Professional Documents
Culture Documents
net/publication/330069201
CITATIONS READS
56 1,357
3 authors, including:
All content following this page was uploaded by Yunseong Nam on 29 March 2019.
0018-9219 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
two basis states of the quantum It is experimentally verified that care must be taken to isolate the
system that represents it, |0 and one can take advantage of the qubits from their environment and
|1. In contrast to a classical bit, superposition and entanglement to ensure that the logic gate operations
which may be in a state of 0 or 1, perform certain tasks for which no modify the quantum system as close
a qubit may exist in a superposition classical methods with comparable to the desired outcome as possible.
of the two states |ψ = x0 |0 + efficiency are known. Such demon- Unlike classical digital logic gates
x1 |1, where x0 and x1 are the com- strations include the following. where the output is rather insensitive
plex coefficients. These coefficients 1) Superposition: The generation to small noise at the input, quantum
satisfy the normalization condition of random numbers, relying on logic gates manipulate the complex
|x0 |2 + |x1 |2 = 1, owing to the fact the true probabilistic nature of coefficients of the superposition states
that |x0 |2 and |x1 |2 represent the quantum mechanics, and reduc- and any error or noise in the gates
probabilities of finding the qubit in ing to the preparation and mea- can accumulate in an analog fash-
the state |0 and |1 upon mea- surement of the quantum state ion. To address this problem, quan-
surement, respectively, and the prob- √1 |0 + √1 |1 to obtain the tum error correcting codes (QECCs)
2 2
abilities must sum up to 1. With desired random number. were developed that can be used to
n qubits, the composite qubit sys- 2) Entanglement: A game (known correct these errors, in a manner sim-
tem is described by a superposi- as the CHSH game [6], named ilar to how classical error correction
as |ψn =
tion of all possible 2n basis states
2n −1
i=0 xi |i, where the
after its inventors Clauser,
Horne, Shimony, and Holt, and
works in an analog communication
channel. In a QECC, the quantum
summation runs over all binary rep- equivalent to Bell’s inequality) data (referred to as logical qubits)
resentations of n-qubit states (from that can be played using is redundantly encoded into a larger
|00...0 to |11...1), with the nor- entanglement to correlate number of physical qubits, in a way
malization condition for the complex
coefficients xi , i=02n −1
|xi |2 = 1. Such
players’ strategies thus enabling
a better than the best possible
that the errors on the physical qubits
can be corrected so long as they occur
an n-qubit system description illus- classical strategy. sufficiently infrequently. Systematic
trates a direct access to the full com- 3) Superposition and Entangle- approaches to construct fault-tolerant
putational space that is exponentially ment: Testing the properties (FT) quantum computers from faulty
large in n. With xi being complex of small Boolean functions components centered around QECC
rather than real numbers, the compu- given as a black box, including were developed, in close analogy
tational advantage of quantum com- Deutsch–Jozsa algorithm to to early days of digital computers
puters versus classical probabilistic distinguish balanced and con- based on faulty vacuum tube technol-
computers comes into play. Similar stant functions, and Grover’s ogy. Current FT quantum computing
to the classical logic gate operations, algorithm to find a satisfying approaches incur a very large over-
quantum logic gates compute the out- assignment. The advantage head, which is expected to improve
put states from the input states but over classical algorithms as the performance of the component
allow the input states to be an arbi- manifests itself through the technology advances.
trary quantum state (such as superpo- ability to query black boxes Similar to the situation in the
sition states). Quantum logic gates are by superposed and entangled early 20th century when the tech-
implemented by designing the time states, therefore extracting the nology to put together a compu-
evolution of the input quantum states desired information faster than tational machine at modest scales
to achieve the desired output states, it is possible to do classically. remained elusive, the current sta-
often by manipulating the external tus of efforts toward constructing
II. Q U A N T U M C O M P U T I N G
control signals. Mathematically, quan- a functional quantum computer is
H A R D WA R E
tum gates are described by the uni- in its early stages. Early research
tary matrices, and their application A. Implementation Challenges effort emerged in the mid-1990s,
is accomplished through multiplica- The access to unconventional com- in search of the adequate physical sys-
tion of the respective matrix by the putational power offered by quan- tem in which to implement the qubits.
state vector. During the computation, tum computers comes at the price Over the course of the ensuing two
the state of the qubits develops cor- of the fragility of quantum systems. decades, a few physical platforms for
relations among them where the state Quantum systems tend to quickly qubits, most notably trapped atomic
of one qubit cannot be described lose their quantum nature (quantum ions and superconducting circuits,
independently of the state of other coherence) and revert to classical have emerged as the leading candi-
qubits, known as quantum entangle- descriptions as the system becomes dates to construct functional quan-
ment. Superposition and entangle- larger and more complex, due to tum computers. Some small-scale but
ment are two unique properties of mainly the interaction with the envi- already fully programmable and uni-
quantum physics that do not have ronment that tends to destroy super- versal quantum computers are avail-
classical counterparts. position and entanglement. Utmost able to a broader public via cloud
mechanical phenomena, akin to how dynamics simulation appear to computer, but many more are known.
a classical computer is useful for sim- be particularly promising [5]: We refer the interested reader
ulating various classical mechanical Suzuki–Trotter formulas [2], [23] to [13].
phenomena. A host of interesting that rely on breaking down the
problems that are not otherwise target evolution into small pieces and
known how to address efficiently expressing the individual evolution IV. O U T L O O K
with classical computers may be terms as quantum circuits, and an The usefulness of a computer is mea-
successfully solved with quantum asymptotically optimal approach by sured by its ability to successfully
simulations performed on a quantum Low and Chuang [16] that extends execute desired computational tasks.
computer. A few examples include, signal processing techniques to the In a conventional digital computer,
in increasing order of difficulty and quantum scenario. the performance typically boils down
impact, the following: Despite the advantage provided to parameters such as the memory
1) deeper understanding of many- by quantum algorithms, simulating size, processing speed, and proces-
body physics and strongly the Hamiltonian evolution is still a sor architecture, as the gate errors
correlated matter, with possible highly demanding task, as illustrated are generally negligible. In a quantum
applications in the design in Fig. 2 containing some of the best- computer, where the logic gate oper-
of room-temperature super- known quantum computing resource ation features a nonnegligible prob-
conductors or materials with estimates. To address this difficulty, ability of error, the performance will
favorable electrical properties; a quantum/classical hybrid approach be constrained by both the size of the
2) high-precision chemistry was developed, known as the vari- computer (measured by the number
calculations for developing ational quantum eigensolver (VQE). of qubits in the system, for example)
new catalysts and chemical VQE seeks to find the lowest eigen- and the size of quantum circuit (mea-
processes, with important values and eigenstates of a matrix H sured by the total number of quantum
applications such as finding corresponding to the desired physical gates in the algorithm) that the system
a replacement for the Haber Hamiltonian. The approach is based can execute before the errors accumu-
process for nitrogen fixation on variational method, which con- late to result in a meaningless out-
(NF) used in the synthesis of sists of preparing a trial quantum come. Fig. 2 shows the performance
fertilizers; state with some variational parame- space of quantum computers charac-
3) large-scale, highly accurate ters, evaluating the energy of the terized by these two metrics. As the
molecular dynamics simu- state, and updating the state prepara- gate error probability is reduced for
lations to study problems tion with new variational parameters a given number of qubits, a wider
in protein folding and drug to reduce the corresponding energy. class of algorithms can be executed
design. The procedure is run in an itera- on such a processor making it more
In quantum theory, the Hamiltonian tive fashion, guided by a classical useful. For NISQ systems, the abil-
is a formula describing interactions optimization strategy, until a solu- ity to operate the two-qubit gates
between constituent particles in a tion, such as the lowest energy state, between a given pair of qubits within
complex quantum system, based is found. VQE utilizes quantum com- the system [connectivity described in
on the energy associated with each puters to prepare the trial states from Fig. 1(b)] makes a big difference in
interaction term. Tracking the actual the variational parameters and can performance [15]; a two-qubit gate
time-dependent evolution of the dramatically simplify the portion of between qubits that are not directly
system subject to these interacting the simulation task run on a quan- connected has to be replaced by mul-
Hamiltonian terms can be extremely tum computer. However, it must be tiple gates allowed by the connectiv-
complicated due to the exponen- repeated a large number of times per ity, incurring a large overhead cost in
tially growing number of possible trial and needs to be supplied with the error probability by the accumu-
interaction configurations among highly optimized (and often empiri- lation of gate errors. For FT quan-
the particles. Methods for simulating cal) procedures for designing and iter- tum computers, the connectivity is
Hamiltonian dynamics on a quantum atively improving the trial states. VQE largely driven by the choice of QECC
computer have been developed, which also runs the risk of being stuck in a and its implementation. A much lower
take the Hamiltonian description local minimum. We note that VQE is error probability can be achieved at
as the input and simulate the time a set of strategies rather than a com- the logical qubit level in these sys-
evolution of the system to extract pletely specified algorithm, and the tems, at the cost of the fault-tolerance
a relevant quantity of interest rigorous efficacy of this method for overhead (in physical qubit numbers
(such as the lowest overall energy achieving a useful answer is not yet and logical-qubit level gate execution
of the system and corresponding known. times) that depends on the choice of
quantum state) to within the specified Here, we highlighted some of QECC.
tolerance. Studies show that two the algorithms and approaches to The quantum computing systems
kinds of algorithms for Hamiltonian solving problems on a quantum available today (green colored region
Fig. 2. Performance space of quantum computers, measured by the error probability of each entangling gate in the horizontal axis
(roughly inversely proportional to the total number of gates that can be executed on a NISQ machine), and the number of qubits in the
system in the vertical axis. Blue dotted line approximately demarcates quantum systems that can be simulated using best classical
computers, while the green colored region shows where the existing quantum computing systems with verified performance numbers lie
(as of September 2018). Purple shaded region indicates computational tasks that accomplish the so-called “quantum supremacy,” where the
computation carried out by the quantum computer defies classical simulation regardless of its usefulness. The different shapes illustrate
resource counts for solving various problems, with solid symbols corresponding to the exact entangling gate counts and number of qubits in
NISQ machines, and shaded regions showing approximate gate error requirements and number of qubits for an FT implementation (not
pictured are the regions where the error gets too close to the known fault-tolerance thresholds): cyan diamond and shaded region
correspond to factoring a 1024-bit number using Shor’s algorithm [14], magenta circle and shaded region represent simulation of a 72-spin
Heisenberg model [20], and orange shaded region illustrates NF simulation [21].
in Fig. 2) are quite limited in their tum computer capable of exploring of quantum computers and encourage
performance and can be simulated complex quantum systems beyond further investments.
with classical computers (blue dashed classical capability is accomplished. Once quantum computers are
line roughly separates classically sim- The current trajectory of quantum developed to such practical utility,
ulable quantum computers from those computer development raises the it is highly likely that it will lead
impossible to simulate). A “quantum hope that quantum supremacy can be to applications of commercial via-
supremacy” experiment is defined to reached in the next few years. bility. The development of quantum
be a task that one can accomplish The next major step is to demon- computers beyond this point could
using a more advanced quantum strate that a quantum computer can mirror the development history of
computer that cannot be simulated be used to solve a problem of prac- classical digital computers, where the
with available classical computational tical utility that cannot otherwise be range of applications expanded as
resources [4], regardless of its use- addressed. This will likely start from the commercial value created in early
fulness. Although the exact boundary a problem of scientific interest, such applications stimulated technology
for quantum supremacy is currently as various kinds of quantum sim- investments. Commercial quantum
a moving target, it is expected that ulations, including the dynamics of computers produced at this stage
the number of qubits (50), quantum closed quantum mechanical systems are envisioned to run applications
gate error probabilities (10−3 ) and useful in the understanding of many- including those based on Shor’s and
coherence times (103 fold the gate body physics [20], or studying the Grover’s algorithms, and ultimately
times) required for such an exper- structure of or interaction between solving most complex and resource
iment is within reach. While such complex chemical molecules [1]. It is demanding problems such as those in
a demonstration may not be suf- conceivable that such demonstrations drug design. Therefore, the demon-
ficient to signal the practical util- can be done before fully FT computers stration of the practical utility of
ity of quantum computers, it proves are constructed, which would signifi- quantum computers discussed in the
the fundamental notion that a quan- cantly speed up the growth of utility previous paragraph can be a trigger
point for a new industry based on for selected applications (application- devices have been demonstrated, yet
quantum computers. We anticipate specific integrated circuit-style quan- the practical utility of quantum com-
this point may be reached within the tum computers). Drawing from the puters has not been established.
next decade or two. history of classical computers dating The transition of the proof-of-concept
Crucial to the development of to the early stage in their develop- devices to useful computational sys-
quantum computing technology is not ment, it is likely that the applica- tems faces a set of new technical chal-
only continued improvements in the tions of quantum computing that we lenges, ranging from improving and
qubit hardware technology but also: can think of now and highlighted expanding qubit hardware to devel-
1) new advances in quantum com- in this paper will be eclipsed by oping control/operating systems to
puter architectures that can accom- new applications found while fur- innovations in algorithms and appli-
modate larger and more complex ther developing quantum computers. cations. Pursuing the solutions to
computational tasks within the avail- A growing ecosystem of quantum these technical challenges defines the
able hardware; 2) developments in computer developers, users, and an demand for the new generation of
algorithms and methodologies that educational system training necessary hardware, software, and application
map useful problems to quantum workforce will be critical in enabling engineers to create and sustain the
computers; 3) the search for and nar- a vibrant future quantum computing upcoming quantum computing indus-
rowing down the set of practical appli- industry. try. There is a major opportunity for
cations (short and long term; this creative minds to be a part of this
requires bringing in domain experts); V. C O N C L U S I O N future.
and 4) optimization of software and Now is a very exciting time when pro-
hardware to boost the performance grammable quantum computational
REFERENCES
[1] R. Babbush, D. W. Berry, J. R. McClean, and [8] D. P. DiVincenzo, “Two-bit gates are universal for [17] J. T. Merrill et al., “Demonstration of integrated
H. Neven. (2018). “Quantum simulation of quantum computation,” Phys. Rev. A, Gen. Phys., microscale optics in surface-electrode ion traps,”
chemistry with sublinear scaling to the continuum.” vol. 51, no. 2, p. 1015, 1995. New J. Phys., vol. 13, p. 103005, Oct. 2011.
[Online]. Available: https://arxiv.org/abs/ [9] R. P. Feynman, “Simulating physics with [18] D. L. Moehring et al., “Design, fabrication and
1807.09802 computers,” Int. J. Theor. Phys., vol. 21, nos. 6–7, experimental demonstration of junction surface ion
[2] D. W. Berry, G. Ahokas, R. Cleve, and B. C. Sanders, pp. 467–488, 1982. traps,” New J. Phys., vol. 13, p. 075018, Jul. 2011.
“Efficient quantum algorithms for simulating sparse [10] J. P. Gaebler et al., “High-fidelity universal gate set [19] C. Monroe and J. Kim, “Scaling the ion trap
Hamiltonians,” Commun. Math. Phys., vol. 270, for 9 Be+ ion qubits,” Phys. Rev. Lett., vol. 117, quantum processor,” Science, vol. 339, no. 6124,
no. 2, pp. 359–371, 2007. no. 6, p. 060505, 2016. pp. 1164–1169, 2013.
[3] R. Babbush et al., “Encoding electronic spectra in [11] L. K. Grover, “A fast quantum mechanical algorithm [20] Y. Nam and D. Maslov. (2018). “Low cost quantum
quantum circuits with linear T complexity,” Phys. for database search,” in Proc. 28th STOC, 1996, circuits for classically intractable instances of the
Rev. X, vol. 8, no. 4, p. 041015, 2018. pp. 212–219. Hamiltonian dynamics simulation problem.”
[4] S. Boixo et al., “Characterizing quantum supremacy [12] T. P. Harty et al., “High-fidelity preparation, gates, [Online]. Available: https://arxiv.org/abs/
in near-term devices,” Nature Phys., vol. 14, memory, and readout of a trapped-ion quantum 1805.04645
pp. 595–600, Apr. 2018. bit,” Phys. Rev. Lett., vol. 113, no. 22, p. 220501, [21] M. Reiher, N. Wiebe, K. M. Svore, D. Wecker, and
[5] A. M. Childs, D. Maslov, Y. Nam, N. J. Ross, and 2014. M. Troyer, “Elucidating reaction mechanisms on
Y. Su, “Toward the first quantum simulation with [13] S. Jordan. Quantum Algorithm Zoo. Accessed: quantum computers,” Proc. Nat. Acad. Sci. USA,
quantum speedup,” Proc. Nat. Acad. Sci., vol. 115, Jul. 7, 2018. [Online]. Available: https://math. vol. 114, pp. 7555–7560, Jul. 2017.
no. 38, pp. 9456–9461, 2018. nist.gov/quantum/zoo/ [22] P. W. Shor, “Algorithms for quantum computation:
[6] J. F. Clauser, M. A. Horne, A. Shimony, and [14] S. A. Kutin. (2006). “Shor’s algorithm on a Discrete logarithms and factoring,” in Proc. 35th
R. A. Holt, “Proposed experiment to test local nearest-neighbor machine.” [Online]. Available: FOCS, 1994, pp. 124–134.
hidden-variable theories,” Phys. Rev. Lett., vol. 23, https://arxiv.org/abs/quant-ph/0609001 [23] M. Suzuki, “General theory of fractal path integrals
no. 15, pp. 880–884, 1969. [15] N. M. Linke et al., “Experimental comparison of two with applications to many-body theories and
[7] S. Debnath, N. M. Linke, C. Figgatt, quantum computing architectures,” Proc. Nat. Acad. statistical physics,” J. Math. Phys., vol. 32, no. 2,
K. A. Landsman, K. Wright, and C. Monroe, Sci. USA, vol. 114, no. 13, pp. 3305–3310, 2017. pp. 400–407, 1991.
“Demonstration of a small programmable quantum [16] G. H. Low and I. L. Chuang, “Optimal Hamiltonian [24] Y. I. Manin, “In Vychislimoe i nevychislimoe
computer with atomic qubits,” Nature, vol. 536, simulation by quantum signal processing,” Phys. [computable and noncomputable],” Sov. Radiol.,
pp. 63–66, Aug. 2016. Rev. Lett., vol. 118, no. 1, p. 010501, 2017. pp. 13–15, 1980.