
ways to engage our members.”
He notes that those concerns
have been studied and debated, and
initiatives have been undertaken
to explore alternative membership
models. But, he says, there remains
more work to do to develop consensus
on the future of membership.
A hopeful future
Welby says he’ll miss the concentrated
chaos of an IEEE Board meeting series,
where many decisions—great and
small—are debated and decided in a
short period of time. He’ll also miss
talking to audiences of thousands
of people about important technical
topics, he says, and meeting with
leaders from government and industry.
“I have had the opportunity to
see the amazing diversity of our
membership,” he says, “and the
fantastic work that they do, and their
impact on the world. I spent a lot of
time talking to as many volunteers
in different roles as I could. I have
particularly enjoyed engaging with
students around the world and seeing
their enthusiasm, their creativity, and
their optimism. They give me great
hope for our future.”
Welby adds that he could not
have accomplished as much as he did
without the support and confidence
of the IEEE presidents and boards
he served, the thousands of IEEE
volunteers, his senior management
team, and the professional staff
members around the world “who
brought their creativity, commitment,
and technical skills to every task.
“I will leave it to others to assess
how effective I have been,” he says.
“But I have woken up every day for
BRAD TRENT
the last five years thinking about what
I could do to improve and advance
IEEE’s mission.”
 NEWS
Sophia Muirhead Is IEEE’s
Next Executive Director
BY KATHY PRETZ
SOPHIA “SOPHIE” MUIRHEAD is the
IEEE’s next executive director and chief
operating officer. She is the first woman
to hold the position. She is set to start
her new job on 1 January, succeeding
Stephen Welby.
Muirhead has been the IEEE’s general
counsel and chief compliance officer since
2019. During that time, she improved
the company’s standards, finance,
and publishing protocols; oversaw
the evolution of the organization’s
contracting process; and deftly navigated
export controls, sanctions, and other
compliance regulations.
“Sophia has a proven track record
of working in collaboration with
IEEE Members Get a
Free Pass to Switzerland’s
Museum ENTER
BY JOANNA GOODRICH
For more than a decade Museum
ENTER, in Solothurn, Switzerland,
has been a place where history buffs
can explore and learn about the
development and growth of computer
and consumer electronics in
Switzerland and the rest of the world.
On display are computers,
TOP: SOPHIA MUIRHEAD; BOTTOM: MUSEUM ENTER
calculators, floppy disks,
phonographs, radios, video game
consoles, and related objects. The
museum is set to move in May to
a larger building in the village of
Derendingen.
Thanks to a new four-year
partnership between the museum
and the IEEE Switzerland Section,
IEEE members may visit the facility
for free. They also can donate
their time to help create exhibits;
the Board of Directors and a deep
understanding of our organization,”
K.J. Ray Liu, IEEE president and CEO,
said in a news release announcing her
appointment.
“Her breadth of experience will
enhance the Board’s continuing
work to provide a professional
home for our members, and to serve
the engineering and technology
community worldwide.”
Muirhead is quoted in the release
as saying she was “attracted to IEEE
in 2019 because it is a mission-based
organization designed to benefit the
public interest [and] its members, and
serve a greater purpose.”
translate pamphlets, display
cards, and other written media;
and present science, technology,
engineering, and math workshops.
IEEE and Museum ENTER have
“many of the same goals,” says
IEEE Member Violetta Vitacca,
chief executive of the museum.
They both aim to inspire the next
generation of engineers, promote
the history of technology, and bring
together engineers from academia
and industry to collaborate, she
Museum ENTER claims to have the
largest collection of working
Apple computers in Europe.
“I am truly honored and excited to
be selected as IEEE’s next executive
director and COO,” she said. “I am
looking forward to collaborating with
IEEE’s leadership, members, and staff
to further IEEE’s mission around
the world.”
Prior to joining IEEE, she was
senior vice president, chief legal
officer, and corporate secretary at The
Conference Board, a global business
membership organization.
Muirhead earned a juris doctor
degree from Harvard Law School and
holds a bachelor’s degree in political
science from Hunter College, in New
York City.
says. The section and museum
created a long-term partnership to
help each other succeed.
In addition to the free visits,
IEEE members receive a 10 percent
discount on services offered by
the museum—which includes
digitizing books and other materials
and repairing broken equipment
such as radios and vintage record
players. Members can donate
historical artifacts too. In addition,
IEEE groups are welcome to host
conferences and meetings at
the facility.
The IEEE Switzerland Section as
well as members of student branches
and the local IEEE Life Members
Affinity Group have agreed to speak
at events held at the museum and
teach STEM classes there.
“The museum is a space where
both professional engineers and
young people can network and learn
from each other,” Vitacca says. “I
think this partnership is a win-win
for both IEEE and the museum.”
DECEMBER 2022  THE INSTITUTE  59

NEWS
Thomas M. Coughlin is 2023
IEEE President-Elect
BY JOANNA GOODRICH
IEEE LIFE FELLOW Thomas M.
Coughlin has been elected as the
2023 IEEE president-elect. He is set
to begin serving as president on
1 January 2024.
Coughlin, who was nominated by
the IEEE Board of Directors, received
10,908 votes in the election. Senior
Member Kathleen Kramer received
10,769 votes, Life Fellow Kazuhiro
Kosuge received 8,682, and Senior
Member Maike Luiken received 4,365.
At press time, the results were
unofficial. The IEEE Board of Directors
was scheduled to accept the IEEE
Tellers’ Committee report last month.
Coughlin is founder and president
of Coughlin Associates, in San Jose,
Calif., which provides market and
technology analysis as well as data
storage, memory technology, and
60  THE INSTITUTE  DECEMBER 2022
Coughlin’s Top
Three Goals as
President-Elect
1 Increase IEEE’s engagement
with members and enhance
the value of membership for
all member grades.
2 Create greater partnerships
across and outside of IEEE
to increase the organization’s
public impact.
3 Ensure that IEEE creates a
vibrant and safe environment
that supports its diverse
members.
business consulting services. He has
more than 40 years of experience in
the data storage industry and has been
a consultant for over 20 years. He has
been granted six patents.
Before starting his own company,
Coughlin held senior leadership
positions at Ampex, Micropolis, and
SyQuest.
He is the author of Digital Storage
in Consumer Electronics: The Essential
Guide, which is in its second edition.
He is a regular contributor on digital
storage for the Forbes blog and other
news outlets.
In 2019 he was IEEE-USA
president as well as the 2015–2016
IEEE Region 6 director. He also was
chair of the IEEE New Initiatives
and Public Visibility committees.
He was vice president of operations
and planning for the IEEE Consumer
Technology Society and served as
general chair of the 2011 Sections
Congress in San Francisco.
He is an active member of the IEEE
Santa Clara Valley (Calif.) Section,
which he chaired, and has been
involved with several societies and
standards groups, as well as the IEEE
Future Directions Committee.
As a distinguished lecturer for the
IEEE Consumer Technology Society
and IEEE Student Activities, he has
spoken on digital storage in consumer
electronics, digital storage and
memory for artificial intelligence, and
how students can make IEEE their
“professional home.”
Coughlin is a member of the
IEEE–Eta Kappa Nu (IEEE-HKN)
honor society.
He has received several
recognitions including the 2020 IEEE
Member and Geographic Activities
Leadership Award.
Coughlin is active in several other
professional organizations including
the Society of Motion Picture and
Television Engineers and the Storage
HARRY WHO PHOTOGRAPHY
Networking Industry Association.
To find out who was chosen as
IEEE-USA president–elect, IEEE
Technical Activities vice president-
elect, IEEE Standards Association
board of governors members-at-large,
and more, read the full results at
ieee.org/election.

NEWS
A Look at IEEE’s
Most-Cited Journals
EACH YEAR, THE Journal Citation Reports from
Clarivate examines the influence and impact of
scholarly research journals. Here is a look at some
of the top IEEE journals in their fields of interest,
according to the 2021 report released in June.
IEEE journals are No. 1 in:
• Automation and control systems (IEEE
Transactions on Cybernetics).
• Hardware and architecture (IEEE
Transactions on Neural Networks and
Learning Systems).
• Imaging science and photographic
technology (IEEE Geoscience and Remote
Sensing Magazine).
IEEE published eight of the top 10 journals in
electrical and electronic engineering.
1 Nature Electronics
2 IEEE Transactions on Pattern Analysis
and Machine Intelligence
3 IEEE Signal Processing Magazine
4 Proceedings of the IEEE
5 IEEE Transactions on Neural Networks
and Learning Systems
6 eTransportation
7 IEEE Vehicular Technology Magazine
8 IEEE Journal on Selected Areas in
Communications
9 IEEE Wireless Communications Magazine
10 IEEE Transactions on Fuzzy Systems
IEEE has nine of the top 10 journals in
telecommunications.
1 IEEE Communications Surveys and
Tutorials
2 IEEE Vehicular Technology Magazine
3 IEEE Journal on Selected Areas in
Communications
4 IEEE Wireless Communications Magazine
5 IEEE Network
6 IEEE Internet of Things Journal
7 IEEE Communications Magazine
8 Vehicular Communications
9 IEEE Transactions on Wireless
Communications
10 IEEE Transactions on Multimedia
SOURCE: The Journal Citation Reports from Clarivate
IEEE SERVICES
New Website
Helps Life Members
Stay Connected
BY KATHY PRETZ
THE NEW IEEE Life Members website is a one-stop resource to
keep the most experienced of members connected.
The life member designation is for those who have reached
the age of 65 and have been with IEEE for such a period of time
that the sum of their age and their years of membership equals
or exceeds 100. There are currently 35,000 life members.
There are many benefits to being a life member. Dues,
regional assessments, and society fees are waived. Life
members also receive reduced registration fees at conferences
for which IEEE is the sole sponsor. In addition, they can
continue to participate in the IEEE Member Discounts
program and retain all the other benefits of membership.
The new website is organized into five sections:
• Ways to get involved. There are many opportunities to
participate, including helping others connect with senior
leaders in the corporate and academic worlds, mentoring
the next generation of engineers, and getting involved in
technical and nontechnical interest groups.
• News and events. Members are kept informed through a
newsletter, announcements, and a calendar of upcoming
meetings.
• Awards and recognitions. Learn about the four awards that
the Life Members Committee bestows to honor the work of
its members. Also included is information about nominating
a colleague.
• Volunteers. For those who want to donate their time,
there are several ways to do so. One is to join a section’s
or council’s life member affinity group. The groups let life
members contribute to the social good in their community,
advance the professional interests of IEEE, and plan social
and technical events.
• Resources. This section of the website offers books,
publications, videos, and continuing-education materials to
help members stay technically current.
An online community for life members has been launched on
IEEE Collabratec, the organization’s professional networking
and collaboration platform. The IEEE Life Members and
Friends community provides opportunities to connect and
engage in conversations of mutual interest with others,
regardless of their physical location.
DECEMBER 2022  THE INSTITUTE  61
 PROFILE ARTI AGRAWAL, WHO is gay, didn’t

This Engineer
have access to LGBTQ+ support groups
when growing up in India, where
homosexuality was a crime until 2018.
“I had to live under the radar,” she
says. “It was very difficult and quite

Champions
traumatizing to live that way.”
The IEEE senior member found
solace in advocacy and support
organizations in London after moving

LGBTQ+
there in 2005. She went on to form her
own groups and lead diversity, equity,
and inclusion efforts at the universities
she worked for in London and Sydney.

Inclusivity
This year she started a consulting
firm, where she helps businesses
around the world improve their
DEI programs. Agrawal also has
spearheaded DEI campaigns for the
Arti Agrawal is helping companies become IEEE Photonics Society.
For her efforts, she received the
more diverse society’s 2020 Distinguished Service
Award.
BY JOANNA GOODRICH
Fighting for DEI
After graduating in 2005 from the
Indian Institute of Technology Delhi
with a doctorate in physics, Agrawal
decided to move to the United
Kingdom. “I didn’t think I could be
my true self at home,” she says, “and I
thought I could live more openly.”
She joined City, University of London,
that same year as a research fellow
working on computational photonics.
But she discovered there
were fewer women in science,
technology, engineering, and math
at the university than at her alma
mater. And, she says, the workplace
atmosphere wasn’t welcoming, and
she faced homophobia.
“Even though the public opinion
on gay rights in the U.K. was shifting,
it wasn’t like that within STEM,”
Agrawal says.
Agrawal got more involved in
support groups for members of
the LGBTQ+ community as well as
women in STEM. She joined the
London Gay Women’s Network, which
offers professional development
and networking opportunities
to community members who are
ARTI AGRAWAL

pursuing a college degree.

But, she says, she still felt isolated
because the groups had few members
from underrepresented ethnicities

62  THE INSTITUTE  DECEMBER 2022

and cultures. So she created her own
subgroup in 2009: the Gay Women’s
Network Multicultural. It provides
women of color a safe space to
connect with each other.
In 2011 she took a position at the
university as a lecturer and started her
own LGBTQ+ network there. The group
got school officials to install gender-
neutral bathrooms, and it won the right
for students and faculty members to
use their preferred pronouns on human
resource forms and in email.
In 2018 she left to join the
University of Technology in Sydney as
director of its Women in Engineering
and IT program. It works to increase
the enrollment of women in the
school’s engineering program and
offers mentoring, scholarships, and
professional development workshops.
While there, she shook up the
engineering school’s admissions
process. The percentage of women in
the undergraduate program “had been
stuck at around 16 percent for
more than 40 years,” she says. Employer
“No number of mentoring Vividhataa, in
programs, scholarships, Sydney
Title Chief
or industry outreach had
executive
changed this statistic.” Member grade
Agrawal pushed to Senior member
change the admissions Alma mater City,
policy. Prospective students University of
are ranked based on their London
Higher Secondary Certificate
scores. HSC is a subject-based
qualification whereby students take
three or four courses of their choice in
their last year of secondary school.
The top-ranking 100 students, no
matter their gender, were admitted
to the engineering program. Agrawal
recommended adding 10 points to
every female applicant’s rank.
“It doesn’t change their actual
marks,” she says. “It just changes
their rank.”
The policy change was approved
by several university committees
as well as by the New South Wales
Anti-Discrimination Board, which
handles citizen complaints. The board
approved the change for 10 years.
The number of women in the
program has increased by 10 percent
since the policy was implemented in
2019, Agrawal says.
“If you want to make a change in
a very difficult and unchanging
environment, you must be
disruptive. You have to do
something bold and brave and
have the courage to stand up to the
backlash that will come eventually.”
The change created a huge
backlash, Agrawal says. People said
“unqualified women would receive
degrees.” Many women were against
the change, saying it was doing them
a disservice.
“But there were people who
understood that the administration
wasn’t changing anyone’s
scores,” Agrawal says. Every
student who was admitted
met the minimum criteria,
she says; the change just
bumped up the ranking of
some qualified women.
“If you want to make a
change in a very difficult and
unchanging environment,
you must be disruptive,”
she says. “You have to do something
bold and brave and have the courage to
stand up to the backlash that will come
eventually.”
Launching a DEI startup
Agrawal left the university last year, and
this year she started her own diversity,
equity, and inclusion consulting
business, Vividhataa, in Sydney.
Vividhataa aims to help companies
set goals and create strategies to
achieve them.
It also is educating managers and
employees on gender identity, cultural
sensitivity, disability inclusion, and
related topics, Agrawal says.
Creating a safe space at IEEE
Agrawal has been involved in similar
efforts at IEEE and is a member of the
IEEE Photonics Society.
She became involved in the
IEEE Women in Photonics group
as its associate vice president. She
spearheaded a campaign to increase
the number of local chapters.
The society also held events
worldwide that addressed the needs
of established chapters. The campaign
grew to include hosting panels on DEI
policies at conferences and holding
leadership training.
Agrawal founded the society’s
diversity oversight committee in 2017
and became associate vice president.
She explored how to attract more
members from underrepresented
groups. She also focused on increasing
cultural, racial, and religious diversity.
In 2018 she helped organize the
society’s first Pride in Photonics
symposium. During the event,
LGBTQ+ engineers trained attendees
on how to be a good ally to members
of the community.
“It brought the community together
and helped reduce the sense of isolation
that people can feel,” she says.
The committee collaborated with
the U.S. National Society of Hispanic
Physicists, the National Society of
Black Physicists, and other groups
on joint conferences at historically
Black colleges and universities. The
groups are also working to increase
representation on IEEE’s committees
and councils.
“I thought the Photonics Society
was, and still is, a marvelous society,”
she says. “It has embraced diversity
and inclusion and made a lot of
headway in that area.”
DECEMBER 2022  THE INSTITUTE  63
 NEWS
IEEE Celebrates 75 Years
of the Transistor
BY JOANNA GOODRICH
THE TRANSISTOR, THE basic building
block of the electronics industry, was
developed in 1947 by John Bardeen,
Walter Brattain, and William Shockley at
Bell Labs in Murray Hill, N.J., to replace
the vacuum tube. The three honorary
IEEE members received the 1956 Nobel
Prize in physics for their work.
To mark the transistor’s
75th anniversary, the IEEE Electron
Devices Society is holding seminars
and panel discussions at IEEE
conferences, as well as presenting
webinars, giving away prizes, and
organizing other events from now until
the end of next year.
“These events are being done
at each conference to remind the
community how important the mighty
transistor has been over the past
75 years,” says IEEE Fellow Ravi Todi,
the EDS president.
Todi says he was excited about the
celebrations held at this year’s IEEE
International Electron Devices Meeting,
the society’s flagship conference, that
took place from 3 to 7 December. The
conference has been bringing together
IEEE members partnered with the Luminarias Community Center, in
Guatemala, to teach Indigenous students about the transistor.
64  THE INSTITUTE  DECEMBER 2022
engineers from academia and industry
for more than 70 years.
“A lot of the key inventions
that have to do with the transistor
have been first presented at this
conference,” Todi says. “IEEE IEDM
has played a key role in driving the
progress of the transistor.”
Several society members are
working on a book about the transistor
and its history. In addition, technical
articles are scheduled to run in IEEE
Spectrum and in the EDS Newsletter.
“Our aim is to teach and enlighten
members on the importance of the
transistor,” says IEEE Senior Member
Manoj Saxena, who is overseeing the
anniversary celebrations. “The events
shall be a tribute to the men and
women who have contributed to the
development of the transistor, which
has had a lasting impact on people’s
lives and has benefited mankind.”
Webinars and historical accounts
EDS is collaborating with the IEEE
Solid-State Circuits Society to mark
the anniversary. Since January, the
two societies have been presenting
a webinar series on the effect the
transistor has had on society and
technological advances. Every month,
a member who works in the field of
electron device engineering gives a
presentation on a technology impacted
by the transistor. Presentations so
far have covered 2D semiconductors,
chemical sensors, and 5G cellular
technology. The talks are available on
demand on the EDS website: eds.ieee.
org/education/webinars.
EDS is compiling a
commemorative book designed
to explore the development of the
transistor and how it has changed
during the past 75 years. Transistor
pioneers including Digh Hisamoto,
Eric Fossum, and Chenming Hu are
writing chapters for the book, which
is scheduled to be published early
next year. Hisamoto, an IEEE Fellow,
developed the 3D double-gate metal-
oxide-silicon field effect transistor in
1989. Three years later, Fossum, also
an IEEE Fellow, invented the CMOS
image sensor. An IEEE Medal of
Honor recipient and IEEE Life Fellow,
Hu is known as the “father of the 3D
transistor” for inventing the FinFET
in 1999.
“We want to capture historical
accounts of the device’s development
and growth,” Todi says.
IEEE EDS chapters are holding
events to celebrate the anniversary.
The activities include social gatherings,
technical talks, and historical seminars.
To learn more about upcoming
events in your area and how to
participate, visit eds.ieee.org.
The IEEE Electron Devices Society
EDS was established in 1952 as a
committee of the Institute of Radio
Engineers, one of IEEE’s predecessor
societies. After IRE’s merger in
1963 with the American Institute of
Electrical Engineers—which formed
IEEE—EDS was made an IEEE technical
group. It became a society 13 years
later.
EDS now has about 9,000 members
and 250 chapters worldwide.
“We pride ourselves on the fact
that a lot of the core inventions of
transistors were developed by IEEE
EDS members,” says Ravi Todi, the
IEEE
society’s president.

IEEE STANDARDS
What’s New in the
2023 National Electrical
Safety Code
BY THE IEEE STANDARDS ASSOCIATION
SINCE 1914, THE National Electrical
Safety Code has been a go-to standard
for electric and telecom utility
companies. The code is updated every
five years. The 2023 edition, which
was released on 1 August, will become
effective on 1 February.
The NESC updates better protect
workers, the public, and facilities
during the installation, operation,
and maintenance of power and
communications supplies. The
revisions are designed to help the code
remain relevant as new technologies
start to be used and safer ways of
working are discovered.
Edited and published by the IEEE
Standards Association and approved
by the American National Standards
Institute, the NESC has contributed to
major electrical safety codes in most
U.S. states, territories, and military
bases; the Caribbean; and other entities
and countries throughout the world.
Not to be confused with the
National Electrical Code, which
primarily addresses indoor wiring
of homes and businesses, the NESC
focuses on the outdoor lines that
connect to homes and businesses,
the electric supply stations and
telecommunications plants, and
ISTOCKPHOTO
their overhead and below-ground
structures.
Sections of the safety code cover
electrical grounding, substations,
overhead and underground power
lines, and work practices.
Roles dependent on the NESC
include:
• Utility company business
leadership, operations management,
engineering and line design, crew
supervisors, and safety trainers.
• Providers of power,
telecommunications, cable
television, wireless services, and the
Internet.
• Railroads, for their power and
signaling systems.
• Manufacturers of electrical
equipment and associated product
testing agencies.
• Consultants and contractors
that assist in the design and
construction of utility lines and
infrastructure.
• State and federal regulatory agencies
that provide industry oversight.
The updates
• New rules 190 through 195 cover
photovoltaic generating stations.
• Rule 116c adds an exception for
short lengths of insulated power
cables and short-circuit protection
if the situation involves fewer than
1,000 volts.
•
Rule 320B has been revised to
clarify separations that apply to
communications and supply in
different conduit systems.
• Table 410-4 is based on the latest
arc flash testing on live-front
transformers.
• Rule 092A adds an exception allowing
protection, control, and safety battery
systems to not be grounded.
• Rules 234 B1, C1, and D1 were
revised to better present vertical
and horizontal wind clearances, and
to coordinate requirements with the
new Table 234-7.
• Rule 120A was amended to provide
•
correction factors for clearances on
higher elevations.
Table 253-1 has been revised to
reduce the load factor for fiber-
reinforced polymer components
under wire tension—including dead
ends—for Grade C construction.
• Rule 410A now requires a specific
radio-frequency safety program for
employees who might be exposed.
Some long-standing sections have been
made clearer. For tables that include
both imperial and metric values, the
inch-foot-pound system is followed
by the corresponding metric values in
parentheses.
Working groups of NESC
subcommittees have been formed
to investigate up-and-coming topics
such as generating stations and fault-
managed power system cables, which
might be used for 5G networks.
Supplementing the code is the
NESC Handbook, which includes
commentary by contributors to help
users improve their understanding of
the code and how it might be applied.
Topics include code changes and rules
for installing and maintaining electric
supply stations. The handbook also
provides rules for the operation of
electric supply and communications
lines and equipment.
You can preorder the 2023 NESC
from the IEEE Standards store
(techstreet.com).
Also available is the NESC course
program, which offers an in-depth look
at the rules, regulations, and changes
made in the 2023 NESC edition.
You can access the courses
through the IEEE Xplore Digital
Library (for institutional customers)
or the IEEE Learning Network (for
individual learners).
DECEMBER 2022  THE INSTITUTE  65
 PROFILE JACK DONGARRA’S DREAM job

Turing Award
growing up was to teach science at a
public high school in Chicago.
“I was pretty good in math and
science, but I wasn’t a particularly good

Winner on His student,” Dongarra says, laughing.

After he graduated high school,
there was only one university he

Pioneering wanted to attend: Chicago State.

That’s because, he says, it was
known for “churning out teachers.”

Algorithms
Jack Dongarra’s software libraries spurred
the growth of high-performance computing
BY KATHY PRETZ
Chicago State accepted his
application, and he decided to major
in mathematics.
His physics professor suggested
that Dongarra apply for an internship
at the Argonne National Laboratory, in
Lemont, Ill., a nearby U.S. Department
of Energy science and engineering
research center. For 16 weeks he
worked with a group of researchers
designing and developing EISPACK,
a package of Fortran routines
that compute the eigenvalues and
eigenvectors of matrices—calculations
common in scientific computing.
Dongarra acknowledges he didn’t
have a background in or knowledge of
eigenvalues and eigenvectors—or of
linear algebra—but he loved what he
was doing. The experience at Argonne,
he says, was transformative. He had
found his passion.
The IEEE Life Fellow has since
made pioneering contributions to
numerical algorithms and libraries for
linear algebra, which allowed software
to make good use of high-performance
hardware. His open-source software
libraries are used in just about every
computer, from laptops to the world’s
fastest supercomputers.
The libraries include basic linear
algebra subprograms (BLAS), the
linear-algebra package LAPACK,
parallel virtual machines (PVMs),
automatically tuned linear algebra
software (ATLAS), and the high-
performance conjugate gradient
(HPCG) benchmark.
Dongarra is also known for his
Top 500 list, which he launched in
1993. The list provides a standardized
JACK DONGARRA (2)

measure of supercomputers during

the past 25 years.
For his work, he was honored this
year with the 2021 A.M. Turing Award
from the Association for Computing

66  THE INSTITUTE  DECEMBER 2022

Machinery. He received US $1 million
as part of the award, which is known
as the Nobel Prize of computing.
“When I think about previous
Turing Award recipients, I’m humbled
to think about what I’ve learned from
their books and papers,” Dongarra
says. “Their programming languages,
theorems, techniques, and standards
have helped me develop my algorithms.
“It’s a tremendous honor to be
this year’s recipient. The award is a
recognition by the computer-science
community that the contributions
we are making in high-performance
computing are important and have an From 1980 to 1989, Dongarra worked as a senior scientist at the
impact in the broader computer-science Argonne National Laboratory, in Lemont, Ill.
community and science in general.”
Instead of teaching science to high
school students, he became a professor “It was time for me to try out some of high-performance computing
of electrical engineering and computer new things,” he says. “I was ready to try in many areas including artificial
science at the University of Tennessee my hand at academia. The challenge intelligence, data analytics, genomics,
in Knoxville, where he taught for was becoming a university professor.” and health care.
33 years. The university recently University culture is very different “The libraries we designed have
named him professor emeritus. from that at a government laboratory, basic components that are needed in
he says, but he quickly fell into the many areas of science so that users
Entrepreneurial spirit rhythm of the academic setting. can draw on those components to
After graduating from Chicago State Although he loved teaching, he help them solve their computational
in 1972 with a bachelor’s degree in says, he also was attracted to the problems,” he says. “That software
mathematics, Dongarra went on to opportunity the university gave its is portable and efficient. It has all the
pursue a master’s degree in computer instructors to work on technology they attributes that we want in terms of
science at the Illinois Institute of are passionate about. being understandable and providing
Technology, also in Chicago. While “Academia gives you this freedom to reliable results.”
there he worked one day a week do things and not be constrained by a He’s currently working on creating
for Argonne with the same team of company’s drive or its motivation,” he a software library for the world’s
researchers. After he got his says. “Rather, I get to work fastest supercomputer, Frontier,
degree in 1973, the lab hired Employer on what motivates me. That’s which recently was installed at
him full time as a researcher. University of why I’ve stayed in academia the Oak Ridge lab. It is the first
He left the lab to study Tennessee, for so many years.” computer that can process more than
applied mathematics at the Knoxville Dongarra founded the 1 quintillion operations per second.
Title Professor
University of New Mexico in emeritus,
university’s Innovative
Albuquerque. He honed his computer science Computing Laboratory, Computer-science recognition
knowledge of linear algebra Member grade whose mission is to Dongarra has been an IEEE member
there and began working Life Fellow provide tools for high- for more than 30 years.
out algorithms and writing Alma mater performance computing to “I enjoy interacting with the
Chicago State
software. University
the scientific community. community,” he says in explaining why
He returned to Argonne He also directs the school’s he continues to belong. “Also I enjoy
after getting his doctorate Center for Information reading IEEE Spectrum and journals
in 1980 and worked there as a senior Technology Research. that are relevant to my specific field.”
scientist until 1989, when he got the He is now a distinguished IEEE is proud of Dongarra’s
opportunity to fulfill his dream of researcher at Oak Ridge. contributions to computing and has
teaching. honored him with several awards over
He was offered a joint position Software for supercomputers the years. Most recently he received
teaching computer science at the It was working in creative the 2020 Computer Pioneer Award.
University of Tennessee and conducting environments that led Dongarra to “I’m very happy and proud to be
research at the nearby Oak Ridge come up with what many describe as a member of IEEE,” he says. “I think
National Laboratory which, like Argonne, world-changing software libraries, it provides a valuable service to the
is a Department of Energy facility. which have contributed to the growth community.”

DECEMBER 2022  THE INSTITUTE  67


IEEE PRODUCTS
Podcast Explores AI’s
Use in Health Care
BY THE IEEE STANDARDS
ASSOCIATION
With the rush toward using artificial
intelligence, machine learning, and
deep learning technologies to analyze
IEEE SERVICES
STEM Activity Kits Are a Hit
in Hundreds of Libraries
BY KATHY PRETZ
MORE THAN 150 public libraries
throughout the central United States
now lend out activity kits that let
children explore just about any aspect
of science, technology, engineering,
and mathematics. The kids can check
These youngsters are checking out
one of their local library’s IEEE-
funded science activity kits.
68  THE INSTITUTE  DECEMBER 2022
health data for insights, questions
are being raised about how good a
job the technologies are doing to
improve outcomes.
Technologists, clinicians,
researchers, scientists, ethicists, policy
stewards, and other experts offer their
thoughts during the third season of the
Re-Think Health Podcast, AI for Good
Medicine. The series is part of the IEEE
Standards Association’s Healthcare
and Life Sciences practice (standards.
ieee.org/practices).
Some of the topics they covered
include whether AI is good for
marginalized patients, using the
technology for precision oncology,
and the use of AI as a pipeline for
targeted drug discovery.
Here are the top five insights from
the experts.
them out just like they would a book.
The kits teach youngsters what
engineers do, as well as how to code,
build robots, design video games, and
create animations.
The collections have been made
possible by the IEEE Region 4 Science
Kits for Public Libraries program with
funding from Region 4 members, IEEE
entities, and corporate sponsors. The
SKPL program is the brainchild of IEEE
Life Senior Member John A. Zulaski,
who chairs the SKPL committee.
Activity kits aren’t new to libraries,
but STEM kits didn’t exist 10 years
ago. Nowadays large, well-funded
libraries might have them, but that’s
not the case for many small and
midsize ones, Zulaski says. The kits
get into the hands of children who may
not have an opportunity to learn about
STEM otherwise.
1 Data is a depreciating asset.
The longer it sits, the less
value it has.
2 Data is an abyss. If you want AI
to make an impact on the health
care system, then make data
reliable by design.
3 Fairness is not a math problem.
Equity in health care is not about
the technology but rather the
approaches taken to make health
care accessible to all.
4 Social determinants of health have
significant, if not equal, value to
diagnostic health data in closing
the health care gap with AI.
5 Make explainable AI transparent
and off the shelf so that
clinicians understand how the
algorithms are addressing the
questions in the data to help them
arrive at the insights needed.
Currently all 80 branches of the
Chicago Public Library have the kits,
as do libraries elsewhere in Illinois,
and in Indiana, Iowa, Minnesota,
Michigan, Nebraska, North Dakota,
Ohio, and Wisconsin. The kits are
tailored to the needs of the local
children, Zulaski says.
Kids aren’t the only ones checking
out the collections. Teachers, Boy Scout
leaders, parents who home-school their
children, and libraries that have in-
house STEM programs do as well.
Zulaski says he is on a mission to
encourage other IEEE entities to make
the kits available at local libraries.
Only public libraries in IEEE Region 4
currently are eligible for an SKPL grant.
“The kits get kids interested
in pursuing a technical career—
which, by the way, helps increase
IEEE membership down the road,”
TOP: ISTOCK; BOTTOM: JOHN A. ZULASKI
Zulaski says.
The 2023 submission period for
SKPL grant applications closes on
15 January.
If your IEEE entity is interested in
offering the SKPL program, complete
the form on the IEEE Region 4
website: r4.ieee.org/skpl. You can
also donate to the program from
the website.
 Store employees
take a picture of a
product on display
using Wisy's platform
and the AI records
information based on
the photo.

STARTUP SHOPPERS AROUND THE WORLD have

Wisy’s AI
experienced going to the store to buy
their favorite shampoo or candy bar
only to find an empty shelf. Because
of inflation and high frequency of
employee resignations, retailers

Keeps Store
are struggling to keep their shelves
stocked. Supply-chain issues caused
in part by the COVID-19 pandemic
have added more challenges in keeping

Shelves
shelves stocked and increased the
product-unavailability rate from
5 percent to 15 percent during the past
three years, according to the Consumer

Stocked
Brands Association.
Startup Wisy developed an AI
platform to make it easier for stores
to track whether there are available
products that haven’t yet been put on
Image recognition lets the platform see display. It uses image recognition to
detect which items need to be restocked.
when products are running low “We are not only solving a
customer-experience problem but
BY JOANNA GOODRICH
also a sustainability problem,” says
IEEE Senior Member Min Chen,
WISY PLATFORMS

Wisy cofounder and CEO. “All those

products that are not sold because
they were not displayed get thrown
away. WisyAI enables store employees

DECEMBER 2022  THE INSTITUTE  69

 to quickly get information about Adding to the stores’ challenges
the stock, reduce losses, and sell were worker shortages.
[products] more effectively.” “It is quite a challenge for our
The startup, headquartered in clients because they have a high
San Francisco, won this year’s CCU number of vacancies, so they have to
(Compania Cervecerias Unidas) Innpacta do more work with less people now,”
Global Open Innovation Challenge in Chen says. “Training employees on
Chile. The competition is for startups how to do in-store placement and
that have designed technologies for the tracking of products also takes time—
consumer goods and retail industry. Wisy even more so with high numbers of
is piloting its AI platform with CCU. The employee turnover.”
company also is included in the startup
portfolio for ChileGlobal Ventures, a Keeping shelves stocked
startup accelerator. Chen says she knew Wisy’s platform
Wisy currently has 16 employees, IEEE Senior Member Min Chen is had to be “available, reliable, fast,
who work in Costa Rica, Panama, Peru, Wisy's cofounder and CEO. and flexible.” It also had to work for
the Philippines, and the United States. large companies as well as mom-and-
Chen is based in Panama City. To learn what the issues were, Wisy pop shops, and without an Internet
Wisy released the platform in April. conducted on-site observations and connection.
interviewed employees about their Wisy developed a platform that
Unreliable Internet connection process for collecting product data. uses AI to process data regardless of
Chen, a software engineer, founded “Usually consulting companies whether there is access to the Internet.
her first company, Alcenit, in 2006. do this,” Chen says, “but we felt it The service itself, which works on a
The consulting business worked with was important for us to do it as the mobile phone or tablet, has the power
companies in the banking, retail, and platform developers.” to process data and produce results in
oil-and-gas industries to help them They found that retail employees milliseconds no matter where the user
manage their IT department. One of manually tracked and verified what is in the store.
Alcenit’s clients— products were available in the store, “With Wisy’s platform, companies
a supermarket chain—was using cloud-based platforms. can be assured that their teams have
having problems collecting Employees counted each a reliable tool to support their work,”
Founded 2017
reliable ground-truth data, or product and recorded Chen says.
Headquarters San
data collected at scale from Francisco information about the item, Images of each shelf and product,
real-world scenarios. Founders Min such as pricing, on a platform. as well as their bar code, are uploaded
Chen discovered that the Chen, Ricardo Manually counting the to the platform. Store associates take a
retailer wasn’t the only client Chen, Nelida products and filling out picture of a product on display, and the
Gomez, Orlando
struggling with product data the information was a slow AI records information based on the
Reyes
collection. Employees 16 process, and employees photo. It takes less than one second.
“We had customers in were prone to making errors, WisyAI can work as a stand-alone
the oil-and-gas industry, Chen’s team found. The cloud system or be integrated with the store’s
as well as construction companies, providers that the retail stores used previous product tracking platform.
dealing with similar problems,” she often were unreliable, she says, and If there is a shortage of a certain
says. She and three of her coworkers the Internet connection was regularly product in the shop, WisyAI notifies
decided in 2016 to launch Wisy to slow or unavailable, especially in stores’ a store associate. Should the platform
develop a solution. basements and warehouses. detect that the product is out of stock
Wisy decided to create the “If employees didn’t have a good or the inventory is running low, it
technology for supermarkets and other Internet connection, they wouldn’t prompts an employee to order more.
retailers, she says, because the tracking be able to see the results until the It also can track when products
process they used caused problems that connection was reestablished,” she were delivered—which is helpful if a
were “big enough, repetitive enough, says. “If their connection was slow, customer is searching for a specific
and complex enough to solve using AI.” then getting the results was slow.” shampoo, for example, but it isn’t on
the shelf. Employees can use WisyAI
to find out whether the shampoo is in
the store’s storage area or is, in fact,
“We are not only solving a out of stock.
The platform uses the data it
customer-experience problem but collects to predict when items will be
LIZ PINTO

delivered and when they will go out of

also a sustainability problem.” stock, based on their popularity.

70  THE INSTITUTE  DECEMBER 2022


The Department of Electrical & Computer Engineering
(ECE) at Boston University (BU) has multiple openings for
a Tenure-Track Assistant Professors with an anticipated
start date of Fall 2023. While all candidates will be
considered, this year’s search is particularly focused on
the following area:
Computer Systems including candidates in operating
systems, compilers, cybersecurity, or software engineering
to build future cloud and edge computing systems
for a smart, secure, and connected society. Potential
application areas include healthcare, communications,
transportation, finance, and scientific computing.
The College of Engineering at Boston University has
embarked on a bold new strategic plan that will pursue
excellence and impact along convergent and collaborative
research themes while remaining committed to our
institutional values regarding diversity, equity, and
inclusion. The College has an opening in a relevant
department for a Tenure-Track Assistant Professor with a
focus on the following convergent theme:
Photonics and Computing including candidates in all
areas of enabling device technologies, photonic integrated
circuits, and optical neural networks for application in
artificial intelligence, computational optics, computer
vision, microscopy, robotics and other computing tasks.
We are looking for outstanding candidates who have
earned or are expected to earn a Ph.D. in the relevant
search area before Fall 2023, demonstrate potential for
leading an independent and vibrant funded research
program in their area of expertise, can teach effectively
at the graduate and undergraduate levels, and can utilize
their expertise to strengthen collaborative research
within the department and beyond.
For more information about BU ECE, and to apply,
please visit: http://www.bu.edu/ece/ Applicants should
submit a brief letter of interest, a statement of research
accomplishments and goals including how collaborations
can enhance their research impact, teaching and diversity
statements, a current CV, and contact information
of three references using the appropriate link below.
We encourage candidates to apply early. Applications
received by December 15, 2022 will be given full
consideration. Boston University is an Equal Opportunity
Affirmative Action Employer.
Boston University is an equal opportunity employer and
all qualified applicants will receive consideration for
employment without regard to race, color, religion, sex,
sexual orientation, gender identity, national origin, disability
status, protected veteran status, or any other characteristic
protected by law. We are a VEVRAA Federal Contractor.
Tufts University: School of Engineering:
Electrical and Computer Engineering Associate Professor in
Electrical or Computer Engineering
The Department of Electrical and Computer Engineering at Tufts
University invites applications for tenure-track faculty positions in
Electrical and Computer Engineering to begin in September 2023.
We are seeking candidates at the rank of Associate Professor, but
exceptional candidates at the rank of Assistant or Full Professor will
also be considered.
This search is part of a broader hiring campaign by the School of
Engineering to grow its research and educational capabilities in
engineering for sustainability. The School of Engineering is deeply
committed to developing pragmatic, innovative solutions for
environmentally and socially responsible use of energy and materials
applying methods that minimize environmental damage to the natural
environment. Currently, tenure-stream faculty searches are ongoing
in four of the six academic departments centered on the common
themes of climate change and sustainable materials. Candidates
applying for this position should express how their research goals
align with this sustainability theme.
Candidates are sought primarily in one of the following areas or
closely related fields:
• Computer Engineering with focus including, but not limited
to, security; embedded and real-time systems; computer
architecture; and computer systems. In these areas, we are
interested in new emerging applications such as mobile and
wearable computing; quantum computing; machine learning;
autonomous vehicles and robotics.
• Electronic materials with focus on epitaxial growth (e.g. MBE)
and fabrication of novel devices for emerging applications:
e.g. renewable energy, power electronics, millimeter wave and
terahertz, photonics, quantum computing, spintronics, and
other next generation technologies.
Electrical and Computer Engineering has grown significantly in the
past twenty years. Located in the Boston area, the department benefits
from outstanding undergraduate and graduate students, collaborative
faculty, and cross-disciplinary research opportunities. Tufts
University is one of the smallest universities ranked as a Research 1
university, and it offers the best of a liberal arts college atmosphere,
coupled with the intellectual and technological resources of a major
research university. Tufts University supports and encourages a
culture of interdisciplinary research and there are numerous such
opportunities within the School of Engineering, the School of Arts
and Sciences, and through graduate and professional schools.
Full details see: http://apply.interfolio.com/114174
Tufts University is an Equal Opportunity/Affirmative Action
Employer. We are committed to increasing the diversity of our
faculty and staff and fostering their success when hired. Members
of underrepresented groups are welcome and strongly encouraged
to apply. See the University’s Non-Discrimination statement
and policy here https://oeo.tufts.edu/policies-procedures/non-
discrimination/. If you are an applicant with a disability who is
unable to use our online tools to search and apply for jobs, please
contact us by calling the Office of Equal Opportunity (OEO) at 617-
627-3298 or at oeo@tufts.edu. Applicants can learn more about
requesting reasonable accommodations at https://oeo.tufts.edu/
DECEMBER 2022  SPECTRUM.IEEE.ORG  71
 Assistant, Associate, or Full Professor (three positions: 84016, 84022, and 84035).
University of Hawai‘i at Mānoa (UHM), Department of Electrical and Computer Engi-
neering, invites applications for three full-time, tenure-track faculty positions, pending
position clearance and availability of funds. To begin approximately August 1, 2023 or
soon thereafter.
The University of Hawai‘i at Mānoa, a Carnegie R1 research university, is a top-50
Tenure Track Faculty Positions
public university dedicated to providing world-class teaching, research, and service in a The UNC Charlotte Department of Electrical
multicultural and inclusive environment. Collaboration, funding opportunities, resourc- and Computer Engineering (ECE) invites
es, and research exposure may be found through the department’s involvement and applications for two tenure-track Assistant
affiliation with the $5 million NSF CyberCorps Scholarship for Service Program (SFS); or Associate Professor positions starting
the NSA/DHS National Center of Academic Excellence in Cyber Defense Research (CAE-R); the NSF AI from Fall 2023. Candidates are sought
Research Institute (focusing on dynamic systems); the NSF Center for Science of Information (focusing on
big data, information theory, and machine learning); and Ike Wai, a $20 million NSF EPSCOR project for bio/ to fill positions in (1) Communications
nano sensing. Further information is available at https://ee.hawaii.edu. networks (position #004414) with
expertise in computer networks, mobile and
Search Area(s): We are seeking candidates with a strong research record in the area of cybersecurity, in- wireless networks, with special interests
cluding system and hardware security, software and application security, network security, cyber forensics,
critical infrastructure security, secure and private data services, or related fields. Or strong research record in in current and emerging topics such as
the area of computer engineering and applications in cyber-physical/IoT systems, Internet/mobile/embedded IoT, 5G networks and beyond, machine
systems and applications, computer systems, robotic systems and applications, autonomous vehicles, cloud/ learning for communications, networking
edge computing, big-data system and applications, data science/artificial intelligence, next-generation wire- infrastructure for smart cities, software
less systems, or related fields. For all positions, exceptional candidates in all areas are encouraged to apply. defined networks, and related areas; and
Duties: Teach and develop courses in electrical or computer engineering, develop an extramurally funded (2) Semiconductor devices (position
research program, publish outstanding work in leading scholarly journals, supervise graduate students, and #004453) with expertise in electronic,
provide departmental, college, and university service. optoelectronic, power electronic, nanoscale
Minimum qualifications: An earned Ph.D. in Electrical or Computer Engineering, Computer Science, or a and novel devices. Exceptional candidates
closely related discipline, with a strong research track record. All-But-Dissertation cases will be considered in other areas are encouraged to apply.
but the degree must be earned before the date of hire. Candidates should be committed to
Apply: For complete duties, qualifications, and application instructions, refer to excellence in teaching at the undergraduate
and graduate levels, development of
sponsored research programs, supervision
of student research, student mentoring,
and academic services. Priority will be
https://go.hawaii.edu/et2 given to candidates having original and
promising research as demonstrated by
Continuous Review: Application reviews will begin on Jan. 10, 2023, and will continue until the positions are peer-reviewed publications, potential for
filled. The University of Hawai‘i is an equal opportunity/affirmative action institution and encourages applica- success in extramural funding, a plan for
tions from women and minority candidates. future research that integrates well with
existing department strengths, and the
ability to contribute to diversity initiatives.
A Ph.D. degree in Electrical or Computer
Engineering or equivalent is required by the
time of contract.
The ECE department offers B.S. and M.S.
degrees in both Electrical Engineering
The Global Institute of Future Technology (GIFT and Computer Engineering, and Ph.D.
Faculty Position in Electrical Engineering degrees in Electrical Engineering, with
Department of Electrical, Computer, hereafter) at Shanghai Jiao Tong University
sincerely invites applications for tenure-track approximately 650 undergraduate and
and Systems Engineering 175 graduate students; see http://ece.
Case Western Reserve University, or tenured professoriate and research scientist
positions, mainly at Associate and Full levels, uncc.edu. The ECE Department currently
Cleveland, Ohio comprises 37 full time faculty members
related to Sustainable Energy.
The Department of Electrical, Computer, and who are actively engaged in research and
Systems Engineering at Case Western Reserve
GIFT manages endowments from a variety of teaching in the broad areas of computer
University (CWRU) invites applications for
sources, including a recent donation of more systems, communications, controls, signal/
a tenure-track faculty position in Electrical than 200 million US dollars. GIFT’s new build- image processing, electronics devices, and
Engineering at the Assistant Professor level. ing will be completed within two years, adding power and energy systems.
50,000 square meters to the current space.
Appointments will be considered for starting Applications must be made electronically
dates as early as July 1, 2023. Candidates must With strong government and university support,
at https://jobs.uncc.edu for Position
have a Ph.D. degree in Electrical Engineering or we will rapidly expand our faculty, including #004414 (communication networks) or
a related field. high-level hiring (e.g. NAE/NAS members and Position #004453 (semiconductor devices).
The faculty search is focused on the broader Fellows of multiple professional societies). At Applicants should submit a complete
area of robotics. The department is particularly present, five research centers have been estab- resume/curriculum vitae, a cover letter/
interested in candidates in the field of robotics lished: Large-scale Energy Storage and Reliability letter of intent, a statement of teaching and
whose focus is in, or at the intersection of, robot Center, Solid-state Battery Center, Advanced Sus-
perception, machine vision, machine learning,
research experience/goals, and contact
tainable Power Systems Center, Green Energy information of at least three references. The
and robot autonomy. Candidates whose research and Modern Agriculture Center, and Perovskite
has applicability to areas such as shared candidate chosen for this position will be
Photovoltaics Center. required to provide an official transcript of
autonomy, human-in-the-loop systems, assistive
robotics, or intelligent multimodal robotic Applicants must hold a doctorate degree in a rel- their highest degree earned. The candidate
perception and manipulation are highly desirable. evant field. Candidates are expected to establish a chosen for this position will be subject to
Additional information about the position, vigorous research program and contribute to un- a criminal background check. Review of
department, and application package is dergraduate and graduate education. For full con- applications will begin December 1, 2022,
available at https://engineering.case.edu/ecse/ sideration, please send a CV, teaching plan and and the position will remain open until
employment. research plan, copies of three representative pub- filled. Women and minority candidates are
CWRU provides reasonable accommodations to lications and contact information of five referees encouraged to apply. Candidates having
applicants with disabilities. Applicants requiring as a single PDF file to the GIFT Search Committee demonstrated ability to contribute to
a reasonable accommodation for any part of the (email: gift-facultysearch@sjtu.edu.cn). For more diversity initiatives are encouraged to apply.
application and hiring process should call 216-
information, please visit http://gift.sjtu.edu.cn. The University of North Carolina at Charlotte is
368-3066.
an EOE/AA employer and an ADVANCE Institution.

72  SPECTRUM.IEEE.ORG  DECEMBER 2022
Applications are invited for:-
Department of Mechanical and Automation Engineering
Professors / Associate Professors / Assistant Professors
(Ref: 220002KJ)
The Chinese University of Hong Kong (CUHK) is ranked one of the
top 50 universities worldwide according to the QS and THE World
University Rankings of 2022/23. It is also named the Most Innovative
University in Hong Kong by Thomson Reuters in four consecutive
years (2016-19). In the 2020 Research Assessment Exercise,
100% of our impact cases achieved the highest rating 4*, meaning
“outstanding impacts in terms of their reach and significance”. The
overall quality in mechanical and production engineering of CUHK is
also the highest, up to 94%, among all universities in Hong Kong by
adding both categories of 4* (world leading) and 3* (internationally
excellent). Further information about the Department is available at
http://www.mae.cuhk.edu.hk.
The Department of Mechanical and Automation Engineering (MAE)
at CUHK is seeking excellent candidates in the following areas:
• Robotics, in particular with expertise of medical sensors, robot
actuators, and soft robotics with connection to the CUHK T Stone
Robotics Institute, which focuses its primary research on medical
and service robotics through collaboration among experts in
engineering, medicine and social science;
• Design and manufacturing: CAD/design optimization and
automation, digital manufacturing;
• Dynamic systems and control;
• Engineering materials and structures;
• Green building and building automation, environmental
engineering.
Applicants should have (i) a PhD degree in Mechanical Engineering or
a related discipline; and (ii) a proven record of academic scholarship
and high potential for excellence in teaching and research.
The appointees will (a) teach undergraduate and postgraduate courses;
(b) develop an externally funded high impact research programme;
(c) supervise postgraduate students; and (d) provide service to the
Department/Faculty/University, professional organizations and the
community.
Similar to tenure track positions at universities in USA, appointments
will normally be made on contract basis for up to three years initially
commencing August 2023, which, subject to mutual agreement,
may lead to longer-term appointment or substantiation. Outstanding
candidates with substantial experience for Professor rank may be
considered for substantive appointment forthwith. The rank and exact
start date will be negotiated with the successful applicants.
[Those who have responded to the previous advertisement for the
posts (Ref. 210002DC) are under consideration and need not re-apply
in this instance.]
Application Procedure
Applicants please upload the full resume with a cover letter, copies
of academic credentials, publication list with abstracts of selected
published papers, details of courses taught and evaluation results
(if available), a research plan, a teaching statement, together with
names, addresses and e-mail addresses of three to five referees for
providing references.
The University only accepts and considers applications submitted
online for the post above. For more information, please contact
Ms. YL Kan at ylkan@mae.cuhk.edu.hk, or visit http://career.cuhk.edu.hk
to apply online.
Ming Hsieh Department of
Electrical and Computer Engineering
University of Southern California
Faculty Positions
Ming Hsieh Department of Electrical and Computer Engineering
The University of Southern California, one of the nation’s top research universities,
invites applications for tenured and tenure-track positions in the Ming Hsieh
Department of Electrical and Computer Engineering (https://minghsiehece.usc.
edu/) in the USC Viterbi School of Engineering. We are looking for outstanding
faculty candidates in all areas of Electrical and Computer Engineering at all ranks.
The USC Viterbi School of Engineering is committed to increasing the diversity of our
faculty and welcome applications from women, those of African, Hispanic and Native
American descent, veterans, and individuals with disabilities.
While outstanding candidates from all areas of electrical and computer
engineering will be considered, candidates with research interests in the following
areas are especially encouraged to apply: trust/privacy/security, experimental
quantum engineering, circuits and systems for AI at the edge, computing
for ML and AI at scale, energy-efficient sensing and computing, bio-sensors
and bio-interface circuits and systems, computational imaging systems,
foundations of machine learning, and human-centered signal processing.
Faculty members are expected to teach undergraduate and graduate courses, mentor
undergraduate, graduate, and post-doctoral researchers, and develop a strong
funded research program. Interdisciplinary and collaborative research is strongly
encouraged. Applicants with interests in research and outreach activities to increase
student and faculty diversity are particularly encouraged to apply. Applicants must
have a Ph.D. degree, or the equivalent, in electrical and computer engineering or a
related field and a strong research and publication record. Applications must include a
letter clearly indicating area(s) of specialization, a detailed curriculum vitae, a concise
statement of current and future research directions, a teaching statement, and contact
information for at least four professional references. Applicants are encouraged to
include a succinct statement on fostering an environment of diversity and inclusion.
This material should be submitted electronically at https://usccareers.usc.edu/
job/los-angeles/open-rank-assistant-associate-full-professor-of-electrical-and-
computer-engineering/1209/37961465472/. Review of applications will begin
immediately. Applications submitted after January 15th, 2023, may not be considered.
The USC Viterbi School of Engineering is among the top tier of engineering schools
in the world with 189 full-time, tenure-track faculty members. The school is home to
the Information Sciences Institute and affiliated with the Alfred E. Mann Institute for
Biomedical Engineering, the Institute for Creative Technologies, and the USC Stevens
Center for Innovation. Research expenditures typically exceed $210 million annually.
USC is an equal opportunity, affirmative action employer. All qualified applicants will receive
consideration for employment without regard to race, color, religion, sex, sexual orientation,
gender identity, national origin, protected veteran status, disability, or any other characteristic
protected by law or USC policy. USC will consider for employment all qualified applicants with
criminal histories in a manner consistent with the requirements of the Los Angeles Fair Chance
Initiative for Hiring ordinance.
DECEMBER 2022  SPECTRUM.IEEE.ORG  73

The Department of Electrical and Computer Engineering (ECE) in the
Norm Asbjornson College of Engineering (NACOE) at Montana State
University invites applications for a full-time tenure-track position
at the level of assistant professor as part of the NACOE’s College-
wide Resilient and Sustainable Communities Initiative (https://coe.
montana.edu/cohort_hire.html). Candidates are encouraged to
explain a vision of how their research aligns with topics of the initiative.
The department has strength in both traditional and emerging areas,
and we seek individuals who are excited about opportunities for
applied research, hands-on teaching, and professional service as part
of the Land-Grant mission. As an integral part of our R1 university,
competitively-awarded research expenditures in the ECE Department
from external sources last year topped $5.3M, with funding from
NSF, DoD, NASA, DoE, and many other public and private agencies.
Enrolling 450 students, the Department offers undergraduate degrees
in Electrical Engineering and in Computer Engineering, and graduate
MS and PhD degrees in EE. https://ece.montana.edu/
About Montana State University: MSU-Bozeman is an R1 university
minutes away from public trails and blue-ribbon trout streams, less
than an hour from Big Sky and Bridger Bowl ski resorts, and just
90 miles from Yellowstone National Park. Bozeman also is the home
of more than 100 small high-tech businesses and 650 non-profit
community organizations. You can build an outstanding academic
career in the fresh air of the northern Rocky Mountains.
Please visit MSU’s Job Page https://jobs.montana.edu/ for the
complete announcement and application procedures.
Connect with thousands
of your colleagues in the
IEEE AuthorLab today!
collabratec.ieee.org
74  SPECTRUM.IEEE.ORG  DECEMBER 2022
Tenure-Track Faculty Positions in Engineering and Science
Zhejiang University -University of Illinois at Urbana-Champaign Institute
The Zhejiang University/University of Illinois at Urbana-Champaign Institute
(the ZJU-UIUC Institute) invites highly qualified candidates for multiple ten-
ure-track faculty positions at all levels from Assistant Professor to Full Profes-
sorship, and in areas of engineering and science that match its multidisciplinary
mission. The Institute is an engineering college of Zhejiang University (ZJU),
China. Candidates should have exceptional academic record or demonstrate
strong potential in the cutting-edge research areas of engineering and science
multidisciplinary technologies.
Duties: Established as a world-class research institute, it conducts undergradu-
ate and postgraduate teaching and research, and conducts classes and student
activities in English.
Applications are particularly encouraged from candidates whose interests ad-
dress interdisciplinary topics exemplified by Computer and Digital Engineering,
Artificial Intelligence, System and Networking, Intelligent Transportation Sys-
tem, Environmental Sustainability, Intelligent Manufacturing, Biomechanics,
Control, Power and Energy, Robotics, Materials Science and Engineering.
Successful candidates will initiate and lead collaborative research and perform
academic and professional service duties associated with the ZJU-UIUC Insti-
tute. They will be leaders for teaching and research innovation, giving students
a meaningful and interactive engineering education.
Faculty at the institute will serve as Adjunct faculty in University of Illinois at
Urbana-Champaign (UIUC).
Compensation and benefits: Salary and research initiation support will be com-
mensurate with qualifications and competitive with international norms, includ-
ing housing benefits.
Application materials should include a cover letter with current contact informa-
tion including email address, as well as complete curriculum vitae, statements
of research and teaching goals, and contact information for three or more refer-
ences. Please submit applications at https://my.zjui.illinois.edu/submit/login.
asp or at zjuihr@zju.edu.cn. For more information, please visit job opportunities
on http://zjui.zju.edu.cn.
One of the most
influential reference
resources for engineers
around the world.
For over 100 years, Proceedings of the IEEE
has been the leading journal for engineers
looking for in-depth tutorial, survey, and
review coverage of the technical
developments that shape our world.
To learn more and start
your subscription today, visit
ieee.org/proceedings-subscribe

Harness the
publishing power
of IEEE Access.
®
• Receive high-quality, rigorous peer
review in only 4 to 6 weeks
• Reach millions of global users through
the IEEE Xplore® digital library by
publishing open access
• Submit multidisciplinary articles that
may not fit in narrowly focused journals
Learn more at ieeeaccess.ieee.org
Faculty Positions in Electrical and Computer Engineering
The Department of Electrical and Computer Engineering at San Diego
State University has recently received a $3.5M gift from Eric and Peg-
gy Johnson to establish The fred harris Endowed Chair in Digital
Signal Processing. This endowment is to honor emeritus profes-
sor fred harris and his legacy of excellence and teaching in digital
signal processing. Applications are invited for a tenured, endowed
full professor position in the broad area of digital signal processing,
with an anticipated start date of August 2023. The areas of interest
include but are not limited to, digital signal processing, artificial intel-
ligence, machine learning, internet of things, audio, image, and video
processing, signal processing over networks, signal processing in
communication systems, biomedical signal processing, and signal
processing in cyber physical systems. The applicants must hold a
tenured associate or full professor appointment with a Ph.D. in Elec-
trical Engineering, Computer Engineering, or closely related disci-
pline, with an outstanding track record of scholarship and externally
funded research. Additional information and application procedure
are available at http://apply.interfolio.com/114232. Inquiries can be
sent to Professor Sunil Kumar, Search Committee Chair, at skumar@
sdsu.edu.
The Department of Electrical and Computer Engineering at San Diego
State University also invites applications for a full-time tenure-track
faculty position in Optics/Photonics at the rank of Assistant Profes-
sor, with an anticipated start date of August 2023. Qualified appli-
cants must have expertise in optics and photonics including but not
limited to the following areas: optical communication networks; inte-
grated photonics and optoelectronics; imaging, sensing, and display;
optical AI and computational photonics. Applicants must hold a Ph.D.
in Electrical Engineering or closely related discipline. The ability to in-
teract with other research groups and create/maintain ties with major
industrial players within the broader optics and photonics community
will be an asset. Additional information and application procedures
are available at https://apply.interfolio.com/114462. Inquiries can
be sent to Professor Ege Engin, Search Committee Chair, at aengin@
sdsu.edu.
The successful candidate will be expected to establish and maintain
a strong externally funded research program, achieve excellence in
teaching at the undergraduate and graduate levels, advise students,
and participate in departmental governance.
The Department of Electrical and Computer Engineering is strongly
committed to excellence in both research and teaching at the grad-
uate and undergraduate levels. The department offers ABET-accred-
ited B.S. degree programs in Electrical Engineering and Computer
Engineering, M.S. programs in Electrical Engineering and Computer
Engineering, and a joint Ph.D. program in Electrical Engineering. The
ECE Department has over 30 full-time and part-time faculty members,
including 3 IEEE Fellows. Research areas in the department include
analog and digital integrated circuits, computer networks, embedded
systems, signal processing and communication systems, antennas,
RF and electromagnetic compatibility, machine learning and artificial
intelligence, power electronics and smart grid. Additional information
about the department and university can be found at http://electrical.
sdsu.edu/ and http://www.sdsu.edu.
We encourage candidates to send applications as soon as possible.
Application review will start from January 18, 2023, and will continue
until the position is filled. Candidates should submit a cover letter,
curriculum vitae, teaching and research statements, diversity state-
ment, and names and contact information of three (3) references.
SDSU is a Title IX and equal opportunity employer.
DECEMBER 2022  SPECTRUM.IEEE.ORG  75
 HISTORY IN AN OBJECT BY ALLISON MARSH
This music box demonstrated
the portability and
responsiveness of the
point-contact transistor.
John Bardeen’s
Marvelous
Music Box
In 1949 an engineer at Bell Labs built
three music boxes that could electron-
ically produce five distinct notes. Each
box contained an oscillator-amplifier
76  SPECTRUM.IEEE.ORG  DECEMBER 2022
circuit and two point-contact tran-
sistors powered by a B-type battery.
One box was used to demonstrate the
portability and responsiveness of
Bell Labs’ recently invented point-
contact transistor. The other two went
to the transistor’s co-inventors, John
Bardeen and Walter Brattain. Only
Bardeen’s survives. He brought his
music box to the University of Illinois
at Urbana-Champaign when he joined
THE SPURLOCK MUSEUM/UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN
the faculty in 1951. He often got a laugh
from his students when he used
the box to play the Prohibition-era
song “How Dry I Am.” Although the
box had stopped working by the
1990s, a careful restoration in 2019
allowed it to play again.
FOR MORE ON THE HISTORY OF BARDEEN’S
MUSIC BOX, see spectrum.ieee.org/
pastforward-dec2022
YOU’VE
INSURED
YOUR
HOUSE
AND CAR.
NOW
WHAT
ABOUT
YOUR LIFE?
1-800-493-IEEE (4333)
To learn more*, visit IEEEinsurance.com/future LIFE INSURANCE
*For information on features, costs, eligibility, renewability, limitations and exclusions.
Group Term Life Insurance is available only for residents of the U.S. (except territories), Puerto Rico and Canada (except Quebec). This is
underwritten by New York Life Insurance Company, 51 Madison Ave., New York, NY 10010 on Policy Form GMR. AMBA does not as a broker
with respect to Canadian residents and acts solely as an Administrator on behalf of New York Life.
Association Member Benefits Advisors, LLC.
In CA d/b/a Association Member Benefits & Insurance Agency
CA License #0I96562 • AR License #100114462
Program Administered by AMBA Administrators, Inc.
98147 (12/22) Copyright 2022 AMBA. All rights reserved
MATLAB SPEAKS
MACHINE
LEARNING
With MATLAB® you can use clustering,
regression, classification, and deep
learning to build predictive models
and put them into production.

mathworks.com/machinelearning

©2022 The MathWorks, Inc.