Chapter

Practical Applications of
Quantum Computing
Esam El-Araby, Manu Chaudhary, Ishraq Ul Islam,
David Levy, Dylan Kneidel, Mingyoung Jeng, Alvir Nobel and
Vinayak Jha

Abstract

With the rapid advancement of quantum computing technology, there is a strong

motivation to explore suitable practical applications for quantum algorithms and
quantum computers. This chapter focuses on reviewing recent research on practical
applications of quantum computing, specifically dimension reduction, pattern recog-
nition, quantum sorting, and quantum communications for which optimized/modi-
fied versions of the quantum wavelet transform (QWT) and Grover’s algorithm are
presented. For practical implementations of the presented algorithms, this chapter
also includes methodologies for classical-to-quantum (C2Q) data encoding and
quantum-to-classical (Q2C) data decoding. Additionally, the chapter presents an effi-
cient quantum sorting technique by combining perfect-shuffle and bitonic networks.
In the domain of quantum communications, the chapter reviews a technique that
combines chaotic communications with quantum key distribution (QKD) to enhance
both security and communication range. The effectiveness of these techniques is
validated through practical results obtained from experiments conducted on IBM-Q
simulators and hardware, as well as on high-performance-reconfigurable-computers
(HPRCs). This chapter aims to provide readers with a comprehensive understanding
of these applications, covering the necessary prerequisites by showcasing the potential
of quantum computing in various domains for innovative problem-solving.

Keywords: quantum computing, quantum algorithms, quantum encoding, quantum

decoding, quantum communications

1. Introduction

Quantum computing is a field at the forefront of technological innovation. It

promises significant speedup over classical computation in certain workloads, largely
due to its inherent parallelism [1], enabled by quantum mechanical properties such as
superposition and entanglement [2]. State-of-the-art quantum processors, called
Noisy Intermediate-Scale Quantum (NISQ) devices, are limited by their relatively few
qubits and susceptibility to noise and decoherence effects [3]. Decoherence is the
process through which the state of a quantum computer is destroyed by unintended
1
Quantum Computers - From Fundamentals to Technology

interactions with the environment [4] and is, therefore, an especially critical challenge
for current NISQ devices. Moreover, quantum algorithms that require deep quantum
circuits take a longer time to run and exacerbate the impact of decoherence [5]. These
limitations make practical implementations of complex quantum algorithms quite
challenging.
This chapter sets out to present and review existing research efforts that demon-
strate some practical applications of quantum computing. On a high level, quantum
computations generally consist of three stages: (1) data input/encoding, (2) data
processing, and (3) data output/decoding. This chapter has thus been organized
roughly in that same flow. More specifically, Section 2 describes a technique for
encoding data into the quantum domain from the classical domain, termed as
classical-to-quantum (C2Q) data encoding, while Section 3 describes algorithm
development and optimization for quantum systems, and Section 4 discusses some
techniques for quantum-to-classical (Q2C) data decoding. In addition, Section 5
explores research in quantum communications that combines chaotic communica-
tions with quantum key distribution (QKD) for enhancing both security and commu-
nication range. Finally, Section 6 concludes the chapter with closing remarks.

2. Classical data encoding

Prior to executing quantum algorithms, particularly for data-intensive applica-

tions, data must be encoded from the classical domain into the quantum domain. The
process of initializing qubits with a quantum state corresponding to arbitrary classical
data is called data encoding, state preparation, and/or arbitrary state synthesis [6–8].
There are three main techniques for encoding classical data into quantum domain: (1)
basis encoding, (2) angle encoding, and (3) amplitude encoding [9]. Among the different
data encoding methods, amplitude encoding requires the least number of qubits [10].
This technique encodes data as the probability amplitudes (coefficients) of the basis
states into a quantum superposition. This section presents a depth optimized ampli-
tude encoding technique with up to 50% reduction in circuit depth and total gate
count compared with closest competing alternative [6]. This technique is termed as
classical-to-quantum (C2Q) data encoding [10].

2.1 Classical-to-quantum (C2Q) data encoding

For a classical dataset of N ¼ 2n elements, where n is the number of required

qubits, an n-qubit quantum state jψi can encode each data element as a probability
amplitude/coefficient Ci of a basis state jii. The C2Q method [11] can be used to
synthesize jψi from the ground state jψ 0 i ¼ j0i ⊗ n using a unitary operator U C2Q , see
Eq. (1).

jψi ¼ U C2Q  jψ 0 i ¼ U C2Q  j0i ⊗ n

X
N 1 X
N 1
(1)
¼ Ci jii, where jCi 2 j ¼ 1
i¼0 i¼0

The state jψi of a single qubit can be visualized using a Bloch sphere of global scale
r ¼ 1, global phase t, azimuth angle ϕ, and elevation angle θ. To initialize jψi from the
ground state j0i, the ZYZ or Pauli decomposition can be used to apply a rotation of
2
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

Figure 1.
Pauli (ZYZ) decomposition for single-qubit state synthesis [10].

angle t around the z-axis, followed by another rotation of angle θ around the y-axis,
and finally a rotation of angle ϕ around the z-axis, see Figure 1 and Eq. (2). Con-
versely, the parameters t, θ, and ϕ in Eq. (2) can be derived from the state coefficients
α and β using Eq. (3),
jψi ¼ C0 j0i þ C1 j1i ¼ αj0i þ βj1i
t
¼ Rz ðϕÞ  Ry ðθÞ  rei2  j0i (2)
¼ Rz ðϕÞ  Ry ðθÞ  Rz ðtÞ  r  j0i
 
∣β∣
t ¼ ∠β þ ∠α, θ ¼ 2 tan 1 , ϕ ¼ ∠β  ∠α, (3)
∣α∣
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where ∣α∣ ¼ Re 2 ðαÞ þ Im2 ðαÞ, ∠α ¼ cos 1 Re∣α∣ðαÞ , ∣β∣ ¼ Re 2 ðβÞ þ Im2 ðβÞ and
 
∠β ¼ cos 1 Re∣β∣ðβÞ .
To synthesize an arbitrary multi-qubit state from the ground state jψ 0 i ¼ j0i ⊗ n ,
an operation U C2Q
j , see Figure 2, can be used to iteratively synthesize the entangled
state of the j-th qubit in the output state, where 0 ≤ j < n. U C2Q
j is composed of kj ¼
2ðn1jÞ conditional Δi,j rotation operations, where 0 ≤ i < kj . Thus, U C2Q
j can be
represented by a block-diagonal matrix, where each diagonal block is a 2  2 trans-
formation matrix Δi,j , see Eq. (4). Block-diagonal matrices such as U C2Q j can be
implemented using a uniformly-controlled circuit or a quantum multiplexer [6]. In
this approach, the target qubit, i.e., the least-significant qubit, is subjected to various
gates or operations for every possible combination of the control qubits. In Figure 2,
the ‘square box’ notation [6] is used for the control bits, and the parameterized
operations on the data qubit are replaced by a single box denoting the operation. The
overall transformation U C2Q from the ground state jψ 0 i ¼ j0i ⊗ n to jψi can be
expressed by Eq. (5), also see Figure 3.

Figure 2.
Multiplexer (conditional-logic) quantum circuit [10].

3
Quantum Computers - From Fundamentals to Technology

Figure 3.
Quantum circuits for unitary C2Q data encoding [10].

U C2Q
j ¼ Δ0,j ⊕Δ1,j ⊕⋯Δi,j ⋯⊕Δðkj 1Þ,j
  (4)
¼ diag Δ0,j , Δ1,j , ⋯, Δi,j , ⋯, Δðkj 1Þ,j
!
Y
n1
C2Q C2Q ⊗j
U C2Q ¼ U 0 ðt, θ, ϕÞ  U j ðθÞ ⊗ I
j¼1 (5)

¼ U C2Q
0 ðt, θ, ϕÞ  U rem ðθÞ
C2Q

As shown in Figure 1, each Δi,j operation requires calculating a 2-tuple with

parameters (αi,j , βi,j ) from the given classical data set jψi. These parameters are
determined by following the steps detailed in Eq. (6) to Eq. (8). Subsequently, they
are utilized in Eq. (3) to derive the 4-tuple parameters (ri,j , ti,j , θi,j , ϕi,j ). It is important
to emphasize from Eq. (6) to Eq. (8), that the transformation retains its unitarity,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where ri,j ¼ jαi,j j2 þ jβi,j j2 ¼ 1.
8
>
> jC2i j2 þ jC2iþ1 j2 , j ¼ 0, 0 ≤ i < 2ðn1Þ
>
<
Pi,j ¼ P2i,j1 þ P2iþ1,j1 , 1 ≤ j < n, 0 ≤ i < 2ðn1jÞ (6)
>
>
>
:
0, 2ðn1jÞ ≤ i < 2ðn1Þ
8
> C2i
>
> pffiffiffiffiffiffi , Pi,j 6¼ 0, j ¼ 0, 0 ≤ i < 2ðn1Þ
>
> Pi,j
>
>
>
< sffiffiffiffiffiffiffiffiffiffiffiffi
αi,j ¼ P2i,j1 (7)
>
>
> , Pi,j 6¼ 0, 1 ≤ j < n, 0 ≤ i < 2ðn1jÞ
>
> Pi,j
>
>
>
:
1, Pi,j ¼ 0

4
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

8
> C2iþ1
>
> pffiffiffiffiffiffi , Pi,j 6¼ 0, j ¼ 0, 0 ≤ i < 2ðn1Þ
>
> Pi,j
>
< sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
βi,j ¼ P2iþ1,j1 (8)
>
>
> , Pi,j 6¼ 0, 1 ≤ j < n, 0 ≤ i < 2ðn1jÞ
>
> Pi,j
>
:
0, Pi,j ¼ 0

where 0 ≤ j < n, 0 ≤ i < kj , and kj ¼ 2ðn1jÞ : The derivations of circuit depths and
experimental results can be found in [10].

3. Algorithm development and optimization

This section introduces the readers to recent research work on some practical
applications of quantum computing. The techniques presented here are (1) dimension
reduction, (2) pattern recognition, and (3) quantum sorting.

3.1 Dimension reduction

The dimension reduction of multi-dimensional data is an area of great interest but

it presents a formidable computational challenge for current classical systems. For
example, wavelet-based techniques have proven effective for dimension reduction
[12, 13], but they are still computationally expensive. The quantum wavelet transform
(QWT), specifically the quantum Haar transform (QHT) [12, 14], offers a potential
solution for dimension reduction due to its inherent parallelism. Presented here are
multidimensional and multilevel-decomposable QHT techniques used for dimension
reduction in both packet and pyramidal form.

3.1.1 Quantum wavelet transform (QWT)

The wavelet transform (WT) decomposes signals/data into its spatio-temporal

spectral components [13]. The first and the simplest wavelet transform, called the
Haar wavelet transform [15], is constructed using a unit step function uðtÞ, see Eq. (9).
The discretized version of the Haar wavelet function, see Eq. (10), can also be
implemented in the quantum domain. The general discrete quantum wavelet trans-
form (QWT) [12] can be expressed by (12), where ΨD is the discretized Haar mother
wavelet function, Δt denotes the sampling period, K is the window size of the wavelet
in samples, N ¼ 2n is the total number of data samples expressed as quantum basis-
states, n represents the number of qubits, ∣ψi stands for the input state, and ∣ψiQWT
represents the output state.
 
1
ΨðtÞHaar ¼ uðtÞ  2u t  þ uðt  1Þ (9)
2
8
>
> N
> þ1, 0 ≤ i ≤
  > >
< 2
i
ΨD ¼ N (10)
N >
> 1, ≤i≤N
>
> 2
>
:
0, otherwise

5
Quantum Computers - From Fundamentals to Technology

X
N 1 X
N 1
∣ψi ¼ f ðq  ΔtÞ∣qi, where jf ðq  ΔtÞj2 ¼ 1 (11)
q¼0 q¼0
 
1 X N1 X
N1
qj
∣ψiQHT ¼ pffiffiffiffi f ðq  ΔtÞΨD ∣ji (12)
N j¼0 q¼0 K

3.1.2 Rotate-left (RoL) and rotate-right (RoR) operations

Quantum RoL and RoR gates [16] are specifically designed permutation operations
constructed using SWAP gates. Each gate executes a cyclic rotation, i.e., perfect-
shuffle, of input qubits, through a series of n  1 SWAP gates over n qubits, see
Figure 4.

3.1.3 QHT circuits

The QHT algorithm is multilevel-decomposable [16], represented by a generalized

d-dimensional operation, denoted as U dDQHT , see Figure 5a. When multi-
dimensional data is encoded into the state amplitudes, a subset of continuous ni qubits
is required to represent the ith dimension of the data, where 0 ≤ i < d. U dDQHT
performs a single level of decomposition over all d dimensions in parallel, see
Figure 5a. In this algorithm, a Hadamard (H) gate is applied to the least-significant
qubit of each dimension to retrieve the low and high frequency components from the
input data. Subsequently, an RoR operation is applied to separate the low-frequency
components from those with high-frequencies [18].
The depth δ of the U dDQHT operation is determined by the depth of the critical
path across all dimensions, see Eq. (13). The execution time t taken by U dDQHT to
run on a quantum hardware can be approximated by using the gate delay times of the
H and SWAP gates, denoted by τH and τSWAP , respectively. In terms of space com-
plexity metric, the total gate count γ can be expressed by Eq. (15), derived from
Figure 5a.

δ ¼ maxðf1 þ ðni  1Þ : i ∈ ℤ, 0 ≤ i < dgÞ ¼ nmax (13)

t ¼ τH þ ðδ  1Þ  τSWAP (14)

Figure 4.
Rotate-left (RoL) and rotate-right (RoR) gates [16].

6
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

Figure 5.
Circuits for QHT decomposition [17]. (a) Single-level decomposition of d-dimensional QHT. (b) ℓ-level,
d-dimensional packet decomposition.

X
d1
γ ¼ ð1 H  gate þ ðni  1ÞSWAP  gateÞ
i¼0 (15)
¼ d H  gate þ ðn  dÞSWAP  gate ¼ n gates

Interleaved packet decomposition: In the multilevel packet decomposition variant of

QHT, the U dDQHT operation is applied repeatedly to all qubits at each decomposi-
tion level, see Figure 5b. The equations used to determine the hardware execution
time and the circuit depth by applying U dDQHT operations in series are presented in
[16] and [18]. However, the circuit depth can be further minimized by the
overlapping the H gates and the SWAP gates across multiple decomposition levels.
This technique, known as interleaving the U dDQHT operations, allows these gates to
be executed concurrently, resulting in an overall reduction in circuit depth [17]. The
optimized packet decomposition circuit employs two extra layers of SWAP gates for
each added interleaved level of decomposition, as shown in Eqs. (16) and (17). The
expression for total gate count γ pkt for the multilevel packet decomposition QHT
circuit, see Eq. (18), is derived from Eq. (15) and Figure 5b.

δpkt ¼ nmax þ 2ðℓ  1Þ (16)

tpkt ¼ τH þ ðnmax þ ℓ  2Þ  τSWAP þ ðℓ  1Þ  maxðτH , τSWAP Þ
  (17)
¼ τH þ δpkt  1  τSWAP
γ pkt ¼nℓ (18)

Interleaved pyramidal decomposition: In pyramdial decomposition, the U dDQHT

operation is applied to a reduced number of data qubits (d fewer qubits), reducing by
one qubit for each dimension. Although this reduction in qubits can lead to shallower
circuit compared to packet decomposition, it necessitates additional interlevel
7
Quantum Computers - From Fundamentals to Technology

Figure 6.
ℓ-level, d-dimensional pyramidal decomposition [17]. (a) Structure of pyramidal decomposition. (b) Interlevel
permutations.

permutations to maintain data locality among the different decomposition levels, as

shown in Figure 6b.
The pyramidal decomposition operation also supports interleaving. When inter-
leaving is applied at the second decomposition level (ℓ ¼ 2), additional layers of gates,
i.e., n  nmax  d þ 2 layers, are appended to the first decomposition level. The first
level is composed of U dDQHT operation and interlevel permutations. On adding each
subsequent decomposition level, i.e., ℓ > 2, an extra d gate layers are added to the total
circuit depth. The overall depth of interleaved pyramidal decomposition is described
by Eq. (19) and the execution time is expressed by Eq. (20).
(
nmax , ℓ¼1
δpyr ¼ (19)
n þ dðℓ  1Þ  2ðd  1Þ, ℓ > 1
 
tpyr ¼ τH þ δpyr  1  τSWAP (20)

The total gate count, denoted as γ pyr , for the multilevel pyramidal decomposition is
determined using Figure 6 and is expressed by Eq. (21), where n0 denotes the number
of required qubits for the first dimension. However, the pyramidal structure reduces
P dℓðℓ1Þ
the gate count required for the packet decomposition by a factor ℓ1 i¼0 ðd  iÞ ¼ 2 ,
see Figure 6a, while requiring additional gates γ pyrperm , for interlevel permutations as
shown in Figure 6b and expressed by Eq. (22).

d  ℓ  ðℓ  1 Þ
γ pyr ¼ γ pkt þ γ pyrperm  , where (21)
2
 
d  ðℓ  1Þ ℓ  ðd  1Þ
γ pyrperm ¼ n  n0  (22)
2 2

Figure 7a and b show a ð64  64  3Þpixel input image and the corresponding
output image after 1-level of 3D-QHT, respectively. Figure 7c presents the output
8
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

Figure 7.
Dimension reduction on ð64  64  3Þ-pixel RGB images using 1-level 3D-QHT [16]. (a) Original RGB images.
(b) Output images from MATLAB simulations. (c) Output images from IBM-Q simulations.

image reconstructed from IBM-Q simulations of 1-level 3D-QHT. For comprehensive

experimental evaluations, please refer to [16].

3.2 Pattern recognition

Pattern recognition on multi-dimensional data is a challenging problem to

solve using classical computers. In domains dealing with multi-dimensional data, not all
measured components are relevant for detecting the area of interest. An effective pre-
processing method would involve techniques of dimension reduction [19] of the data
for faster processing and matching. Wavelet-based dimension reduction techniques
[13, 15] are effective in data pre-processing, reducing computation overhead, and
improving classification accuracy. This section presents a pattern recognition technique
that uses optimized versions of QWT and quantum Grover’s search (QGS) for time-
efficient single-pattern/multi-pattern search in high-dimensional datasets.

3.2.1 Pattern recognition using quantum dimension reduction

The methodology of pattern recognition using multiple decomposition levels of

2D-QHT as a pre-processing step, see Figure 8a, is implemented as cascaded packet
wavelet transforms, see also Section 3.1.1. A pattern matching search is then performed
on the dataset with low spatial resolution using multi-pattern Grover’s search. In this
approach, the input classical data is encoded into the N basis states of a superimposed
quantum state [11] utilizing a set of n ¼ log 2 N qubits. Using L decomposition levels
of 2D-QHT, the input dataset is converted to a low spatial resolution data consisting of

Figure 8.
Pattern recognition using quantum dimension reduction [20]. (a) Overview of methodology for pattern recognition
using dimension reduction [20]. (b) Pattern recognition using quantum dimension reduction on a (512  512)-
pixel image.

9
Quantum Computers - From Fundamentals to Technology

N r quantum basis states. Here, L ¼ ⌊21 log 2 NNr ⌋, N r is pre-determined quantum data
resolution less than N. The N r reduced data is represented with nr qubits, on which a
search to find a number of patterns/basis states N patterns is performed. Thus, the
multilevel 2D-QHT reduces the dimensionality of the data and the multipattern QGS
returns the indices of the searched pattern, see Figure 8b. The use of this technique
yields faster results compared to classical dimension reduction and search methods [20].

3.2.2 Quantum Grover’s search algorithm

The quantum Grover’s search algorithm is used to find an element s ∗ in an

unordered N element set S ¼ fs1 , s2 , s3 , … , sN g, that makes f ðs ∗ Þ ¼ 1. Here, N is the
cardinality of S, and f is a boolean function such that f ðxÞ ! f0, 1g. A quantum
computer running Grover’s algorithm can provide a quadratic speedup [21] compared
to classical counterparts.
Grover’s oracle and Grover’s diffusion: In QGS algorithm, in order to search an
element in an unordered set, the essential steps involve performing the operations of
phase inversion (oracle) and inversion about the mean (diffusion). These operations are
performed for an optimal number of iterations. The oracle operation flips the sign of
the target pattern’s state while leaving the other states unchanged. It is represented as
U oracle ∣xi ¼ ð1Þf ðxÞ ∣xi. The diffusion operation amplifies any inverted coefficients and
attenuates other coefficients [22]. This operation is represented as
U diffusion ¼ I  2∣xihx∣. The U diffusion is performed by calculating the mean of all the
amplitudes, followed by inverting the value of each amplitude with respect to the
mean. This results in reducing the amplitudes that are larger than the mean while
amplifying the ones that are smaller [11].

3.2.3 Multi-pattern quantum search

The design and methodology of the modified/optimized multi-pattern Grover’s

search algorithm is shown in Figure 9. The first key modification involves the addi-
tion of a dynamic oracle circuit denoted as U oracle , which efficiently identifies items

Figure 9.
Modified quantum circuit for multi-pattern Grover’s algorithm [23].

10
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

positioned at the first N patterns indices in the search list. This is followed by the
application of a diffusion circuit U diffusion , which increases the probabilities of locating
the pattern(s). The second modification incorporates a permutation operation,
referred to as U permute , which utilizes ancilla qubits for encoding and assigning proba-
bility coefficients to the corresponding basis states. This step is critical for successfully
locating the target pattern(s).
The modified quantum circuit for multi-pattern Grover’s algorithm, see Figure 9
has four inputs:

• a collection of n compute qubits that are all initialized to the state ∣ψi ¼ ∣0i,

• a flag ancilla qubit, configured to the ground state ∣0i,

• a collection of N patterns entries of statically initialized n ancilla qubits, i.e.,

S ¼ j0i, j1i, … , jN patterns1 i , and

• a collection of N patterns entries consisting of n ancilla qubits that can be

dynamically changed. These qubits represent n the input patternsothat needs to be
sought and amplified in ∣ψ in i, i.e., P ¼ jP0 i, jP1 i, … , jPN patterns1 i .

After applying the H gate, the input qubit state vector ∣ψ in i is expressed as
PN1
∣ψ in i ¼ p1ffiffiffi
N i¼0 ∣ii. After putting the input qubits in superposition, a modified
dynamic oracle operator U oracle and an unmodified diffusion operator U diffusion are
applied m times to amplify the first N patterns states. The optimal number of iterations,
m is determined using Eq. (23) where k ¼ 1,3,5,7, … is an odd number. Since, only the
initial N patterns are amplified, a permutation operation, denoted as U permute is required
to assign higher amplitudes to the target states based on input patterns P. The final
output state is derived by performing these repeated iterations and the complete
process can be represented by a single unitary matrix, denoted as
 m
U G ¼ U diffusion  U oracle . The probability Psuccess of successfully finding a desired pat-
tern in the final output state ∣ψ out i is expressed in Eq. (24) [24], where N patterns ≤ N.
6 7
6 7
6 7
6 πk 7
m¼6
4 qffiffiffiffiffiffiffiffiffiffiffiffi 7
5 (23)
N patterns
4 sin 1 N

Psuccess ¼ sin 2 ðð2m þ 1Þ  θÞ,

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
1 N patterns π (24)
θ ¼ sin , and 0 < θ ≤
N 2

Modified oracle and diffusion circuits: The modified oracle circuit U oracle uses CNOT
gates to dynamically modify the target patterns as seen in Figure 10a and b.
This dynamic search pattern modification extends the algorithm’s capabilities to
search for any pattern using a single quantum circuit. The CNOT gates within each
oracle are efficiently controlled using ancilla qubits by adjusting them to the current
pattern ∣ii, see Figure 10a. For multi-pattern search, cascaded and incremental single-

11
Quantum Computers - From Fundamentals to Technology

Figure 10.
Modified Grover’s oracle for single and multiple solutions/pattern [23]. (a) Single solution/pattern. (b) Multiple
solutions/patterns.

pattern oracle quantum circuits are applied to invert the first N patterns amplitudes as
depicted in Figure 10b. The output of the oracle is provided to the diffusion circuit for
amplification. The U oracle and U diffusion are then iterated over m times to maximize the
target amplitudes.
Quantum state permutation: In the modified design of Grover’s algorithm, the
amplitudes of the first N patterns states are amplified, and an additional permutation step
assigns these amplitudes to the target basis states within the output state ∣ψ out i. For a
detailed study of the permutation circuit, please refer to [23].

3.3 Quantum sorting

Another interesting application of quantum computing is sorting. This section

presents the quantum implementation of the bitonic sort [25]. The bitonic sort is a
comparison-based sorting algorithm that leverages bitonic sequences and is suitable for
parallel implementations [26]. For an unordered sequence of N elements, it has a
 
spatio-temporal complexity of O Nlog 4 N [26]. However, the optimized
implementations using the perfect shuffle technique have shown to have a spatio-
 
temporal complexity of O Nlog 2 N on a sequential processor and a complexity of
 2 
O log N on a system with N parallel processors [26]. When implemented on a
 
quantum computer, the complexity improves to O nlog 2 n [25], where n ¼ log 2 N
and n is the number of qubits.

3.3.1 Bitonic sort and perfect shuffle

Bitonic sequences [27] are sequences of elements which first increase monotonically
in value and then decrease monotonically. A bitonic sequence can be expressed as
x0 ≤ ⋯ ≤ xi ≥ ⋯xn1 where 0 ≤ i < n. Sequences sorted in ascending order or
descending order are also considered bitonic sequences with zero elements on the
descending side and ascending side respectively.
The classical version of the algorithm works by taking in an unsorted input
sequence and recursively creating and merging smaller bitonic sequences into larger
sequences until the whole list is eventually sorted. The bitonic sort algorithm is
modeled as a sorting network [27] which uses comparator circuits that compare two
input elements ðx0 , x1 Þ and produces an output sequence ðmaxðx0 , x1 Þ, minðx0 , x1 ÞÞ or
ðminðx0 , x1 Þ, maxðx0 , x1 ÞÞ based on how the comparator is configured.
A quantum implementation of the comparator [25] takes two qubits, ∣q0 i and ∣q1 i,
 
and a mode qubit as inputs and if the mode qubit is ∣0i sets ∣q0 i ¼ min q0 , q1 and
12
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033
   
∣q1 i ¼ max q0 , q1 . Similarly, if the mode qubit is set to ∣1i, it sets ∣q0 i ¼ max q0 , q1
 
and ∣q1 i ¼ min q0 , q1 . An implementation of such a comparator circuit is provided in
Figure 11.
The perfect shuffle [28] technique improves the temporal complexity of a bitonic
sort, which helps against the decoherence constraints on a quantum computer. A
perfect shuffle intersperses elements of two equal length sequences, so that each
element in the first group is followed by the corresponding element at the same
position in the second group. The quantum perfect shuffle (QPS) [25] is a quantum
implementation of a perfect shuffle that leverages the fact that a cyclical left shift on
the indexes of a sequence of elements, expressed in their binary form, performs the
equivalent of a perfect shuffle on the elements of that sequence. The quantum perfect
shuffle can be implemented by a series of SWAP gates, identical to the RoL circuit
shown in Figure 4.

3.3.2 Quantum bitonic sort with perfect shuffle

Data is encoded into a quantum circuit with n qubits using amplitude encoding
[11] and follows Algorithm 1.1. The algorithm operates in m stages, where
m ¼ log 2 ðnÞ. At each stage t, m  t QPS operations are performed, followed by t QPS-
comparison pairs, see Figure 12. A QPS-comparison pair is a QPS operation followed
by a comparison operation, which is the application of a set of comparators between
two adjacent qubits, see Figure 11.
For t ¼ 1, the comparators are applied with their modes following an alternating
min-max (mode 0) and max-min (mode 1) pattern, which can be represented as
½0,1,0,1, ⋯. For each subsequent stage 1 < t < m, the pattern begins as ½0,1,0,1, ⋯ for
the first QPS-comparison pair and is quantum perfect-shuffled before each of the
following QPS-comparison pairs in the same stage. Finally, when t ¼ m, the compar-
ator pattern is ½0,0,0,0, ⋯ and no perfect shuffle of the pattern is necessary. The

Figure 11.
Quantum comparator circuits [25]. (a) Quantum Comparator Circuit, Mode 0. (b) Quantum Comparator
Circuit, Mode 1.

Figure 12.
Quantum circuit for combined bitonic sort with perfect shuffle permutation [25].

13
Quantum Computers - From Fundamentals to Technology

Figure 13.
Example of an 8-qubit quantum bitonic sorter [25].

general structure of the algorithm is provided in Figure 12 and an example circuit for
an 8-qubit bitonic sorter is given in Figure 13.

Algorithm 1.1: Bitonic sort with perfect shuffle.

1 for t in range (1, m):
2 mode(t)
3 for i in range (1, m  t):
4 QPS(qubits)
5 end for
6 for i in range (m  t + 1,m):
7 QPS(qubits)
8 comparator (qubits, mode)
9 QPS(mode)
10 end for
11 end for

4. Quantum-to-classical (Q2C) data decoding

The process of decoding classical information from a quantum computer output

can be a complex and time-consuming process. This critical challenge [3] limits the
practical applications of existing Noisy Intermediate-Scale Quantum (NISQ) devices
[29]. For example, in applications like quantum image processing data is usually
represented using state amplitudes [30]. In such cases, to retrieve the processed image
data from its quantum representation, it is necessary to sample the quantum circuit
repeatedly, creating a probability distribution [31]. This operation, known as
quantum-to-classical (Q2C) data decoding, introduces a substantial time overhead
during circuit execution, promoting the need for further exploration of time-efficient
methods for data decoding.
Traditional data decoding methods, as depicted in Figure 14a acquire the entire
quantum state by repeated circuit sampling, often referred to as circuit ‘shots’. In
14
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

Figure 14.
Q2C data decoding using conventional and QHT-based methods [17]. (a) Conventional Q2C data decoding. (b)
QHT-based Q2C data decoding.

general, a substantial number of repeated circuit samples are required to enhance the
measurement accuracy and to reduce the impact of statistical noise. Consequently,
this significantly increases the total time required for circuit execution.
To minimize the overhead of repeated circuit sampling, algorithms which
decrease either the number of measured qubits or the number of required shots
can be added to the circuit immediately prior to measurement. One effective
technique for reducing the number of measured qubits is the multilevel-decomposable
QHT [17]. By using this approach, data initially represented by n qubits can be
transformed into a representation using fewer qubits, specifically k ¼ n  ðℓ  dÞ,
where 0 ≤ k ≤ n, 0 ≤ ℓ ≤ ðn=dÞ represents the number of decomposition levels, and

Figure 15.
ℓ-level, d-dimensional measurement-based decomposition [17]. (a) Single-gate depth with H gates.
(b) Zero-depth circuit.

15
Quantum Computers - From Fundamentals to Technology

d ≥ 1 is the data dimensionality, see Figure 14b. Another depth-optimized method,

known as the ‘measurement-based’ QHT decomposition, operates effectively without
using the additional quantum circuit [17]. In this method, the measurement of selec-
tively lower number of qubits enables the sampling of output data within a lower-
dimensional space. The two QHT-based circuits which can be used for Q2C data-
decoding are multilevel packet and multilevel pyramidal decomposition circuits of
QHT, see Figures 5 and 6 and Section 3:1:3 for details. The ‘measurement-based’
QHT decomposition circuit is shown in Figure 15. Please refer to [17] for a detailed
study of different Q2C circuits, their mathematical expressions and experimental
evaluations.

5. Chaotic and quantum communications

In today’s interconnected world, ensuring the security of transmitted data remains

a challenging problem. Current communication systems rely on classical crypto-
graphic methods to safeguard data privacy. However, these approaches are suscepti-
ble to potential threats posed by future quantum computers capable of breaking
cryptographic algorithms. This concern has spurred significant interest in exploring
quantum communication solutions. This section presents a Free-space optical (FSO)
communication scheme that combines chaotic communication and Quantum Key
Distribution (QKD) to enhance both range and security compared to currently used
FSO methods. This approach employs auto-synchronizable Lorenz chaotic transmitter
and receiver models to produce chaotic signals as data carriers. Utilizing the chaotic
communication techniques, the data is sent securely over the classical channel, while
QKD ensures a secure exchange of critical parameters used for synchronization over
the quantum channel. This section illustrates the combination of chaotic
communication and QKD to establish an end-to-end encrypted Deep-Space optical
communications link.

5.1 Chaotic communications

In conventional communication systems, a carrier signal is required to carry infor-

mation. For successful communication, the transmitter and receiver need to be syn-
chronized. However, chaotic communication systems use a different approach.
Instead of using a periodic carrier signal, they use chaotic signals. These systems can
self-synchronize when driven by a common signal [32]. One example of such a chaotic
system is the Lorenz chaotic system [33], represented as the Lorenz attractor. It can be
split into two stable response subsystems, and both subsystems can synchronize with
the original system, facilitating robust and secure communication.

5.2 Quantum key distribution (QKD)

QKD is a highly secure method of distributing cryptographic keys between

two parties using the principles of quantum mechanics. Unlike conventional crypto-
graphic methods that rely on the complexity of algorithms, QKD’s security is based on
fundamental laws of physics. QKD works by encoding a private key as a quantum
state (qubits) and sending it over a quantum channel [34]. If an eavesdropper tries to
intercept the qubits, it will alter their states due to the fundamental characteristics of
quantum mechanics [34]. The BB84 protocol [35], introduced by Bennett and
16
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

Brassard in 1984, is one of the earliest and most widely used QKD protocols. It uses
photons’ polarization properties to transmit the key information, establishing a secure
key between two communicators.

5.3 Communication system combining chaotic systems with QKD

The communication system presented consists of a transmitter (TX) and a receiver

(RX), communicating through classical and quantum channels. The process begins
when the QKD transmitter (TX) and receiver (RX) exchange chaotic synchronization
parameters using a two-way BB84-like protocol employing both quantum and classi-
cal channels. The QKD RX model retrieves these chaotic parameters and provides
them to the Lorenz chaotic receiver, as shown in Figure 16. The system uses digital
modulation, error correction, and decoding techniques to achieve reliable communi-
cation over the Additive White Gaussian Noise (AWGN) channel. The hardware
circuits of the transmitter and receiver are presented in Figure 17. For details about
the preshaped codebook used for encrypting/decrypting data, see Figure 18 and refer
to [36]. In the study of cryptographic methods for secure communication, some

Figure 16.
Chaotic communication system secured by QKD [36].

Figure 17.
The communication scheme’s transmitter and receiver hardware models [36]. (a) Chaotic transmitter model. (b)
Chaotic receiver model.

17
Quantum Computers - From Fundamentals to Technology

Figure 18.
Preshared codebook for encrypting/decrypting data [36].

Figure 19.
Transmission of a ð512  512Þ-pixel image between Alice and Bob at SNR ¼ 0:1dB, with interception attempted
by Eve [36]. (a) Original image transmitted by Alice using parameters σ = 10, ρ = 54, β = 4. (b) Reconstructed
image by Bob using recovered parameters σ = 10, ρ = 54, β = 4 and 0% pixel error. (c) Reconstructed image by Eve
using incorrect parameters σ = 10, ρ = 45.6, β = 14 and 98.4652% pixel error.

central characters are defined as Alice (the sender or source of information), Bob
(intended recipient), and Eve (eavesdropper seeking unauthorized access to trans-
mitted data). The effectiveness of this technique is illustrated by the practical exper-
imental results shown in Figure 19, please refer to [16] for complete results.
Figure 19b shows the grayscale image of size ð512  512Þ pixels transmitted by Alice
and reconstructed by Bob showing 0% error in pixels between the original and
reconstructed image. However, when Eve tries to reconstruct the image without
knowing the choatic parameters quantumly shared by Alice and Bob, it results in
98.46% pixel error, see Figure 19c.

6. Conclusions

The rapid advancement of quantum computing technology has opened up a realm

of possibilities for practical applications of quantum computing. Throughout this
chapter, a spectrum of these applications, including dimension reduction, pattern
recognition, quantum sorting, and quantum communications were explored. This
18
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

chapter presented optimized circuits for QWT to achieve efficient dimension reduc-
tion in multi-dimensional high-resolution data and an innovative approach for pattern
recognition using QWT and Grover’s search. For practical implementations of these
algorithms, methodologies for classical-to-quantum (C2Q) data encoding and
quantum-to-classical (Q2C) data decoding were also presented. This chapter also
presented an efficient quantum sorting technique, which amalgamates perfect-shuffle
and bitonic networks. Finally, in the field of quantum communications, a novel free-
space optical (FSO) communication system that combines chaotic communications
with Quantum Key Distribution (QKD) was discussed focusing on enhancing security
and extending the communication range. This chapter serves as an invitation to
readers to explore further, delve deeper, and engage with the ongoing journey of
quantum computing’s practical impact on various domains.

Acknowledgements

This research used resources of the Oak Ridge Leadership Computing Facility,
which is a DOE Office of Science User Facility supported under Contract DE-
AC05-00OR22725.

Nomenclature and Abbreviations

QWT quantum wavelet transform

QHT quantum haar transform
C2Q Classical-to-quantum
Q2C quantum-to-classical
QKD quantum-key-distribution
HPRC high-performance-reconfigurable computers
NISQ Noisy Intermediate Scale Quantum
QGS multi-pattern Grover’s search

Author details

Esam El-Araby*†, Manu Chaudhary†, Ishraq Ul Islam†, David Levy†, Dylan Kneidel†,
Mingyoung Jeng†, Alvir Nobel† and Vinayak Jha†
The University of Kansas, Lawrence, USA

*Address all correspondence to: [email protected]

† These authors contributed equally.

© 2023 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of
the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided
the original work is properly cited.
19
Quantum Computers - From Fundamentals to Technology

References

[1] Shor PW. Algorithms for quantum the 27th Conference on Pattern Languages
computation: Discrete logarithms and of Programs (online). 2020. pp. 1-11
factoring. In: Proceedings 35th Annual
Symposium on Foundations of [10] El-Araby E, Mahmud N, Jeng MJ,
Computer Science. New Mexico, USA: MacGillivray A, Manu Chaudhary M,
IEEE; 1994. pp. 124-134 Nobel AI, et al. Towards complete and
scalable emulation of quantum
[2] Bennett CH, DiVincenzo DP. algorithms on high-performance
Quantum information and computation. reconfigurable computers. IEEE
Nature. 2000;404(6775):247-255 Transactions on Computers. 2023;72(8):
2350-2364
[3] Preskill J. Quantum computing in the
nisq era and beyond. Quantum. 2018;2: [11] Williams CP. Explorations in
79 Quantum Computing. Berlin, Germany:
Springer Science & Business Media; 2010
[4] Schlosshauer M. Quantum
decoherence. Physics Reports. 2019;831:
[12] Mahmud N, Haase-Divine B,
1-57
MacGillivray A, El-Araby E. Quantum
dimension reduction for pattern
[5] Zhang C, Chen Y, Jin Y, Ahn W,
recognition in high-resolution spatio-
Zhang Y, Zhang EZ. A depth-aware
spectral data. IEEE Transactions on
swap insertion scheme for the qubit
Computers. 2020;71(1):1-12
mapping problem. arXiv preprint arXiv:
2002.07289. 2020
[13] El-Araby E, El-Ghazawi T, Le
Moigne J, Gaj K. Wavelet spectral
[6] Shende VV, Bullock SS, Markov IL.
dimension reduction of hyperspectral
Synthesis of quantum-logic circuits. imagery on a reconfigurable computer.
IEEE Transactions on Computer-Aided In: Proceedings. 2004 IEEE International
Design of Integrated Circuits and Conference on Field-Programmable
Systems. 2006;25(6):1000-1010 Technology (IEEE Cat. No. 04EX921).
Brisbane, Australia: IEEE; 2004.
[7] Mottonen M, Vartiainen JJ, pp. 399-402
Bergholm V, Salomaa MM.
Transformation of quantum states using [14] Li H-S, Fan P, Xia H-y, Song S, He X.
uniformly controlled rotations. arXiv The multi-level and multi-dimensional
preprint quant-ph/0407010. 2004 quantum wavelet packet transforms.
Scientific Reports. 2018;8(1):1-23
[8] Niemann P, Datta R, Wille R. Logic
synthesis for quantum state generation. [15] Wickmann JMG. A Wavelet
In: 2016 IEEE 46th International Approach to Dimension Reduction and
Symposium on Multiple-Valued Logic Classification of Hyperspectral Data.
(ISMVL). Sapporo, Japan: IEEE; 2016. [Thesis]. Faculty of Mathematics and
pp. 247-252 Natural Sciences, University of Oslo.
2007
[9] Weigold M, Barzen J, Leymann F,
Salm M. Data encoding patterns for [16] Mahmud N, Macgillivray A,
quantum computing. In: Proceedings of Chaudhary M, El-Araby E.
20
Practical Applications of Quantum Computing
DOI: http://dx.doi.org/10.5772/intechopen.1003033

Decoherence-optimized circuits for [23] Mahmud N, Haase-Divine B,

multi-dimensional and multi-level MacGillivray A, Srimoungchanh B,
decomposable quantum wavelet Kuhnke A, Blankenau N, et al. Modifying
transform. IEEE Internet Computing quantum grover’s algorithm for dynamic
(IEEE IC) Special Issue on Quantum and multi-pattern search on reconfigurable
Post-Moore’s Law Computing. 2022; hardware. Journal of Computational
26(1):15-25 Electronics. 2020;19:1215-1231

[17] Mingyoung Jeng SM, Islam DL, [24] Boyer M, Brassard G, Høyer P,
Riachi A, Manu Chaudhary M, Nobel AI, Tapp A. Tight bounds on quantum
Kneidel D, et al. Improving quantum-to- searching. Fortschritte der Physik:
classical data decoding using optimized Progress of Physics. 1998;46(4–5):493-505
quantum wavelet transform. The
Journal of Supercomputing. 2023;2023: [25] Mahmud N, Srimoungchanh B,
1-30 Haase-Divine B, Blankenau N,
Kuhnke A, El-Araby E. Combining
[18] Mahmud N, Jeng MJ, et al. Time- perfect shuffle and bitonic networks for
efficient quantum-to-classical data efficient quantum sorting. In: 2019 IEEE/
decoding. In: The International ACM International Workshop on
Conference on Emergent Quantum Heterogeneous High-Performance
Technologies (ICEQT 2022). To Reconfigurable Computing (H2RC).
appear in Transactions on Denver, CO, USA: IEEE; 2019. pp. 42-49
Computational Science &
Computational Intelligence. Las Vegas, [26] Akl SG. Parallel Sorting Algorithms.
Nevada, USA: Springer Nature – Vol. 12. Cambridge, Massachusetts:
Research Book Series; 2022 Academic Press; 2014

[19] Imola K. A Survey of Dimension [27] Kenneth E. Sorting networks and

Reduction Techniques. California: their applications. In: Proceedings of the
Lawrence Livermore National Lab; 2002 April 30–May 2, 1968. Spring Joint
Computer Conference. New York, NY,
[20] Mahmud N, El-Araby E. Dimension USA. 1968. pp. 307-314
reduction for efficient pattern
recognition in high spatial resolution [28] Stone HS. Parallel processing with
data using quantum algorithms. In: 2019 the perfect shuffle. IEEE Transactions on
32nd IEEE International System-on-Chip Computers. 1971;100(2):153-161
Conference (SOCC). California, USA:
IEEE; 2019. pp. 126-131 [29] Guan W, Perdue G, Pesah A,
Schuld M, Terashi K, Vallecorsa S, et al.
[21] Lov K. A fast quantum mechanical Quantum machine learning in high
algorithm for database search. In: energy physics. Machine Learning:
Proceedings of the Twenty-eighth Science and Technology. 2021;2(1):
Annual ACM Symposium on Theory of 011003
Computing. Pennsylvania, USA. 1996.
pp. 212-219 [30] Weigold M, Barzen J, Leymann F,
Salm M. Data encoding patterns for
[22] Yanofsky NS, Mannucci MA. quantum computing. In: HILLSIDE
Quantum Computing for Computer Proceedings of Conference on Pattern
Scientists. Cambridge, England: Languages of Programs’22. Portland,
Cambridge University Press; 2008 Oregon, USA. 2020
21
Quantum Computers - From Fundamentals to Technology

[31] Lanzagorta M, Uhlmann J. Is

quantum parallelism real? In: Quantum
Information and Computation VI. Vol.
6976. Bellingham, WA, USA:
International Society for Optics and
Photonics; 2008. p. 69760W

[32] Pecora LM, Carroll TL.

Synchronization in chaotic systems.
Physical Review Letters. 1990;64(8):821

[33] Lorenz EN. Deterministic

nonperiodic flow. Journal of
Atmospheric Sciences. 1963;20(2):
130-141

[34] Shor PW, Preskill J. Simple proof of

security of the bb84 quantum key
distribution protocol. Physical Review
Letters. 2000;85(2):441

[35] Bennett CH, Brassard G. Quantum

cryptography: Public key distribution
and coin tossing. arXiv preprint arXiv:
2003.06557. 2020

[36] Mahmud N, MacGillivray A, Rai A,

Patterson J, Gharaibeh A, El-Araby E,
et al. Combining quantum key
distribution with chaotic systems for
free-space optical communications.
Quantum Information Processing. 2021;
20:1-25

22