INDIA’S SOLAR THE NEW AN INTERNET THE TRANSISTORS

SHANGRI-LA FUSIONEERS IN THE SKY OF FLATLAND

Two gigawatts and Alternative approaches Can it break the 2D semiconductors in
53 square kilometers to power’s holy grail satcom bottleneck? the twilight of silicon
P. 08 P. 30 P. 38 P. 44

FOR THE TECHNOLOGY INSIDER | 02.20

THE
SHADOW
OF A
BLACK
HOLE
How 200 scientists
and engineers
managed to see
the unseeable
Katherine L. Bouman
explains
Challenge us.
What are your requirements for arbitrary waveform generation?
Our mission: providing instrumentation, software, and advice
to get you the best results with minimal time and effort.

HDAWG
Arbitrary Waveform Generator
∏Up to 144 synchronized channels starting at
∏750 MHz, 16 bit, 2.4 GSa/s, $29,950
5 Vpp maximum amplitude
∏LabOne® AWG Sequencer and Compiler
∏Digital modulation at multiple frequencies
∏ NEW Real-time precompensation

Now hiring in the US –

check our open positions!

Zurich Find out more today

Instruments www.zhinst.com
 CONTENTS_02.20

22 PORTRAIT OF A

5 BIG IDEAS
BLACK HOLE
What does a black hole
look like? Astronomers

FOR FUSION
now know the answer.
By Katherine L.
Bouman

Small-scale fusion energy

projects aim to leapfrog
38 ROUTERS
mainstream efforts. Page 30 IN SPACE
By Tom Cly nes Satellites acting
as Internet routers
can keep orbiting
spacecraft connected
to Earth.
By Mina Mitry

44 ATOM-THICK
TRANSISTORS
Devices with 2D
semiconductors might
be coming quicker than
you’d think.
By Iuliana Radu

06 NEWS
14 HANDS ON
17 CROSSTALK
52 PAST FORWARD

On the cover
Illustration for IEEE Spectrum
by StoryTK

PHOTOGRAPH BY First Light Fusion

IEEE SPECTRUM
(ISSN 0018-9235) is published monthly by The
Institute of Electrical and Electronics Engineers,
Inc. All rights reserved. © 2020 by The Institute of
Electrical and Electronics Engineers, Inc., 3 Park
Avenue, New York, NY 10016-5997, U.S.A.
Volume No. 57, Issue No. 2. 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/web/
aboutus/whatis/policies/p9-26.html.

SPECTRUM.IEEE.ORG | FEB 2020 |  01
 BACK STORY

THE IEEE APP:

Your mobile
gateway to
IEEE.

WHEN ALL REALITY IS VIRTUAL

W
e’r e pleased to a nnounce the debut , in this
issue, of a new column, Macro & Micro. Perhaps
you’ve heard of its author, Mark Pesce. If not, prepare
to be impressed.
An early milestone in his engineering career was
his founding, in 1991, of Ono-Sendai Corp., named
after a fictional company in William Gibson’s science-
fiction classic Neuromancer (Ace, 1984). In the real world, Ono‑Sendai
Download now became the world’s first consumer virtual-reality startup.
Pesce was one of the inventors of the orientation sensor that Sega
and get IEEE at Corp. adopted for its Sega-VR head-mounted display. Also, he and
your fingertips. others developed the Virtual Reality Modeling Language (VRML).
In 1996, Pesce cofounded BlitCom, the first company to use VRML
to deliver streaming 3D entertainment over the Web. Two years later,
Pesce helped create the graduate program in interactive media at the
University of Southern California. Not long afterward, he was invited to
Sydney to develop a postgraduate program in interactive and emerging
media at the Australian Film Television and Radio School. Pesce soon
made his home in Sydney, where he now serves as entrepreneur-in-
residence at the University of Sydney’s Incubate program.
In addition to being an engineer and a teacher, Pesce is also a
popularizer. In 2005, the Australian Broadcasting Corp. invited him
to become a panelist and judge on the television series “The New
Inventors.” In 2012, Pesce published his sixth book, The Next Billion
Seconds (Blurb Books), which explores a world where everyone
is “hyperconnected.” In 2014, he and Jason Calacanis launched the
podcast “This Week in Startups Australia.” Later Pesce started “The
Next Billion Seconds” podcast. And since 2014, he’s been a columnist
for The Register. Somehow, he also finds time to consult on blockchain-
based technologies for banks and fintech firms.
At the end of 2017, the Meanjin Quarterly published Pesce’s essay,
“The Last Days of Reality,” which describes a future in which it
JAMIE MACFADYEN
02.20

becomes impossible to know what is true. Well folks, we’re there. We

hope that Pesce’s columns in IEEE Spectrum will help you to navigate
that new reality. ■

02  | FEB 2020 | SPECTRUM.IEEE.ORG
 CONTRIBUTORS

EDITOR IN CHIEF Susan Hassler, s.hassler@ieee.org ADVERTISING PRODUCTION +1 732 562 6334

Katherine (Katie) Bouman

EXECUTIVE EDITOR Glenn Zorpette, g.zorpette@ieee.org
EDITORIAL DIRECTOR, DIGITAL
Harry Goldstein, h.goldstein@ieee.org
MANAGING EDITOR Elizabeth A. Bretz, e.bretz@ieee.org
SENIOR ART DIRECTOR
Mark Montgomery, m.montgomery@ieee.org
SENIOR EDITORS
Stephen Cass, cass.s@ieee.org
Erico Guizzo (Digital), e.guizzo@ieee.org
Jean Kumagai, j.kumagai@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
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
NEWS MANAGER Amy Nordrum, a.nordrum@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
ADMINISTRATIVE ASSISTANT
Ramona L. Foster, r.foster@ieee.org
CONTRIBUTING EDITORS Evan Ackerman, Mark Anderson,
Robert N. Charette, Peter Fairley, W. Wayt Gibbs, Tam Harbert,
Mark Harris, David Kushner, Robert W. Lucky, Prachi Patel,
Morgen E. Peck, Richard Stevenson, Lawrence Ulrich
EDITOR IN CHIEF, THE INSTITUTE
Kathy Pretz, k.pretz@ieee.org
ASSISTANT EDITOR, THE INSTITUTE
Joanna Goodrich, j.goodrich@ieee.org
DIRECTOR, PERIODICALS PRODUCTION SERVICES Peter Tuohy
EDITORIAL & WEB PRODUCTION MANAGER Roy Carubia
SENIOR ELECTRONIC LAYOUT SPECIALIST Bonnie Nani
PRODUCT MANAGER, DIGITAL Shannan Dunlap
WEB PRODUCTION COORDINATOR Jacqueline L. Parker
MULTIMEDIA PRODUCTION SPECIALIST Michael Spector
ADVERTISING PRODUCTION MANAGER
Felicia Spagnoli, f.spagnoli@ieee.org
SENIOR ADVERTISING PRODUCTION COORDINATOR
Nicole Evans Gyimah, n.gyimah@ieee.org
Bouman is an assistant professor in the
departments of electrical engineering and of
EDITORIAL ADVISORY BOARD, IEEE SPECTRUM
computing and mathematical sciences at Caltech.
Susan Hassler, Chair; Steve Blank, David C. Brock, Ronald
F. DeMara, Shahin Farshchi, Lawrence O. Hall, Jason K. Hui,
Her work in computational imaging includes the
Leah Jamieson, Mary Lou Jepsen, Deepa Kundur, Gianluca development of algorithms used to create the first
Lazzi, Allison Marsh, Carmen Menoni, Sofia Olhede, Maurizio picture of a black hole, as she describes in this
Vecchione, Edward Zyszkowski issue [p. 22]. Although she has many research
EDITORIAL ADVISORY BOARD, THE INSTITUTE interests, this area has long been one of her
Kathy Pretz, Chair; Qusi Alqarqaz, Philip Chen, Roberto Graglia, favorites. As Bouman says, “What’s cooler than
working on black holes?”
Shashank Gaur, Susan Hassler, Cecilia Metra, San Murugesan,
Mirela Sechi Annoni Notare, Tapan K. Sarkar, Joel Trussell,
Hon K. Tsang, Chonggang Wang
MANAGING DIRECTOR, PUBLICATIONS
Michael B. Forster Tom Clynes
EDITORIAL CORRESPONDENCE Clynes is a freelance writer and photojournalist
IEEE Spectrum, 3 Park Ave., 17th Floor, who covers science and environmental issues. In
New York, NY 10016-5997
this issue, he writes about new efforts to build
TEL: +1 212 419 7555 FAX: +1 212 419 7570
practical nuclear fusion reactors [p. 30]. His 2015
BUREAU Palo Alto, Calif.; Tekla S. Perry +1 650 752 6661
book, The Boy Who Played With Fusion, tells the
DIRECTOR, BUSINESS DEVELOPMENT,
tale of a 14-year-old who built a working fusion
MEDIA & ADVERTISING Mark David, m.david@ieee.org
reactor. “But the challenge of harnessing a fusion
ADVERTISING INQUIRIESNaylor Association Solutions, reaction to produce electricity for the grid will
Erik Henson +1 352 333 3443, ehenson@naylor.com
likely continue to vex the best engineering minds
REPRINT SALES +1 212 221 9595, ext. 319 for many years to come,” Clynes says.
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. Allison Marsh
COPYRIGHTS AND TRADEMARKS IEEE Spectrum is a Marsh is a historian at the University of South
registered trademark owned by The Institute of Electrical and Carolina, where she codirects the Ann Johnson
Electronics Engineers Inc. Responsibility for the substance Institute for Science, Technology & Society. In
of articles rests upon the authors, not IEEE, its organizational researching this month’s Past Forward, about
units, or its members. Articles do not represent official
a radioactive science kit for kids [p. 52], Marsh
positions of IEEE. Readers may post comments online;
ordered a copy of the 1966 movie U-238 and the
comments may be excerpted for publication. IEEE reserves
the right to reject any advertising.
Witch Doctor. “It’s about a group of American
mining engineers trying to negotiate uranium
rights in Africa,” Marsh explains. “It was too
painful to watch.”

IEEE  BOARD OF DIRECTORS PUBLICATIONS Michael B. Forster

PRESIDENT & CEO Toshio Fukuda, president@ieee.org
+1 732 562 3928 FAX: +1 732 981 9515
PRESIDENT-ELECT Susan K. “Kathy” Land
TREASURER Joseph V. Lillie SECRETARY Kathleen A. Kramer
José M.F. Moura
PAST PRESIDENT
VICE PRESIDENTS
Stephen M. Phillips, Educational Activities; Tapan K. Sarkar,
Publication Services & Products; Kukjin Chun, Member &
Geographic Activities; Kazuhiro Kosuge, Technical Activities;
Robert S. Fish, President, Standards Association; James M.
Conrad, President, IEEE-USA
DIVISION DIRECTORS
Alfred E. “Al” Dunlop (I); David B. Durocher (II); Sergio
Benedetto (III); John P. Verboncoeur (IV); Thomas M. Conte (V);
Manuel Castro (VI); Miriam P. Sanders (VII); Elizabeth L. “Liz”
Burd (VIII); Rabab Kreidieh Ward (IX); Ljiljana Trajkovic (X)
REGION DIRECTORS
Eduardo F. Palacio (1); Wolfram Bettermann (2);
Jill I. Gostin (3); David Alan Koehler (4); James R.
Look (5);  Keith A. Moore (6); Jason Jianjun Gu (7);
Magdalena Salazar-Palma (8); Alberto Sanchez (9);
Akinori Nishihara (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
+1 732 562 3998, m.b.forster@ieee.org
CHIEF MARKETING OFFICER Karen L. Hawkins
+1 732 562 3964, k.hawkins@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
MANAGING DIRECTOR, IEEE-USA Chris Brantley
+1 202 530 8349, c.brantley@ieee.org
IEEE  PUBLICATION SERVICES & PRODUCTS BOARD
Tapan K. Sarkar, Chair; Sergio Benedetto, Edhem Custovic,
Stefano Galli, Lorena Garcia, Ron B. Goldfarb,
Lawrence O. Hall, W. Clem Karl, Hulya Kirkici, Paolo Montuschi,
Sorel Reisman, Gaurav Sharma, Maria Elena Valcher,
John P. Verboncouer, John Vig, Bin Zhao
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
Mina Mitry
Mitry is an aerospace engineer and cofounder
and CEO of Kepler Communications. With the
rest of Kepler’s engineering team, he’s building
a constellation of satellites for low Earth orbit
that will function as “routers” to move data from
other satellites back to Earth. In this issue [p. 38],
Mitry explains how such a space-based Internet
will replace aging infrastructure and allow new
generations of Earth-imaging and communications
satellites to reach their full potential.
Iuliana Radu
Radu is program director at Imec, in Leuven,
Belgium, where she leads the Beyond CMOS
program and quantum-computing activities. The
2D semiconductors she writes about in “Atom-Thick
Transistors” [p. 44] are hardly the strangest things
she’s worked with. In grad school, she searched
for mysterious, and as yet unseen, quasiparticles
that appear only in two-dimensional systems.
Compared to that, Radu says, “the physics of
transistors with 2D materials is a walk in the park.”

SPECTRUM.IEEE.ORG | FEB 2020 |  03
 SPECTRAL LINES 02.20

JOHN DEERE’S
millions of pictures of kernels. Kernels harvested in
central Illinois might have one color. But kernels of
the same hybrid from a different farm 5 miles away

QUEST TO SOLVE might look slightly different. Now imagine solving

that challenge for dozens of grain varieties—some of

AGRICULTURE’S
which, like canola, are nearly microscopic.
Rostapshov points out that the scale of agricultural
data is also daunting. John Deere is one of the larg-

DEEP-LEARNING est users of cloud computing in the world. It gath-

ers 5 million to 15 million measurements per second

PROBLEMS
from 130,000 globally connected machines. The
company has over 150 million acres in its databases,
and it processes more data than Twitter.
Much of this information is so-called dirty data that
Farming poses unique doesn’t share the same format or structure because
and difficult AI challenges, it’s coming from some 100 other companies that
according to the green- have access to the John Deere platform, in addition
machine maker to the wide variety of John Deere machines. Those
companies add data about weather conditions, aerial

W
hat’s the world’s hardest machine-learning challenge?
Autonomous vehicles? Robots that can fall over and get
back up? Cancer detection? ¶ Julian Sanchez believes it’s
agriculture. ¶ He may be a little biased. Sanchez is the
director of precision agriculture for John Deere, and he’s
in charge of adding intelligence to traditional farm vehi-
cles. I met with Sanchez and Alexey Rostapshov, head
of digital innovation at John Deere Labs, in San Francisco, where John
Deere launched the spin-off in 2017 to take advantage of Silicon Valley’s
tech expertise. ¶ Sanchez believes agriculture is the biggest challenge for
artificial intelligence because it’s not just about driving tractors around,
although autonomous driving is certainly part of the mix. According to
Sanchez, the more complex problems revolve around issues such as crop
classification. John Deere would like to create an AI system that allows
farmers to know, for example, whether a grain being harvested is high
quality or low quality. The many differences between grain types, and
between grains grown under different conditions, make this a tough
imagery, soil analyses, and so on. As a result, Deere
has had to make tremendous investments in back-
end data cleanup.
Company-wide, Deere employs thousands of soft-
ware engineers, with many using machine-learning
tools in their work, and about the same number of
mechanical and electrical engineers. Ten years ago,
Deere’s hiring was heavily weighted toward mechani-
cal engineers, but now the large majority of Deere
engineers are working in the software space. As San-
chez says, mechanical engineers are still critical—
Deere builds green machines, after all—but if you look
at Deere’s tech-talent footprint, it’s pretty safe to call
John Deere a software company. —Tekla S. Perry
An extended version of this article appears in our View
From the Valley blog.
POST YOUR COMMENTS AT spectrum.ieee.org/spectrallines-feb 2020

task for machine learning. ¶ Sanchez uses corn as an example. To build CORRECTION: In “Transgenic Salmon Hits U.S. Shelves” (January 2020), we misstated
the timing of AquaBounty Technologies’ first harvest of its U.S.-raised transgenic salmon.
a deep-learning algorithm to analyze corn kernel quality, you’d feed it The fish will debut during the final months of this year.

STOP THAT TRACTOR!

RIGHT: SCHMUECKER RENEWABLE ENERGY SYSTEM

The cover of our November 2019 issue [left] featured an illustra-

tion of a rogue tractor rolling over tender crops in a field. Many
readers took the time to call our attention to this incongruity. It’s
the kind of mistake that can occur when well-intentioned people—
in this case, magazine editors and art directors—work well out-
side their area of expertise. Photos of the actual hydrogen- and
ammonia-powered tractor [right] described in the November cover
article are available on our website. Historically, many engineers
grew up on farms and first grew interested in p ­ roblem-solving
and innovation within that setting. That’s the spirit we hoped to
highlight in our November issue, and we regret that our e­ rroneous
cover illustration overshadowed a worthy project.

04  | FEB 2020 | SPECTRUM.IEEE.ORG
Antenna design optimization is crucial to
launching the Internet of Space.

Visualization of the electric field, power flow, and sharp far-field

radiation pattern of a parabolic reflector antenna.

The wired and wireless networks that currently connect

people around the world cannot reach everywhere on
Earth. To solve the problem, engineers are turning their eyes
toward space. The goal is to form a suborbital high-data-rate
communications network to revolutionize how data is shared
and collected. Before this Internet of Space can be built, design
engineers need to optimize their antenna designs.
The COMSOL Multiphysics® software is used for simulating
designs, devices, and processes in all fields of engineering,
manufacturing, and scientific research. See how you can apply
it to advanced antenna designs.
comsol.blog/IoS
NEPAL
News BHUTAN
This 765-kilovolt circuit could
boost power imports to Bangladesh,
including hydropower generated in
Bhutan and Nepal.
ELECTRIC THREE-WHEELERS
ferry people and goods around
Bangladesh but are banned in its capital.
INDIA
Batteries and motors could accelerate the
PROJECT STATUS bicycle rickshaws that gum up D ­ haka’s
AC Substation Completed
traffic and eliminate exhaust from
RANGPUR
Ongoing/planned tuk tuks, gas-powered three-­wheelers.
Under discussion But charging such EVs would further bur-
den already strained power lines.
That’s just one of many opportunity
costs that Bangladesh pays for a weak
INDIA
SY L H E T
MYMENSINGH
electrical grid. Frequent power outages
hurt businesses and deter foreign invest-
I NDI A
ment. A sweeping grid-­modernization
RAJSHAHI

HVDC back-to-back DHAKA

program promises to alleviate such
converter station
troubles.
Dhaka
In 2018, the government-run Power
Grid Company of Bangladesh (PGCB) dou-
This 400-kilovolt bled the capacity of its first ­international
line will reach more transmission link—a high-voltage DC con-
than 80 kilometers
into India, delivering nection delivering 1 gigawatt from India.
1.6 gigawatts of C H I TTAGO NG This month, it hopes to finalize require-
KHULNA
coal-fired power ments for generators that promise to sta-
from a plant built to BARISAL

supply Bangladesh. bilize the voltage and frequency of the

grid’s alternating current.
And next year, Bangladesh expects to
achieve universal electricity access for
Bangladesh
the country’s 160 million people, only
MYANM AR half of whom had electricity a decade
ago. “It’s a whole social transformation,”
says Tawfiq-e-Elahi Chowdhury, special
advisor on energy to Bangladesh prime
minister Sheikh Hasina.
However, it’s not clear what the grid

BANGLADESH SCRAMBLES
revamp will mean for Bangladesh’s
energy mix. Domestic natural gas is
running out, and the country is scram-
bling to replace it and maintain rapid

TO GROW POWER SUPPLY economic growth.

A nuclear power plant is now under
ISTOCKPHOTO

construction, and Bangladesh is import-

The country has promised universal electrification in 2021 ing liquefied natural gas. But the gov-

06  | FEB 2020 | SPECTRUM.IEEE.ORG
 ernment sees coal-fired and imported
electricity as its cheapest options, and
both come with challenges and risks.
Coal delivered less than 2 percent of
Bangladesh’s electricity last year, but
plants burning imported coal could soon
match the scale of its gas-fired genera-
tion. Three coal plants under construc-
tion are each capable of serving about
­10 percent of the country’s current 13-GW
peak power demand. And Chowdhury
expects similar projects in development
to lift total coal capacity to about 10 GW
by 2030.
The government expects to boost
imports fivefold, to 5 GW, by 2030.
Importing more electricity will provide
access to relatively low-cost and renew-
able hydropower. A deal struck with
Nepal should provide 500 megawatts,
and more interconnections to India, as
well as Bhutan, China, and Myanmar, are
under discussion.
To convey these new power flows
around the country, PGCB is building
a network of 400-kilovolt lines atop its
existing 230-kV and 130-kV lines, with
several 765-kV circuits on the drawing
board [see map]. The firm is simultane-
ously improving power quality—which
will allow Bangladesh to accommodate
more imported power and operate the
nuclear plant.
Imports will be costlier if high-voltage
DC converter stations must be erected at
each border crossing. Instead, the gov-
ernment has agreed to synchronize its
AC grid with India’s, enabling power to
flow freely between the two. Synchroni-
zation will not be possible, however, until
PETER FAIRLEY
PGCB eliminates its grid’s large voltage
and frequency deviations.
OPPORTUNITY COST: Bangladesh’s economy has grown by an impressive 6 percent per
year for the past decade despite a nagging electricity supply gap.
Sahbun Nur Rahman, PGCB’s execu- off-grid solar systems, and millions still
tive engineer for grid planning, says most do. But Mahmood Malik, CEO of the
private generators don’t properly adjust country’s Infrastructure Development
the power they produce to maintain the Company, says the expanding national
grid’s voltage and frequency. Stability has grid means there’s “not much need” to
improved over the last two years, how- build more.
ever, as government plants have stepped Off-grid solar still contributes more
up. He says the grid could be ready for than half of Bangladesh’s renewable
synchronization in as little as five years. electricity, which makes up less than
Coal power will push the country’s 3 percent of its power supply. Mean-
annual per capita greenhouse gas emis- while on-grid solar is growing slowly, and
sions up to about 1 metric ton—still tiny, wind development has barely begun. As
Chowdhury says, since the average devel- a result, the government will miss its
oped economy generates 12 metric tons. commitment to source 10 percent of
Still, betting on coal is controversial for the nation’s electricity from renewable
a low-lying country contending with cli- sources by 2021.
mate change. By some estimates, global Abdul Hasib Chowdhury, a grid expert
coal use needs to drop by 80 percent at the Bangladesh University of Engi-
within a decade to hold global warming neering and Technology, in Dhaka, says
to 1.5 °C this century. And one of Bangla- the best long-term bet for Bangladesh
desh’s first coal-plant projects is 14 kilo- is imported power from beyond South
meters upstream from the Sundarbans, Asia. He looks to the rich winds and sun-
the world’s largest contiguous mangrove shine in sparsely populated Central Asia.
forest, which serves as a buffer against “South Asia is nearly 2 billion people
cyclones and sea level rise. crammed into this small space,” says
What’s missing from the grid push, say A.H. Chowdhury. “They will require a
critics, is wind and solar. ­Bangladesh lot of energy in the next 50 years.”
pioneered the use of solar power to elec- —Peter Fairley
trify rural communities. At the peak, at
POST YOUR COMMENTS AT
least 20 million Bangladeshis relied on spectrum.ieee.org/bangladesh-feb2020
NEWS
SPECTRUM.IEEE.ORG | FEB 2020 |  07
 JOURNAL WATCH

Smart Headphones Warn

of Nearby Cars
HOW CAN PEDESTRIANS safely tune out
the world? Perhaps with a pair of intelligent
headphones that alert them to oncoming vehicles.
The number of pedestrians killed on U.S.
roads reached a three-decade high in 2018.
Smart headphones are unlikely to prevent most
pedestrian deaths—but a few seconds’ warning
might spare some lives.
Xiaofan Jiang, an assistant professor of
electrical engineering at Columbia University,

INDIA GRAPPLES WITH

developed the Pedestrian Audio Wearable System
(PAWS) with collaborators at the University of
North Carolina at Chapel Hill and Barnard College.
They published their work in the October 2019
IEEE Internet of Things Journal.
Many cars with collision warning systems rely on
cameras, radar, or lidar to detect objects. But Jiang
and his colleagues decided to use inexpensive
microphones to serve as low-power sensors for
their system.
The group placed four microphones in different
spots in a pair of commercial headphones. The
system’s main hardware fits inside the left ear
housing and draws power from a rechargeable
lithium-ion battery. A custom integrated circuit
saves on power by extracting only the most
relevant sound features from captured audio.
It transmits that information to an app, which
hosts machine-learning algorithms trained
on audio from 60 types of vehicles in different
environments.
The latest prototype can pinpoint the location
of a single vehicle from up to 60 meters away—
but it can’t plot the vehicle’s or pedestrian’s
trajectories. The system also has trouble
estimating how many vehicles are present and
VAST SOLAR PARK
The pros and cons of the
2-gigawatt Pavagada installation
I T ’ S 10 A . M . A N D
Indian peanut farmer
Venkeapream is relax-
ing at his family compound in
Pavagada, an arid area north
of Bangalore. The 67-year-old
retired three years ago upon
leasing his land to the ­Karnataka
state government. That land
is now part of a 53-square-­
kilometer area festooned with
millions of solar panels. As his
fields yield carbon-free electric-
ity, Venkeapream pursues his
passion full time: playing the
electric harmonium, a portable
whereby governments provide
multiple companies land and
transmission—two big hurdles
that slow solar development.
Solar parks account for much
of the 25.5 GW of solar capacity
India has added in the last five
years. The states of Rajasthan
and Gujarat have, respectively,
2.25-GW and 5.29-GW solar
parks under way, and Egypt’s
1.8-GW installation is one of sev-
eral new international projects.
Alas, even as they speed the
growth of renewable energy,
solar parks also concentrate
MANJUNATH KIRAN/AFP/GETTY IMAGES

tends to focus on the loudest vehicle, which may
not be the one closest to the pedestrian.
“We have to take into account other information
to make this collision detection more useful,” Jiang
explains. The project has received US $1.3 million
from the National Science Foundation.
—JEREMY HSU
reed organ.
With a capacity of 2 giga-
watts and counting, Pavagada’s
arrays represent the world’s
largest cluster of photovoltaics.
It’s also one of the most success-
ful examples of a solar “park,”
some of solar energy’s liabilities.
Sheshagiri Rao, an agricul-
tural researcher and farmer
based near Pavagada, says lease
payments give peanut farmers
such as Venkeapream a steadier
income. But Rao says shep-

An extended version of this article appears on our

website in the Journal Watch section. NEWS

08  | FEB 2020 | SPECTRUM.IEEE.ORG

BIG SOLAR: Construction on Pavagada
began in October 2016. The land was
leased for 28 years from 2,300 farmers.
herds who held traditional rights
to graze their fields were fenced out
without compensation, and many
have sold out. In Venkeapream’s vil-
lage, flocks once totaled 2,000 to
3,000 sheep. There are now only
about 600 left.
The constant need to keep dust
off the panels, meanwhile, has
put more strain on already over-
tapped groundwater supplies. Local
farmers bring water to clean the
more than 400,000 panels at the
­Pava­gada site of Indian energy devel-
oper Acme Cleantech Solutions. “At
least 2 liters of water is required to
clean one panel. This is huge,” says
B. Prabhakar, Acme’s site manager.
Robotic dusters allow Acme to clean
just twice a month, but most opera-
tors lack such equipment.
Then there are the power surges
and drops created as clouds pass
over Pavagada—generation swings
that must be countered with coal-
fired and hydropower plants. Bal-
ancing renewable energy swings is
a growing challenge for grid opera-
tors in Karnataka, which leads India
in solar capacity and also has more
than 4 GW of variable wind power.
Karnataka capped new solar
parks at 0.2 GW after launching
Pava­gada. Analysts heralded the
state’s apparent shift toward distrib-
uted installations, such as rooftop
solar systems, during a November
2019 meeting on sustainable energy
in neighboring state Tamil Nadu. As
Saptak Ghosh, who leads renewable
energy programs at the Bangalore-
based Center for Study of Science,
Technology & Policy (CSTEP), put
it: “Pavagada will be the end of big
solar parks in Karnataka. Smaller
is the future.”
Just a few days later, though,
news broke that Karnataka’s renew-
able energy arm was acquiring land
for three 2.5-GW solar megaparks.
The state’s move may reflect pres-
sure from the national government
to accelerate solar installations, as
well as confidence that Pavagada’s
shortcomings can be fixed.
Instead of harming shepherds,
for example, solar operators could
open their gates. Grass and weeds
growing amidst the panels pose
a serious fire risk, according to
Acme’s Prabhakar. Increasingly,
operators in other countries rely
on sheep to keep vegetation down.
Higher-tech solutions may ulti-
mately address Pavagada’s water
consumption and cloud-induced
power swings. Israeli robotics firm
Ecoppia is already providing what
it calls “water free” cleaning at the
Pavagada site operated by Fortum,
a Finnish energy company.
Karnataka’s solution for power
swings at its new megaparks, mean-
while, is to plug the parks straight
into the national grid’s biggest
power lines. The trio of plants are
a joint project with the national-
government-owned Solar Energy
Corporation of India, and designed
to export renewable electricity to
other states. Power stations out-
side of Karnataka will balance the
solar parks’ generation, according
to Ghosh’s colleague, CSTEP senior
research engineer and power-grid
specialist Milind R.
India’s government is eager to
help, having promised to boost
renewable capacity to 175 GW by
March 2022 and to 450 GW by 2030.
As Thomas Spencer, research fellow
at the Energy and Resources Insti-
tute, a New Delhi–based nonprofit,
noted at the November meeting in
Tamil Nadu, India is “well off the
track” for meeting either target.
—Peter Fa ir ley
POST YOUR COMMENTS AT
spectrum.ieee.org/india-feb2020
4 WAYS TO
HANDLE MORE
QUBITS
Cryochips, microrelays,
and other tech could help
quantum computers scale
AS RESEARCHERS STRIVE TO
boost the capacity of quantum
computers, they’ve run into a problem
that many people have after a big holi-
day: There’s just not enough room in
the fridge.
Today’s quantum-computer processors
must operate inside cryogenic enclosures
at near absolute zero, but the electron-
ics needed for readout and control don’t
work at such temperatures. So those cir-
cuits must reside outside the refrigera-
tor. For today’s sub-100-qubit systems,
there’s still enough space for specialized
cabling to make the connection. But for
future million-qubit systems, there just
won’t be enough room. Such systems
will need ultralow-power control chips
that can operate inside the refrigera-
tor. Engineers unveiled some potential
solutions in December during the IEEE
Inter­national Electron Devices Meeting
(IEDM), in San Francisco. They ranged
from the familiar to the truly exotic.
CryoCMOS
Perhaps the most straightforward way
to make cryogenic controls for quantum
computers is to modify CMOS technol-
ogy. Unsurprisingly, that’s Intel’s solution.
The company unveiled a cryogenic CMOS
chip called Horse Ridge that translates
quantum-computer instructions into
basic qubit operations, which it delivers
to the processor as microwave signals.
Horse Ridge is designed to work at
­4  kelvins, a slightly higher tempera-
ture than the qubit chip itself, but low
enough to sit inside the refrigerator with
it. The company used its 22-nanometer
FinFET manufacturing process to build
SPECTRUM.IEEE.ORG | FEB 2020 |  09
 the chip, but the transistors that make
up the control circuitry needed substan-
tial reengineering.
“If you take a transistor and cool it to
4 K, it’s not a foregone conclusion that
it will work,” says Jim Clarke, director of
quantum hardware at Intel. “There are
a lot of fundamental characteristics of
devices that are temperature dependent.”
Others are working along the same
lines. Google presented a cryogenic
CMOS control circuit earlier in 2019.
In research that was not yet peer-
reviewed at press time, Microsoft and
its collaborators say they have built a
­100,000-transistor CMOS control chip
that operates at 100 millikelvins.
Microrelays
In logic circuits, transistors act as
switches, but they aren’t the only
devices that do so. Engineers in Tsu-Jae
King Liu’s laboratory at the University
of California, Berkeley, have developed
­m icrometer-scale electromechanical
relays as ultralow-power alternatives to
transistors. They were surprised to dis-
cover that their devices operate better at
4 K than at room temperature.
At room temperature, the devices
DEEP FREEZE: Intel Labs principal engineer Stefano Pellerano holds
Horse Ridge, a new cryogenic control chip for quantum computers.
NEWS
10  | FEB 2020 | SPECTRUM.IEEE.ORG
suffer some mechanical peculiarities.
First, ambient oxygen can react with
the relay’s electrode surfaces. Over time,
this reaction can form a high-resistance
layer, limiting the device’s ability to con-
duct current. But at cryogenic tempera-
tures, oxygen freezes out of the air, so
that problem doesn’t exist.
Second, the contacts in microscale
relays tend to stick together. This shows
up as a hysteresis effect: The relay opens at
a slightly different voltage than the one at
which it closes. But because the adhesive
forces are weaker at cryogenic tempera-
tures, the hysteresis is less than 5 percent
of what it is at room temperature.
“We didn’t suspect ahead of time that
these devices would operate so well at
cryogenic temperatures,” says Liu, who
led the research presented at IEDM by
her graduate student Xiaoer Hu. ­“ In
­retrospect, we should have.”
Single-flux quantum logic
Hypres, in Elmsford, N.Y., has been com-
mercializing cryogenic ICs for several
years. Seeking to steer its rapid single-flux
quantum (RSFQ) logic tech into the realm
of quantum computing, the company
recently spun out a startup called Seeqc.
In RSFQ and its quantum version,
SFQuClass logic, quantized pulses of
voltage are blocked, passed, or routed
by Josephson junctions, the same type
of superconducting devices that make
up most of today’s quantum computer
chips. In 2014 physicists at University of
Wisconsin–Madison first suggested that
these pulses could be used to program
qubits, and Seeqc’s scientists have been
collaborating with them and Syracuse
University scientists since 2016.
Seeqc is now designing an entire system
using the technology: a ­digital-control,
error-correction, and readout chip
designed to work at 3 to 4 K and a separate
chip designed to work at 20 millikelvins
to interface with the quantum processor.
Weyl semimetals
Quantum computing is already strange,
but it might take some even stranger
tech to make it work. Scientists at Lund
University, in Sweden, and at IBM
Research–Zurich have designed a new
device called a Weyl semimetal ampli-
fier that they say could bring readout
electronics closer to the qubits. Don’t
worry if you don’t know what a Weyl
semimetal is. There are things about
these materials that even the scientists
trying to make devices from them don’t
fully understand.
What they do know is that these materi-
als, such as tungsten diphosphide, exhibit
extremely strong, t­ emperature-dependent
magnetoresistance when chilled to below
about 50 K. The device they simulated has
a gate electrode that produces a magnetic
field inside the Weyl semimetal, causing
its resistance to go from tiny to huge in a
matter of picoseconds. Connecting the
input from a qubit to the device could
make a high-gain amplifier that dissipates
a mere ­40 microwatts. That could be low
enough for the amplifier to live in the part
of the fridge close to where the qubits
themselves reside.
—Sa muel K. Moor e
INTEL
POST YOUR COMMENTS AT
spectrum.ieee.org/qubits-feb2020
 QUANTUM-DOT
TATTOOS STORE
VACCINE HISTORY
Invisible to the eye, the dots glow
under infrared light
I R E M E M B E R A FA D E D
yellow booklet about the
size of a wallet that my mother used
to pull out at the doctor’s office to
record my vaccines. Today, nurses
document my children’s vaccina-
tions in electronic health records
that will likely follow them to
adulthood.
To eradicate diseases such as
polio or measles, health care work-
ers need to know who was vacci-
nated and when. Yet in developing
countries, vaccination records
are sparse and, in some cases,
nonexistent.
Now, a team of MIT bioengineers
has developed a way to keep invis-
ible vaccine records under the skin.
Delivered through a microneedle
patch, biocompatible quantum
dots embed in the skin and fluo-
resce under near-infrared light—
creating a glowing trace that can
be detected at least five years after
vaccination. The work was recently
described in Science Translational
Medicine and is funded by the Bill &
Melinda Gates Foundation.
K.J. MCHUGH/SCIENCE TRANSLATIONAL MEDICINE (2)
“We started thinking about using
a dye that’s not visible by the naked
eye” but that would be persistent
and inexpensive to detect, says
Ana Jaklenec, a research scientist
at MIT’s Koch Institute for Integra-
tive Cancer Research. Together with
MIT’s Robert Langer, she came up
with a solution: quantum dots.
Quantum dots are small, semi-
conducting nanoparticles, rang-
ing in diameter from 2 to 10
­nanometers, with unique chemi-
SKIN GLOW: With a microneedle array
[top], researchers placed quantum dots
under the skin of rats. A smartphone filter
detects light from the dots [bottom].
cal and physical properties. Notably,
they absorb light of one wavelength
and efficiently convert it to light of
another wavelength. Quantum dots
are being explored for medical use
as biological sensors and probes,
and commercially in solar panels,
displays, and televisions.
To create a safe, lasting dye to
inject under the skin, the scien-
tists rejected traditional lead or
cadmium quantum dots, which
can be toxic, and instead created
a nanoparticle with a copper core and
a shell of aluminum and zinc sulfide,
which is safe and stable in the skin and
resists bleaching from the sun, says
Jaklenec.
The dots were then loaded into a
microneedle patch, a square with
an array of tiny needles that deliver
nanoparticles into the skin in a desired
pattern and then dissolve.
The scientists applied the patch to the
skin of rats and were able to detect the
dots using a smartphone modified with
a filter to sense near-infrared light. The
quantum-dot signal remained strong and
detectable in the rats for nine months after
application (when the experiment ended),
and the rats showed no side effects.
The team also applied the patches to
pieces of pig and human cadaver skin of
various tones. The light from the dots was
stable and detectable even after a stint
in a solar simulator, which imitated five
years of sun exposure.
Finally, the researchers codelivered
a polio vaccine with the quantum dots
in rats, and found that the dye did not
interfere with the vaccine’s function—
the rats still produced protective immune
responses.
The team is now working on a way to
encode data, such as the date of appli-
cation, into the quantum-dot array, says
Jaklenec. The group is also planning a
study, expected to begin early this year,
to survey people in Bangladesh, Kenya,
and Malawi about how and where this
method could be most useful and if par-
ents will be on board.
But before quantum dots can ever
be used in an actual clinic, a toxicol-
ogy study in rodents and human trials
must demonstrate the long-term safety
of the technology. “If funding is there,
that could happen in the next one to two
years,” says Jaklenec.
While Jaklenec and the team are devel-
oping the quantum-dot technology for
medical use, she admits there is probably
a commercial market too: When people
hear about the technology, they often ask
about getting an invisible tattoo.
—Mega n Scudell a r i
POST YOUR COMMENTS AT
spectrum.ieee.org/quantumdots-feb2020
SPECTRUM.IEEE.ORG | FEB 2020 |  11
12 | FEB 2020 | SPECTRUM.IEEE.ORG PHOTOGRAPH BY Andrey Rudakov/Bloomberg/Getty Images
 LOCATION,
LOCATION
BITCOIN MINING
is energy-intensive
work. Pictured is part
of the BitRiver facility
in Bratsk, Russia—one
of the world’s largest
data centers dedicated
to doing the complex
calculations that keep
the blockchain humming
along. There were two
solid reasons for BitRiver
to open up shop there.
Keeping this huge
facility running requires
nearly 100 megawatts
of electric power. But
because the site is a
stone’s throw from the
Bratskaya hydropower
plant on the Angara
River, it takes advantage
of some of the cheapest
electricity rates in the
world. The data center
depends on dozens of
industrial fans that blow
on the racks to keep the
20,000 mining devices
cool. But Siberia’s bitter
cold climate means
that BitRiver can get
away with just those
fans instead of a fancy
and expensive cooling
system, further slashing
its energy budget.

THE BIG
THE BIG PICTURE
PICTURE news
NEWS

SPECTRUM.IEEE.ORG | FEB 2020 | 13
Hands On

HANDS ON
HANDS ON DAVID
BY STEPHEN CASS
SCHNEIDER

14  | FEB 2020 | SPECTRUM.IEEE.ORG ILLUSTRATIONS BY James Provost

A DIY AUDIO
INDUCTION LOOP
HERE’S A SIMPLE
WAY TO TRANSMIT
AUDIO TO PEOPLE
WHO WEAR
T-COIL-EQUIPPED
HEARING AIDS
I REGULARLY ATTEND AN
unprogrammed Quaker meeting,
where a lot of the folks are older
and have hearing impairments. This style of
worship poses a special challenge for these
people because, unlike the convention at
many other religious gatherings, there is no
one person who addresses the congregation.
At an unprogrammed Quaker meeting, any- IN THE LOOP: A multiturn loop of wire [blue line] energizes the T-coils in suitably equipped hearing aids
by virtue of the magnetic field it creates in its vicinity [blue circles]. Voices are picked up by microphones
one may speak from anywhere in the room. whose signals are sent to an ordinary stereo amplifier, to which this 8-ohm wire loop is attached.
Some years ago, the congregation in-
stalled microphones in the room, which pick
up voices and amplify them. These signals
are then transmitted by FM radio over special
frequencies near the radio broadcast band. pick up audio-frequency signals electron- and-white sign that looks like an ear with a
The thing is, a person who needs help hear- ically instead of using their built-in micro- “T” next to it.
ing others must use a small FM receiver and phones. This system for passing the signal Many of the hearing-aid wearers who at-
earphones, which is a little awkward, in part wirelessly doesn’t involve radio transmis- tend our Quaker meeting live in a nearby re-
because it makes you stand out. sions—it just uses magnetic induction. tirement community where induction loops
One way to mitigate this problem is to Suitable audio induction loops for energiz- have been installed to help hearing-impaired
take advantage of the fact that some of ing T-coils are found in all sorts of places, in- residents enjoy lectures and films. These
these people wear hearing aids equipped cluding museums and theaters; even some people have had good experiences in such
with a telecoil (or T-coil). These hearing aids taxicabs are equipped with them. You may settings, and so they have been hoping that
can be switched to a mode whereby they see their presence advertised with a blue- an audio induction loop would help them to

DEPARTMENTS

SPECTRUM.IEEE.ORG | FEB 2020 |  15
 HANDS ON BY DAVID SCHNEIDER

AMP IT UP: This project augments an existing

microphone multiplexer and FM transmitter [smaller
units] with a stereo receiver attached to the wire loop.

hear what’s being said at our Quaker meet- ceiver, one that was collecting dust in the ing showed the signal-pickup range to be
ing, too. back of my garage. about 5 meters.
I worried, though, that their disappoint- Using that arrangement, I was able to con- Encouraged by those initial tests, anoth-
ment with our current system reflected the vey audio to the T-coil in his hearing aid with er friend and I recently placed a similar wire
difficulty of picking up voices cleanly in a his head about a meter and a half away (on loop in the attic of the building—which was a
reverberant room using microphones po- axis) from my coil. This test was tougher than lot harder than laying it on the floor because
sitioned at the periphery. And commercial in- needed to assess requirements, because the we had to snake the wire around an obstacle
stallers of this equipment charge thousands magnetic field at my friend’s head was m ­ ostly course of roof trusses and HVAC ducting. We
of dollars—a lot for something that’s proba- horizontal, and my understanding is that ended up with a loop that’s got a rather com-
bly not going to solve our particular problem. ­T-coils are positioned to pick up vertical fields. plicated geometry and doesn’t lay flat on the
But for very little money, I figured, we could Based on that test (and the Biot-Savart joists in many places.
install a DIY audio induction loop and at least Law), I figured out that I should be able to Nevertheless, this cobbled-together in-
let people try it out. create a magnetic field of the same strength duction loop appears to convey signals just
A little searching online uncovered about 3 meters away from a long wire, which fine to hearing-aid wearers below. At least
some brief commentary posted by some- is exactly what you would need if said long that’s what my retired electrical engineer
one who installs such systems at music wire were part of a loop installed in the at- friend reported when I set the stereo receiv-
festivals. What he described was quite tic of the building where the meeting is held. er attached to the loop to output a program
simple—­creating a multiturn loop using a Next I used a wire-resistance table to from our local public-radio station. The only
multiconductor wire (one for which the total determine that 240 meters of 20-gauge dead spots he found as he walked around
resistance is between 4 and 8 ohms), and at- wire should be around 8 ohms. That would the room were in the far corners.
taching it to a 200-watt audio amplifier, just be long enough for me to create a six- That said, I don’t expect the addition of this
as you would an 8-ohm speaker. turn loop in the attic. So I bought a roll of informally constructed induction loop—or
Before diving into this, I enlisted an old- 1,000 feet (304 meters) of two-conductor even a professionally installed one—could
er friend (a retired electrical engineer) who 20-gauge wire for about US $125 on Ama- ever make up for the weak link in our system:
wears a T-coil hearing aid to perform an ex- zon. Using two conductors would allow me the mounting of the microphones on the walls
periment. I constructed an induction coil to create two parallel loops, powering each of the room, far from the people speaking. Un-
from a six-turn square loop of magnet wire through one channel of my stereo receiver. fortunately, there’s not much of an alternative,
that was about a half meter on a side (I taped Because each channel of the receiver can short of giving everyone TED-talk-style micro-
the wire to a flattened cardboard box), using output as much as 80 watts, I figured I’d be phones to wear, which wouldn’t really mesh
wire of the right diameter to make the loop close to the 200-watt figure I saw quoted with the Quaker penchant for simplicity. n
resistance 8 ohms. I then attached it to the online. Preliminary tests with six turns of
POST YOUR COMMENTS AT
speaker terminals of an ordinary stereo re- this wire laid out on the floor of the build- spectrum.ieee.org/tcoil-feb2020

16  | FEB 2020 | SPECTRUM.IEEE.ORG
 CrossTalk
ELECTRICITY PRICES:
A CHANGING BARGAIN
TH E G E N E R ATI O N S - O LD TR E N D
toward lower electricity prices now
appears to have ended. In many afflu-
TONY CORDOZA/GETTY IMAGES
ent countries, prices tilted upward at the
NUMBERS DON’T LIE BY VACLAV SMIL
turn of the century, and they continue
to rise, even after adjusting for inflation.
Even so, the price we pay for electricity
is an extraordinary bargain, and that’s why
this form of power has become ubiquitous
in the modern world. When expressed in
constant 2019 dollars, the average price
of electricity in the United States fell from
$4.79 per kilowatt-hour in 1902 (the first
year for which the national mean is avail-
able) to 32 cents in 1950. The price had
dropped to 10 cents by 2000, and in late
2019 it was just marginally higher, at about
11 cents per kilowatt-hour. This represents
a relative decline of more than 97 percent.
CROSSTALK
SPECTRUM.IEEE.ORG | FEB 2020 |  17
A dollar now buys nearly 44 times as much
electricity as it did in 1902.
Because average inflation-adjusted
manufacturing wages have quadrupled
since 1902, blue-collar households now
find electricity about 175 times as afford-
able as it was nearly 120 years ago. And
it gets better: We buy electricity in order
to convert it into light, kinetic energy, or
heat, and the improved efficiency of such
conversions have made the end uses of
electricity an even greater bargain.
Lighting offers the most impressive
gain: In 1902, a lightbulb with a tanta-
lum filament produced 7 lumens per
watt; in 2019 a dimmable LED light deliv-
ered 89 lm/W (see “Luminous Efficacy,”
IEEE Spectrum, April 2019). That means a
lumen of electric light is now three orders
of magnitude (approximately 2,220
times) more affordable for a working-
class household than it was in the early
20th century. Lower but still impressive
reductions in end-use costs apply in the
case of electric motors that run kitchen
appliances and force hot air into the ducts
to heat houses using natural-gas furnaces.
An international perspective shows
some surprising differences. The United
States has cheaper residential electric-
ity than other affluent nations, with the
exception of Canada and Norway, which
derive high shares of their power from
hydroelectric generation (60 percent
and 95 percent, respectively).
When using prevailing exchange rates,
the U.S. residential price is about 45 per-
cent of the European Union’s mean,
about half the Japanese average, and
about a third of the German rate. Electric-
ity prices in India, Mexico, Turkey, and
South Africa are lower than in the United
States in terms of the official exchange
rates, but they are considerably higher
in terms of purchasing power parity—
more than twice the U.S. level in India
and nearly three times as much in Turkey.
A naive observer, reading the reports
of falling prices for photovoltaic cells and
wind turbines, might conclude that the
rising shares of solar and wind power
NUMBERS DON’T LIE BY VACLAV SMIL
18  | FEB 2020 | SPECTRUM.IEEE.ORG
will bring a new era of falling electricity over an 18-year period (2000 to 2018)
prices. Just the opposite has been true. electricity prices more than doubled, to
Before the year 2000, when Germany €0.31/ kWh. The E.U.’s largest economy
embarked on its large-scale and expen- thus has the continent’s highest elec-
sive Energiewende (energy transition), tricity prices, followed by heavily wind-
that country’s residential electricity dependent Denmark, at €0.3/kWh.
prices were low and declining, bottoming A similar contrast can be seen in the
at less than €0.14/kWh ($.13/kWh, using United States. In California, where the
the prevailing exchange rate) in 2000. new renewables have taken an increas-
By 2015, Germany’s combined solar ing share, electricity prices have been
and wind capacity of nearly 84 gigawatts rising five times as fast as the national
had surpassed the total installed in fossil- mean and are now nearly 60 percent
fueled plants, and by March 2019 more higher than the countrywide average. n
than 20 percent of all electricity came
POST YOUR COMMENTS AT
from the new 10,000
renewables. However, spectrum.ieee.org/electricityprices-feb2020
2,300
1,000
ELECTRICITY
PRICE (2019 U.S.
479 cents per kilowatt-hour)
100
89
LUMINOUS
EFFICACY 28
(lumens per watt)
10 10.5
6.88
7
MANUFACTURING
WAGE
(2019 dollars per hour)
AFFORDABILITY
INDEX
11
1900 2020
CROSSTALK
 INTERNET OF EVERYTHING BY STACEY HIGGINBOTHAM
WI-FI’S LONG GOODBYE
IN TEN YEARS, we won’t
need Wi-Fi.
At least, that’s what Azhar
Hussain, the CEO of IoT
company Hanhaa, told me on a phone
call late last year. He thinks the end of
Wi-Fi is nigh because he believes that
allocating spectrum in smaller chunks
will let municipalities, universities, and
companies create private 5G cellular
networks. The convenience of those net-
works will impel companies to choose
cellular connections over Wi-Fi for their
IoT devices.
There’s reason to think Hussain is right,
at least for higher-value devices, such as
medical devices, home appliances, and
outdoor gear like pool-­cleaning robots.
Zach Supalla, the CEO of Particle, a com-
pany that supplies IoT components to
businesses with little experience build-
ing connected products, says more than
ILLUSTRATION BY Dan Page
half of the IoT devices in Particle’s cloud
that use cellular connections are also
within range of a Wi-Fi network. Supalla
says that companies choose cellular
modules over Wi-Fi because the modules
are easier to set up and businesses can
better control the consumer experience.
Wi-Fi devices are notoriously difficult
to connect to one another, or pair. To
get a connected product on their home
Wi-Fi network, consumers must often
pair with a software-based access point
before switching the device over to their
own network.
This process can be fraught with
errors. Even I, a reporter who has tested
hundreds of connected devices, fail to
get a device on my network on the first
try roughly a third of the time. To make
it easier, Amazon and Google have both
created proprietary onboarding pro-
cesses that handle the setup on behalf of
CROSSTALK
the user, so that when consumers power
their devices on, they automatically try
to join their network.
However, device manufacturers still
have to implement both Amazon’s and
Google’s programs separately, and that
requires know-how that some compa-
nies don’t possess. Thankfully, Amazon,
Apple, and Google are now working on a
smart-home standard that may simplify
things. But the details are scant, and any
solution they develop won’t be available
until 2021 at the earliest.
When you’re faced with multiple
Wi-Fi ecosystems, cellular is just easier,
­Hussain says. Cellular networks cost
more now because you have to install
radios on the devices and pay a subscrip-
tion to use the cellular network. Hussain
sees those costs coming down, poten-
tially even disappearing, given time.
That’s because he’s anticipating a
future where universities, businesses,
and municipalities set up their own
­c ellular net works using spectrum
obt a i ned t h roug h new spec t r u m
­auctions, such as the Citizens Broadband
Radio ­Services (CBRS) auctions occur-
ring in the United States in June. Cellular
­equipment makers are already building
gear and testing these private networks
in factories and offices. If new roaming
plans are developed to allow devices
to come onto these local networks eas-
ily, similar to joining a Wi-Fi hotspot,
­cellular connectivity will become prac-
tically free.
Even if Hussain’s vision doesn’t come
to pass in the next 10 years, the costs
of low-data-rate cellular contracts will
continue to drop, and that could still
eventually put the nail in the coffin for
Wi-Fi. And I mostly agree: I think there
are plenty of reasons to believe that
Wi-Fi will never disappear entirely, but
I do think small cellular networks will
take its place in our lives. n
POST YOUR COMMENTS AT
spectrum.ieee.org/goodbyewifi-feb2020
SPECTRUM.IEEE.ORG | FEB 2020 |  19
 MACRO & MICRO BY MARK PESCE
SENSE AND SENSORIBILITY
I’M WRITING A BOOK about
augmented reality, which forced
me to confront a central question: When
will this technology truly arrive? I’m not
talking about the smartphone-screen ver-
sions offered up by the likes of Pokémon
Go and Minecraft Earth, but in that long-
promised form that will require nothing
more cumbersome than what feels like
a pair of sunglasses.
Virtual reality is easier. It can now be
delivered, in reasonable quality, for a few
hundred dollars. The nearest equivalent
for AR, Microsoft’s second-generation
HoloLens, costs an order of magnitude
more while visually delivering a lot less.
Ivan Sutherland’s pioneering Sword of
Damocles AR system, built in 1968, is
more than a half-century old, so you
might expect that we’d be further along.
Why aren’t we?
Computation proved to be less of a
barrier to AR than anyone believed back
in the 1960s, as general-purpose proces-
sors evolved into application-specific ICs
and graphics processing units. But the
20  | FEB 2020 | SPECTRUM.IEEE.ORG
essence of augmented reality—the manip-
ulation of a person’s perception—cannot
be achieved by brute computation alone.
Connecting what’s inside our heads
to what is outside our bodies requires a
holistic approach, one that knits into a
seamless cloth the warp of the compu-
tational and the weft of the sensory. VR
and AR have always lived at this intersec-
tion, limited by electronic sensors and
their imperfections—all the way back to
the mechanical arm that dangled from
the ceiling and connected to the headgear
in Sutherland’s first AR system, inspir-
ing its name.
Today’s AR technology is much more
sophisticated than Sutherland’s con-
traption, of course. To sense the user’s
surroundings, modern systems employ
photon-measuring time-of-flight lidar or
process images from multiple cameras
in real time—computationally expen-
sive solutions even now. But much more
is required.
Human cognition integrates various
forms of perception to provide our sense
CROSSTALK
of what is real. To reproduce that sense,
an AR system must hitch a ride on the
mind’s innate workings. AR systems focus
on vision and hearing. Stimulating our
eyes and ears is easy enough to do with
a display panel or a speaker situated
meters away, where it occupies just a
corner of our awareness. The difficulty
increases exponentially as we place these
synthetic information sources closer to
our eyes and ears.
Although virtual reality can now trans-
port us to another world, it does so by
effectively amputating our bodies, leav-
ing us to explore these ersatz universes
as little more than a head on a stick. The
person doing so feels stranded, isolated,
alone, and all too frequently motion sick.
We can network participants together in
these simulations—the much-promised
“social VR” experience—but bringing even
a second person into a virtual world is
still beyond the capabilities of broadly
available gear.
Augmented reality is even harder. It
doesn’t ask us to sacrifice our bodies or
our connection to others. An AR system
must measure and maintain a model
of the real world sufficient to enable
a smooth fusion of the real with the
synthetic. Today’s technology can just
barely do this, and not at a scale of bil-
lions of units.
Like autonomous vehicles (another
blend of sensors and computation that
looks easier on paper than it proves in
practice), augmented reality continues
to surprise us with its difficulties and
dilemmas. That’s all to the good. We
need hard problems, ones that can’t be
solved with a straightforward technologi-
cal fix but require deep thought, reflec-
tion, insight, even a touch of wisdom.
Getting to a solution means more than
building a circuit. It means deepening
our understanding of ourselves, which
is always a good thing. n
POST YOUR COMMENTS AT
spectrum.ieee.org/macromicro-feb2020
ILLUSTRATION BY Greg Mably
 ILLUMINATE EDUCATE ENGAGE ENERGIZE
the possibilities of the next generation a wider audience innovation by
technology by using of innovators and in appreciating the celebrating
it to address global engineers value and importance technological
challenges of engineering and excellence
technology

The world’s most daunting challenges The IEEE Foundation is leading a special
require innovations in engineering, and campaign to raise awareness, create
IEEE is committed to finding the solutions. partnerships, and generate financial resources
needed to combat these global challenges.

Our goal is to raise $30 million by 2020.

DONATE NOW
ieeefoundation.org
 SPECTRUM.IEEE.ORG | FEB 2020 | 22

PORTR A IT OF A BL ACK HOLE

HERE’S HOW THE EVENT HORIZON TELESCOPE TEAM PIECED TOGETHER
A NOW-FAMOUS IMAGE B y Ka t h er i n e L . B o u m a n
SEEING BY RADIO: Last year, scientists published images of a black hole called M87* [left]. The data that went into constructing those images
came from seven radio telescopes spread over the globe, including the James Clerk Maxwell Telescope, in Hawaii [top]. Another telescope, at
the South Pole [bottom], aided in the calibration of the telescope network and is used for observing other astronomical sources.
PEOPLE AND ­EQUIPMENT:
The scientific
collaboration that
resulted in the first
images of a black hole
involved hundreds of
people, giant radio
telescopes, and
supercomputer facilities
spread around the world.
In addition to the James
Clerk Maxwell Telescope
and the South Pole
Telescope [shown on the
previous page], six more
radio telescopes were
involved in the effort to
produce pictures of a
black hole.

L
AST APRIL, A RESEARCH TEAM THAT I’M PART OF UNVEILED A PICTURE
that most astronomers never dreamed they would see:
one of a massive black hole at the center of a distant gal-
axy. Many were shocked that we had pulled off this feat.
To accomplish it, our team had to build a virtual tele-
scope the size of the globe and pioneer new techniques
in radio astronomy.
Our group—made up of more than 200 scientists
and engineers in 20 countries—combined eight of the
world’s most sensitive radio telescopes, a network of synchro-
Breakthrough Prize in Fundamental Physics was
awarded to the entire collaboration.
Now we are adding more giant dish antennas to
sharpen our view of such objects. Thanks to an infu-
sion of new funding from the U.S. National Science
Foundation and others, we have set an ambitious goal:
to make a movie of the swirling gravitational monster
that forms the black heart of our own Milky Way galaxy.
THE EXISTENCE OF BLACK HOLES was a dubious prediction of
PREVIOUS PAGES, CLOCKWISE FROM LEFT: EHT COLLABORATION; WILLIAM

nized atomic clocks, two custom-built supercomputers, and sev- the general theory of relativity, which Einstein devel-
MONTGOMERIE/JCMT; DANIEL MICHALIK/SOUTH POLE TELESCOPE

eral new algorithms in computational imaging. After more than
10 years of work, this collective effort, known as the Event Hori-
zon Telescope (EHT) project, was finally able to illuminate one
of the greatest mysteries of nature.
Within weeks of our announcement (and the publication of six
journal articles), an estimated billion-plus people had seen the pic-
ture of the light-fringed shadow cast by M87*, a black hole at the
center of Messier 87, a galaxy in the Virgo constellation. It is likely
among the most labor-intensive scientific images yet created. In
recognition of the global teamwork required to combine efforts
in black-hole theory and modeling, electronic instrumentation
and calibration, and image reconstruction and analysis, the 2020
oped a little over a century ago. Astronomers debated
for decades later whether massive stars would cre-
ate black holes when they collapse under their own
weight at the end of their lives. Even more mysteri-
ous were supermassive black holes, hypothesized to
lurk in the hearts of galaxies. Astronomers observed
extraordinarily bright, compact objects and powerful
galactic-scale jets beaming from the centers of many
galaxies—including Messier 87—as well as stars and gas
clouds orbiting hidden central objects. Supermassive
black holes could account for such observations, and
no other explanation seemed very plausible.

24  | FEB 2020 | SPECTRUM.IEEE.ORG

By the turn of the 21st century, astronomers had
come to believe that black holes must be common.
Most black holes are probably far too small and dis-
tant for us ever to observe. These “ordinary” black
holes form when a star 10 times as massive as the
sun or heavier collapses into something so dense
that it bends the very fabric of space-time enough to
form a spherical trap. Any matter, light, or radiation
crossing the edge of this trap, known as the event
CLOCKWISE FROM TOP LEFT: HELGE ROTTMAN; ANA TORRES CAMPOS; SALVADOR
horizon, disappears forever.
SANCHEZ; JUNHAN KIM/ESO; JONATHAN WEINTROUB; BABAK A. TAFRESHI
But the gargantuan black holes that we now think
inhabit the centers of most galaxies—like the 4-million-
solar-mass Sagittarius A* (Sgr A* for short) cloaked
behind a veil of dust in the Milky Way and the 6.5-billion-
solar-mass M87*—are different beasts altogether. Fed by
matter and energy spiraling in from their host galaxies
over eons, these are the most massive objects known,
and they create around them the most extreme con-
ditions found anywhere in the universe.
So says the math of general relativity. And the
Nobel Prize–winning detection in 2015 of gravita-
tional waves—essentially, a chirp of rippling space-
time created by the whirling merger of two ordinary
YET MORE TELESCOPES:
These were the Atacama
Large Millimeter/
submillimeter Array
(ALMA), in Chile [top
left]; the Large Millimeter
Telescope (LMT), in
Mexico [top center];
the IRAM Pico Veleta
telescope, in Spain [top
right]; the Atacama
Pathfinder EXperiment
(APEX), in Chile [bottom
left]; the Submillimeter
Array (SMA), in Hawaii
[bottom center]; and the
Submillimeter Telescope
(SMT) [bottom right], in
Arizona.
black holes more than a billion light-years away—offered yet another
piece of evidence that black holes are real. It took an incredible
feat of technology to “hear” the vibrations of that cosmic crash:
two laser interferometers, each 4 kilometers on a side and able to
detect a change in the lengths of their arms less than 0.01 percent
the width of a proton. But hearing is not seeing.
Encouraged by the success of initial pilot studies in the late 1990s
and early 2000s, an international team of astronomers proposed
a bold plan to make an image of a supermassive black hole. They
believed that advances in electronics now made it possible to see
these bizarre objects for the first time and to open a new window
on the study of general relativity. Teams from six continents came
together to form the Event Horizon Telescope project.
Creating such a picture would require another giant leap in astro-
nomical interferometry. We would have to virtually link observa-
tories across the planet to function together as one giant virtual
telescope. It would have the resolving power of a dish almost as
wide as Earth itself. With such an instrument, it was hoped, we
could finally see a black hole.
WHILE M87* ITSELF IS BLACK, it is backlit by a blinding glow of radiation
emanating from the material swirling around it. Friction and mag-
netic forces heat that matter to hundreds of billions of degrees. The
SPECTRUM.IEEE.ORG | FEB 2020 |  25
 RINGING SUCCESS: Four
teams of researchers, working
independently, produced
the first images of the M87
black hole. Two of the four
imaging teams used traditional
algorithms from radio
astronomy to produce images
of the black hole [top two]. The
other two teams used a more
modern class of algorithms
developed for the Event Horizon
Telescope data [bottom
two]. While the four images
produced by these teams differ
in many details, they each
show the same fundamental
structure—a luminous ring of
about 40 microarcseconds
in diameter, one that is
asymmetrical with the
brighter portion at the bottom.
The diameter of the black
hole shadow is particularly
important to astronomers in
confirming an earlier estimate
of the mass of the object, which
appears to be about 6.5 billion
times the mass of the sun.

result is an incandescent plasma that emits light and radio waves.
The massive object bends those rays, and some of them head our way.
A shadow of the black hole, just bigger than the object and its
event horizon, is imprinted on that radiation pattern. Measuring
the size and shape of this eerie shadow could tell us a lot about
M87*, settling arguments about its mass and whether it is a black
hole or something more exotic still. Unfortunately, the cores of gal-
axies are obscured by giant, diffuse clouds of gas and dust, which
leave us no hope of seeing M87* with visible light or at the very
long wavelengths typically used in radio astronomy.
Those clouds become more transparent, however, at short radio
wavelengths of around 1 millimeter; we chose to observe at 1.3 mm.
Such radio waves, at the extremely high frequency of 230 gigahertz,
also pass through air largely unimpeded, although atmospheric
water vapor does attenuate and delay them somewhat in the last
observed to the instrument’s diameter. To resolve
M87* using 1.3-mm radio waves, we would need a
radio dish 13,000 km across.
Although that might seem like a showstopper, inter-
ferometry offers a way around that problem—but only
if you can collect pieces of the same radio wave front
as it arrives at different times at all our telescopes in
far-flung locations, stretching from Hawaii and the
Americas to Europe. To do that, we have to digitally
time-stamp our measurements and then combine
them with enough precision to extract the relevant sig-
nals and convert them into an image. This technique
is called very-long-baseline interferometry (VLBI), and
the members of the Event Horizon Telescope project
believed it could be used to image a black hole.

T
few miles of their 55-million-year journey from the periphery of
LEFT: EHT COLLABORATION; RIGHT: CHI-KWAN CHAN

M87* to our radio telescopes on Earth.
The brightness and relative proximity of M87* worked in our
favor, but on the other side were some formidable technical chal-
lenges, starting with that wobbly signal delay caused by water vapor.
Then there was the size problem. Although M87* is ultramassive,
it is comparatively small—the shadow this black hole casts is only
around the size of our solar system. Resolving it from Earth is
like trying to read a newspaper from 5,000 km away. The resolv-
ing power of a lens or dish is set by the ratio of the wavelength
HE DETAILS PROVED TRULY DEVILISH, HOWEVER. By the
time the radio signals from M87* bounce
into a receiver connected to one of our 6-
to 50-meter dish antennas on Earth, the
signal power has dropped to roughly 10-16 W
(0.1 femtowatt). That’s about a billionth of
the strength of the signals a satellite TV
dish typically picks up. So the very low signal-to-
noise ratio posed one major problem, exacerbated

26  | FEB 2020 | SPECTRUM.IEEE.ORG
by the fact that our largest single-dish telescope, the
50-meter Large Millimeter Telescope, in Mexico, was
still being completed and wasn’t yet fully operational
when we used it in 2017.
The 1.3-mm wavelength, considerably shorter
than the norm in VLBI, also pushed the limits of our
technology. The receivers we used converted the
230-GHz signals down to a more manageable fre-
quency of around 2 GHz. But to get as much infor-
mation as we could about M87*, we recorded both
right- and left-hand circular polarizations at two
frequencies centered around 230 GHz. As a result,
our instrumentation had to sample and store four
separate data streams pouring in at the prodigious
rate of 32 gigabits per second at each of the partici-
pating telescopes.
Interferometry works only if you can precisely
align the peaks in the signal recorded at each pair
of telescopes, so the short wavelength also required
us to install hydrogen-maser atomic clocks at each
site that could sample the signal with subpicosec-
ond accuracy. We used GPS signals to time-stamp
the observations.
On four nights in April 2017, everything had to
come together. Seven giant telescopes (some of
them multidish arrays) pointed at the same minus-
cule point in the sky. Seven maser clocks locked into
sync. A half ton of helium-filled, 6- to ­8-terabyte
hard drives started spinning. I along with a few
dozen other bleary-eyed scientists sat at our screens
in mountaintop observatories hoping that clouds
would not roll in. Because of the way interferom-
etry works, we would immediately lose 40 percent
of our data if cloud cover or technical issues forced even one of
the telescopes to drop out.
But the heavens smiled on us. By the end of the week, we were
preparing 5 petabytes of raw data for shipment to MIT Haystack
Observatory and the Max Planck Institute for Radio Astronomy,
in Germany. There, researchers, using specially designed super-
computers to correlate the signals, aligned data segments in
time. Then, to counter the phase-distorting influence of turbu-
lent atmosphere above each telescope, we used purpose-built
adaptive algorithms to perform even finer alignment, matching
signals to within a trillionth of a second.
Now we faced another giant challenge: distilling all those qua-
drillions of bytes of data down to kilobytes of actual information
that would go into an image we could show the world.
NOWHERE IN THOSE PETABYTES OF DATA were numbers we could simply
plot as a picture. The “lens” of our telescope was a tremendous
amount of software, which drew heavily on open-source pack-
ages and now is available online so that anyone can replicate or
improve on our results.
Radio interferometry is relatively straightforward when you
have many telescopes close together, aimed at a bright source, and
observing at long wavelengths. The rotation of Earth during the
night causes the baselines connecting pairs of the telescopes to
sweep through a range of angles and effective lengths, filling in the
space of possible measurements. After the data is collected, you
line up the signals, extract a two-dimensional spatial-frequency
pattern from the variations in amplitude and phase among them,
and then do an inverse Fourier transform to convert the 2D fre-
quency pattern into a 2D picture.
VLBI is a lot harder, especially when observing with just a hand-
ful of dishes at a short wavelength, as we were for M87*. Perfect
calibration of the system was impossible, though we used an
eighth telescope at the South Pole to
help with that. Most problematic were
differences in weather, altitude, and
humidity at each telescope. The atmo-
spheric noise scrambles the phase of
the incoming signals.
The problem we faced in observing
with just seven telescopes and scram-
bled phases is a bit like trying to make
out the tune of a duet played on a piano
on which most of the keys are broken
and the two players start out of sync
with each other. That’s hard—but not
impossible. If you know what songs
ON THE SAME PAGE: The author reveals
initial results [above left] of the four
imaging teams, which to their great
satisfaction produced similar images.
SPECTRUM.IEEE.ORG | FEB 2020 |  27
typically sound like, you can often still CRUNCHING NUMBERS INTO PICTURES
work out the tune. The measurements taken to construct these images of a black hole came from seven radio
It also helps that the noise scram- telescopes spread around the world. An eighth (at the South Pole) aided in the calibration
bles the signal in an organized way of these measurements. Newly developed algorithms and supercomputers were used to
that allows us to exploit a terrific trick correlate the observed signals to make measurements and reconstruct the images from
called closure quantities. By multiply- these measurements.
ing correlated data from each pair in
a trio of telescopes in the right order, Submillimeter Array Atacama Large Millimeter/
(SMA), Hawaii submillimeter Array (ALMA), Chile
we are able to cancel out a big chunk
IRAM Pico Veleta
of the noise, though at the cost of add- Submillimeter Telescope Atacama Pathfinder telescope, Spain
ing some complexity to the problem. (SMT), Arizona EXperiment (APEX),
Chile
The longest and shortest baselines
in our telescope network set the limits
of our resolution and field of view, and
they were limited indeed. In effect, we
could reconstruct a picture 160 micro-
arcseconds wide (equivalent to 44 bil-
lionths of a degree on the sky) with
roughly 20 microarcseconds of reso-
lution. Literally an infinite number of
images could fit such a data pattern. James Clerk Large Millimeter South Pole Telescope
Somehow we would have to pick— Maxwell Telescope Telescope (LMT), Mexico
(JCMT) Hawaii
and decide how confident to be in
our choice.

T
O AVOID FOOLING OURSELVES, we cre-
–– Measurements Measurements
ated lots of images from the collected by that, theoretically,
M87* data, in lots of different pairs of radio could be collected
telescopes in the with a radio dish
ways, and developed a rig- Event Horizon the size of Earth
orous process—well beyond Telescope project

what is typically done in

radio astronomy—to deter-
mine whether our reconstructions
were reliable. Every step of this pro- Image created from Image created from Image created from
cess, from initial correlation to final processing pipeline 1 processing pipeline 2 processing pipeline 3
interpretation, was tested in multiple
ways, including by using multiple soft-
ware pipelines.
Before we ever started collecting
data, we created a computer simula-
tion of our telescope network and all
the various sources of error that would
affect it. Then we fed into this simula-
tion a variety of synthetic images—some
derived from astrophysical models of Blurred and averaged

black holes and others we had com- April 5 April 6 April 10 April 11
pletely made up, including one loosely
based on Frosty the Snowman.
Next, we asked various groups to
IMAGES: EHT COLLABORATION

reconstruct images from the synthetic

observations generated by the simula-
tion. So we turned images into obser-
vations and then let others turn those Final images
back into images. The groups all pro-

28  | FEB 2020 | SPECTRUM.IEEE.ORG
duced pictures that were fuzzy and a bit off, but in
the ballpark. The similarities and differences among
those pictures taught us which aspects of the image
were reliable and which spurious.
In June 2018, researchers on the imaging team
split into four squads to work in complete isolation
on the real data we had collected in 2017. For seven
weeks, each squad worked incommunicado to make
the best picture it could of M87*.
Two squads primarily used an updated version of
an iterative procedure, known as the CLEAN algo-
rithm, which was developed in the 1970s and has
since been the standard tool for VLBI. Radio astron-
omers trust it, but in cases like ours, where the data
is very sparse, the image-generation process often
requires a lot of manual intervention.
Drawing on my experience with image reconstruc-
tion in other fields, my collaborators and I developed
a different approach for the other two squads. It is a
kind of forward modeling that starts with an image—
say, a fuzzy blob—and uses the observational data
to modify this starting guess until it finds an image
that looks like an astronomical picture and has a
high probability of producing the measurements
we observed.
I’d seen this technique work well in other contexts,
and we had tested it countless times with synthetic
EHT data. Still, I was stunned when I fed the M87*
data into our software and, within minutes, an image
appeared: a fuzzy, asymmetrical ring, brighter on
the bottom. I couldn’t help worrying, though, that
the other groups might come up with something
quite different.
On 24 July 2018, a group of about 40 EHT mem-
bers reconvened in a conference room in Cambridge,
Mass. We each put up our best images—and every-
one immediately started clapping and laughing. All
four were rings of about the same diameter, asym-
metrical, and brighter on the bottom.
We knew then that we were going to succeed, but
we still had to demonstrate that we hadn’t all just
injected a human bias favoring a ring into our soft-
ware. So we ran the data through three separate imag-
ing pipelines and performed image reconstruction
with a wide range of prototype images of the kind
that we worried might fool us. In one of the imaging
pipelines, we ran hundreds of thousands of simula-
tions to systematically select roughly 1,500 of the
best settings.
At the end, we took a high-scoring image from
each of the three pipelines, blurred them to the
same resolution, and took the average. The result-
ing picture made the front page of newspapers
around the world.
A DECADE OR TWO FROM NOW, astronomers will no doubt look back at
this first snapshot of a black hole and consider it a milestone, but
they’ll also smile at how indistinct and uncertain—and unmoving—it
is compared with what they will probably be able to do. Although
we are confident in the asymmetry of the ring and its size—roughly
40 microarcseconds in diameter—the fine structure in that image
should be taken with a grain of salt.
But we did see a ring and the shadow of the black hole! That in
itself is astonishing.
Measurements of that shadow add a lot of weight to the argu-
ment that M87* has a mass equal to 6.5 billion suns, consistent
with the estimate astronomers had set by measuring the speed
of stars circling the black hole. (In contrast, estimates made from
the complicated effects of M87* on nearby gas were much lower:
around 3.5 billion solar masses.) The size of the ring is also large
enough to rule out speculation that M87* is not a supermassive
black hole but rather a wormhole or a naked singularity—even
stranger objects that appear to be consistent with general relativ-
ity but have never been observed.
Perhaps equally important, our initial success
POST YOUR
gives us good reason to believe that with fur- COMMENTS AT
spectrum.ieee.org/
ther improvements to both the telescope net- blackhole-feb2020
work and the software, we will be able to image
Sgr A* at the center of the Milky Way. Our nearest s­ upermassive
black hole is only a few hundred times as bright as the sun, and
it is less than a thousandth as massive as M87*. But because it is
2,000 times closer than M87*, it would appear a little larger to
us than M87*.
The biggest challenge in imaging Sgr A* is the speed at which it
evolves. Blobs of plasma orbit M87* every couple of days, whereas
those around Sgr A* complete an orbit every few minutes. So our
goal is not to snap a still image of Sgr A* but to make a crude movie
of it spinning like a billion-degree dervish at the center of the gal-
axy. This could be the next milestone in our quest to further con-
strain Einstein’s theory of gravity—or point to physics beyond it.
And there could be practical spinoffs. The methods we are devel-
oping to make movies of Sgr A* are strikingly similar to those needed
to make a medical MRI of a child squirming in a scanner or to image
subterranean movements during an earthquake.
For future observations, we expect to use 11 or more telescopes—
including a 12-meter dish in Greenland, an array of a dozen 15-meter
dishes in the French Alps, and one of Caltech’s radio dishes in Owens
Valley, Calif.—to increase the number of baselines. We have also dou-
bled the data-sampling rate from 32 Gb/s to 64 Gb/s by expanding
the range of radio frequencies we record, which will strengthen our
signals and eventually allow us to connect smaller dishes to the net-
work. Together, these upgrades will boost the amount of data we
collect by an order of magnitude, to about 100 petabytes a session.
And if all continues to go well, we hope that in the years or decades
ahead the reach of our computational telescope will grow beyond
the bounds of Earth itself, to include space-based radio telescopes
in orbit. Adding fast-moving spacecraft into our existing VLBI net-
work would be a tremendous challenge, of course. But for me, that
is part of the appeal. n
SPECTRUM.IEEE.ORG | FEB 2020 |  29
FUSION VORTEX: General
Fusion’s magnetized target
reactor injects pulses of plasma
into a sphere filled with swirling
molten lead and lithium.

30  | FEB 2020 SPECTRUM.IEEE.ORG
 5
Big Ideas for Fusion Power

Startups, universities, and major

companies are vying to commercialize
a nuclear fusion reactor • By Tom Clynes

THE JOKE HAS BEEN AROUND almost as long as the dream:

Nuclear fusion energy is 30 years away...and always will be. But
now, more than 80 years after Australian physicist Mark ­Oliphant
first observed deuterium atoms fusing and releasing dollops of
energy, it may finally be time to update the punch line.
Over the past several years, more than two dozen research
groups—impressively staffed and well-funded startups, university
programs, and corporate projects—have achieved eye-opening
advances in controlled nuclear fusion. They’re building fusion
reactors based on radically different designs that challenge the two
mainstream approaches, which use either a huge, doughnut-shaped
magnetic vessel called a tokamak or enormously powerful lasers.
What’s more, some of these groups are predicting significant
fusion milestones within the next five years, including reaching
the breakeven point at which the energy produced surpasses the
energy used to spark the reaction. That’s shockingly soon, con-
sidering that the mainstream projects pursuing the conventional
tokamak and laser-based approaches have been laboring for
decades and spent billions of dollars without achieving breakeven.
In Cambridge, Mass., MIT-affiliated researchers at Common-
wealth Fusion Systems say their latest reactor design is on track

I L LU S T R AT I O N S BY C H R I S P H I L P OT

SPECTRUM.IEEE.ORG | FEB 2020 |  31
to exceed breakeven by 2025. In the United Kingdom, a
University of Oxford spin-off called First Light Fusion claims
it will demonstrate breakeven in 2024. And in Southern
California, the startup TAE Technologies has issued a breath-
takingly ambitious five-year timeline for commercialization
of its fusion reactor.
Irrational exuberance? Maybe. Fusion research is among
the most costly of endeavors, depending on high inflows of
cash just to pay a lab’s electricity bills. In the pursuit of fund-
ing, the temptation to overstate future achievements is strong.
And past expectations of impending breakthroughs have
repeatedly been dashed. What’s changed now is that advances
in high-speed computing, materials science, and modeling
and simulation are helping to topple once-recalcitrant tech-
nical hurdles, and significant amounts of money are flowing
into the field.
Some of the new fusion projects are putting the newest
generation of supercomputers to work to better under-
stand and tweak the behavior of the ultrahigh-temperature
plasma in which hydrogen nuclei fuse to form helium. Others
have reopened promising lines of inquiry that were shelved
decades ago. Still others are exploiting new superconduc-
tors or hybridizing the mainstream concepts.
Despite their powerful tools and creative approaches, many
of these new ventures will fail. But if just one succeeds in build-
ing a reactor capable of producing electricity economically,
it could fundamentally transform the course of human civi-
lization. In a fusion reaction, a single gram of the hydrogen
isotopes that are most commonly used could theoretically
yield the same energy as 11 metric tons of coal, with helium
as the only lasting by-product.
As climate change accelerates and demand for electricity
soars, nuclear fusion promises a zero-carbon, low-waste base-
load source of power, one that is relatively clean and comes
with no risk of meltdowns or weaponization. This tantaliz-
ing possibility has kept the fusion dream alive for decades.
Could one of these scrappy startups finally succeed in mak-
ing fusion a practical reality?
NOT SO LONG AGO, the outlook for fusion power was pretty
bleak, with two of the biggest projects seemingly stalled.
In 2016, the U.S. Department of Energy admitted that its
US $3.5 billion National Ignition Facility (NIF) had failed to
meet its goal of using lasers to “ignite” a self-sustaining fusion
Key Fusion 1934 Australian
physicist Mark Oliphant
1958 The first tokamak,
the Soviet Union’s T-1,
Energy observes atoms fusing begins operation.
and emitting energy
Milestones in his University of
Cambridge laboratory. 1974 KMS Fusion, a
private-sector company,
fires an array of lasers
1920 British astrono- 1958 Los Alamos at a deuterium-tritium
mer Arthur Eddington researchers pellet, achieving the
theorizes that the sun demonstrate the first successful laser-
and other stars are first controlled induced fusion.
powered by the fusion thermonuclear fusion.
of hydrogen atoms.
32  | FEB 2020 | SPECTRUM.IEEE.ORG
reaction. A DOE report suggested that NIF’s research should
shift from investigating laser-sparked ignition to determin-
ing whether such ignition is even possible.
The same year, the U.S. and several other governments
began debating whether to pull their support from the Inter-
national Thermonuclear Experimental Reactor (ITER). First
proposed in 1985 and now under construction in southern
France, ITER is the world’s biggest fusion experiment. It is
a type of tokamak, which uses magnetic forces to confine
and isolate the ferociously hot, energetic plasma needed to
initiate and sustain fusion. But the project has been plagued
by delays and cost overruns that have quintupled its original
$5 billion price tag and pushed its projected completion date
to 2035. (And even if it makes that date, it could be decades
after that before commercial plants based on the design are
in operation.) The setbacks and enormous expense of NIF
and ITER had the effect of draining not just money but also
enthusiasm from the field.
Even as the government-backed megaprojects foundered,
alternative fusion-energy research began to gain momentum.
The hope of those pursuing these new efforts is that their
novel and smaller-scale approaches can accelerate past the
decades-long incremental slog. Investors are finally taking
notice and pouring money into the field. Over the past five
years, private capitalists have injected about $1.5 billion into
small-scale fusion-energy companies. Among those who
have made significant bets on fusion are Amazon’s Jeff Bezos,
Microsoft’s Bill Gates, and venture capitalist Peter Thiel. A
few major corporations, including Lockheed Martin, have
launched their own small-fusion projects.
Jesse Treu, a Ph.D. physicist who spent much of his career
investing in biotech and med-tech startups, says he realized in
2016 that “wonderful things were starting to happen in fusion
energy, but funding wasn’t catching up. It’s clear that private
equity and venture capital are part of the solution to develop
this technology, which is clearly the best energy answer for
the planet.” He cofounded the Stellar Energy Foundation to
connect fusion researchers with funding sources and to pro-
vide support and advocacy.
And public money has started to follow private: U.S.
Department of Energy grant makers, who for decades fun-
neled most nondefense fusion allocations to ITER, are now
channeling some funding to projects at the fringes of main-
stream research. The federal budget includes a $107 million
1985 Mikhail
Gorbachev and Ronald
1997 The Joint
European Torus (JET)
2013 National Ignition
Facility (NIF) implosion
Reagan agree to a joint reactor in England yields more energy than
PREVIOUS PAGES: GENERAL FUSION
collaboration on fusion outputs 16 megawatts the energy absorbed by
research, which leads to of fusion power, still the the fuel.
the ITER experiment. world record.
2019 Construction
1995 Princeton 2013 Construction of ITER is two-thirds
Plasma Physics begins on ITER, in complete. It is expected
Laboratory’s tokamak southern France. to produce 10 times the
achieves a record input energy.
plasma temperature
of 510 million °C.
increase for fusion projects in fiscal year 2020, including a
research partnership program that allows small companies to
conduct major experiments at the DOE’s national laboratories.
The U.S. government’s renewed interest stems in part from
a perceived need to keep up with China, which recently
restarted its fusion-energy program after a three-year mora­
torium. The Chinese government plans to switch on a new
fusion reactor in Sichuan province this year. Meanwhile, the
Chinese energy company ENN Energy Holdings has been
investing in research programs abroad and is building a
duplicate of Princeton Fusion Systems’ compact reactor in
central China, with help from top U.S. scientists.
“Now that it’s looking like China will gobble up every idea
the U.S. has failed to fund,” says Matthew J. Moynihan, a
nuclear engineer and fusion consultant to investors, “that’s
serving as a wake-up for the U.S. government.”
FOR ALL THIS ACTIVITY and investment, fusion power
remains as tough a problem as ever.
Unlike nuclear fission, in which a large, unstable nucleus
is split into smaller elements, a fusion reaction occurs when
the nuclei of a lightweight element, typically hydrogen, col-
lide with enough force to fuse and form a heavier element. In
the process, some of the mass is released and converted into
energy, as laid out in Albert Einstein’s famous formula: E = mc2.
There’s an abundance of fusion energy in our universe—
the sun and other stable stars are powered by thermo­
nuclear fusion—but the task of triggering and controlling a
self-­sustaining fusion reaction and harnessing its power is
arguably the most difficult engineering challenge humans
have ever attempted.
To fuse hydrogen nuclei, earthbound reactor designers need
to find ways to overcome the positively charged ions’ mutual
repulsion—the Coulomb force—and get them close enough
to bind via what’s known as the strong nuclear force. Most
methods involve temperatures that are so high—­several orders
of magnitude hotter than the sun’s core temperature of
15 million °C—that matter can exist only in the plasma state,
in which electrons break free of their atomic nuclei and cir-
culate freely in gaslike clouds.
But a high-energy-density plasma is notoriously unstable
and difficult to control. It wriggles and writhes and attempts to
break free, migrating to the edges of the field that contains it,
where it quickly cools and dissipates. Most of the challenges
surrounding fusion energy center around plasma: how to
heat it, how to contain it, how to shape it and control it. The

Cryostat Vacuum vessel

Superconducting

1. Magnetic-
magnets

Confinement Fusion First wall

(MCF) Heaters

Monitoring
THE BIG IDEA: Powerful electro­ sensors
magnetic fields confine and heat
plasma inside a doughnut-shaped
reactor called a tokamak, a Russian
acronym for “toroidal chamber with
axial magnetic field.” Since the
1960s, more than 200 functional
tokamaks have been built, and the
plasma physics fundamentals are well
established. The most ambitious of
these is the US $25 billion ITER, now
under construction in southern France.

REALITY CHECK: Scientists

are a long way from achieving a
self-sustaining reaction, and from
preventing neutron activation from
destroying the reactor’s walls.

PROJECTS TO WATCH:
Commonwealth Fusion Systems,
Tokamak Energy

SPECTRUM.IEEE.ORG | FEB 2020 |  33
 two mainstream approaches are magnetic confinement and
inertial confinement. Magnetic-confinement reactors such
as ITER attempt to hold the plasma steady within a tokamak,
by means of powerful magnetic fields. Inertial-confinement
approaches, such as NIF’s, generally use lasers to compress
and implode the plasma so quickly that it’s held in place long
enough for the reaction to get going.
Scientists have long thought that bigger is better when it
comes to creating stable and energy-dense plasma fields.
But with recent advances in supercomputing and complex
modeling, researchers are unraveling more of the mysteries
underlying plasma behavior and developing new tricks for
handling it without huge, complex machinery.
Among the researchers at the forefront of this work is phys-
icist C. Wendell Horton Jr. of the University of Texas Insti-
tute of Fusion Studies. He uses the university’s ­Stampede
supercomputer to build simulations of plasma flow and tur-
bulence inside magnetic-confinement reactors. “We’re mak-
ing calculations that were impossible just a few years ago
and modeling data about plasma in three dimensions and
in time,” Horton says. “Now we can see what’s happening
with much more nuance and detail than we would get with
analytic theories and even the most advanced probes and
diagnostic measurements. That’s giving us a more holistic
picture of what’s needed to improve reactor design.”
Horton’s findings have informed the design of large-scale
experiments such as ITER, as well as small-scale projects.
“The problem with ITER is that no matter how well they
get the plasma to behave, they haven’t figured out how to
2. Inertial-Confinement
Fusion (ICF)
THE BIG IDEA: Powerful pulsed laser or
ion beams (or other methods) compress a
small fuel pellet to extremely high densities,
and the resulting shock wave heats the
plasma before it has time to dissipate.
REALITY CHECK: Forces exerted on
the fuel pellet result in laser-plasma
instabilities that produce high-energy
electrons, which heat and scatter much
of the fuel before it can fuse. In addition,
the high cost and complexity of the laser
drivers may make traditional approaches
to ICF unsuitable for energy production.
PROJECTS TO WATCH: First Light
Fusion, General Atomics
34  | FEB 2020 | SPECTRUM.IEEE.ORG
get the reaction to self-sustain,” he says. “It’s still going to
burn out in a matter of minutes, and that’s obviously not
solving the energy problem.” He and other researchers
believe that some of the small-scale efforts are much closer
to achieving a steady-state reaction that could generate
baseload electricity.
Among the most mature of the fusion startups is California-
based TAE Technologies (formerly Tri Alpha Energy), which
launched in 1998.
The TAE reactor is designed to make use of what’s called a
field-reversed configuration (FRC) to create a swirling ring of
plasma that contains itself in its own magnetic field. ­(Princeton
Fusion Systems’ design is also an FRC.) Instead of using
­deuterium and tritium—the hydrogen-isotope blend that
fuels most fusion reactors—the TAE reactor injects beams
of high-energy neutral hydrogen particles into hydrogen-
boron fuel, forcing a reaction that produces alpha particles
(ionized helium nuclei). Heat generated in the containment
vessel caused by the deposit of soft X-ray energy will be con-
verted into electricity using a conventional thermal conver-
sion system, which heats water into steam to drive a turbine.
Hydrogen-boron fusion is aneutronic, meaning that the
primary reaction does not produce damaging neutron radia-
tion. The drawback is that burning the fuel requires extraor-
dinary temperatures, as high as 3 billion °C. “When you’re
that hot, the electrons are radiating like crazy,” says William
Dorland, a physics professor at the University of Maryland.
“They’re going to cool off the plasma faster than you can heat
it.” Although FRC machines seem to be less prone to plasma
Vacuum chamber Lenses
Laser beams
Fuel pellet Coolant
 Plasma injector Rotational axis
Pneumatic
pistons
Cylindrical
vortex cavity
Plasma toroid
Rotating liquid
lead and lithium
Plasma injector
instabilities than some other magnetic-confinement meth-
ods, no one has yet demonstrated an FRC reactor that can
create a stable plasma.
TAE cofounder and CEO Michl Binderbauer says the com-
pany’s latest machine, dubbed Norman (in honor of com-
pany cofounder Norman Rostoker), is achieving “significant
improvements in plasma containment and stability over the
previous-generation machine.” What’s driving the improve-
ments are advances in artificial intelligence and machine
learning, enabled by a cutting-edge algorithm developed
by Google called Optometrist. TAE adapted the algorithm
in partnership with Google to analyze the plasma-behavior
data and home in on the combination of variables that will
create the most ideal conditions for fusion. The researchers
described it in a Nature paper published in 2017.
“We’re doing things we could have never done 10 years
ago, and that’s driving faster and faster cycles of learning,”
says Binderbauer.
ADVANCED COMPUTING is also breathing new life into
promising lines of inquiry that were abandoned years ago
due to budget cuts or technical roadblocks. General Fusion,
based near Vancouver, was founded by Canadian plasma
physicist Michel Laberge. He quit a lucrative job develop-
ing laser printers to pursue an approach called magnetized
target fusion (MTF). The company has attracted more than
$200 million, including investments from Jeff Bezos and the
governments of Canada and Malaysia.
General Fusion’s design combines features of magnetic-
confinement and inertial-confinement fusion. It injects pulses
3. Magnetized Target
Fusion (MTF)
THE BIG IDEA: Sometimes called
magneto-inertial fusion (MIF), this
hybrid approach uses magnetic fields
to confine a lower-density plasma (as in
magnetic-confinement fusion), which is
then heated and compressed using an
inertial-confinement method such as lasers
or pistons (as in inertial-confinement fusion).
REALITY CHECK: Scientists have yet to
increase the plasma density to a working
level and keep it there long enough for a
significant fraction of the fuel mass to fuse.
PROJECTS TO WATCH: General Fusion,
HyperJet Fusion, Magneto-Inertial Fusion
Technologies
of magnetically confined plasma fuel into a sphere filled with
a vortex of molten lead and lithium. Pistons surrounding
the reactor drive shock waves toward the center, compress-
ing the fuel and forcing the particles into a fusion reaction.
The resulting heat is absorbed in the liquid metal and used
to produce steam to spin a turbine and generate electricity.
“You can think of it in some ways as the opposite of a toka-
mak,” says Laberge. “Tokamaks work with a big plasma
field that’s [relatively] low density. We’re trying to make a
mini-size plasma that’s extremely high density, by squash-
ing it in with the shock waves. Because the field is so dense
and small, we only need to keep it together for a millisec-
ond for it to react.”
In the 1970s, the U.S. Naval Research Laboratory experi-
mented with a piston system to trigger nuclear fusion. Those
experiments failed, due in large part to an inability to pre-
cisely control the timing of the shock waves. Laberge’s team
has developed advanced algorithms and highly precise con-
trol systems to fine-tune the speed and timing of the shock
waves and compression.
“In those experiments in the 1970s, the problem was sym-
metry,” says Laberge. “We’ve now achieved the accuracy
and force we need, so that part’s solved.”
Using liquid metal could solve another of fusion energy’s
primary challenges: Neutron radiation erodes a reactor’s walls,
which must be replaced frequently and disposed of as low-
level radioactive waste. The liquid metal protects the solid
outer wall from damage. There’s some irradiation of the liquid
metal, but there’s no need to regularly replace it, and so the
reactor doesn’t produce a steady stream of low-level waste.
SPECTRUM.IEEE.ORG | FEB 2020 |  35
 Neutral beam
injectors

DC magnets

Formation chamber

Plasma gun

Confinement chamber

4. Field-Reversed Configuration (FRC)

THE BIG IDEA: An FRC reactor
contains plasma in its own magnetic
field by inducing a toroidal electric
current inside a cylindrical plasma.
Compared to the direction of an
externally applied magnetic field, the
axial field inside the reactor is reversed
by eddy currents in the plasma.
TAE Technologies’ reactor [pictured]
uses plasma guns to accelerate two
plasmas into each other and then
heats them with particle beams.
REALITY CHECK: Although
FRC machines are less prone to
instabilities than are some other
magnetic-confinement methods, no
lab has yet demonstrated a working
FRC reactor that can create a
sufficiently dense and stable plasma.
PROJECTS TO WATCH: Helion
Energy, Princeton Fusion Systems,
TAE Technologies

General Fusion’s newest reactor, which generated plasma
for the first time in late 2018, is the centerpiece of a facility
that Laberge says will demonstrate an end-to-end capability
to produce electricity from nuclear fusion. “Now that we’ve
successfully created a stable, long-lived plasma, we can see
that we have a viable path toward having the plasma generate
more energy than it consumes,” he says. “In terms of commer-
cialization, our timeline is now a matter of years, not decades.”
V I R G I N I A- BAS E D H Y P E R J E T F U S I O N C O R P. has an
approach similar to General Fusion’s, but instead of pistons,
some 600 plasma guns fire jets of plasma into the reactor.
The merging of the jets forms a plasma shell, or liner, which
then implodes and ignites a magnetized target plasma. The
system doesn’t need a heating system to bring the fuel to
fusion temperatures, says HyperJet CEO and chief scientist
F. Douglas Witherspoon. “The imploding plasma liner con-
tains the target plasma and provides the energy to elevate the
temperature to fusion conditions. And because we’re using
a much higher-density plasma than a magnetic-confinement
system would, it reduces the size of the fusing plasma from
meter scale to centimeter scale.”
Witherspoon says the advantage of the HyperJet approach
over tokamaks is that it doesn’t require expensive supercon-
ducting magnets to generate the enormous magnetic fields
needed to confine the fusion-burning plasma.
Tokamaks themselves are also getting a reboot, thanks to
the use of different superconducting materials that could
make magnetic confinement more viable. MIT spin-off
Commonwealth Fusion Systems is employing yttrium-barium-
copper oxide (YBCO), a high-temperature superconductor,
in the magnets on its Sparc reactor.
Commonwealth cofounder Martin Greenwald, who is also
the deputy director of MIT’s Plasma Science and Fusion
Center, calculates that the Sparc reactor’s YBCO magnets
will be able to generate a field of about 21 teslas at their sur-
face and 12 T at the center of the plasma, roughly doubling
the field strength of tokamak magnets made of niobium-tin.
Stronger magnetic fields produce a stronger confining force
on the charged particles in the plasma, improving insulation
and enabling a much smaller, cheaper, and potentially bet-
ter performing fusion device.
“If you can double the magnetic field and cut the size of
the device in half, with identical performance, that will be a
game changer,” Greenwald says.
Indeed, one advantage of the newer, small-scale fusion
projects is that they can concentrate on the novel aspects
of their designs, while taking advantage of decades of hard-
won knowledge about the fundamentals of fusion science. As
Greenwald puts it, “We think we can get to commercial deploy-
ment of fusion power plants faster by accepting the conven-
tional physics basis developed around the ITER experiment

36  | FEB 2020 | SPECTRUM.IEEE.ORG

Magnetic coil
Vacuum vessel
Twisted
magnetic field
Plasma
and focusing on our collaborations between physicists and
magnet engineers who have been setting records for decades.”
SOME PROMISING STARTUPS, though, aren’t content to
accept the conventional wisdom, and they’re tackling the
underlying physics of fusion in new ways. One of the more rad-
ical approaches is that of First Light Fusion. The British com-
pany intends to produce fusion using an inertial-confinement
reactor design inspired by a very noisy crustacean.
The pistol shrimp’s defining feature is its oversize pistol-
like claw, which it uses to stun prey. After drawing back the
“hammer” part of its claw, the shrimp snaps it against the
opposite side of the claw, creating a rapid pressure change
that produces vapor-filled voids in the water called cavita-
tion bubbles. As these bubbles collapse, shock waves pulse
through the water at 25 meters per second, enough to take
out small marine animals.
POST YOUR
“The shrimp just wants to use the pressure wave COMMENTS AT
spectrum.ieee.
to stun its prey,” says Nicholas Hawker, First Light’s org/fusionpower-
feb2020
cofounder and CEO. “It doesn’t care that as the cavity
implodes, the vapor inside is compressed so forcefully that it
causes plasma to form—or that it has created the Earth’s only
example of inertial-confinement fusion.” The plasma reaches
temperatures of over 4,700 °C, and it creates a 218-decibel bang.
Hawker focused on the pistol shrimp’s extraordinary claw
in his doctoral dissertation at the University of Oxford, and
he began studying whether it might be possible to mimic and
scale up the shrimp’s physiology to spark a fusion reaction
that could produce electricity.
After raising £25 million (about $33 million) and teaming up
with international engineering group Mott MacDonald, First
Light is building an ICF reactor in which the “claw” consists of
5. Stellarator
THE BIG IDEA: The
stellarator’s spiraling ribbon
shape produces high-density
plasma that’s symmetrical and
more stable than a tokamak’s,
allowing the reactor to run for
long periods of time.
REALITY CHECK: The
stellarator’s challenging
geometry makes it
complicated to build and
extremely sensitive to
imperfect conditions.
PROJECT TO WATCH:
Wendelstein 7-X at Max Planck
Institute for Plasma Physics
a metal disk-shaped projectile and a cube with a cavity filled
with deuterium-tritium fuel. The projectile’s impact creates
shock waves, which produce cavitation bubbles in the fuel.
As the bubbles collapse, the fuel within them is compressed
long enough and forcefully enough to fuse.
Hawker says First Light hopes to initiate its first fusion
reaction this year and to demonstrate net energy gain by
2024. But he acknowledges that those achievements won’t
be enough. “Fusion energy doesn’t just need to be scientifi-
cally feasible,” he says. “It needs to be commercially viable.”
No one believes it will be easy, but the extraordinary chal-
lenge of fusion energy—not to mention the pressing need—
is part of the attraction for the many scientists and engineers
who’ve recently been drawn to the field. And increasingly,
they have the resources to finance their work.
“This notion that you hear about fusion being another 30 or
40 or 50 years away is wrong,” says TAE’s Binderbauer,
whose company has raised more than $600 million.
“We’re going to see commercialization of this technol-
ogy in time frames of a half decade.”
Veteran fusion researchers such as Dorland and Horton
tend to have a more tempered outlook. They worry that
grand promises that fall short may undercut public and
investor support, as has happened in the past. Any claims
of commercialization within the decade “are just not true,”
says Dorland. “We’re still a lot more than one breakthrough
away from having a pathway to fusion power.”
What few will argue with, though, is the dire need for
nuclear fusion in the near future.
“I think it’s not going too far to say that fusion is having its
Kitty Hawk moment,” says MIT’s Greenwald. “We don’t have
a 747 jet, but we’re flying.” n
SPECTRUM.IEEE.ORG | FEB 2020 |  37
38  | FEB 2020 | SPECTRUM.IEEE.ORG
 ROUTERS
IN
SPACE

Kepler
Communications’
CubeSats will
create an
Internet for
other satellites
By Mina Mitry

ILLUSTRATION BY Eric Frommelt SPECTRUM.IEEE.ORG | FEB 2020 |  39

F
OR DECADES, THE ASTRONOMICAL COST OF LAUNCHING A
satellite meant that only government agencies
and large corporations ever undertook such a
herculean task. But over the last two decades or
so, newer, commercial rocket designs that accom-
modate multiple payloads have reduced launch
costs dramatically—from about US $54,000 per
kilogram in 2000 to about $2,720 in 2018. That trend in turn
has fostered a boom in the private satellite industry. Since
2012, the number of small satellites—roughly speaking, those
under 50 kilograms—being launched into low Earth orbit
(LEO) has increased 30 percent every year.
One huge problem with this proliferation of small satellites
is communicating with the ground. Satellites in low Earth orbit
circle the planet about once every 90 minutes, and so they
usually have only about a 10-minute window during which to
communicate with any given ground station. If the satellite
can’t communicate with that ground station—because it’s on
the other side of the planet, for example—any valuable data the
satellite needs to send won’t arrive on Earth in a timely way.
At present, NASA’s Tracking and Data Relay Satellite System
(TDRSS) is the only network that can help route signals from
satellites to the correct ground stations. However, TDRSS is
rarely accessible to companies, prohibitively expensive to
use, and over 25 years old. It’s simply unable to handle the
traffic created by all the new satellites. Getting data back to
Earth from a satellite is oftentimes one of the bottlenecks
that limits an observation system’s capabilities.
With three other engineers, I started Kepler Communica-
tions in 2015 to break this bottleneck. Our goal is to create a
commercial replacement for TDRSS by building a constel-
lation of many tiny satellites in LEO. The satellites will form
the backbone of a space-based mesh network, sending data
back and forth between Earth and space in real time. Each
40  | FEB 2020 | SPECTRUM.IEEE.ORG
of our satellites, roughly the size of a loaf of bread, will oper-
ate much like an Internet router—except in space. Our first
satellite, nicknamed KIPP after the companion robot from
the 2014 sci-fi epic Interstellar, launched in January 2018.
When fully deployed by 2022, Kepler’s network will include
140 satellites spread equally among seven orbital planes. In
essence, we’re building an Internet service provider high
above Earth’s surface, to allow other satellites to stay in con-
tact with one another and with ground stations, even if two
satellites, or a satellite and a ground station, are on opposite
sides of the planet. Our customers will include companies
operating satellites or using satellite communications to
transfer data, as well as government agencies like the Cana-
dian Department of National Defense, the European Space
Agency, and NASA. None of this would be possible without
the ongoing developments in building tiny satellites.
KEPLER’S SATELLITES ARE what the aerospace community calls
CubeSats. In the early 2000s, CubeSats were developed to try
to reduce the cost of satellites by simplifying and standard-
izing their design and manufacture. At the time, each new
satellite was a one-off, custom-built spacecraft, created by
teams of highly specialized engineers using bespoke materi-
als and fabrication methods. In contrast, a CubeSat is made
up of a standardized multiple of 10- by 10- by 10-centimeter
units. The fixed units allow manufacturers to develop essen-
tial CubeSat parts like batteries, solar panels, and comput-
ers as commercial, off-the-shelf components.
KEPLER COMMUNICATIONS
Thanks to CubeSats, space startups like Kepler can
design, build, and launch a satellite, from napkin sketch
to orbital insertion, in as little as 12 months. For compari-
son, a ­traditional satellite program can take three to seven
years to complete.
 The rise of CubeSats and the falling costs of launch have
led to a surge in commercial satellite services. Companies
around the world are building constellations of simultane-
ously operating spacecraft, with some planned constella-
tions numbering in the hundreds. Companies like Planet
are focused on delivering images of Earth, while others, like
Spire Global, aim to monitor the weather.
So how are all those satellites getting all the data they col-
lect back to customers on the ground? The short answer is,
they’re not. A single Earth-imaging CubeSat, for example,
can collect something like 2 gigabytes in one orbit, or 26 GB
per day. Realistically, the CubeSat will be able to send back
only a fraction of that data during its short window above
a particular ground station. And that’s the case for all the
companies now operating satellites to collect data on agri-
culture, climate, natural disasters, natural-resource man-
agement, and other topics. There is simply too much data
for the communications infrastructure to handle efficiently.
To hand off its data, an Earth-observation satellite sends its
imagery and other measurements by contacting its ground
station when one is in sight. Such satellites are almost always
in low Earth orbits to improve their image resolution, but,
as I mentioned, that means they orbit the planet roughly
once every 90 minutes. On average, the satellite has a line of
sight—and therefore a line of communication—with a specific
ground station for about 10 minutes. In that 10-minute win-
dow, the satellite must transmit all the data it has collected
so that the ground station can then relay it through terres-
trial networks to its final destination, such as a data center.
The result is that satellite operators often collect far more
information than they can ever hope to send back to Earth, so
they are constantly throwing away valuable data, or retriev-
ing it only after a delay of hours or even days.
One recent solution is to operate ground stations as a ser-
vice in order to increase the total number of ground stations
available for use by any one company. Historically, when a
company or government agency launched a satellite, it would
also be responsible for developing its own ground stations—a
very costly proposition. Imagine how expensive and compli-
cated it would be if all cellphone users also had to purchase
Number of small satellites launched
their own towers and operate their own network just to make
a call. A cheaper alternative is for companies to build ground
stations that anyone—for a price—can use to connect with their
satellites, like Amazon’s AWS Ground Stations.
But there’s still a catch. To ensure that a LEO satellite can
continuously communicate with a ground station, you basically
need ground stations all over the globe. For continuous cover-
age, you would need several thousand ground stations—one
every few hundred kilometers, though more closely spaced
ground stations would ensure more reliable connections.
That can be difficult at best in remote areas on land. It’s even
more difficult to maintain a connection over oceans, where
SOURCE: SPACEWORKS
islands to build on are few and far between, and those islands
rarely, if ever, have robust fiber connections to the Internet.
That’s why Kepler plans to move more of the communica-
tions infrastructure into orbit. Rather than creating a globe-
spanning network of ground stations, we think it makes more
sense to build a constellation of CubeSat routers, which can
keep satellites connected to ground stations regardless of
where the satellite or the ground station is.
A
T THE HEART OF EACH 5-kilogram Kepler satellite
is a software-defined radio (SDR) and a pro-
prietary antenna. SDRs have been around
since the 1990s. At the most fundamental level,
they replace analog radio components like
modulators (which convert analog signals to
ones and zeros) and filters (which limit what
part of the analog signal gets converted) with software. In
Kepler’s SDR, these elements are implemented using soft-
ware running on a field-programmable gate array, or FPGA.
The result is a radio that’s cheaper to develop and easier to
configure. The use of SDRs has in turn allowed us to shrink
our spacecraft to CubeSat scale. It’s also one of the reasons
our satellites cost one-hundredth as much as a traditional
communication satellite.
To understand how the Kepler constellation will work, it
helps to know how conventional satellite connections are
made: with the “bent pipe” method. Imagine the pipe as two
straight lengths of pipe joined together at an angle; the satel-
lite sits where the two lengths meet so it has a continuous line
of sight with both ends of the connection, whether they’re
two ground stations on different continents or a ground sta-
tion and another spacecraft. The satellite essentially acts
as a relay, receiving the signal at one end of the connection
and transmitting it in a different direction to the other end.
In Kepler’s network, when each satellite passes over a
ground station it will receive data that has been routed to that
ground station from elsewhere in terrestrial networks. The
satellite will store the data and then transmit it later when
the destination ground station becomes visible. Kepler’s
800
700
600
500
400
300
200
100
0
’13 ’14 ’15 ’16 ’17 ’18 ’19 ’20 ’21 ’22 ’23
SKYROCKETING SATELLITES: The number of newly launched
small satellites weighing 50 kilograms or less rises every year, with
nearly 500 such satellites expected to be launched in 2020.
SPECTRUM.IEEE.ORG | FEB 2020 |  41
SOME ASSEMBLY
REQUIRED: A Kepler
Communications
engineer works to hand
assemble KIPP [top
left], the company’s
first satellite in orbit.
Software-defined
radio components [top
right] replace many
analog parts and make
it possible to build
satellites that are the
size of a loaf of bread. A
mostly completed KIPP
waits for the finishing
touches [left]. Kepler
has also built ground
stations [bottom left] so
that its satellites in orbit
can communicate with
terrestrial networks. The
SatOps team [bottom
right] keeps an eye
on Kepler’s orbiting
satellites, which will
grow in importance
as the constellation
becomes larger.
 network will include five ground-station sites spread across
five continents to connect all of our satellites. Unfortunately,
this method doesn’t allow for real-time communications. But
those will become possible as our satellite constellation grows.
Here’s how: Future iterations of our network will add the
ability to send data between our satellites to create a real-
time connection between two ground stations located any-
where on the planet, as well as between a ground station and
an orbiting satellite. We’re also planning to include new fea-
tures such as transcoding—essentially a way to translate the
data into different formats—and queueing the data accord-
ing to what needs to be delivered most urgently.
We can make these big changes to how our satellites com-
municate relatively quickly, thanks to SDR. New code, for
example, can be uploaded to an orbiting satellite like KIPP
for testing. If the code passes muster, it can be deployed to
the rest of the constellation without having to replace any of
the hardware. Much like CubeSat standardization, SDR short-
ens development cycles and lets us prototype more ideas.
K
EPLER IS CURRENTLY IN the process of deploying
our constellation. KIPP has been operating
successfully for over two years and is support-
ing the communication needs of our ground
users. The MOSAiC expedition, for example,
is a yearlong effort to measure the arctic cli-
mate from an icebreaker ship near the North
Pole. It’s the largest polar expedition in history. Since the
start of the mission, KIPP’s high-bandwidth communica-
tion payload has been regularly transferring gigabytes of
data from MOSAiC’s vessel to the project’s headquarters in
Bremerhaven, Germany.
In December 2018, our second satellite, CASE (named after
another robot companion from Interstellar), joined KIPP in
orbit. Even with just two satellites in operation, we’re able to
provide some level of service to our customers, mostly by tak-
ing up data from one ground station and delivering it to another,
in the method I previously described. That has allowed us to
avoid the fate of some other satellite-­constellation compa-
nies, which went bankrupt in the process of trying to deploy
a complete network prior to delivering service.
While we’ve been successful so far, establishing a constel-
lation of 140 satellites is not without challenges. When two
fast-moving objects—such as satellites—try to talk to each other,
their communications are affected by a Doppler shift. This
phenomenon causes the frequencies of radio waves transmit-
ted between the two objects to change as their relative posi-
tions change. Specifically, the frequency is compressed as the
objects approach each other and stretched as they grow more
distant. It’s the same phenomenon that causes an ambulance
KEPLER COMMUNICATIONS (5)
siren to change in pitch as it speeds past you.
With our satellites traveling at over 7 kilometers per sec-
ond relative to the ground or potentially communicating
with another satellite moving in the opposite direction at the
same speed, we end up with a very compressed or stretched
signal. To deal with this issue, we’ve created a proprietary
network architecture in which adjacent satellites will com-
municate with each other only when they’re traveling in
the same direction. We’ve also installed software on KIPP
and CASE to manage Doppler shift by tracking the change
in frequency caused by its relative motion. At this point, we
believe we’re able to compensate for any Doppler shifts, and
we expect to improve upon that capability in future itera-
tions of our network and software.
As the number of satellites in the constellation increases,
we must also ensure that data is routed efficiently. We don’t
want to beam data among, say, 30 satellites when just 3 or 4
will do the job. To solve this problem, our satellites will run
an algorithm in orbit that uses something called a two-line
element set to determine the position of each satellite. A
two-line element set operates in a way that’s similar to how
GPS identifies locations on Earth. Knowing every satellite’s
position, we can run an optimization algorithm to figure out
the route with the shortest transit time.
Of course, all these challenges will be moot if we can’t actu-
ally build the 140 satellites and place them in orbit. One thing
we discovered early on is that supply chains for producing
hundreds of spacecraft—even small ones—don’t yet exist, even
if the components are standardized. Ultimately, we’ve had
to do most of the production of our satellites ­in-house. Our
manufacturing facility in downtown Toronto can produce
up to 10 satellites per month by automating what were pre-
viously manual processes, such as testing circuit boards to
ensure they meet our requirements.
As I’ve said before, Kepler’s constellation will be possible
because of the drastic size and cost reductions in satellite
components in recent years. But there is one area where effi-
ciency has limited miniaturization: solar panels. Our CubeSats’
ability to generate power is still constrained by the surface
area on which we can mount solar panels.
We’re also seeing limitations in antenna size, as antennas
reach theoretical limits in efficiency. That means a certain
amount of each satellite’s surface area must be reserved for
the antennas. Such limitations will make it hard to further
shrink our satellites. The upside is that it’s forcing us to be
creative in finding new computational methods and software,
and even develop foldable components inspired by origami.
By the end of 2020, we plan to have at least 10 more sat-
ellites operating in orbit—enough to run early tests on our
in-space router network. If everything stays on schedule,
by 2021 we will have 50 satellites in operation, and by 2022
all 140 satellites will be available to both users on Earth and
other satellites in space.
Space is the new commercial frontier. With the increased
level of access that the entrepreneurial space race has brought,
upstart groups are imagining new opportunities that a space-
craft in orbit, and its data, can provide. By creating an Inter-
net for space, Kepler’s network will give those opportunities
a route to success. n
POST YOUR COMMENTS AT spectrum.ieee.org/kepler-feb2020
SPECTRUM.IEEE.ORG | FEB 2020 |  43
GUTTER CREDIT GOES HERE
 ATOM-THICK TRANSISTORS

If there’s one thing about With the ability to

Moore’s Law that’s obvious to anyone, it’s grow materials like
that transistors have been made smaller
tungsten disulfide on
300-mm wafers, 2D
and smaller as the years went on. Scien-
semiconductors might
tists and engineers have taken that trend find a role in advanced
to an almost absurd limit during the past silicon chips
decade, creating devices that are made
of one-atom-thick layers of material. • By Iuliana Radu

The most famous of these materials is,

of course, graphene, a hexagonal honeycomb-shaped sheet of carbon with
outstanding conductivity for both heat and electricity, odd optical abilities,
and incredible mechanical strength. But as a substance with which to make
transistors, graphene hasn’t really delivered. With no natural ­bandgap—
the property that makes a semiconductor a semiconductor—it’s just not
built for the job. • Instead, scientists and engineers have been exploring
the universe of transition metal dichalcogenides, which all have the chemi-
cal formula MX2. These are made up of one of more than a dozen transition
metals (M) along with one of the three chalcogenides (X): sulfur, selenium,
or tellurium. Tungsten disulfide, molybdenum diselenide, and a few ­others
can be made in single-atom layers that (unlike graphene) are natural semi-
conductors. These materials offer the enticing prospect that we will be able
to scale down transistors all the way to atom-thin components long after
today’s silicon technology has run its course.

ILLUSTRATION BY Chad Hagen SPECTRUM.IEEE.ORG | FEB 2020 |  45

ATOM-THICK TRANSISTORS
While this idea is really exciting, I and my colleagues at Imec
believe 2D materials could actually show up much sooner, even
while silicon still remains king. We’ve been developing a tech-
nology that could put 2D semiconductors to work in silicon
chips, enhancing their abilities and simplifying their designs.
Dev ices m a de w i t h 2D m at er i a l s are worth all the
scientific and engineering work we and other researchers
around the world have put into them because they could
eliminate one of the biggest problems with today’s transis-
tors. The issue, the result of what are called short-channel
effects, is a consequence of the continual shrinking of the
transistor over the decades.
A metal-oxide semiconductor field-effect transistor (MOSFET)­,
the type of device in all digital things, is made up of five basic
parts: The source and drain electrodes; the channel region
that connects them; the gate dielectric, which covers the
channel on one or more sides; and the gate electrode, which
contacts the dielectric. Applying a voltage at the gate relative
to the source creates a layer of mobile charge carriers in the
channel region that forms a conductive bridge between the
source and drain, allowing current to flow.
But as the channel was made smaller and smaller, current
would increasingly leak across it even when there was no volt-
age on the gate, wasting power. The change from the planar
designs of the 20th century to the FinFET transistor structure
used in today’s most advanced processors was an attempt to
counter this important short-channel effect by making the
channel region thinner and having the gate surround it on
more sides. The resulting fin-shaped structure provides bet-
ter electrostatic control. (The coming move to the nanosheet
transistor is a furthering of this same idea. See “The Last Sili-
con Transistor,” IEEE Spectrum, August 2019.)
GETTING 2D
SEMICONDUCTORS
on a S I L I C O N W A F E R
Glass Laser Adhesive
carrier release
layer
Tungsten Growth
disulfide (WS2 ) wafer
Chemical vapor A specialized glass
1 2
deposition grows carrier wafer is
WS2 on a silicon wafer. bonded to the WS2 .
46  | FEB 2020 | SPECTRUM.IEEE.ORG
Certain 2D semiconductors could circumvent short-channel
effects, we think, by replacing the silicon in the device chan-
nel. A 2D semiconductor provides a very thin channel region—
as thin as a single atom if only one layer of semiconductor is
used. With such a restricted pathway for current to flow, there
is little opportunity for charge carriers to sneak across when
the device is meant to be off. That means the transistor could
continue to be shrunk down further with less worry about the
consequences of short-channel effects.
These 2D materials are not only useful as semiconductors,
though. Some, such as hexagonal boron nitride, can act as gate
dielectrics, having a dielectric constant similar to that of sili-
con dioxide, which was routinely used for that job until about
a decade ago. Add graphene in place of the transistor’s metal
parts and you’ve got a combination of 2D materials that forms
a complete transistor. Indeed, separate groups of researchers
built such devices as far back as 2014. While these prototypes
were much larger, you could imagine scaling them down to
the size of just a few nanometers.
As amazing as an all-2D transistor that’s a fraction of the
size of today’s devices might be, that won’t be the first imple-
mentation of 2D materials in electronic circuits. Instead, 2D
materials will probably arrive in low-power circuits that have
more relaxed performance requirements and area constraints.
The set of circuits we’re targeting at Imec are built in the so-
called back-end-of-line. Chipmaking is divided into two parts:
the front-end-of-line part consists of processes—many of them
requiring high temperatures—that alter the silicon itself, such
as implanting dopants to define the parts of a transistor. The
back-end-of-line part builds the many layers of interconnects
that link the transistors to form circuits and deliver power.
With traditional transistor scaling becoming more and more
difficult, engineers have been looking for ways to add func-
tionality to the interconnect layers. You can’t do this simply
by using ordinary silicon processes because the heat involved
would damage the devices and interconnects beneath them.
So, many of these schemes rely on materials that can be made
into devices at relatively low temperatures.
Growth Target Laser
wafer silicon
wafer
FLIP FLIP
The growth wafer is A laser melts away the
3 4
peeled away, leaving release layer and the
the WS2 atop the carrier. carrier wafer is peeled off.
 A specific advantage of using 2D semiconductors instead of
some other candidates is the potential ability to build both
p-type (carrying positive charges) and n-type (carrying elec-
trons) devices, a necessity in CMOS logic. CMOS circuits are
the backbone of today’s logic because, ideally, they consume
power only when switching from one state to the other. In our
preferred 2D semiconductor, we’ve demonstrated n-type tran-
sistors but not yet p-type. However, the physics underlying these
materials strongly suggests we can get there through engineer-
ing the dielectrics and metals that contact the semiconductor.
Being able to produce both p- and n-type devices would
allow the development of compact back-end logic circuits
such as repeaters. Repeaters essentially relay data that must
travel relatively far across a chip. Ordinarily, the transistors
involved reside on the silicon, but that means signals must
climb up the stack of interconnects until they reach a layer
where they can travel part of the distance to their destination,
then go back down to the silicon to be repeated and up again
to the long-distance interconnect layer. It’s a bit like having to
exit the highway and drive into the center of a crowded city to
buy petrol before getting back on the highway.
A repeater up near the long-distance interconnect layer is
more akin to a motorway petrol station. It saves the time it
would take the signal to make the two-way vertical trip and
also prevents the loss of power due to the resistance of the
vertical interconnects. What’s more, moving the repeater to
the interconnect layer saves space on the silicon for more logic.
Repeaters aren’t the only potential use. A 2D material could
also be used to build other circuits, such as on-chip power-
management systems, signal buffers, and memory selectors.
One thing these circuits all have in common is that they don’t
require the device to drive a lot of current, so one layer of 2D
material would probably be sufficient.
N either f u t u r e su per sm a ll 2D devices nor the less
demanding back-end-of-line circuits will be possible with-
out a fabrication process compatible with industry-standard
300-­millimeter silicon wafers. So our team at Imec is work-
ing on just that, hoping to develop a process that will serve
for all applications.
The first step is identify-
ing the most promising 2D
material and device archi-
tecture. We have therefore
benchmarked a variety of
2D semiconductors and 2D
FET architectures against
an advanced silicon FinFET
device.
Because researchers have
Chemicals and
5 the most experience with
plasma scrub off the
adhesive to reveal the WS2 . molybdenum disulfide (MoS2),
experimental devices made
using it have advanced furthest. Indeed, at the IEEE Inter­
national Electron Device Meeting last December, Imec unveiled
an MoS2 transistor with a channel just 30 nanometers across
and source and drain contacts only 13 nm long. But after exam-
ining the possibilities, we’ve decided that MoS 2 is not the
answer. Instead, we concluded that among all the materials
compatible with 300-mm ­silicon-wafer technology, tungsten
disulfide (WS2 ) in the form of a stacked nanosheet device has
the highest performance potential, meaning it can drive the
most current. For less demanding, back-end-of-line applica-
tions, we also concluded that a FET architecture with a gate
both below and above the semiconductor channel region
works better than one with only a single gate.
We already knew one important thing about WS2 before
we reached that conclusion: We can make a high-quality ver-
sion of it on a 300-mm silicon wafer. We demonstrated that
for the first time in 2018 by growing the material on a wafer
using metal-organic chemical vapor deposition (MOCVD), a
common process that grows crystals on a surface by means of
a chemical reaction. The approach we took results in thickness
control down to a single-molecule layer, or monolayer, over
the full 300-mm wafer. The benefits of the MOCVD growth
come, however, at the price of a high temperature—and recall
that high temperatures are forbidden in back-end processes
because they could damage the silicon devices below.
To get around this problem, we grow the WS2 on a separate
wafer and then transfer it to the already partially fabricated
silicon wafer. The Imec team developed a unique transfer pro-
cess that allows a single layer of WS2—as thin as 0.7 nm—to be
moved to a silicon target wafer with negligible degradation in
the 2D material’s electrical properties.
The process starts by growing the WS2 on an oxide-­covered
silicon wafer. That’s then placed in contact with a specially pre-
pared wafer. This wafer has a layer of material that melts away
when illuminated by a laser. It also has a coating of adhesive.
The adhesive side is pressed to the WS2-covered wafer, and the
2D material peels away from the growth wafer and sticks to the
adhesive. Then the adhesive wafer with its 2D cargo is flipped
over onto the target silicon wafer, which in a real chipmaking
effort would already have transistors and several layers of inter-
connect on it. Next, a laser is shone through the wafer to break
the bulk of it away, leaving only the adhesive and the WS2 atop
the target wafer. The adhesive is removed with chemicals and
plasma. What’s left is just the processed silicon with the WS2
attached to it, held in place by Van der Waals forces.
The process is complicated, but it works. There is, of course,
room for improvement, most importantly in mitigating defects
caused by unwanted particles on the wafer surface and in
eliminating some defects that occur at the edges.
Once the 2D semiconductor has been deposited, build-
ing devices can begin. On that front there have been triumphs,
but some major challenges remain.
SPECTRUM.IEEE.ORG | FEB 2020 |  47
 scattering problem is more pronounced because the interface
is the channel.
Sulfur vacancies are the most common defects that affect
device channel regions. Imec is investigating how different
Source Drain plasma treatments might make those vacancies less chemi-
Top gate Top cally reactive and therefore less prone to alter the transistor’s
(metal) dielectric behavior. We also need to prevent more defects from forming
after we’ve grown the monolayer. WS2 and other 2D materials
are known to age quickly and degrade further if already defec-
tive. Oxygen attacking a sulfur vacancy can cause more vacan-
cies nearby, making the defect area grow larger and larger. But
Bottom gate WS2 Bottom we’ve found that storing the samples in an inert environment
(silicon) semiconductor dielectric makes a difference in preventing that spread.
Defects in the semiconductor aren’t the only problems we’ve
2D TRANSISTOR: A two-gate device structure works best for devices encountered trying to make 2D devices. Depositing insulating
meant to exist in the interconnect layers of a chip. The tungsten disulfide
semiconductor is barely visible between the metal source and drain. materials on top of the 2D surface to form the gate dielectric is
Different dielectrics separate the semiconductor from the two gates. a true challenge. WS2 and similar materials lack dangling bonds
that would otherwise help fasten the dielectric to the surface.
Perhaps the most crucial issue to tackle is the creation of Our team is currently exploring two routes that might help:
defects in the WS2. Imperfections profoundly degrade the per- One is atomic layer deposition (ALD) at a reduced growth tem-
formance of a 2D device. In ordinary silicon devices, charge perature. In ALD, a gaseous molecule adsorbs to the semicon-
can get caught in imperfections at the interface between the ductor’s exposed surface to form a single layer. Then a second
gate dielectric and the channel region. These can scatter elec- gas is added, reacting with the adsorbed first one to leave an
trons or holes near the interface as they try to move through atomically precise layer of material, such as the dielectric

IMEC
the device, slowing things down. With 2D semiconductors the ­hafnium dioxide. Doing this at a reduced temperature increases

w
Best Ne
in
Journal
15
STM 20

Become a published author in 4 to 6 weeks.

IEEE Access is a multidisciplinary journal that allows you to:
• Reach millions of global users through the IEEE Xplore® digital library with free access to all
• Submit multidisciplinary articles that do not fit neatly in traditional journals
• Expect a rapid yet rigorous peer review—a key factor why IEEE Access is included in Web of Science
(and has an Impact Factor)
• Establish yourself as an industry pioneer by contributing to trending, interdisciplinary topics in one
of the Special Sections
• Integrate multimedia and track usage and citation data for each published article
• Connect with readers through commenting
• Publish without a page limit for only $1,750 per article

IEEE Access...a multidisciplinary open access journal

that’s worthy of the IEEE.
17-PUB-013 3/17

Learn more at: ieeeaccess.ieee.org

48  | FEB 2020 | SPECTRUM.IEEE.ORG
the ability of the gas molecules to stick to the surface of the tric; and we’re on a path to boost the mobility of charge carri-
WS2 even when no chemical bonds are available. ers in devices toward a level that could compare with silicon.
The other option is to enhance ALD by using a very thin oxi- But, as we’ve laid out, there are still significant problems
dized layer, such as silicon oxide, to help nucleate the growth remaining. These will require an intensive engineering effort
of the ALD layer. A very thin layer of silicon is deposited by and an even better fundamental understanding of this new class
a physical deposition method such as sputtering or evapo- of intriguing 2D materials. Solving these challenges will enable
ration; it’s then oxidized before a regular ALD deposition of high-performance devices that are scaled down to atomic layers,
gate oxide is done. We’ve achieved particularly good results but they might first bring new capabilities that need less demand-
with evaporation. ing specifications even as we continue to scale down silicon. n
A further challenge in making superior 2D devices is in choos-
ing the right metals to use as source and drain contacts. Metals POST YOUR COMMENTS AT spectrum.ieee.org/2dtransistors-feb2020
can alter the characteristics of the device,
depending on their work function. That
parameter, the minimum energy needed
to extract an electron from the metal, can
mean the difference between a contact
that can easily inject electrons and one
that can inject holes. So the Imec team
has screened a variety of metals to put
in contact with the WS2 nanosheet. We
found that the highest on-current in an
n-type device was obtained using a mag-
nesium contact, but other metals such
as nickel or tungsten work well. We’ll be
searching for a different metal for future
p-type devices.
Despite these challenges, we’ve been
able to estimate the upper limits of device
performance, and we’ve mapped out
what roads to follow to get there.
As a benchmark, the Imec team used
dual-gated devices like those we described
earlier. We built them with small, natu-
rally exfoliated flakes of WS2, which have
fewer defects than wafer-scale semicon-
ductors. For these lab-scale devices, we
were able to measure electron mobility
values up to a few hundred square centi­
meters per volt-second, which nearly
matches crystalline silicon and is close
to the theoretically predicted maximum
for the 2D material. Because this excellent Over 50 New Features & Apps in this New Version!
mobility can be found in natural mate-
rial, we are confident that it should also Over 500,000 registered users worldwide in:
be possible to get there with materials
synthesized on 300‑mm­wafers, which ■ 6,000+ Companies including 20+ Fortune Global 500
currently reach just a few square centi- ■ 6,500+ Colleges & Universities
meters per volt-second. ■ 3,000+ Government Agencies & Research Labs

F or s om e of t h e main challenges
®

ahead in 2D semiconductor develop-

ment, our team has a clear view of the www.originlab.com
solutions. We know, for example, how
25+ years serving the scientific and engineering community.
to grow and transfer the material onto a
300-mm target wafer; we’ve got an idea
For a 60-day FREE TRIAL, go to OriginLab.Com/demo and enter code: 8547
of how to integrate the crucial gate dielec-

SPECTRUM.IEEE.ORG | FEB 2020 |  49
IS IT TIME TO DRAG
Careers Rank Metro Area
Population, Eligibility
TECH JOBS OUT OF 2018 Index

SILICON VALLEY? 1 Madison, Wis. 660,422 1.63

SPREADING EMPLOYMENT
Minneapolis–St. Paul–
AROUND WILL LIKELY REQUIRE U.S. 2
Bloomington, Minn.–Wis.
3,629,190 0.68
GOVERNMENT INVOLVEMENT
3 Albany–Schenectady–Troy, N.Y. 883,169 0.66
SILICON VALLEY CONTINUES TO DOMINATE.
In every study I’ve seen, it has the lion’s share of tech 4 Lexington–Fayette, Ky. 516,697 0.58
jobs. Its engineers command the highest salaries. This is de-
spite the fact that over the years, various regions in the United 5 Rochester, N.Y. 1,071,082 0.53

States and worldwide have pitched themselves as the “next

6 Provo–Orem, Utah 633,768 0.47
­Silicon Valley” (or “Silicon Glen” or “Silicon Fen.” ) Some have
indeed increased their pool of tech jobs. But none have really Portland–Vancouver–
7 2,478,810 0.47
become a serious technology hub, outside of Boston, ­Seattle, Hillsboro, Ore.–Wash.
San Diego, and North Carolina’s ­Research Triangle.
8 Tucson, Ariz. 1,039,073 0.45
In fact, notes a recent study by the Brookings Institution,
Boston, Seattle, and San Diego plus Silicon Valley accounted
9 Pittsburgh, Pa. 2,324,743 0.40
for more than 90 percent of tech-job growth from 2005 to
2017. Other “superstar” metro areas are growing quickly. By 10 Salt Lake City, Utah 1,222,540 0.34
contrast, much of the rest of the United States is starving for
tech jobs—and losing ground. 11 Columbus, Ohio 2,106,541 0.30
The solution the Brookings’ researchers propose? Govern-
ment intervention. They suggest the federal government cre- Chicago–Naperville–Elgin, Ill.–
12 9,498,716 0.29
Ind.–Wis.
ate eight to 10 regional “growth centers” in the U.S. heartland.
Each area, they argue, should annually receive US $700 mil- Nashville–Davidson–
13 1,930,961 0.22
Murfreesboro–Franklin, Tenn.
lion in direct R&D funding for the next 10 years, workforce
development funding of $5 million per year, exemptions from 14 Akron, Ohio 704,845 0.19
certain regulations, and other benefits, for a total 10-year cost
of about $69 billion to $100 billion. 15 St. Louis, Mo.–Ill. 2,805,465 0.19
Where are these potential Silicon Valleys? The report iden-
tified 35 possibilities (we show 20 here), calculating an eli- 16 Boise City, Idaho 730,426 0.18
gibility index that considers factors like population and the
Milwaukee–Waukesha–West
presence of a university. Madison, Wis., came out on top, fol- 17 1,576,113 0.18
Allis, Wis.
lowed by the Minneapolis–St. Paul metropolitan area, the
­Albany area of New York, and the Lexington area of Kentucky. 18 Cincinnati, Ohio–Ky.–Ind. 2,190,209 0.16
—TEKLA S. PERRY
Buffalo–Cheektowaga–
An extended version of this article appears in our View 19 1,130,152 0.15
SOURCE: BROOKINGS INSTITUTION

Niagara Falls, N.Y.

From the Valley blog.
20 Kansas City, Mo.–Kan. 2,143,651 0.14
POST YOUR COMMENTS AT spectrum.ieee.org/techjobs-feb2020

DEPARTMENTS

50  | FEB 2020 | SPECTRUM.IEEE.ORG
Information Technology Data Engineer (Regions Bank –
Hoover, AL) Utilize minimum requirements to analyze, design,
develop, test complex data transformation solutions for an
Enterprise Data Warehouse system. Develop ETL programs
using Informatica ETL technology. Work with data stored in
DB2, SQL Server, VSAM, ISAM, Oracle and flat files. Write/
deploy stored procedures, tune SQL scripts/queries. Architect
scalable performing ETL solutions. Build ETL solutions to
extract and transform data from mainframe systems. Design
systems that are recoverable, auditable, parameter driven.
Build processes that meet Data Lifecycle management
standards. Work with Structured and Unstructured data.
Implement ETL solutions. Processing and error handling
techniques. Minimum Requirements: Must have a Bachelor’s
degree or equivalent in Computer Science, Information
Systems, Electronics Engineering or related field and 5 years
of experience in the job offered or 5 years in a data warehouse
engineering position. Will accept educational equivalence
evaluation prepared by qualified evaluation service or in
accordance with 8 CFR 214.2(h)(4)(iii)(D). Education or
work experience may have been obtained concurrently and
must include 5 years of experience in each of the following:
Informatica PowerExchange and PowerCenter tools and
architecture, Oracle databases procedures, structures and
data manipulation; Unix or LINUX scripting and navigation;
data warehouse modeling concepts; ETL standards,
methodologies, guidelines and techniques; DBMS concepts of
data structures and normalization. Must have legal authority to
work in U.S. Any suitable combination of education, training
or experience is acceptable. Resume to Regions Bank, Attn:
Ethan Dorman, Talent Acquisition Partner, Riverchase Center,
2090 Parkway Office Circle, Hoover AL 35244.
The Department of Electrical
and Computer Engineering in
the College of Engineering at
Texas A&M University invites applications for up to five full-time, tenured or
tenure-track faculty positions with a 9-month academic appointment, and the
possibility of an additional summer appointment contingent upon need and
availability of funds, beginning the fall of 2020. Applicants will be considered for
the faculty titles of assistant professor, associate professor, and full professor
depending on qualifications. The successful applicant will be required to teach
at the graduate and undergraduate levels; advise and mentor graduate students;
develop an independent, externally funded research program; participate in
all aspects of the department’s activities; and serve the profession. Applicants
must have earned a doctorate in Electrical and Computer Engineering or a
closely related engineering or science discipline. Strong written and verbal
communication skills are required. Applicants should submit a cover letter,
curriculum vitae, teaching statement, research statement, diversity statement
(optional) and a list of four references (including postal addresses, phone
numbers and email addresses) by applying for this specific position at http://
apply.interfolio.com/71388. For more information, contact Dr. Thomas
Overbye at overbye@tamu.edu.
Texas A&M University is committed to enriching the learning and working
environment for all visitors, students, faculty, and staff by promoting a
culture that embraces inclusion, diversity, equity, and accountability. Diverse
perspectives, talents, and identities are vital to accomplishing our mission and
living our core values. Equal Opportunity/Affirmative Action/Veterans/Disability
Employer committed to diversity.

College of Computing Open Rank Faculty Search for the

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
Robotics/Mechatronics Assistant or
Associate Professor Positions (2)
Michigan Technological University’s College of
Computing invites applications for a tenure-track faculty
position at the Assistant or Associate Professor level
in the ETAC-ABET-accredited Electrical Engineering
Technology (BS) program and the Mechatronics (MS)
program. The successful candidate will teach in the
areas of programmable logic controllers, robotics,
mechatronics, industrial control and automation, and
related topics in support of a new MS in Mechatronics,
which began accepting students in Fall 2019.
Review of applications will begin immediately and
continue until the position is filled. Applications should
be submitted online at http://www.jobs.mtu.edu/
postings/8530.
Michigan Technological University is an Equal
Opportunity Educational Institution/Equal Opportunity
Employer, which includes providing equal opportunity
for protected veterans and individuals with disabilities.
Michigan Tech does not discriminate on the basis of
race, creed, color, national origin, sex, marital status,
sexual orientation, gender identity, age, disability,
religion, honorably discharged veteran or military
status, or the use of a trained guide dog or service
animal by a person with a disability. Persons requiring
reasonable accommodation in the application process
or requiring information in an alternative format should
contact Human Resources at mtujobs@mtu.edu or
906/487-2280.  edu/hr/resources/er/workplacepostings/.
Bruce Eisenstein Endowed Chair in
Electrical and Computer Engineering
The Electrical and Computer Engineering (ECE)
Department at Drexel University, Philadelphia, PA
invites applicants from all ranks (assistant, associate
and full professor) for the newly established Bruce
Eisenstein Endowed Chair faculty position. Qualified
candidates with an outstanding scholarly record,
commensurate with their academic rank, in any field
of ECE, as well as interdisciplinary and cross-cutting
areas, will be given full consideration.
Applications should be made through the Drexel
Careers website for consideration. The application
process is detailed at the following site: http://
drexel.edu/engineering/ece-jobs. General inquiries
should be directed to Dr. Baris Taskin, Search
Committee Chair, at ece.search@drexel.edu.
DREXEL UNIVERSITY is an Equal Opportunity/
Affirmative Action employer, welcomes individuals
from diverse backgrounds and perspectives,
and believes that an inclusive and respectful
environment enriches the University community
and the educational and employment experience
of its members. Please visit our website to view all
University Policies https://drexel.edu/hr/resources/
policies/ and Workplace Postings https://drexel.

SPECTRUM.IEEE.ORG | FEB 2020 | 51
 PAST FORWARD BY ALLISON MARSH

OAK RIDGE ASSOCIATED UNIVERSITIES

FUN WITH “Users should not take ore samples out of their jars, for they tend to flake and crumble and you would run the
risk of having radioactive ore spread out in your laboratory.” Such was the warning that came with the Gilbert U-238
URANIUM! Atomic Energy Lab, a 1950s science kit that included four small jars of actual uranium. Budding young nuclear
scientists were encouraged to use the enclosed instruments to measure the samples’ radioactivity, observe radioactive
decay, and even go prospecting for radioactive ores. Yes, the Gilbert company definitely intended for kids to try this
at home. And so the company’s warning was couched not in terms of health risk but rather as bad scientific practice:
Removing the ore from its jar would raise the background radiation, thereby invalidating your experimental results. ■
↗ For more on the history of radioactive toys, see spectrum.ieee.org/pastforward-feb2020

52  | FEB 2020 | SPECTRUM.IEEE.ORG
The World’s
OP2
Courtesy of
ROBOTIS

Best
ROBOTS
GUIDE
Is Here!
ROBOTS.IEEE.ORG
IEEE Spectrum’s new ROBOTS site features more
than 200 robots from around the world.
• Spin, swipe and tap to make • Rate robots and check • Play Faceoff, an interactive
robots move. their ranking. question game.
• Read up-to-date robotics • View photography, videos
news. and technical specs.

Check out
Robots.ieee.org
on your desktop,
tablet, or phone now!
MATLAB SPEAKS
DEEP
LEARNING
With MATLAB,® you can build and deploy deep learning
models for signal processing, reinforcement learning,
automated driving, and other applications. Preprocess
data, train models, generate code for GPUs, and deploy
to production systems.

©2020 The MathWorks, Inc.

mathworks.com/deeplearning