of 21
PRX QUANTUM
2,
017002 (2021)
Roadmap
Development of Quantum Interconnects (QuICs) for Next-Generation
Information Technologies
David Awschalom,
1
Karl K. Berggren,
2
Hannes Bernien,
1
Sunil Bhave,
3
Lincoln D. Carr,
4
Paul Davids,
5
Sophia E. Economou,
6
Dirk Englund,
2
Andrei Faraon,
7,8
Martin Fejer,
9
Saikat Guha,
10,11,12
Martin V. Gustafsson,
13
Evelyn Hu,
14,15
Liang Jiang,
1
Jungsang Kim,
16,17
Boris Korzh,
18
Prem Kumar,
19,20
Paul G. Kwiat,
21,22
Marko Lon
ˇ
car,
14,15,
*
Mikhail D. Lukin,
15,23
David A.B. Miller,
9
Christopher Monroe,
24,25,26
Sae Woo Nam,
27
Prineha Narang,
14,15
Jason S. Orcutt,
28
Michael G. Raymer,
29,30,
Amir H. Safavi-Naeini,
9
Maria Spiropulu,
31
Kartik Srinivasan,
25,32
Shuo Sun,
33
Jelena Vu
ˇ
ckovi
́
c,
9
Edo Waks,
25,34
Ronald Walsworth,
24,34,35,36
Andrew M. Weiner,
3,37
and Zheshen Zhang
10,38
1
Pritzker School of Molecular Engineering, University of Chicago, Chicago, Illinois 60637, USA
2
Department of Electrical Engineering and Computer Science, MIT, Cambridge, Massachusetts 02139, USA
3
School of Electrical and Computer Engineering, Purdue University, West Lafayette, Indiana 47907, USA
4
Department of Physics, Colorado School of Mines, 1500 Illinois Street, Golden, Colorado 80401, USA
5
Photonic & Phononic Microsystems, Sandia National Laboratory, Albuquerque, New Mexico 87185, USA
6
Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA
7
T .J. Watson Laboratory of Applied Physics, California Institute of Technology, Pasadena, California 91125, USA
8
Kavli Nanoscience Institute, California Institute of Technology, Pasadena, California 91125, USA
9
E.L. Ginzton Laboratory, Stanford University, Stanford, California 94305, USA
10
College of Optical Sciences, The University of Arizona, Tucson, Arizona 85721, USA
11
Department of Electrical and Computer Engineering, The University of Arizona, Tucson, Arizona 85721, USA
12
Department of Applied Mathematics, The University of Arizona, Tucson, Arizona 85721, USA
13
Raytheon BBN Technologies, Cambridge, Massachusetts 02138, USA
14
John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts
02138, USA
15
Harvard Quantum Initiative (HQI), Harvard University, Cambridge, Massachusetts 02138, USA
16
Department of Electrical and Computer Engineering, Duke University, Durham, North Carolina 27708, USA
17
IonQ Inc., College Park, Maryland 20740, USA
18
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California 91109, USA
19
Department of Electrical and Computer Engineering, Northwestern University, Evanston, Illinois 60208, USA
20
Department of Physics and Astronomy, Northwestern University, Evanston, Illinois 60208, USA
21
Department of Physics, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, USA
22
IQUIST, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, USA
23
Department of Physics, Harvard University, Cambridge, Massachusetts 02138, USA
24
Department of Physics, University of Maryland, College Park, Maryland 20742, USA
25
Joint Quantum Institute, University of Maryland, College Park, Maryland 20742, USA
26
Joint Center for Quantum Information and Computer Science, University of Maryland, College Park, Maryland
20742, USA
27
National Institute of Standards and Technology, Boulder, Colorado 80305, USA
28
IBM T. J. Watson Research Center, Yorktown Heights, New York 10598, USA
29
Oregon Center for Optical, Molecular, and Quantum Science, University of Oregon, Eugene, Oregon 97403,
USA
30
Department of Physics, University of Oregon, Eugene, Oregon 97403, USA
*
loncar@seas.harvard.edu
raymer@uoregon.edu
Published by the American Physical Society under the terms of the
Creative Commons Attribution 4.0 International
license. Further
distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.
2691-3399/21/2(1)/017002(21)
017002-1
Published by the American Physical Society
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
31
Division of Physics Mathematics and Astronomy, California Institute of Technology, Pasadena, California
91125, USA
32
National Institute of Standards and Technology, Gaithersburg, Maryland 20899, USA
33
JILA and Department of Physics, University of Colorado, Boulder, Colorado 80309, USA
34
Department of Electrical and Computer Engineering, University of Maryland, College Park, Maryland 20742,
USA
35
Quantum Technology Center University of Maryland, College Park, Maryland 20742, USA
36
Harvard - Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, USA
37
Purdue Quantum Science and Engineering Institute, Purdue University, West Lafayette, Indiana 47907, USA
38
Department of Materials Science and Engineering, The University of Arizona, Tucson, Arizona 85721, USA
(Received 18 January 2020; accepted 21 October 2020; published 24 February 2021)
Just as “classical” information technology rests on a foundation built of interconnected information-
processing systems, quantum information technology (QIT) must do the same. A critical component of
such systems is the “interconnect,” a device or process that allows transfer of information between dis-
parate physical media, for example, semiconductor electronics, individual atoms, light pulses in optical
fiber, or microwave fields. While interconnects have been well engineered for decades in the realm of
classical information technology, quantum interconnects (QuICs) present special challenges, as they must
allow the transfer of fragile
quantum states
between different physical parts or degrees of freedom of the
system. The diversity of QIT platforms (superconducting, atomic, solid-state color center, optical, etc.)
that will form a “quantum internet” poses additional challenges. As quantum systems scale to larger size,
the quantum interconnect bottleneck is imminent, and is emerging as a grand challenge for QIT. For these
reasons, it is the position of the community represented by participants of the NSF workshop on “Quantum
Interconnects” that accelerating QuIC research is crucial for sustained development of a national quantum
science and technology program. Given the diversity of QIT platforms, materials used, applications, and
infrastructure required, a convergent research program including partnership between academia, industry,
and national laboratories is required.
DOI:
10.1103/PRXQuantum.2.017002
I. EXECUTIVE SUMMARY
A quantum science and technology revolution is cur-
rently in the making, which is widely expected to bring
a myriad of scientific and societal benefits. Commensu-
rate with this promise, large challenges exist in seeing
the vision become a reality, one of which is the engineer-
ing of an essential class of components of any quantum
information system—quantum interconnects (QuICs).
QuICs are devices or processes that allow the transfer of
quantum states between two specified physical degrees of
freedom (material, electromagnetic, etc.), or, more broadly,
connect a quantum system with a classical one. Figure
1
shows a schematic of a network of diverse quantum infor-
mation systems, central to which is a quantum switch
(QS)—a device that can route optical signals between dif-
ferent channels while maintaining quantum coherence and
entanglement.
An entangled quantum state of two or more quantum
objects or fields describes their joint state (condition) and
thus their statistically correlated measured properties, with
correlations that are stronger than possible according to
classical physics. Entanglement is the essential resource
that enables nearly all quantum technology, but is very
fragile, making it hard to create and maintain over long
times and across large distances.
QuICs are an integral part of nearly all conceivable
quantum information processing systems, including quan-
tum computing, quantum sensing, and quantum communi-
cation. For example, it can be argued that modular quan-
tum computing schemes provide the only viable approach
that will enable scaling up to truly large numbers of error-
corrected qubits. Since modular approaches are crucially
dependent on efficient QuICs, substantial and focused
investment in this vital next stage of quantum comput-
ing is timely. Similarly, the ability to transmit information
securely by leveraging the laws of quantum physics, in a
way that it is “future proof” against even the most powerful
quantum computers, is of great national importance. How-
ever, the reach of secure fiber-based quantum networks,
and the communication rates that they currently allow,
are severely limited by the optical losses in the existing
quantum interconnects (transmission drops exponentially
in conventional optical fibers). Enhancing these intercon-
nects with quantum repeaters will extend the reach and
the rate of quantum communication systems. With recent
017002-2
DEVELOPMENT OF QUANTUM INTERCONNECTS. . .
PRX QUANTUM
2,
017002 (2021)
Switching and
Repeater Network
FIG. 1. The
broad
role
of
QuICs
in
quantum
informa-
tion
technology.
QS, quantum
switch;
QR, quantum
repeater
(a device that can relay an
entangled state from one set of
qubits to a distant set without
physically
sending
an
entan-
gled qubit the entire distance);
QMod, modular
quantum
pro-
cessor; QFC, quantum frequency
converter; RNG, random-number
generator. The QuICs are indi-
cated by bold red arrows or
by wave packets representing
photons.
proof-of-principle demonstrations at hand, this effort is
ready to be accelerated.
Large technical hurdles exist to implementing QuICs:
they must transfer the quantum information (quantum
states) with high fidelity, fast rates, and low loss, often
across a wide range of energies, and do so in a scalable
fashion. In some cases a viable candidate QuIC approach
is well understood, but dedicated engineering effort is
needed to implement it, while in others new physical phe-
nomena need to be explored to implement a given QuIC.
An acceleration of research toward the invention and
implementation of QuICs will also greatly boost progress
in development of materials, devices, systems, and sup-
porting infrastructure in critical-path areas that support
the development of practical quantum technologies. Such
research would enable quantum information science and
technology across a wide range of specialties, with ensuing
scientific and societal benefits as described in Appendix
A
.
This document is a summary of an NSF-sponsored
two-day “QuICs Accelerator Workshop,” which brought
together a representative group of over 30 scientists and
engineers from academia, industry, and national labora-
tories to identify the present roadblocks that need to be
overcome to create functioning QuICs across the necessary
range of QIT platforms. The consensus of the participants
is that there are concepts and technologies whose devel-
opment warrants a large, synergistic, and convergent effort
involving a range of expertise on a national scale.
II. INTRODUCTION
As quantum technology progresses to real-world appli-
cations, a major identified hurdle needs to be overcome:
the development of quantum interconnects (QuICs). Just
as “classical” information technology rests on a foundation
built of interconnected information-processing systems,
quantum information technology (QIT) must do the same.
Quantum interconnects include a wide range of systems
and processes that allow the transfer of quantum states
between two specified physical degrees of freedom (mate-
rial, electromagnetic, etc.). They may also include compo-
nents that connect a quantum system with a system that is
well described by classical physics for purposes of control-
ling or reading out information from the quantum system.
Quantum interconnects present specific challenges, as they
must allow the transfer of fragile
quantum states
between
different physical parts or degrees of freedom of the sys-
tem. With the recent dramatic progress in individual QIT
systems for quantum computation, communication, and
sensing, an urgent need is to push rapidly toward the inte-
gration of such subsystems to create core technologies that
will revolutionize the economy and society in many ways.
See Appendix
A
.
Examples of quantum interconnects include the follow-
ing:
communication channel
(optical, acoustic, microwave,
etc.) between two quantum systems that can be on
the same chip or separated by large distance. Exam-
ples include an optical cavity, waveguide, or fiber
connecting two quantum emitters, or cold microwave
waveguide connecting two superconducting-qubit
processors;
quantum memory
(e.g., color center, trapped ion,
all-photonic cluster-state-based) and the associated
interface to the communication channel;
quantum transducer
used to connect qubits of differ-
ent kinds (acousto-optical, spin-photon, spin-phonon,
etc.), or of the same kind but at different energy
(microwave-optical photon, visible-telecom photon);
017002-3
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
converter
between different qubit-encoding schemes
or degrees of freedom (e.g., polarization, temporal,
spectral encodings of photons);
small scale and application-specific quantum computer,
e.g., quantum repeater, to extend the reach of quantum
communication channels;
entanglement sources
—physical processes that cre-
ate quantum-entangled states of two or more matter-
based or photonic qubits.
As quantum systems scale to larger size, a quantum inter-
connect bottleneck becomes imminent, and surmounting
it is emerging as a central goal for QIT. In the context
of quantum communication networks, a challenging but
extremely important purpose of an interconnect will be to
enable the transfer of quantum information (that is, quan-
tum states) across a distance—long or short, depending
on the application needs. A prime example of a long-
sought-after but elusive subsystem of long-range commu-
nication networks (over distances exceeding hundreds of
kilometers) is the
quantum repeater
, which would relay an
entangled quantum state across a distance that is not acces-
sible using optical fibers only, due to unavoidable signal
losses in the communication channels. At shorter length
scales, modular quantum computing schemes, which are
likely the
only
viable many-qubit near-term approaches,
depend crucially on quantum transducers—devices that
convert variations in a physical quantity, such as spin
state or superconducting flux, into a transmittable signal.
Finally, at the chip-scale level, large numbers of quantum
memories—devices or systems that can maintain a quan-
tum state over long periods of time—implemented, e.g.,
using trapped atoms or spin systems in solid state, need to
be interfaced using integrated, low-loss and fast on-chip
optical networks in order to realize integrated quantum
repeaters.
A consensus in the scientific community is that the
technologies needed for quantum computing and quantum
networking are closely intertwined, indicating that con-
vergent approaches to these challenges will be the most
productive. For these reasons, it is the position of the
community represented by participants of the NSF work-
shop on “Quantum Interconnects” that accelerating QuIC
research is crucial for sustained development of a national
quantum science and technology program.
An important affiliated technology is quantum-enhanced
sensing of a wide range of physical factors: gravitation,
electromagnetism, and environmental factors as well as
biomedical structure and function. For quantum sensors
to reach full capability, in many cases, interfacing them
with quantum memories and processors and distributing
them across space for collective sensing will be required.
Quantum interconnects will play a crucial role in such
distributed sensing applications.
Combination of different elements of QuIC would
enable, for example, links between different processing
regions in a quantum computing system in which data
qubits are stored in memories (based on, e.g., trapped ions),
and transferred into an alternate form (e.g., supercon-
ducting qubits) for fast quantum processing. Such hybrid
systems will benefit from integrated approaches to con-
necting classical systems with quantum systems, e.g., for
delivery of optical signals to trapped ion- and atom-based
quantum computers, clocks, sensors, etc., and to enable
efficient data read out.
Finally, it is important to recognize that many infor-
mation channels, such as an optical fiber or a metallic
stripline, largely act as a conduit that can carry both clas-
sical signals and quantum states of the signaling medium
under appropriate conditions. Thus, classical technolo-
gies and quantum-enabled technologies live in a com-
mon technological ecosystem with large positive feed-
back in both directions. For example, classical telecom
technology has already provided enormous acceleration
of quantum-optical-communication research; at the same
time, the stringent needs of all-optical quantum processors
have driven advances in building on-chip reconfigurable
multimode optical networks, which may benefit classical
approaches to information technology. Thus, the
dual-use
paradigm of technology innovation applies to quantum-
inspired developments.
III. MODULAR QUANTUM PROCESSORS AND
COMPUTERS
Constructing a large-scale quantum processor is chal-
lenging because of the errors and noise that are inherent in
real-world quantum systems, as well as the practical engi-
neering challenges that emerge. One promising approach
to addressing this challenge is to utilize modularity—a
strategy used frequently in nature and engineering to build
complex systems robustly. Such an approach manages
complexity and uncertainty by assembling relatively small,
specialized modules into a larger architecture. Modern
high-performance classical computers and data centers are
constructed by connecting thousands of computers, memo-
ries, and storage units into an interconnected network, over
which complex computational tasks are distributed. These
considerations have motivated the vision of a quantum
modular architecture, in which separate quantum systems
are incorporated into a quantum network via quantum
interconnects [
4
,
6
].
In a modular architecture, the essential building block
is the teleportation-based quantum gate, which uses quan-
tum entanglement to connect different modules and thereby
implement nonlocal quantum operations [
7
10
]. In order
to connect the modules with each other to perform dis-
tributed quantum computation, one has to be able to gen-
erate quantum entanglement between pairs of modules to
017002-4
DEVELOPMENT OF QUANTUM INTERCONNECTS. . .
PRX QUANTUM
2,
017002 (2021)
teleport quantum states or quantum gates. Critical figures
of merit of such intermodule entanglement generation are
(1) the rate of entanglement generation, (2) the fidelity of
the generated entanglement, and (3) the reconfiguration of
the pairs of modules between which the entanglement is
generated.
A. Review of protocols and progress
Several protocols have been proposed and demonstrated
for transporting quantum information between two nodes.
The first is the so-called pitch-and-catch protocol, where
a flying qubit, such as a photon, emitted by a stationary
qubit (or reflected off a cavity holding a qubit) on the
transmitting end, would carry the quantum state over the
communication channel and transfer it to another qubit
on the receiving end [
11
]. Heroic experiments have been
performed using atomic qubits in high-finesse optical cav-
ities demonstrating this process [
12
]. However, the loss in
the photonic channel rapidly degrades the performance of
this scheme, which makes it impractical at optical frequen-
cies. In superconducting circuits, it is possible to create
very strong coupling between the transmitting and receiv-
ing qubits with a microwave photon in a transmission
line connecting the two modules and featuring negligible
loss over the short communication distances involved [
13
16
]. Therefore, such a pitch-and-catch protocol is more
practical in these systems.
The second is a heralded entanglement generation pro-
tocol, where a pair of entangled qubits in the two modules
is first generated probabilistically using photon emission
from the qubits and the detection of emitted photons, then
a deterministic teleportation of the qubit (or quantum gate)
is accomplished using the generated entanglement as a
resource. In this protocol, first the communication qubit on
each module (such as a trapped ion, neutral atom, atomlike
color center in solid-state or quantum dot) emits a pho-
ton in such a way that a degree of freedom of the photon
(such as polarization, frequency, phase or time bin, etc.) is
entangled with the qubit. The emitted photons are collected
(with finite loss), interfere on a 50:50 beam splitter, and are
detected at the outputs. The detection event signals a suc-
cessful generation of entanglement between the two qubits
that emitted the photons. Although the successful execu-
tion of the protocol occurs only probabilistically, success
is heralded (i.e., confirmed) by detection of two photons at
the output of the beam splitters, and reliable entanglement
can be generated at low-to-moderate rates [
9
,
17
].
There have been significant advances in generating
entanglement between different modules with improved
efficiency and fidelity. In trapped-ion systems, the entan-
glement generation rate has significantly improved from
10
3
[
18
] to approximately 200 events per second over
the course of the past 12 years [
19
,
20
], which enables
quantum teleportation between different quantum modules
[
21
,
22
]. The advances come from improving the efficiency
of photon collection from atoms, reducing photon loss
in the channels, and using single-photon detectors with
higher detection efficiencies. Similar protocols have been
demonstrated in neutral atoms [
12
,
23
], nitrogen-vacancy
(NV) color centers in diamond [
24
] and quantum dots
[
25
]. In order to ensure that a modular quantum computer
can be constructed, it is important to have fully functional
quantum computers as the modules, and the entanglement
generation rate (quantum communication rate) between
the modules must be fast compared to the decoherence
rate of the qubits in the modules. Furthermore, efficient
optical interconnects to the modules have to be compat-
ible with quantum computing within the modules. For
instance, optical cavities can provide an optical interface
to atomic quantum computing modules [
26
] but it remains
a challenge to integrate cavities with neutral-atom quantum
computing architectures based on Rydberg interactions
[
27
] or trapped-ion quantum computing architectures [
28
].
Recently, efficient quantum optical interfaces have been
realized using integrated nanophotonic devices for both
trapped neutral atoms [
29
] and diamond color centers [
30
].
In superconducting circuits, the pitch-and-catch pro-
tocol is indeed practical using a microwave photon as
an information carrier. The communication between two
superconducting qubit modules has been demonstrated by
several research groups [
13
16
]. As long as the communi-
cation channel has high quality, it should be possible to
send quantum states, even when the number of thermal
photons in the channel is much larger than one [
31
,
32
].
Therefore, the current demonstrated approaches can be
extended to connecting different dilution fridges using high
quality thermal microwave links.
B. Challenges and research opportunities: modular
quantum processors
Recently, proof-of-principle demonstrations of deter-
ministic teleportation-based quantum gates have been car-
ried out in both superconducting-circuit and trapped-ion
platforms [
33
,
34
]. These demonstrations show a promising
path towards scalable modular quantum computing. How-
ever, finding a technical development path to fully modular
quantum computers interconnected via quantum commu-
nication channels is an extremely challenging task, which
requires substantial advances in basic physical principles,
device (qubit)-level advances, new protocols, integration
of modules and interfaces, and coherent operation across
the modules. Here, we outline some of the research direc-
tions towards the realization of scalable, modular quantum
computers. A timeline and set of milestones for modular
processors is presented in Appendix
B
.
(1) Improving quantum interfaces:
While the
existing quantum interfaces between modules have seen
017002-5
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
dramatic improvements, most systems still have not
reached the regime where connection between the mod-
ules can be utilized for reliable transfer of qubits within
the timescale required for distributed quantum computa-
tion. For the heralded scheme, we have to continue to
improve the entanglement generation rate so that it is
comparable to the local entangling gate operation rate
within a module. While this is not a strict requirement
for efficient quantum computation, it means that the cost
of distributing a quantum task across the modules would
not substantially constrain the execution of the computa-
tional task. Another topic worth noting is that all quantum
interconnects are not perfect in terms of the fidelity of
the distributed entanglement, or the success probability
of the pitch-and-catch scheme. The errors in the quan-
tum interconnects must be minimized or corrected, so that
the distributed quantum computation can succeed with
viable probabilities, i.e., so that distributing the computa-
tion actually improves performance rather than degrading
it. New protocols and implementation strategies to over-
come the errors in the communication channel need to be
developed.
(2) Integration of modules and interfaces:
Seamless
integration of the communication interfaces with the com-
putational functions of the modules can introduce some
challenges. For example, in heralded entanglement gen-
eration protocols, the qubit-photon entanglement genera-
tion protocols can lead to decoherence of nearby qubits
storing information. For these systems, novel integration
approaches must be developed so that the communica-
tion and local data processing can coexist. For solid-state
qubits (such as superconducting qubits) that use photons
in the microwave range of the electromagnetic spectrum,
communication over room-temperature channels becomes
impractical. In order to take advantage of modules real-
ized outside the cryogenic environment, frequency up-
conversion of the photonic qubit to the optical spectrum
is necessary. Quantum transduction techniques to reli-
ably convert microwave photons to optical photons are an
important area of research for these applications.
(3) Hybrid modular architectures and intercon-
nects:
The need for modularity can also be driven by the
computational functions, where various qubit technologies
provide opportunities for executing tasks with different
performance requirements. For instance, memory modules
that contain qubits with very long coherence times could
be implemented on a different platform than processing
modules where fast gate times are essential. This potential
tremendous advantage comes with additional challenges.
In order to take full advantage of such a hybrid modular
architecture it is important to develop interconnects capa-
ble of distributing entanglement between different qubit
implementations, for example, superconducting currents
or charges, color centers, neutral atoms, ions, or pho-
tons. The spectral characteristics of the photons that couple
to each of these systems—including the wavelength and
bandwidth—can be very different, by several orders of
magnitude, leading to extremely inefficient interspecies
conversion in the absence of suitable quantum transducers.
(4) Coherent operation of modular quantum com-
puter and distributed algorithms
: Even if local quantum
computer modules and the needed quantum interconnects
are adequately integrated, distributed quantum computa-
tion will require operating every module in the system
with full quantum coherence among them. This poses chal-
lenges in designing and operating phase-coherent control
systems across the modules, as well as tracking the quan-
tum phase of every module in the system. Algorithm-level
strategies for efficiently distributing the computational task
over the modular quantum computer based on the per-
formance specifications of various functional components
that constitute the system is therefore an important area of
research.
IV. QUANTUM INTERNET
The quantum internet describes a collection of dis-
tributed quantum nodes, separated by a range of distances
over which one desires to perform some quantum com-
munication protocol that can support, for example, dis-
tributed quantum computation (Sec.
III A
) or distributed
sensing (Sec.
III C
). For an accessible overview, see
Ref. [
35
]. There are now numerous quantum communica-
tion and cryptographic protocols identified, including secu-
rity distribution for encryption [
36
43
], quantum-certified
random-number generation in the form of random-number
beacons and personal devices, secret sharing [
44
,
45
],
quantum fingerprinting [
46
48
], and other multiparty com-
putation protocols, such as secure quantum voting, byzan-
tine agreements, and multiparty private auctions [
49
]. Of
particular relevance is the possibility of “blind” quantum
computation [
50
,
51
], whereby a remote user can program
a quantum computer without revealing to its owner the
algorithm that is run or the computational result, and dis-
tributed quantum processing, whereby two or more quan-
tum computers share entanglement to enable them to act as
a single larger processor. Because of the distances involved
(0.1–1000 km), optical photons must be used.
Another key aspect of a fully functioning quantum
internet is the potential for unconditional information secu-
rity—a feature of using quantum information that is not
possible with classical information processing. A further
benefit of using quantum-secured information will be that
the lifetime of the security is “infinite”; it will be secure
against any advances in computation capability that may
occur in the future. There have been many cryptographic
tasks in which quantum-secured versions have been con-
ceived. For all of these tasks, quantum interconnects are
required because of the need to preserve entangled quan-
tum states.
017002-6
DEVELOPMENT OF QUANTUM INTERCONNECTS. . .
PRX QUANTUM
2,
017002 (2021)
To realize fully the potential of a quantum internet,
significant convergent work is still needed to improve
the physical hardware. Theoretical work is also required
to develop efficient information-processing techniques to
preserve the quantum information and determine the most
robust and secure network connectivity. The development
of quantum-secured devices and protocols could transform
the cryptographic landscape.
There are two primary channels over which to transmit
the photons: optical fiber and free space. Each of these
has challenges and opportunities. The former can lever-
age the enormous existing network of telecommunication
fibers, though then the photons need to be in the telecom-
munications band to avoid excessive losses. Even still,
the transmission through such a fiber will drop exponen-
tially with length, with a typical loss in telecom fiber of
0.2 dB/km. Thus direct transmission of quantum states
becomes highly inefficient beyond about 100 km, where
20 dB loss implies only 1% transmission, so that direct
transmission of quantum states becomes highly inefficient
beyond about 100 km. Free-space optical communication
is far less well developed, but has the advantage that it can
operate over a much larger range of wavelengths, and the
losses (due to diffraction) grow only quadratically. Typi-
cally, greater care is needed to reduce background light in
free-space quantum communication channels; also, there is
typically the added challenge of stabilizing the free-space
coupling using pointing and tracking methods, and pos-
sibly adaptive optics to reduce the effects of turbulence.
Nevertheless, many of these challenges have been over-
come in a series of free-space quantum communication
demonstrations, between mountains [
52
,
53
], over water
[
54
] within cities [
55
57
], from airplanes [
58
], balloons,
and drones [
59
] and even using satellites in low-earth orbit
[
60
,
61
]. While the achieved transmission rates in these
experiments might have greatly exceeded what would
have been possible using fiber channels—in one case by
nearly 20 orders of magnitude [
60
,
61
], they are still often
very low, and methods such as multiplexing or employing
higher-dimensional states (see below) may be needed to
achieve practical rates.
A. Challenges and research opportunities
To build a fiber-based global network capable of dis-
tributing quantum entanglement, there are two main chal-
lenges that have to be overcome. First, optical attenuation
during fiber transmission leads to an exponential decrease
in the entangled-pair distribution
rate
. Second, operational
errors such as channel errors, gate errors, measurement
errors, and qubit memory errors can severely degrade
the
quality
of the distributed entanglement, which at best
reduces the quantum advantage and at worst completely
eliminates it, e.g., a quantum cryptographic key may be
completely insecure! A timeline and set of milestones for
a quantum internet is presented in Appendix
C
.
(1) Quantum repeaters:
To overcome these chal-
lenges and extend the range of fiber-based entanglement
distribution beyond a few hundred kilometers, quantum
repeaters (QRs) are required, but are not yet available.
Depending on the tools used for suppressing these imper-
fections, the quantum information community has iden-
tified the following three generations of QRs: The first
generation of QRs [
62
,
63
] uses heralded entanglement
generation and heralded entanglement purification, which
can tolerate more errors but requires two-way classical
signaling over the entire chain of QRs; such signaling
then implies that the requisite quantum memory lifetimes
and coherence times must be substantially longer than the
round-trip communication times. The second generation
of QRs introduces quantum encoding and classical error
correction to replace the entanglement purification with
classical error correction, handling all operational errors
[
64
,
65
], which is more demanding in physical resources
but requires only two-way classical signaling between
neighboring repeater stations, and consequently further
improves the quantum communication rate. The third gen-
eration of QRs would use quantum encoding to determin-
istically correct both photon losses and operation errors
[
66
,
67
]. By entirely eliminating two-way classical signal-
ing, the third generation of QRs would promise extremely
high entanglement distribution rates that can be close to
classical communication rates, limited only by the speed
of local operations, in turn limited by, e.g., photon source
rates, detector saturation rates, and timing jitter, etc.
One important benchmark for QRs is the repeaterless
bound [
68
,
69
], which imposes the fundamental limit of
the direct quantum communication protocols. Recently,
there have been significant advances in experimentally
demonstrating key elements of a QR in an integrated sys-
tem. An important recent highlight is the experimental
demonstration of memory-enhanced quantum communica-
tion surpassing a repeaterless bound in a proof-of-concept
laboratory setting, using a solid-state spin memory associ-
ated with silicon-vacancy (SiV) color center integrated in
a diamond nanophotonic resonator [
30
,
70
]. This paves the
way towards the demonstration of a full quantum repeater,
which in turn will enable scalable large-scale quantum
networks.
(2) Quantum memories:
The major challenge for the
first generation of quantum repeaters is the development of
long-lived quantum memories with efficient optical inter-
faces, such as addressable color-center nuclear spins with
integrated nanophotonics [
71
], trapped-atomic qubits with
Purcell-enhanced emission [
12
,
19
,
23
], or superconducting
circuits with microwave-to-optical transduction. In addi-
tion, the availability of efficient photon detectors with low
017002-7
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
dark counts is crucial, with significant advances needed in
reducing the cost, integration, etc.
(3) Spectral-temporal encoding:
It is now generally
recognized that practical rates of entanglement distribution
can likely be achieved only by employing high levels
of channel multiplexing (e.g., spectral, temporal, spatial)
to enhance success probabilities; for instance, quantum
signals are simultaneously sent at multiple nearby wave-
lengths or in multiple time bins. Although each spectral or
temporal channel has some probability of failure or loss,
the likelihood that all would be unsuccessful decreases
with the number of multiplexed channels. However, one
needs a mechanism to demultiplex into a single spectral-
temporal mode; alternatively, they are each coupled to their
own quantum-memory qubit, but then some mechanism
for identifying and coupling a particular pair of success-
fully “loaded” quantum registers is needed. The use of
such temporal multiplexing has recently enabled a 30
×
enhancement in the success rate of a two-photon quan-
tum communication protocol [
72
]; the advantages become
exponentially larger for protocols requiring higher num-
bers of qubits. The benefits of multiplexing arise only if
the quantum interconnects that implement the multiplexing
and demultiplexing have high fidelity and low loss.
Another emerging strategy is to use qudits, the higher-
dimensional counterparts to qubits, e.g., using three time
bins to encode numbers 0, 1, and 2, and arbitrary superpo-
sitions thereof. Just as it does for classical communication,
such encoding increases the information-carrying capac-
ity of a photon by log(
d
), where
d
is the dimensionality, at
the expense of more complex measurements and manipula-
tions. Recent research shows that such higher-dimensional
encoding can enable more efficient quantum error correc-
tion [
73
]. Finally, encoding multiple qubits (or even their
higher-dimensional counterparts, qudits) onto a single pho-
ton can yield intrinsic robustness to loss: because all of
them are guaranteed to be lost or transmitted together,
the net success probability can be greatly enhanced. For
instance, the probability that a channel with 99% loss
will successfully transmit a three-photon three-qubit state
is only 10
6
; in comparison, a single-photon three-qubit
state experiences the loss only once, i.e., with a 1% suc-
cess probability. The concept of qubit entanglement also
generalizes to hybrid entanglement [
74
], between different
degrees of freedom of a single photon, e.g., polarization
and spatial mode, and hyperentanglement [
75
], between
multiple corresponding degrees of freedom of two pho-
tons, e.g., polarization and time bin [
71
], or time bin
and frequency bin [
76
,
77
]. One critical need is a method
to transduce such higher-dimensional quantum states into
qubit memories and qubit-based quantum processors [
78
]
(4) Efficient measurements:
Finally, all three gen-
erations of QRs can be greatly enhanced by including
efficient quantum nondemolition (QND) measurements
[
79
]—a measurement that records the successful passing
of a photon without observing it or changing its quan-
tum state. In this way, any memory can be converted
to a
heralded
quantum memory, which enables one to
know whether a photon has successfully been transmitted
down the entire length of a communication channel; such
knowledge greatly reduces the required number of quan-
tum memories, since one is only needed in cases where
the quantum signal is successfully transmitted through the
optical channel.
With the emerging demonstrations of quantum
repeaters, it will be important to optimize them to over-
come realistic imperfections through use of robust archi-
tecture and encoding. It is also urgently needed to develop
novel quantum-network applications and appropriate cor-
responding performance metrics, such as entanglement
fidelity, throughput, latency, resource overhead, etc. These
performance metrics should also guide the device design
and fundamental investigation of relevant physical plat-
forms.
V. QUANTUM-ENHANCED SENSORS
Quantum sensing technology has made significant
progress over the last few decades and has given rise
to atomic clocks [
80
], magnetometers [
81
], and iner-
tial sensors [
82
] that operate at the standard quantum
limit (SQL), where the relative uncertainty in measur-
ing some parameter scales as 1
/
N
, where
N
is the
number of copies of the system, e.g., the number of pho-
tons being detected; at the SQL systematic errors have
been suppressed to the extent that the measurement is
dominated by quantum-mechanical uncertainty. With the
tremendous advances in the theoretical and experimen-
tal aspects of quantum information science over the last
decade, new quantum resources, such as quantum mem-
ories and entangled particles, can now be harnessed to
enhance further the performance of quantum sensors. Also
known as quantum metrology, quantum-enhanced sensing
is aimed at taking advantage of these emerging quantum
resources to outperform the SQL and achieve unprece-
dented sensing performance. As a remarkable instance
of quantum-enhanced sensing, the Laser Interferometer
Gravitational-Wave Observatory (LIGO) utilizes nonclas-
sical squeezed light to enable a measurement sensitivity
below the SQL [
83
]. Quantum-enhanced sensing has also
been proven to be a powerful paradigm for a variety of
scenarios, including magnetic sensing with quantum mem-
ories [
84
], quantum-illumination target detection [
85
],
sub-SQL atomic clocks [
86
], and nanomechanical sensors
[
87
].
Most existing quantum-enhanced sensing demonstra-
tions leverage nonclassical resources to improve the mea-
surement performance at a single sensor, but many real-
world applications rest upon a network of sensors that
work collectively to undertake measurement tasks. Notable
017002-8
DEVELOPMENT OF QUANTUM INTERCONNECTS. . .
PRX QUANTUM
2,
017002 (2021)
examples for such a setting include wireless sensor net-
works [
88
], phased arrays [
89
], and long-baseline tele-
scopes [
90
]. In this regard, the quantum internet presents
unique opportunities for quantum sensors to utilize shared
entanglement to boost the performance in networked sens-
ing tasks. The following section discusses the concept,
promising research avenues, and application space for
interconnected quantum sensors.
A. Interconnected quantum sensors
Extensive studies have been dedicated to using bipar-
tite (two-party) entanglement as a resource to overcome the
SQL at a single sensor. In one step forward, recent theoret-
ical works on quantum-enhanced sensing based on multi-
partite entanglement show that interconnecting distributed
quantum sensors to form an entangled sensor network can
probe global parameters at the Heisenberg limit, i.e., at
an estimate uncertainty that scales favorably compared to
the scaling for a network of independent sensors. Specifi-
cally, Ref. [
91
] proposed a quantum network of clocks that
enjoys boosted precision and security over conventional
classical clock networks. More generally, two theoreti-
cal frameworks for distributed quantum sensing based
on, respectively, discrete-variable [
92
,
93
] and continuous-
variable multipartite entanglement have been formulated
[
94
]. On the experimental front, a proof-of-concept dis-
tributed quantum sensing experiment demonstrated the
utility of multipartite continuous-variable entanglement
for enhancing the measurement sensitivity for estimat-
ing global phase shifts. To demonstrate the prospect for
interconnected quantum sensors in real-world applications,
Ref. [
95
] reported an entangled rf-photonic sensor network
in which distributed rf sensors harness their shared multi-
partite entanglement to enhance the precision of estimating
the properties, e.g., the angle of arrival, of an incident rf
wave across all sensor nodes.
In the context of a quantum internet, quantum sensors
distributed over a distance will be able to establish high-
fidelity entanglement to achieve measurement sensitivities
beyond the SQL. Potential application scenarios for large-
scale entangled quantum sensor networks would encom-
pass high-precision astronomical observation [
90
,
96
] envi-
ronmental and health monitoring, positioning, navigation,
and timing. Two possible means of building up entangle-
ment shared by quantum sensors are the following: (1) A
matter-based quantum sensor first entangles with a pho-
tonic mode, which is then transmitted through the quan-
tum internet equipped with quantum repeaters to ensure
high-rate long-distance entanglement distribution. Entan-
gling photonic quantum measurements are performed at
the destination quantum repeater nodes to establish mul-
tipartite entanglement between matter-based quantum sen-
sors. (2) As an alternative method to form an entangled
quantum sensor network, photonic multipartite entangle-
ment tailored for a specific networked sensing task is first
produced by a photonic quantum chip at a central node.
Each arm of the photonic entangled state is then transmit-
ted to a quantum sensor located in the quantum internet. As
in the matter-based quantum sensor network, the quantum
internet takes advantage of quantum repeaters to compen-
sate for entanglement distribution loss so that high-fidelity
photonic multipartite entanglement is maintained. At each
quantum sensor node, a high-efficiency low-noise quantum
transducer converts the information carried by the object of
interest into the photonic domain so that quantum measure-
ments on the photonic multipartite entanglement unveil the
global property of the interrogated object.
B. Challenges and research opportunities
Apart from the need for a quantum internet as a back-
bone, a number of technical accelerations will be critical
for the construction of entangled quantum sensor net-
works. A timeline and set of milestones for quantum
sensors is presented in Appendix
D
.
(1) Device concepts:
Matter-based quantum sensor
networks are composed of any of a diverse range of useful
sensors. Examples, not exhaustive, include sensors of mas-
sive particles, photons, magnetic fields, electric fields, tem-
perature, gravity, pressure, and chemical processes. Such
sensor networks require efficient light-matter interfaces
or interconnects to create entanglement between quantum
sensors and photonic modes. In an ideal situation, estab-
lishing entanglement between multiple matter-based quan-
tum sensors at the quantum repeaters calls for deterministic
multipartite Bell measurements with near-unity efficiency.
Such a measurement can be realized by first transferring
the quantum states of photons into those of solid-state
qubits, followed by fault-tolerant quantum computation on
a special-purpose small-scale quantum computer. As a nec-
essary ingredient, the outcomes of the Bell measurements
need to be communicated to different quantum sensor
nodes in real time, by fast electronic processing and a
low-latency classical communication network. Since most
matter-based quantum sensors operate with read-out in the
visible to the near-infrared spectral range, high-efficiency
low-loss quantum frequency converters are required to
shift the wavelengths of photons into the telecommunica-
tion window for long-haul communication via a quantum
internet.
The device requirement for the photonic quantum
sensor network encompasses envisaged programmable
photonic quantum chips (PQCs) to generate appropriate
photonic multipartite entangled states. Each PQC would
entail low-loss waveguides and couplers, high
Q
-factor
ring resonators, and single quantum emitters that provide
needed (“non-Gaussian”) resources for universal quantum
017002-9
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
information processing. The produced photonic multipar-
tite entangled states need to be inserted into optical fibers
through couplers with near-unity transmissivity. The PQCs
should also provide classical controls to precompensate the
dispersion and other imperfections incurred in the trans-
mission. At each quantum sensor node, high-efficiency
quantum transducers convert the physical information con-
tained in the microwave, mechanical, or magnetic domains
into modulations on the visible photonic quantum states.
To ensure high performance for the quantum sensor net-
work, it is important for the quantum transducers to
achieve high efficiency while minimizing additional loss
and noise.
To extract information carried by the photons, high-
efficiency quantum-limited homodyne, heterodyne, or
direct measurement detectors are subsequently utilized.
These detectors ideally will be integrated on the same
PQC as are the quantum transducers to obviate addi-
tional coupling and conversion losses. Recent advances
in the fabrication and integration of quantum devices
based on widely used optoelectronic materials such as
lithium niobate [
97
] point to a promising platform for inter-
connected quantum sensors. Further technology accelera-
tions would lead to a versatile photonic quantum sensing
platform capable of accommodating hundreds to thou-
sands of elements with different functionalities on the
same PQC.
(2) Sensor network architectures:
architectural per-
spective
,
the engineering of multipartite entangled states
for a large-scale quantum sensor network consisting of a
large number of sensors remains an open problem, due
to the complexity of multipartite entanglement. In this
regard, machine-learning tools would be useful for identi-
fying near-optimum entangled states for networked sens-
ing problems. A recent theoretical study shows that the
optimum entangled state and measurement configuration
can be found by training photonic quantum circuits by a
support-vector machine and a principal component ana-
lyzer, for data classification and data compression tasks
at a physical layer [
98
]. Further investigations would
incorporate the machine-learning framework into quan-
tum devices and the quantum internet under develop-
ment to accelerate the performance, scale, and application
scope.
VI. CONVERGENT ACCELERATION
OPPORTUNITIES
An acceleration of technical research toward invention
and implementation of quantum interconnects will greatly
boost progress in quantum information science and tech-
nology across a wide range of specialties. Particularly
important needs and goals are summarized here.
VII. DEVICES AND SYSTEMS
Future quantum computers and networks will require
unprecedented connectivity over distances ranging from
micrometers to hundreds or even thousands of kilometers.
Such diverse connectivity will put significant demand on
quantum interconnects, requiring, for example, the scal-
able fabrication and integration of a large number of
components in compact optoelectronic chips.
In particular, integrated (on-chip) device technologies
are likely to play a number of important roles in imple-
menting quantum interconnects. The general ability to
enhance interactions through control of the electromag-
netic density of states in suitably engineered geometries
enables a wide variety of physical resources to be real-
ized, while the manufacturing technologies used to create
such geometries can be predicted to reach the level of
scaling and integration required. Here we outline the dif-
ferent QuIC technologies needed, and then discuss specific
device and material platforms in which these technologies
should be developed.
Quantum interconnects will provide the links between
various quantum devices to realize large-scale systems. We
organize interconnects into three primary categories:
(1)
Interconnects between bosonic and atomic systems
:
The role of such interconnects is to interface bosonic
fields with atoms for a variety of applications. Here
“bosonic fields” include a variety of harmonic-
oscillator systems that are suitable to carry quantum
information across distance, such as optical photons,
mm waves, microwaves, and acoustic phonons.
“Atomic” systems should be broadly interpreted to
encompass all forms of matter systems including
neutral cold atoms, trapped ions, Rydberg-excited
atoms, cold molecules, quantum dots, color centers,
impurity bound excitons, superconducting Joseph-
son devices, etc. Examples include transfer of quan-
tum information from atomic memories to photons
for quantum networks, read-out of phase informa-
tion in quantum sensors, transfer of microwave exci-
tations from transmon qubits to microwave cavities,
etc.
(2)
Interconnects between two bosonic photonic plat-
forms
: The role of these interconnects is to
impedance match two bosonic platforms or
information-encoding schemes in order to achieve
interconversion or to combine various quantum sys-
tems. Examples include quantum frequency conver-
sion of optical and microwave signal to the telecom,
temporal-waveform conversion that enables opti-
mal interfacing of heterogeneous quantum nodes,
quantum transduction between optical photons and
acoustic phonons, and connections between pho-
tonic systems that use different encodings (e.g.,
017002-10
DEVELOPMENT OF QUANTUM INTERCONNECTS. . .
PRX QUANTUM
2,
017002 (2021)
time-bin, frequency-bin, wave-packet shape, coher-
ent or squeezed-state, or polarization-state).
(3)
Interconnects between two atomic platforms
: Here,
quantum interconnects mediate long-range interac-
tions between atomic systems. Examples include
entanglement of quantum memories separated by
large distances, hybrid quantum systems com-
posed of different matter qubits [
99
], microwave
interconnection of transmon qubits in superconduct-
ing devices, development of low-loss switches and
architectures to connect arrays of quantum systems
to each other in a scalable fashion, etc.
Below we describe the various requirements for these
different QUIC applications.
A. Atomic-photonic interconnects
Quantum information relies on a broad array of matter-
based quantum systems to store and manipulate quan-
tum coherence. Such matter systems include single atoms
[
27
], quantum dots [
100
], color centers [
24
,
101
,
102
], rare-
earth ions [
103
108
], defect-bound excitons, and super-
conducting Josephson devices [
109
], to list a few. These
systems provide a variety of essential functionalities in
quantum information that include single-photon sources
[
110
], quantum memories [
103
,
111
], quantum sensors, and
photon storage devices.
An essential role of quantum interconnects is to inter-
face these matter systems with optical photons, the unique
carriers of quantum signals across long distances. Intercon-
nects used to form shorter-distance networks, e.g., inside
a room, or on a chip, may use lower-frequency pho-
tons (e.g., mm-wave [
112
] or microwave [
113
]) or even
acoustic phonons [
114
] to accomplish the analogous con-
nectivity function. The quantum interconnect must provide
strong interactions in order to mediate quantum state trans-
fer, entanglement, or other uniquely quantum resources.
Typically, such functionality entails using optical cavities
or waveguides to enhance light-matter interactions into
the single-photon regime. These photonic structures must
support high quality factors and small mode volumes to
attain the desired interaction strengths. Furthermore, these
devices often have to operate at short wavelengths (vis-
ible and near-IR), which puts additional constraints on
the materials used. Another important requirement of such
interconnects is low insertion loss. Low-loss operation is
particularly important in quantum applications where the
loss of a single photon can destroy the quantum state of
the entire system. Finally, there is the issue of compatibil-
ity of the interconnect hardware with the atomic systems.
Many atomic systems operate in millikelvin cryogenic
environments with miniscule acceptable heat loads, and
extreme sensitivity to quasiparticles generated by optical
absorption. Scalable atomic-photonic interconnects must
be able to scale within these constraints.
B. Photonic-photonic interconnects
The future quantum internet will likely be composed of
a broad range of disparate systems that must interact and
exchange quantum signals. The most viable candidate for
interacting different quantum systems across a distance is
via photonic channels. But these diverse quantum systems
emit photons with diverse frequencies and temporal shapes
and durations, necessitating quantum interconnects that
couple photons with different properties. Important exam-
ples include quantum frequency conversion to telecom,
microwave-to-optical conversion, and spectral bandwidth
conversion.
The choice of interconnects largely depends on the
distance scale over which such quantum interconnects
operate. Microwave photons may be suitable for intra-
chip and interchip links within a mK environment where
thermal background is suppressed, but are unlikely to
be used for connections over much longer length scales.
Telecommunication-band photons remain the information
carrier of choice for long-distance networks based on opti-
cal fiber, while long-distance free-space links and shorter
metro-area networks may be amenable to optical photons
in different frequency bands. While most visible photon
frequencies will be suitable for relatively short distance
networks (e.g., to connect nodes within a distributed quan-
tum computer), there is also the potential to work with
millimeter-wave and terahertz frequency photons for suffi-
ciently short links.
Quantum frequency conversion (QFC) devices, e.g.,
based on three- or four-wave nonlinear optical mixing
[
115
118
] or direct frequency shifting using electro-optic
modulation [
119
], enable the spectral translation of a quan-
tum state of light to targeted frequencies with high effi-
ciency, low added noise, and sufficient bandwidth. QFC
devices are needed to enable quantum interconnects that
link the most suitable matter qubits for a given application
across the relevant length scales and photonic communi-
cation medium [
120
]. For networks consisting of homo-
geneous nodes, these QFC devices may primarily consist
of down-conversion and up-conversion units that are ide-
ally seamlessly integrated with the photonic qubits that are
directly coupled to the matter qubits. For networks con-
sisting of heterogeneous nodes, it is likely that QFC needs
to be combined with coherent temporal-spectral wave-
form manipulation, to ensure optimal coupling to matter
qubits that may have significantly different acceptance
bandwidths and lineshapes.
Another approach to linking matter qubits over distance
is through the use of intermediate entangled-photon-pair
sources that are engineered to create one photon at a fre-
quency suitable for direct interaction with the matter qubit
(e.g., at 637 nm for a NV
center in diamond), and another
at the relevant frequency for transmission across the physi-
cal interconnect channel (e.g., 1550 nm for a long-distance
fiber link). Through entanglement swapping, such sources
017002-11
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
can be used to entangle distant quantum nodes, though if
those nodes are heterogeneous, it is likely that some form
of waveform reshaping will also be needed, e.g., using
“time-lensing” methods [
121
].
Multiplexing and demultiplexing of optical pulses based
on their temporal mode identity (wave-packet shape) can
also play an important role [
122
]. As mentioned above, by
providing a high-dimensional state space for single-photon
packets, temporal modes offer higher information content
per photon. The ability to demultiplex light into temporal-
mode components offers increased signal-to-noise ratio
in photon-starved communication links, such as can be
envisioned in deep-space communication [
123
].
C. Atomic-atomic interconnects
Given that near-future quantum computers and networks
will likely be composed of modular units of a few tens
to hundreds of matter qubits (atoms, solid-state defects,
superconducting devices, photonic, etc.) that can perform
small-scale quantum information tasks, scaling to larger
systems will require interconnection of multiple modu-
lar components. Photonics provides an ideal approach to
achieve this interconnectivity. For example, the current
state of the art in fully controlling trapped-ion qubits
involves of the order of 30 ions; scaling up to hundreds
of ions will require using light for interconnecting sepa-
rate modules each containing around this number of ions.
Industrial quantum computers have recently reached 53
qubits [
124
], but scaling up to hundreds to thousands of
qubits on a single chip is presently well out of reach.
A scalable optical interconnect for atomic modular
nodes requires the ability to route photons efficiently
and provide high-fidelity multiphoton interference. This
multiphoton interference provides the necessary quantum
step to generate effective atom-atom interactions at a
distance. Interconnects should ideally combine photonics
with detectors and other components to provide a com-
plete on-chip solution for large-scale modular quantum
information processing. Additional functionality such as
filtering, on-chip quantum frequency conversion, and mul-
tiplexing will significantly enhance the scalability of the
total system. In many other circumstances where the atoms
possess large material-strain susceptibility [
125
], acoustic
phonons might be particularly suitable to mediate inter-
actions between two or multiple atomic qubits on a chip
[
126
].
D. Integrated quantum photonics platforms
1. Atomic-photonic interconnects
Several different photonic platforms naturally host
atomiclike systems, including diamond (color centers)
[
84
], GaAs and InP (quantum dots), and SiC (color cen-
ters) [
110
,
127
] and various transparent crystals doped with
rare-earth ions [
103
,
106
] or transition metals. Integrated
devices are important for realizing the crucial local inter-
connect between a matter qubit and a photonic qubit. For
solid-state qubits, this often involves engineering of suit-
able photonic cavities or waveguides to ensure that, for
example, emitted photons entangled with spins are effi-
ciently funneled into a single desired collection channel.
It is critically important that the fabrication processes that
create such photonic structures, which sometimes have
features at the 100-nm-length scale, do not induce excess
dephasing or spectral diffusion that will limit the coherence
properties of the matter qubit. Such issues are increas-
ingly being addressed through design constraints on the
separation of the matter qubit from an etched surface,
surface-passivation techniques, and electrical-charge stabi-
lization methods. Hybrid integration of materials that host
quantum emitters with materials that support scalable fab-
rication of photonic devices [
128
] may also be beneficial
to the atomic-photonic interconnects.
For photonic chip integration of trapped neutral atoms,
ions, and cold molecules, protecting qubit coherence in
a platform with a wide enough transparency window to
accommodate the short wavelength photons associated
with these systems is paramount. For trapped ions, this
may require new electromagnetic designs that limit delete-
rious effects on the electrostatic traps; similar approaches
may be needed for systems like Rydberg atoms, which are
unlikely to be brought close (within the evanescent tail)
to a guided-wave photonic structure. Other atomic sys-
tems, in particular single neutral atoms and cold molecules,
can be loaded and trapped within the evanescent field of
photonic devices, enabling an efficient matter-photon qubit
interconnect. Recent progress with optical tweezer traps in
Rydberg chains and small collections of neutral atoms and
molecules offer a key challenge for quantum interconnects.
2. Photonic-photonic interconnects
Several different nonlinear nanophotonic platforms are
being developed within the QuIC community to enable the
QFC and entangled photon-pair interconnect approaches
needed to realize photonic interconnects across different
frequency bands. A table listing the variety of integrated
quantum photonics platforms is presented in Appendix
E
.
In practice, the most mature technology is based on
cm-scale quasi-phase-matched nonlinear media [
129
]such
as periodically poled lithium niobate (PPLN) waveguides,
where internal conversion efficiencies approach 100% and
signal-to-background levels in excess of 100:1 are achiev-
able for single-photon-level inputs. In its current state, this
waveguide technology is not directly amenable to dense
integration within photonic circuits (the optical mode field
diameter and bend radius are similar to those of an opti-
cal fiber), nor is it directly amenable to direct integration
with single quantum emitters or a variety of other inte-
grated photonics technologies. Its further development
017002-12
DEVELOPMENT OF QUANTUM INTERCONNECTS. . .
PRX QUANTUM
2,
017002 (2021)
is important from the perspective of providing hardware
for near-term networking efforts and for implementing
strategies that realize waveform conversion via nonlinear
optics.
Integrated nanophotonic platforms based upon geome-
tries with high refractive-index contrast (typically cre-
ated using thin-film nonlinear optical materials on a low
refractive-index substrate such as SiO
2
) can enable QFC
and waveform-conversion approaches using manufactur-
ing approaches that can be highly scalable and enable
complex integration with other photonic circuit function-
alities, including beam splitters and filters. For second-
order nonlinear processes, thin-film lithium niobate on
insulator (LNOI) [
130
,
131
] and AlN [
132
] have shown
particular promise, given their wide optical transparency
window, very low optical losses [
133
,
134
], appreciable
nonlinear coefficients, and amenability to nanofabrica-
tion processes. For third-order nonlinear processes, silicon
nitride has risen to the forefront of many related classi-
cal applications (e.g., compact frequency comb genera-
tion), and QFC of quantum dot single photons and photon
pairs using silicon-nitride nonlinear resonators has recently
been demonstrated [
118
,
135
]. III-V semiconductors such
as GaP [
136
] and AlGaAs and wide-bandgap materials
such as SiC [
102
] and diamond also possess strong opti-
cal nonlinearities [
137
], but have not yet achieved the
level of performance of the aforementioned systems. How-
ever, their ability to directly host matter qubits (e.g., color
center or quantum dot spins) is of significant benefit to
integration.
For all of these platforms, the basic strategies for
achieving efficient frequency conversion are generally
conceptually well-understood, and fabrication techniques
are developed. However, there are still many challenges
in realizing connections between the ultrawide frequency
separations [
138
] (e.g., UV telecom), and understanding
the relevant noise-generation mechanisms (impurity-based
fluorescence, Raman scattering, spontaneous parametric
processes, to name a few) is an ongoing process [
139
,
140
].
Materials supporting a second-order nonlinearity often
exhibit an appreciable electro-optic effect [
97
], which
enables fast (tens of picoseconds) reconfigurable switch-
ing operations. For scenarios in which slower speeds are
adequate (e.g., MHz switching bandwidths), thermo-optic
and microelectromechanical switches can be considered,
though the former do not function well at cryogenic tem-
peratures.
Microwave-to-optical QFC typically requires access to
physical processes distinct from those described above,
with the exception of electro-optic platforms [
141
] that do
provide natural links between the two frequency bands,
though the extent to which such links can be sufficiently
low noise in practice is not known. Piezoelectric media
[
114
,
142
] such as LNOI, AlN, GaP, and GaAs are being
considered for microwave-to-optical QFC mediated by
nanomechanics, while modular approaches based on free-
space cavities [
143
] coupled to electromechanical systems
have shown the best performance thus far in terms of
efficiency, albeit over moderate bandwidths and without
adequately low noise. Another way to mediate the interac-
tion between optical and microwave photons is via atomic
ensembles simultaneously coupled with high cooperativity
to photonic and microwave resonators. In this context,
rare-earth-doped media are particularly well suited as
microwaves can be coupled to either Zeeman or hyperfine
transitions in ensembles exhibiting very narrow inhomo-
geneous lines [
144
,
145
]. Unlike QFC between optical
wavelengths, there has to this point been no full demon-
stration of QFC between microwave and optical wave-
lengths where, for example, nonclassical photon statistics
or quantum interference are shown to be preserved.
From the above, it is evident that a wide variety of
material platforms are likely needed to address the full
range of quantum-interconnect challenges. One approach
to combining the best attributes of multiple systems is
heterogeneous integration of multiple materials into a com-
mon platform. For example, rather than developing new
QFC resources in III-V materials or diamond, heteroge-
neous integration [
146
] with Si
3
N
4
, AlN, or LN would
enable a direct coupling of the matter-photonic qubit inter-
face with the QFC interface. Several approaches, including
full wafer bonding, die bonding, transfer printing, and
pick-and-place device transfer, are being considered by the
community to realize this functionality.
VIII. SUPPORTING TECHNOLOGY
A wide range of supporting technologies will be critical
to enabling the variety of QuICs discussed in this docu-
ment, and their sustained development is crucial. A few
examples are highlighted in this section.
Almost all quantum computing and quantum commun-
ication approaches—which require the ability to make
measurements of a quantum state—use devices that per-
form best at cryogenic temperatures, where thermal noise
can be avoided. For example, in quantum communication
systems, optical detectors are an essential component, and
cryogenic superconducting nanowire single-photon detec-
tors are currently the state of the art (in terms of efficiency,
noise, and timing jitter), and have demonstrated efficien-
cies as high as 98%, [
147
], timing jitter below 3 ps, [
148
]
and noise below 1/s, [
149
] though notably
not
in a single
device. In addition, superconducting quantum computing
requires operation at ultralow temperatures to maintain
qubit integrity. In most of these systems, achieving such
low temperatures requires the use of helium. For temper-
atures below 0.8 K, the use of helium-3 is also required.
Unfortunately, helium is a strategically important, nonre-
newable natural resource, and is becoming scarcer.
017002-13
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
In the past decade, there has been significant technolog-
ical progress to use mechanical coolers using recyclable
helium gas as a refrigerant to get to sufficiently low tem-
peratures. These new cryocooling systems, however, were
not developed to meet the needs of the quantum informa-
tion community. While the cost, size, weight, and price
(C-SWAP) are sufficient for research purposes, widespread
adoption in commercial applications will be hampered by
the high C-SWAP. Convergent research into the develop-
ment of efficient, long-life, lightweight coolers with low
vibration will provide a critical enabling technology for
accelerating progress in quantum science and technology.
In addition, the research and development of robust,
inexpensive, low-noise, and stable lasers would accelerate
both research and commercialization of quantum infor-
mation science and technology. For optical wavelengths
that overlap with existing large markets (e.g., telecom-
munications), compact lasers already exist. However, for
wavelengths that are of interest for atomic and artificial
atomic systems (e.g., quantum dots, defects in diamond,
SiC), significant effort is spent by the research commu-
nity to optimize and stabilize custom-built lasers to the
level of performance needed to enable quantum appli-
cations. Unfortunately, the reliability, stability, and cost
of these lasers are not at a level for widespread adop-
tion by researchers or early adopters. There is tremendous
opportunity to accelerate progress with multidisciplinary
research into making better-targeted lasers.
A number of additional supporting technologies, includ-
ing high-speed low-power cryocompatible classical digital
and analog electronics, will also be necessary, and thus
warrant development efforts. Similarly, development of
noncryogenic counterparts of currently cryogenic tech-
nologies is important. For example, development of single-
photon counters such as efficient, high-rate linear-mode or
avalanche-mode photodiodes, could substantially simplify
the task of creating scalable repeater technologies [
150
152
]. For example, recent developments in Si avalanche
photodiodes have led to saturation rates exceeding 160
Mcps in a single device, with a timing jitter as low as 32
ps [
153
]. Many of these developments would also bene-
fit classical computation and communication systems, and
as such are examples of the
dual-use
paradigm of technol-
ogy innovation in which quantum-inspired advances assist
classical technologies and vice versa.
IX. CONCLUSIONS
As the size of quantum systems grows, in terms of the
number of qubits in the case of quantum computers, or
physical size and spatial separation in the case of quantum
networks, so do the challenges related to connecting dif-
ferent parts of the system while maintaining quantum
entanglement across it. For example, long-range commu-
nication networks rely on establishing, distributing, and
maintaining entanglement across thousands of kilometers.
This is challenging due to unavoidable signal losses in the
communication channels. At shorter length scales, diffi-
culties associated with connecting hundreds or thousands
of qubits point to the importance of modular quantum-
computing schemes—likely the
only
viable many-qubit
approach in the near term. Therefore, QuICs, which will
support modular and distributed QIT systems, are emerg-
ing as a grand challenge for QIT. Yet, they have received
significantly less attention from the funding agencies and
from the research community than the quantum hardware
systems they are connecting.
It is the position of the community, as represented by
participants of the NSF workshop on QuICs, that invest-
ment in a national-scale QuICs program is a high pri-
ority. Given the diversity of QIT platforms, materials
used, applications, and infrastructure required, a conver-
gent research approach and partnership between academia,
industry, and national laboratories is required for these
efforts.
The focus of the envisioned QuICs program should be:
(1) a small number of well-supported “Convergent Devel-
opment Teams” comprised of specialists from academia,
industry, and national laboratories, to address specific
QuIC challenges, to create prototype quantum intercon-
nects and application developments; (2) a focused inter-
disciplinary effort aimed at the development of scalable
integrated quantum photonic platforms for QuICs—such
an effort should include synthesis of emerging quantum
materials, fabrication, and packaging of integrated quan-
tum photonic devices, and development of novel ultralow
loss optical fibers; (3) a QuIC test bed where researchers
would gain access to the equipment and expertise needed
to test their own hardware (e.g., qubits, optical squeez-
ing modules, frequency conversion modules, entanglement
sources, transducers, sensors, detectors, lasers, etc.), and
thus carry out research in a convergent environment. The
program will drive the advancement of a quantum infor-
mation science and technology ecosystem that combines
research and technology development with commercial
and educational elements. This will result in new uni-
versity degrees (e.g., quantum engineering), creation of
student internships in industry, and retraining of current
industry employees, thus resulting in an appropriately
skilled workforce.
The goals discussed here for accelerating progress in
quantum interconnects resonate with the goals of two other
recent related NSF Accelerator Workshops—Quantum
Simulators and Quantum Computers [
154
,
155
].
ACKNOWLEDGMENTS
The authors acknowledge NSF OIA-1946564 Grant
“Project Scoping Workshop (PSW) on Quantum Inter-
connects (QuIC)” that provided financial support for the
workshop. The participants are thankful to Ms. Kathleen
L. Masse from John A. Paulson School of Engineering
017002-14
DEVELOPMENT OF QUANTUM INTERCONNECTS. . .
PRX QUANTUM
2,
017002 (2021)
at Harvard University for help with organization of the
workshop.
DISCLAIMER
Any subjective views or opinions that might be
expressed in the paper do not necessarily represent the
views of the U.S. Department of Commerce, the U.S.
Department of Energy, or the United States Government.
APPENDIX A: SOCIETAL BENEFITS OF
QUANTUM-ENABLED TECHNOLOGY
In September 2018 the National Science and Tech-
nology Council released a report, “National Strategic
Overview for Quantum Information Science,” which
stated, “Through developments in [quantum information
science], the United States can improve its industrial base,
create jobs, and provide economic and national security
benefits” [
1
]. Among the intentions of the national effort
outlined by the OSTP report are to: “Focus on a science-
first approach that aims to identify and solve Grand
Challenges: problems whose solutions enable transforma-
tive scientific and industrial progress;” and to “Provide
the key infrastructure and support needed to realize the
scientific and technological opportunities.”
A commentary paper in
Science
[
2
] co-authored by two
participants in the QuIC Workshop, along with a co-author
of the NSTC report cited above, summarizes several of
the societal benefits that QIT can bring: “A fully func-
tioning quantum computer would radically enhance our
capabilities in simulating nuclear and high-energy physics;
designing new chemicals, materials, and drugs; breaking
common cryptographic codes; and performing more spec-
ulative tasks such as modeling, machine learning, pattern
recognition, and optimizing hard logistical problems such
as controlling the electric energy grid or traffic control sys-
tems [
3
].” And, “Using qubits instead of conventional bits
makes it possible to create shared randomness between
parties while knowing whether the communication channel
has been compromised by an eavesdropper. This enables
sending information securely. Quantum communication
can also allow secure communication between multiple
parties, and for interconnecting large-scale quantum com-
puters via a quantum internet [
4
,
5
].” Finally, the
Science
paper also states that the next generation of quantum-based
sensors is projected to outperform current sensing tech-
nologies, for example, in geo-exploration and GPS-free
navigation, biological and medical research, and diagnos-
tic technology.
APPENDIX B: TIMELINE AND MILESTONES FOR MODULAR PROCESSORS (see Sec.
III A
)
Three year
Five year
Ten year
Homogeneous
qubit-qubit
Interconnects
Connection of two fully func-
tional quantum computer mod-
ules with a quantum inter-
connect. Intermodule entangle-
ment distribution rate better
than 10
×
decoherence rate and
1% of the local gate rate.
Connection
of
four
fully
functional quantum computer
modules
with
a
reconfig-
urable quantum interconnect.
Intermodule
entanglement
distribution rate better than
100
×
decoherence rate and
10% of the local gate rate.
Manufacturable quantum com-
puter modules with quantum
interfaces that can scale to
over 100 modules. Intermodule
entanglement distribution rate
better than 1000
×
decoherence
rate and 100% of the local gate
rate.
Transduction
with
non-native
photonic channels
Demonstration of tunable quan-
tum interconversion between
disparate photons (e.g., tunable
visible-to-telecom or optical-
to-microwave, including band-
width conversion).
Demonstration
of
intercon-
version between microwave
and optical photons with high
fidelity, SNR, and bandwidth.
Connection to quantum inter-
net.
Heterogenous
qubit-qubit
interconnects
Interface between atom- and
solid-state-based memory to
non-native flexible and tunable
photonic channel.
Entanglement between two dif-
ferent types of quantum proces-
sor (various atomic and solid-
state memory qubits).
QC performance in a multin-
ode cluster that goes beyond
the capability of any individ-
ual node, and also beyond those
individual nodes connected by
a classical network.
017002-15
DAVID AWSCHALOM
et al.
PRX QUANTUM
2,
017002 (2021)
APPENDIX C: TIMELINE AND MILESTONES FOR QUANTUM INTERNET (see Sec.
III B
)
Three year
Five year
Ten year
Major
achievements
Detected photonic entanglement
rate beyond 10
8
ebits/sec
Quantum
repeaters
with
error
correction
against
operation errors
Forward
error-corrected
photonic quantum states
for one-way repeaters
Distance and rates
Entangled quantum memory over
>
10 km distance
Verifiable quantum entan-
glement distribution over
>
100 km at
>
1 M-ebits/sec;
distillable
entanglement
rates
>
100k-ebits/sec
Quantum networks reach-
ing transcontinental scales
of thousands of km
Capability
of
repeater nodes
Quantum repeater node via entan-
glement swapping beyond direct
transmission
Active
error
correction
against
operation
errors;
many-party
protocols
demonstrated
in
fielded
quantum networks
Full
error
correction
against loss and operation
errors; hybrid nodes with
different functions.
Number
of
repeater nodes
Quantum networks with
>
3mem-
ory nodes and
>
10 user nodes
Networks of
>
10 quantum
repeaters and quantum com-
puters in superposition
Networks with
>
100 of
repeater nodes
Free-space
quantum network
Constellation of 3
5 mobile plat-
forms demonstrated
Entanglement
swapping
between space earth
Transcontinental
entan-
glement distribution via
quantum-memory-enabled
satellite
Quantum network
applications
Quantum-secured communication
rate exceeding 1 MB/sec over
100 km
Network-based
quantum
metrology
Blind Quantum Comput-
ing
APPENDIX D: TIMELINE AND MILESTONES FOR QUANTUM SENSORS (see Sec.
III C
)
Three year
Five year
Ten year
Matter-based
quantum sensors
Entanglement-enhanced sensing in
local registers (e.g., multinuclear
or electron-nuclear spin entangle-
ment around or within a color cen-
ter).
Entanglement-based distributed
solid-state sensors (e.g., instan-
tiated via spin-photon entangle-
ment)
Demonstration of entanglement-
assisted clock synchronization
(in one room—easier, across the
world—geography is more chal-
lenging).
Sensors connected to quantum
internet.
Photonic quantum
sensors
Scale up to ten entangled photonic
sensors in an integrated platform.
Development of various trans-
ducers including high-efficiency
rf-Photonic and optomechanical
transducers.
Improvement of chip-to-fiber cou-
pling efficiency to approximately
80%.
Applications include rf, inertial,
mechanical sensing, etc.
Fully
reconfigurable
on-chip
entangled photon sources oper-
ating at different frequencies and
entangled degrees of freedom.
Integrated on-chip transducers
>
95% chip-to-fiber coupling effi-
ciency.
New
entanglement-enhanced
sensing approaches for classical
noise rejection, high resolution,
etc.
Entanglement-enhanced multia-
perture telescopy.
Scale up to approximately 100
entangled sensors.
Integrate with the quantum inter-
net for long-distance entanglement
distribution to sensors.
Use quantum error correction to
enhance sensitivity.
Incorporating entangled sensors
into existing classical sensing
infrastructures.
Long-baseline telescope enabled
by quantum repeaters.