Quantum Computer Architectures: A Survey

Saumya Jain
Georgia State University
Atlanta, USA
Email Id: [email protected]

Abstract – Computers reduce human effort and also focus on The reason behind this emerging curiosity is because of its
increasing the performance to push the technology forward. vast computational power. However, it was complicated to
Many approaches have been devised to increase the build a quantum computer which implements the quantum
performance of the computers. One such way is to reduce the effects. One reason was because of the limited knowledge of
size of the transistors used in the systems. Another very how quantum properties could help to achieve the speed up.
significant tactic is to use quantum computers. It proved to be This changed in 1994, when Peter Shor at Bell Laboratories
very effective when used to factor large numbers. It was amazed everybody by factoring large numbers using a
found that it could decrypt codes in 20 minutes which took quantum algorithm in polynomial time [Rieffel et al. 2000;
billions of years with classical computers. This was a great Ross 2008; Shor 1994]. This encouraged scientists to try to
motivation for focusing on this topic. A quantum computer develop quantum computers and quantum algorithms. This
allows a ‘quantum bit’ or qubit to have three states – 0, 1, and also led to discovering an algorithm of quantum key
0 or 1. The last state is the coherent state. This enables an distribution for transmission of information. This steered
operation to be performed on two diverse values at the same towards viable quantum cryptography techniques. A new
time. However, this brings out a problem of decoherence. It approach of teleportation, which is yet another method of
becomes difficult to perform the computation using quantum communication, was also presented.
computers. A quantum computer is desired to have five Quantum Computation helps in explaining tricky physics
capabilities – scalable system, initializable state, long experiments [Harrow 2012]. There are techniques that help in
decoherence time, universal set of quantum gates, high modeling quantum computation using a functional language
efficiency measurements. Architecture of the quantum [Sabry 2003]. Concepts used in Quantum Computing are
computer is the new research area. It is affected by quantum exhibited as structures and operations in Haskell. It offers
arithmetic, error management, and cluster-state computing. exponential speed-up and power over conventional
Without it, the quantum algorithms would not prove to be as computation. We know that when classical computers are used
efficient. To fully utilize the power of a quantum computer, for computation, the speed can be improved by using
the algorithms used should be based on quantum parallelism. parallelism. This aids in reducing the running time. However,
This paper discusses a few of the basic concepts used in this exponential reduction in time can only be achieved if we
quantum computing briefly. The paper also focuses on augment the number of processors exponentially. This
different architectures of quantum computer available in requires an exponential increase in the physical space as well.
literature. However, the problem of a practical large-scale When contrasted with quantum systems, parallelism is
quantum computer still persists. exponentially increased with the linear increase in the size of
the system [Rieffel 2000]. Parallelism is inbuilt in quantum
Keywords – Cluster-state computing, entanglement, fault- systems because of the properties it inherits. Thus, quantum
tolerance architecture, layered architecture, quantum computing makes it possible for more efficient solutions to a
computing, qubits particular problem than its classical counterpart [Lanzagorta et
al. 2005].
I. INTRODUCTION
II. BACKGROUND
Quantum Computation was first thought of by Richard
Feynman who said that by using the quantum mechanical This section covers all the introductory theories needed to
effects, faster computation can be achieved. This was found better comprehend quantum computing. It starts with the
when scientists tried simulating these effects on a computer. explanation of subjects of quantum mechanics with the help of
Another hint was the exponentially large state spaces that an experiment. After that, the basic concepts used in quantum
quantum mechanics makes available which indicates enormous computing are explained which would help in our
amount of computational resources [Harrow 2012]. Quantum understanding.
computing is a new field which is of interest to the researchers.

978-9-3805-4416-8/15/$31.00 2015
c IEEE 2165
A. Quantum Mechanics Experiment If there are n qubits in a system, the resulting state space has
the dimension of 2n. Thus, the exponential growth in
This experiment requires a strong light source, a projection
computation can be seen by this. State space in a quantum
screen to observe light, and three polaroids or filters which are
systems combine through tensor products, also called outer
polarized horizontally, vertically, and at 45°. The light source
products. Tensor products are represented by a ۪ b. If there
emits light particles called photons which are witnessed on the
are two qubits with the same basis as {|0ۧ, |1ۧ}, then the state
screen. The light source is assumed to be randomly polarized
space for the two qubits has the basis {|0ۧ ۪ |0ۧ, |0ۧ ۪ |1ۧ, |1ۧ
which means that the photon particles are moving in random
۪ |0ۧ, |1ۧ ۪ |1ۧ}. |0ۧ ۪ |1ۧ can also be denoted as |01ۧ.
directions. The polaroids are to be inserted in between the
source of light and the screen. When we add the horizontal
C. Entangled States
filter between the source and the screen, we see that the output
on the screen has half of the intensity of the source. This can be Entanglement is a phenomenon experienced in quantum
mathematically explained by expressing the random systems. It makes the qubits involved dependent on each
polarization of a photon by a| + b| , where | , and | are other. It becomes impossible to obtain one component from
the entangled system. This phenomenon is exclusive only to
basis vectors which are orthogonal to each other, and a and b the quantum systems, and there is no classical equivalent for
are complex numbers. These vectors are chosen such that |a|2 + this. If one system is measured, entanglement makes it
|b|2 = 1. When this passes through a horizontal polaroid, the possible to conclude the measurement of the other system.
vertical component disappears. Thus, the output results in only This is done when no measurement has been applied on the
half the intensity of the incoming source. second system. An entangled state can be fashioned by
allowing a qubit to pass a Hadamard gate and then a CNOT gate
[Nielson et al. 2000]. It is because of entangled states that the
exponential growth in state spaces is accomplished. A
limitation is that to mimic for a small quantum system, it
requires mammoth resources on a traditional computer. This
paves way for motivation for designing quantum computers.

III. ARCHITECTURAL DESIGNS

Scalable quantum computers have emerged which allows the
Fig. 1. The three filters with their polarization are shown. When the incoming presence of faulty elements in the circuit if a certain criteria is
source light passes the three filters, the result at the screen becomes one-eighth met. The restriction imposed on them is to keep the occurrence
the intensity of that source [Riefell et al. 2000]. of error to be below some threshold (estimated to be up to 10-
4). It has been achieved to a certain extent by using improved
error correcting codes. Generally, the architectures developed
B. Quantum Bits
are built using solid-state technology such as quantum dots or
In classical systems, we use a bit to represent information. A other molecules. For a universal architecture, general-purpose
classical bit can contain either binary 0 or binary 1 as the value hardware is considered. For such architecture, the building
[Kasivajhula 2006]. The smallest unit in quantum systems is blocks should be such so as to minimize the error-correction
called as the qubit. It is a two-dimensional unit vector. Its basis overhead. To provide this, there should be reliable data paths
has been decided as |0ۧ and |1ۧ. We may assume that the fixed and enough quantum memory which is also resourceful. It is
basis is corresponding to spins, or the polarization. One qubit noticed that when the qubits that interact are near to each
can be used to characterize a value of ‘0’, ‘1’, or ‘0 or 1’. This other, it is the least error-prone. There are three chief
signifies that a quantum bit can be used to inhabit many states components – quantum ALU, quantum memory, and a
between 0 and 1. A state is the superposition of the two basis dynamic scheduler [Oskin 2002]. This architecture also uses
vectors such as a|0ۧ + b|1ۧ, where a and b are complex numbers teleportation by the way the components are wired together.
with |a|2 + |b|2 = 1. These coefficients indicate the probability of The complete general architecture is shown in Fig. 2.
the basis when they are measured. For example, from the above
equation, we can say that to measure |0ۧ as the result, the
probability is |a|2. Similar is the case for measuring |1ۧ.

2166 2015 2nd International Conference on Computing for Sustainable Global Development (INDIACom)
 B. Fault-Tolerant Architecture
This architecture came into picture because of a fundamental
problem. There was no computer which was tolerant towards
errors. Errors are bound to creep into the system because
perfect isolation can never be achieved. It is impractical to
assume a perfectly isolated system as a requirement.
Interference from the environment can be prevented to some
extent by using polarization when quantum dots are used
[Possignolo et al. 2012]. To overcome this, the architecture
uses electrically controlled semiconductor spins along with
quantum memory and long-distance coupling mechanisms
[Taylor et al. 2005]. The proposed design is modular, and
hierarchical. It allows autonomous control and non-local
coupling using controlled electron transport. The architecture
is still scalable with the error threshold in an allowed range for
fault-tolerance. One task faced by this architecture is to
Fig. 2. Quantum Architecture basic components [Oskin et al. 2002]. alleviate environmental couplings which become harder for
solid states. The most significant function of a fault-tolerant
quantum computer is preparing ancilla (temporary) states for
A. Cluster State Computer circuits implementing fault-tolerance [Jones et al 2012].
Universal set of gates is used for subduing the effects of
Cluster state is a very large entangled state. The computation
hyperfine interactions which are the primary source of errors.
involves large number of resources. For computation, one-qubit
measurements are used on the cluster states [Raussendorf et al.
C. Layered Architecture
2003]. Cluster states allow for retrying the operation if it did
not provide with the desired result. This phenomenon is called Layered architecture supports the development of a framework
heralded [Meter et al. 2006]. The computation has to deal with in a systematic and a hierarchical manner. This allows
error management like all other quantum systems. A proposed confronting the challenges at each individual layer. The
architecture is to take the cluster state as the bottom subsystems can be considered autonomously when layered
architectural layer, middle layer as one performing fault approach is used. Every layer has a set of related
tolerance. The top-most layer would be one which executes all responsibilities assigned. Interfaces are defined between two
the applications. interacting layers. Each lower layer provides services to the
The architecture described here uses photonic qubits. It seems layer above it. The lower layer processes the results when a
practical to use photonic qubits because of their movement and command is given out to it by a higher layer. A complex
low cost. The architecture has focused on using topological system is divided into small and controllable sets of
cluster state quantum computing as its underlying model. The procedures. Management of resources also becomes a trivial
advantage of using this model is that error correction protocols task for such architecture. A framework must address certain
are implicitly incorporated. Photon loss, a limitation of using issues for the creation of an efficient design of a quantum
photonic qubit (discussed in the previous section) is corrected computer. The design must focus on consideration of a faulty
by this model. The building block for this architecture is the quantum hardware, error management, classical processing. In
photonic module and chip (shown in Fig. 6). The chip is a this section, two types of layered architecture are discussed.
quantum device which provides entanglement of photons. It is One is based on optical control of quantum dots (QuDOS),
also assumed that the chip contains the source and detectors of while the other is based on photonic qubits with topological
photons. The chip is considered as the basic block because it is cluster state quantum computation (TCQC).
utilized to develop a lattice preparation network. This lattice is i. Based on QuDOS: It includes five layers depicted in Fig. 3.
where the cluster states are formed. The qubits used for this framework are electron spins within
quantum dots. The layers are physical, virtual, Quantum Error
Correction (QEC), logical, and application [Jones et al. 2012].

2015 2nd International Conference on Computing for Sustainable Global Development (INDIACom) 2167
 Approximations were used for quantum processes because
exact depiction of these processes is not possible on classical
computers. However, using FPGA makes the mapping of
parallel tasks more proficient. To make the FPGA more
efficient, instead of using classical, the focus shifted to
quantum FPGA for making quantum computing
programmable. Programmable structures are more flexible and
clear. The users can program the architecture according to
their own needs.
Quantum FPGA architecture (shown in Fig. 5) is a mixture of
architectures using measurement-based quantum computation
and the qubus system. Qubus is used to generate entanglement
in the system, and is reusable. The improvement by using
qubus is lowering the error probability to 10-2. This consists of
two building blocks, Quantum Logic Blocks (QLB) and
Quantum Routing Channel (QRC). QRC encapsulates multiple
QLBs to form large logic realization. It also consists of
memory qubits and channel (ancilla) qubits (as depicted in
Fig. 3. Layered architecture for a quantum computer architecture [Jones et al. Fig. 11). QLBs are ephemeral qubits with small coherence
2012]. time, and are for realizing quantum logic smaller in size than
QRC. QRC takes longer to execute than QLB. This explains
why QRCs are needed to have long coherence times. Two
ii. Based on Topological Cluster State Computation with buses are also employed. A local bus in QLB is useful for the
Photonic Qubit: This uses the building block as photonic chip generation of cluster states. QRC contains a global bus used
(Fig. 6) described in section 4.1. This proposes to be better than for multiqubit rotations around the Z-axis [Chen et al. 2013].
the one described previously. It guarantees a high threshold 
error rate of 0.75% [Bardhan et al. 2011]. Another reason is
because it uses the photonic chip which has error correcting
capabilities in it, the cost is cut down to an extent. Thus,
instead of five layers, there are just four layers in the
architecture (Fig. 4).

Fig. 5. Quantum FPGA architecture [Chen at al. 2013].

IV. CONCLUSION AND FUTURE SCOPE

Quantum Computing is a new area of study with many
Fig. 4. Layered architecture using photonic qubit in TCQC [Bardhan et al. advantages over classical computing due to entanglement and
2011]. superposition. Many quantum algorithms have been designed
for various applications. These algorithms provide exponential
D. Quantum FPGA Architecture speed up. The algorithms don’t prove to be that effective when
The architecture was introduced because when quantum they are run on classical processors. This called for an
algorithms were modeled on classical computers, the efficiency emergence of quantum architectures. A few technologies have
was lost. The reason is because quantum algorithms include been described which are used in architecture development.
concepts like entanglement and superposition which are Researchers have tried to build architectures based on the
difficult to implement classically [Khalid et al. 2004].

2168 2015 2nd International Conference on Computing for Sustainable Global Development (INDIACom)
technologies available. However, accomplishing scalable Vivien M. Kendon. 2011. Reduce, Reuse, Recycle for Robust
quantum system architecture remains an open issue. Cluster-State Generation. Phys. Rev. A 83, 042327.
Many architectures have been discussed in the paper such as [11] N. Cody Jones, Rodney Van Meter, Austin G. Fowler, Peter
based on cluster state computing, fault-tolerant, layered, and L. McMahon, Jungsang Kim, Thaddeus D. Ladd and
Yoshihisa Yamamoto. 2012. Layered Architecture for
quantum FPGA design. Quantum algorithms must be optimized Quantum Computing. DOI: http://arXiv:1010.5022v3 [quant-
in accordance with the architecture followed. This would result ph], 27 September 2012.
in an effective quantum computation. Another future work that [12] Siddhartha Kasivajhula. 2006. Quantum Computing: A
can be performed is replacing the layers in layered architecture. Survey. ACM SE’06 March 10-12, 2006, Melbourne, Florida,
This is anticipated because of the ongoing change in the USA. ACM 1-59593-315-8/06/0004.
technologies available. Over time new technologies are [13] Ahmed Usman Khalid, Zeljko Zilic and Katarzyna Radecka.
developed. The architecture would become much better if the 2004. FPGA Emulation of Quantum Circuits. Proceedings of
technology followed is the most recent and optimized one. the IEEE International Conference on Computer Design.
When the circuit becomes large, it turns out to be difficult to ICCD’04. DOI: 1063-6404/04
[14] Marco Lanzagorta and Jeffrey K. Uhlmann. 2005. Hybrid
optimize. This issue must be addressed in the future. At the Quantum-Classical Computing with Applications to
programming architecture front, if libraries for quantum gates Computer Graphics. University of Missouri-Columbia.
(reversible) are made available, it would lead to easy emulation [15] Rodney Van Meter and Mark Oskin. 2006. Architectural
of the quantum circuits. Software packages for topological Implications of Quantum Computing Technologies. ACM
quantum circuits are also needed. A new programmable Journal on Emerging Technologies in Computing Systems,
quantum computer has been developed by D-wave systems but Vol. 2, No. 1, January 2006, Pages 31-63.
debates are ongoing about the insides of the system. [16] Mark Oskin, Frederic T. Chong, Isaac L. Chuang. 2002. A
Practical Architecture for Reliable Quantum Computers.
ACKNOWLEDGEMENT IEEE 2002. DOI:0018-9162/02.
[17] Rafael Trapani Possignolo and Cintia Borges Margi. 2012. A
The author would like to thank Dr. Sushil K. Prasad of Georgia Quantum-Classical Hybrid Architecture for Security
State University for giving inspiration and support throughout Algorithms Acceleration. 2012 IEEE 11th International
the duration of scripting the paper. Conference on Trust, Security and Privacy in Computing and
Communications.DOI:10.1109/TrustCom.2012.49
[18] Robert Raussendorf, Daniel E. Browne and Hans J. Briegel.
REFERENCES 2003. Measurement-based Quantum Computation on Cluster
[1] Bhaskar Roy Bardhan, Manish K. Gupta and Jonathon P. States. Phys. Rev. A 68, 022312
Dowling. 2011. Layered Architectural Design with Photonic [19] Eleanor Rieffel and Wolfgang Polak. 2000. An Introduction
Qubit Topological Cluster State Computation. OSA/CLEO to Quantum Computing for Non-Physicists. ACM Computing
2011. Optical Society of America. Surveys, Vol. 32, No. 3, September 2000, pp. 300-335.
[2] Anne Broadbent and Alain Tapp. 2008. Can Quantum [20] Michael Ross. 2008. Quantum Computing. Communications
Mechanics Help Distributed Computing. ACM SIGACT of the ACM, July 2008, Vol. 51, No. 7.
News, September 2008, Vol. 39, No. 3. DOI:10.1145/1364782.1364787
[3] Jilian Chen, Lingli Wang and Bin Wang. 2013. Quantum [21] Amr Sabry. 2003. Modeling Quantum Computing in Haskell.
FPGA Architecture Design. IEEE 2013. DOI:978-1-4799- Haskell’03. 2003 ACM 1-58113-758-3/03/0008
2198-0/13 [22] M. A. Nielson and I. L. Chuang. 2000. Quantum Computation
[4] Vasil S. Denchev and Gopal Pandurangan. 2008. Distributed and Quantum Information. Cambridge University Press.
Quantum Computing: A New Frontier in Distributed Systems [23] Peter Shor. 1994. Algorithms for Quantum Computation:
or Science Fiction? Purdue University. Discrete Logarithms and Factoring. Proceedings 35th Annual
[5] Simon J. Devitt, Austin G. Fowler, Ashley M. Stephens, Symposium on Foundations of Computer Science, 1994.
Andrew D. Greentree, Lloyd C.L. Hollenberg, William J. [24] Peter Shor. 1997. Polynomial-time Algorithms for Prime
Munro and Kae Nemoto. 2009. Architectural Design for a Factorization and Discrete Logarithms on a Quantum
Topological Cluster State Quantum Computer. DOI: Computer. SIAM J. Comput. 26, 1484-1509.
http://arXiv: 0808.1782v2 [quant-ph] 5 June 2009 [25] J.M. Taylor, H. –A. Engel, W. Dur, A. Yacoby, C. M.
[6] Simon Devitt and Kae Nemoto. 2012. Programming a Marcus, P. Zoller and M. D. Lukin. 2005. Fault-Tolerant
Topological Quantum Computer. 2012 IEEE 21st Asian Test Architecture for Quantum Computation Using Electrically
Symposium. DOI:10.1109/ATS.2012.52 Controlled Semiconductor Spins. Nature Physics, Vol. 1,
[7] M. Fujishima, K. Saito, M. Onouchi, and H. Hoh. 2003. High- December 2005.
Speed Processor For Quantum-Computing Emulation and its
Applications. School of Frontier Sciences, University of
Tokyo, CREST. IEEE 2003.
[8] Lov Grover. 1996. A Fast Quantum Mechanical Algorithm for
Database Search. ACM SOTC’96, pp212-219.
[9] Aram Harrow. 2012. Why Now is the Right Time to Study
Quantum Computing. XRDS, Spring 2012, Vol. 18, No. 3.
DOI:10.1145/2090276.2090288
[10] C. Horsman, Katherine L. Brown, Wiiliam J. Munro and

2015 2nd International Conference on Computing for Sustainable Global Development (INDIACom) 2169