IEEE ICC 2017 Wireless Communications Symposium
Outage Performance of Full/Half-Duplex User
Relaying in NOMA Systems
Xinwei Yue∗ , Yuanwei Liu† , Shaoli Kang‡ , Arumugam Nallanathan† , and Zhiguo Ding§
∗ Beihang
University, Beijing, China
†
King’s College London, London, UK
‡ China Academy of Telecommunication Technology, Beijing, China
§ Lancaster University, UK
Abstract—This paper investigates the performance of coopera-
tive non-orthogonal multiple access (NOMA) systems, where one
near user works as a decode-and-forward (DF) full-duplex (FD)
or half-duplex (HD) relaying to help far user. Two cooperative
relay scenarios are considered insightfully. 1) The first scenario is
that no direct link exists between the base station (BS) and the far
user; and 2) The second scenario is that direct link exists between
the BS and the far user. To characterize the performance of FD
NOMA in two considered scenarios, new closed-form expressions
for both exact and asymptotic outage probability as well as delay-
limited throughput are derived for each NOMA user. Based on
the analytical results derived, the diversity orders achieved by
users are obtained. It is confirmed that the use of the direct link
overcomes the zero diversity order of far NOMA user inherent
to FD relaying. Simulation results demonstrate that the outage
performance of FD NOMA is superior to HD NOMA at low SNR
region rather than at high SNR region.
I. I NTRODUCTION
To achieve higher spectral efficiency of the fifth generation
(5G) mobile communication network, non-orthogonal multiple
access (NOMA) has received a great deal of attention [1].
The key characteristics of NOMA is to allow multiple users
to share the same resource elements (i.e., time/fequency/code)
via different power levels. At the receiver side, the succes-
sive interference cancellation (SIC) is carried out. In [2], the
influence of user pairing with the fixed power allocation for
NOMA system was studied in detail. Cooperative communica-
tion is an effective technique for either extending the coverage
of networks or improving the communication reliability [3].
Current NOMA research contributions in terms of cooperative
communication are mainly researched in [4–8]. The authors in
[4] investigated the outage probability of decode-and-forward
(DF) relaying for NOMA. In [5], outage behavior of a variable
gain amplify-and-forward (AF) relaying with NOMA has been
discussed over Nakagami-m fading channels. Furthermore,
cooperative NOMA concept was first proposed in [6], in which
the near users with better channel conditions was regarded as
DF relaying to help the far users. As a further development
with considering energy related issue, simultaneous wireless
information and power transfer (SWIPT) was employed at
near NOMA user which was regarded as DF relaying in [7].
Additionally, the outage performance of NOMA in large-scale
underlay cognitive radio was evaluated by using stochastic-
geometry in [8].
978-1-4673-8999-0/17/$31.00 ©2017 IEEE
Although cooperative NOMA is capable of enhancing per-
formance gains for weak user, but it brings additional slot cost
for the systems. To avoid this issue, one promising solution
is to adopt full-duplex (FD) relay technology. FD relaying
receives and transmits simultaneously in the same frequency
band and time slot has attracted the interest of researchers to
explore more spectral efficient systems [9]. Recently, FD relay
technology has been proposed as a promising technique for
the 5G networks in [10]. In [11], the authors investigated the
performance of FD device-to-device based cooperative NOMA.
However, only the outage performance of weak user was
analyzed. In [6], it is demonstrated that the maximum diversity
order can be obtained for all users, but only cooperative NOMA
with direct link was considered with HD operation mode.
With the above works inspiration, we focus on the outage
performance of FD/HD user relaying in NOMA systems in this
paper. We present a compressive investigation on adopt the user
with strong channel consider as a FD/HD user relaying to keep
connectivity and improve the reliability of weak user. We derive
the closed-form expressions of outage probability for near user
and far user. For obtaining more insights, we further derive the
asymptotic outage probability of two users and the diversity
orders are obtained. We demonstrate that FD NOMA converges
to an error floor and obtain a zero diversity order. Furthermore,
we conclude that the use of the direct link overcomes the zero
diversity order of far NOMA user inherent to conventional FD
relaying. In addition, we analyze the delay-limited transmission
throughput according to the derived outage probability.
II. S YSTEM M ODEL
We consider a FD cooperative NOMA system composed of
one source (i.e the base station (BS)) that intends to communi-
cation with the far user D2 under the assistance of the near user
D1 . Both no direct link and direct link scenarios between the
BS and D2 are considered. D1 is regarded as user relaying and
DF protocol is employed to decode and forward the information
to D2 . To enable FD communication, D1 is equipped with
one transmit antenna and one receive antenna, while the BS
and D2 are single-antenna device. All wireless links in the
network are assumed to be independent non-selective block
Rayleigh fading and are disturbed by additive white Gaussian
noise with mean power N0 . h1 , h2 , and h0 are denoted as
the complex channel coefficient of BS → D1 , D1 → D2 ,
 IEEE ICC 2017 Wireless Communications Symposium

and BS → D2 links, respectively. The channel power gains III. P ERFORMANCE EVALUATION
|h1 |2 , |h2 |2 and |h0 |2 are subjected to exponentially distributed When the target rate of users is determined by its quality of
random variables (RVs) with the parameters Ωi , i ∈ {1, 2, 0}, service (QoS), the outage probability is an important metric for
respectively. The residual loop self-interference (LI) is modeled performance evaluation.
as a Rayleigh fading feedback channel with coefficient hLI , and
ΩLI is the corresponding average power. A. User Relaying without Direct Link
According to [12], D1 receives the superposed signal and In this subsection, the first scenario is investigated in terms
loop interference signal simultaneously. The observation at D1 of outage probability.
can be given by 1) Outage Probability of D1 : According to the NOMA
  
yD1 = h1 ( a1 Ps x1 + a2 Ps x2 ) + hLI Pr xLI + nD1 , protocol, the complementary events of outage at D1 can be
(1) explained as: D1 can detect x2 as well as its own message x1 .
where  is the switching operation factor between HD mode From the above description, the outage probability of D1 can
and FD mode.  = 0 and  = 1 denote D1 working in be expressed as below:
HD and FD mode, respectively. Ps and Pr are the normalized  
PDFD
1 ,nodir
= 1 − Pr γD2 →D1 > γth FD
2
FD
, γD1 > γth 1
, (7)
transmission powers at the BS and D1 , respectively; a1 and a2
are the power allocation coefficient and x1 and x2 are the signal where  = 1. γthFD
1
= 2R1 − 1 with R1 being the target rate at
of D1 and D2 , respectively; xLI denotes loop interference D1 to detect x1 and γthFD
2
= 2R2 − 1 with R2 being the target
signal. Without loss of generality, we assume that a2 > a1 with rate at D1 to detect x2 .
a1 +a2 = 1. Applying NOMA principle, successive interference The following theorem provides the outage probability of D1
cancellation (SIC) [13] is employed at D1 . Therefore, the for FD NOMA.
received signal to interference and noise ratio (SINR) at D1
Theorem 1. The closed-form expression for the outage prob-
to detect the D2 ’s message x2 is given by
ability of D1 is given by
2
|h1 | a2 ρ
γD2 →D1 = , (2) Ω1 θ
2
|h1 | a1 ρ + |hLI | ρ + 1
2 FD
PD =1− e− Ω1 , (8)
1 ,nodir
Ω1 + ρθΩLI
where ρ = N Ps
0
is transmit signal to noise radio (SNR). Note that FD
γth FD
γth
x1 and x2 are supposed to be normalized unity power signals, where θ = max (τ, β), τ =  2
FD
 and β = 1
a1 ρ . Note
ρ a2 −a1 γth
i.e, E{x21 } = E{x22 }. E{·} denotes expectation operation. (8) is derived by assuming the condition a2 >
2
is FD
a1 γth 2
After SIC, the received SNR at D1 to detect its own message
satisfied.
x1 is given by
Proof: By definition, J1 denotes the complementary outage
2
|h1 | a1 ρ event at D1 and can be calculated as follows:
γD1 = 2 . (3)    
|hLI | ρ + 1 2 2
J1 =Pr |h1 | ≥ |hLI | ρ + 1 θ
In the FD
√ mode, the √received signal
 at D2 can be written as
√  ∞ ∞
yD2 = h0 a1 Ps x1 + a2 Ps x2 +h2 Pr x2 +nD2 . However, = f|hLI |2 (x) f|h1 |2 (y)dxdy
the observation
√ at D2 √ for the direct
 link can be written as 0 (xρ+1)θ
yD2 = h0 a1 Ps x1 + a2 Ps x2 + nD2 . The received SINR Ω1 θ

at D2 to detect x2 is given by = e− Ω1 , (9)

Ω1 + ρθΩLI
2
|h0 | a2 ρ where fX (·) denotes the probability density function (PDF)
γ1,D2 = 2 . (4)
|h0 | a1 ρ + 1 of a random variable X. Substituting (9) into (7), (8) can be
Furthermore, the superposition of the signals for two users at obtained and the proof is completed.
√ to D2 . The observation at D2
D1 is decoded and forwarded Corollary 1. The outage probability of D1 for HD NOMA with
can be written as yD2 = h2 Pr x2 +nD2 and the received SNR  = 0 can be given by
is given by θ1

γ2,D2 = |h2 | ρ.
2
(5)
HD
PD1 ,nodir
= 1 − e− Ω1 , (10)

As in [14, 15], the relaying link from D1 to D2 corresponding HD

where γth 1
= 22R1 −1 and γthHD
2
= 22R2 −1 denote the
to the direct link from BS to D2 has small time delay for any target SNR at D1 to detect x1 and x2 with HD operation
HD
γth
transmitted signals. Therefore, we assume that the signals from mode, respectively. θ1 = max(τ1 , β1 ), β1 = 1
and τ1 =
a1 ρ
the relaying link and direct link can be combined by maximal HD
γth HD
2
ratio combining (MRC) at D2 . The received SINR after MRC with a2 >
HD )
ρ(a2 −a1 γth
a1 γth 2
.
2
at D2 can be given by Proof: Based on (7), when  = 0, an outage probability
2 of D1 for HD NOMA can be calculated as follows:
2 |h0 | a2 ρ
M RC
γD 2
= |h2 | ρ + 2 . (6)
|h0 | a1 ρ + 1
 IEEE ICC 2017 Wireless Communications Symposium

 
2 Δ
HD
PD1 ,nodir
=1 − Pr |h1 | ≥ max(τ1 , β1 ) = θ1 a) D1 for FD NOMA case: Based on analytical result in
θ
− Ω1
(8), when ρ → ∞, the asymptotic outage probability of D1
=1 − e 1 . (11) with 1 − ex ≈ x for FD NOMA is given by
The proof is completed. F D,∞ Ω1
PD =1− . (19)
2) Outage Probability of D2 : The outage events of D2 can 1 ,nodir
Ω1 + ρθΩLI
be explained for two reasons. The first is that D1 cannot detect
D1 ,nodir = 0.
Substituting (19) into (18), we can obtain dF D
x2 . The second is that D2 cannot detect its own message x2
on the conditions that D1 can detect x2 successfully. Based on Remark 1. The diversity order of User1 is zero, which is the
this, the outage probability of D2 can be expressed as below: same as the conventional FD relaying.
 
PDFD
2 ,nodir
=Pr γD2 →D1 < γthFD
b) D1 for HD NOMA case: Based on (10), the asymptotic
 2

+ Pr γ2,D2 < γthFD FD
, γD2 →D1 > γth , (12) outage probability of D1 for HD NOMA is given by
2 2

where  = 1. HD,∞ θ1
PD = . (20)
The following theorem provides the outage probability of D2 1 ,nodir
Ω1
for FD NOMA.
D1 ,nodir = 1.
Substituting (20) into (18), we obtain dHD
Theorem 2. The closed-form expression for the outage prob- c) D2 for FD NOMA case: Based on analytical result in
ability of D2 without direct link is given by (13), the asymptotic outage probability of D2 for FD NOMA
 
FD
γth is given by
Ω1 − Ω2 ρ
2 + Ωτ
1
FD
PD =1− e . (13) Ω1 Ω2 ρ − Ω1 γth
FD
− τ ρΩ2
2 ,nodir
(Ω1 + τ ρΩLI ) F D,∞
PD =1− 2
. (21)
Proof: By definition, J2 and J3 denote the first and second
2 ,nodir
Ω2 ρ (Ω1 + τ ρΩLI )
outage events, respectively. The process calculated is given by:
D2 ,nodir = 0.
Substituting (21) into (18), we can obtain dF D
  
2 2
J2 =Pr |h1 | < τ |hLI | ρ + 1 Remark 2. The diversity order of User2 is zero, which is the
 ∞  τ (yρ+1) same as user1 in FD NOMA.
= f|h1 |2 (x)f|hLI |2 (y) dxdy
0 0 d) D2 for HD NOMA case: Based on (16), the asymptotic
Ω1 τ outage probability of D2 for HD NOMA is given by
=1 − e− Ω1 . (14)
(Ω1 + τ ρΩLI ) HD
HD,∞ γth τ1
Applying some manipulations, J3 can be obtained as follows: PD = 2
+ . (22)

1 ,nodir
ρΩ2 Ω1
FD
Ω1 − Ωτ
γth
− Ω ρ2
J3 = e 1 1−e 2 . (15) D2 ,nodir = 1.
Substituting (22) into (18), we can obtain dHD
(Ω1 + τ ρΩLI )
Remark 3. As can be seen from (19) and (21), the error floors
Combining (14) and (15), (13) can be obtained.
of two users are existent at high SNR region for FD NOMA
Corollary 2. The outage probability of D2 for HD NOMA with without direct link.
 = 0 can be given by
4) Throughput Analysis: In this subsection, the delay-
HD
γth
τ
− Ω1 − ρΩ 2 limited transmission mode is considered for FD/HD NOMA.
HD
PD 2 ,nodir
=1−e 1 2 . (16)
a) FD NOMA case: In this mode, the BS transmits the
Proof: Based on (12), when  = 0, an outage probability information at a constant rate R, which is subject to the effect of
of D2 for HD NOMA can be calculated as follows: outage probability due to wireless fading channels. The system

  γ HD sum throughput of FD NOMA without direct link is expressed
2 2 th
HD
PD 2 ,nodir
=1 − Pr |h1 | > τ1 Pr |h2 | > 2
as
ρ
   
τ
HD
γth
FD
Rnodir = 1 − PD FD
1 ,nodir
R1 + 1 − PD FD
2 ,nodir
R2 , (23)
− Ω1 − 2
=1 − e 1 ρΩ2
. (17)
FD FD
where PD 1 ,nodir
and PD 2 ,nodir
are given in (8) and (13),
The proof is completed. respectively.
3) Diversity Analysis: To get deep insights, the asymptotic b) HD NOMA case: Similar to (23), the system sum
analyses are provided in terms of the investigated outage throughput of HD NOMA without direct link is expressed as
probability at high SNR region. The diversity order is defined    
as
HD
Rnodir = 1 − PD HD
1 ,nodir
R1 + 1 − PD HD
2 ,nodir
R2 , (24)
∞
log (PD (ρ)) where PD HD HD
and PD are given in (10) and (16),
d = − lim . (18) 1 ,nodir 2 ,nodir
ρ→∞ log ρ respectively.
 IEEE ICC 2017 Wireless Communications Symposium

B. User Relaying with Direct Link b) D2 for HD NOMA case: The outage performance of
In this subsection, we exploit the direct link between BS and D2 for HD NOMA has been investigated in [6] and we can
D2 to convey information. However, the outage probability of D2 ,dir = 2.
obtain dHD
D1 will not be affected by the direct link. As such, we only 3) Throughput Analysis:
show outage probability of D2 in the following. a) FD NOMA case: As suggested in Section III-A4 in
1) Outage Probability of D2 : For the second scenario, the delay-limited transmission mode, the system sum throughput
outage events of D2 for FD NOMA can be described below. of FD NOMA with direct link is expressed as
One of the events is when x2 can be detected at D1 , but the    
FD
Rdir = 1 − PD FD
1 ,nodir
R1 + 1 − PD
FD
2 ,dir
R2 , (28)
received SINR after MRC at D2 in one slot is less than its target
FD FD
SNR. Another event is that neither D1 nor D2 can detect x2 . where PD 1 ,nodir
and PD 2 ,dir
can be obtained from (8) and
Therefore, the outage probability of D2 can be expressed as (26), respectively.
below: b) HD NOMA case: Similar to (28), the system sum
 M RC  throughput of HD NOMA with direct link is expressed as
PDFD
2 ,dir
=Pr γD FD
< γth , γD2 →D1 > γthFD
   
 2 2 2
 HD
Rdir = 1 − PD HD
R1 + 1 − PDHD
R2 , (29)
+Pr γD2 →D1 < γth FD
2
FD
, γ1,D2 < γth 2
. (25) 1 ,nodir 1 ,dir

HD HD
The following theorem provides the outage probability of D2 where PD1 ,nodir
and PD2 ,dir
can be obtained from (10) and
for FD NOMA. [6].

Theorem 3. The closed-form expression for the outage prob- IV. N UMERICAL R ESULTS
ability of the investigated D2 with direct link is given by In this section, numerical simulation results are provided to

∞
 validate our analytical expressions derived in previous section.
− τ  (−1)
n ϕ
e (−1)
2n+1
φ1 n+1
PDFD
= 1 − e Ω0
− Without loss of generality, we assume that the distance between
2 ,dir
n!φ2 n+1 (n + 1)! BS and D2 is normalized to unity, i.e. ΩSD2 = 1. ΩSD1 = d−α
n=0
−α
 n n+1 ψ k and ΩD1 D2 = (1 − d) , where d is the normalised distance
(1 + a1 ρτ ) e ψ − eφ1 φ 1 k
× (Ei (ψ) − Ei (φ1 )) + between the BS and D1 setting to be d = 0.3 and α is pathloss
(n + 1) n · · · (n + 1 − k)
 k=0
  exponent setting to be α = 2. The power allocation coefficients
τ
× χe− Ω1 + 1 − χe− Ω1
τ τ
1 − e− Ω0 , (26) of NOMA are a1 = 0.2 and a2 = 0.8 for D1 and D2 ,
respectively.
γF D For user relaying without direct link, the target rate is set
where ϕ = ρa11Ω0 − ρΩ th2
− φ1 , φ1 = a−a 2
, φ2 = a1 ρΩ0 ,
−a2
2
Ω1
1 ρΩ2 to be R1 = 3, R2 = 0.5 bit per channel use (BPCU) for D1
ψ = a1 ρΩ2 (1+a1 ρτ ) and χ = Ω1 +τ ρΩLI . Ei (·) is the exponential
and D2 , respectively. The performance of conventional OMA
integral function [16, eq. (8.211.1)].
is shown as a benchmark for comparison, in which the total
Proof: See Appendix A. communication process is finished in three slots. In the first
2) Diversity Analysis: slot, the BS sends the information x1 to user relay D1 and
a) D2 for FD NOMA case: For D2 with direct link, it is sends x2 to D1 in the second slot. In the last slot, D1 decodes
challenging to obtain the diversity order from (26). We can use and forwards the information x2 to D2 .
Gaussian-Chebyshev quadrature to find an approximation from Fig. 1 plots the outage probability of two users versus SNR
(25) and the approximation expression of outage probability for without direct link. The exact theoretical curves for the outage
D2 at high SNR region is given by probability of two users for FD/HD NOMA are plotted ac-
    cording to (8), (13) and (10), (16), respectively. Apparently, the
F D,appro τ Ω2 τ + 2Ω0 τ a2 − a1 γth FD
PD2 ,dir = − 1− 2 exact outage probability curves match precisely with the Monte
Ω0 2Ω0 Ω2 Carlo simulation results. It is observed that the performance of
 
τπ  FD NOMA exceeds the HD NOMA and OMA on the condition
N
(sn + 1) τ a2 sn τ
× 1+ − of low SNR region. This is because loop interference is not the
2N Ω0 n=1 Ω2 ((sn + 1) τ a1 ρ + 2) 2Ω0
  dominant impact factor for FD NOMA in low SNR region.
  Ω1 Ω1 τ The asymptotic outage probability curves of two users for HD
× 1 − s2n + 1− ,
(Ω1 + τ ρΩLI ) (Ω1 + τ ρΩLI ) Ω0 NOMA are plotted according to (20) and (22), respectively.
(27) The asymptotic curves well approximate the exact performance
where N is a parameter curves at high SNR region. It is shown that error floors exist
 to ensure a complexity-accuracy in FD NOMA, which verify the conclusion in Remark 3 and
tradeoff, sn = cos 2n−1 2N π . Substituting (27) into (18), we
obtain the zero diversity order. This is due to the fact that there
can obtain dFD2 ,dir = 1.
D
is loop interference in FD NOMA, which plays a decisive role
Remark 4. From the above explanation, the observation is in high SNR region.
that the direct link (BS → D2 ) to convey information is an Fig.2 plots system throughput versus SNR in delay-limited
effective way to overcome the problem of zero diversity order transmission mode without direct link. The solid curves repre-
for D2 . sent throughput for FD/HD NOMA without direct link which
 IEEE ICC 2017 Wireless Communications Symposium

0 0
10 10

−1
10
User1 −2
10
−2
Outage Probability

Outage Probability
10
Sim
−3 User1 − OMA −4
10 User2 − OMA
10
User1 − Exact HD−NOMA
User1 − Asym HD−NOMA User2
−4
10 User2 − Exact HD−NOMA
−6
User2 − Asym HD−NOMA 10 Sim
−5 User1 − Exact FD−NOMA E{|h |2}= −15 (dB) User1 − Exact HD−NOMA
10 LI
User1 − Error floor FD−NOMA User2 − Exact HD−NOMA
User2 − Exact FD−NOMA User1 − Exact FD−NOMA E{|h |2}= −20, −15, −10 (dB)
User2 − Error floor FD−NOMA User2 − Exact FD−NOMA LI
−8
10
−10 0 10 20 30 40 −10 0 10 20 30 40
SNR (dB) SNR (dB)

Fig. 1: Outage probability versus transmit SNR without direct Fig. 3: Outage probability versus transmit SNR for different
link. values of LI with direct link.

3.5 3
E{|hLI|2}= −20, −15, −10 (dB)
2
E{|hLI| }= −20, −15, −10 (dB)
3
2.5
System Throughput (BPCU)

System Throughput (BPCU)

2.5
2 FD NOMA
gain
2
FD NOMA HD NOMA 1.5
1.5 gain gain

1
1

0.5 OMA 0.5

HD−NOMA HD+NOMA
FD−NOMA FD+NOMA
0 0
−10 0 10 20 30 40 −10 0 10 20 30 40
SNR (dB) SNR /dB

Fig. 2: System throughput in delay-limited transmission mode Fig. 4: System throughput in delay-limited transmission mode
versus SNR with different values of LI without direct link. versus SNR with different LI with direct link.

are obtained from (23) and (24), respectively. We can observe
that the FD NOMA achieves a higher throughput than HD
NOMA and OMA, since FD NOMA has low values of LI
influence. It is worth noting that increasing the values of LI
from −20 dB to −10 dB reduce the system throughput at high
SNR region.
For user relaying with direct link, the target rate is set to
be R1 = 2, R2 = 1 BPCU for D1 and D2 , respectively.
The performance of conventional HD NOMA is shown as a
benchmark for comparison, which has been researched in [6].
Fig. 3 plots the outage probability of the two users versus
different values of LI. The exact and approximate outage
probability curves of D2 for FD NOMA with direct link are
plotted according to (26) and (27), respectively. We observe that
D2 obtains one diversity order by using the direct link, which
overcomes the problem of zero diversity order inherent to FD
cooperative system. It is observed that the superiority of FD
NOMA is no longer apparent with the values of LI increasing.
Therefore, it is essential to consider the influence of LI when
designing practical FD NOMA systems.
Fig. 4 plots system throughput versus SNR in delay-limited
transmission mode with direct link. The solid curves, repre-
senting the FD NOMA, is obtained from (28). The dashdotted
curve, representing the HD NOMA, is obtained from (29). We
observe that FD NOMA also outperform HD NOMA at low
SNR region. The reason is that at low SNR region, the outage
 IEEE ICC 2017 Wireless Communications Symposium

probability is small and has no effect on the throughput, which Substituting (A.3) into (A.2), J11 is written as follows:
only depends on the fixed transmission rates at the BS.
∞

− Ωτ
 n
(−1) eϕ (−1)
2n+1
φ1 n+1
J11 = 1 − e 0 − (Ei (ψ)
V. C ONCLUSION n!φ2 n+1 (n + 1)!
n=0

This paper has investigated FD/HD user relaying in cooper- n
(1 + a1 ρτ )
n+1 ψ k
e ψ − e φ 1 φ1 k
ative NOMA systems and two cooperative relaying scenarios −Ei (φ1 )) + . (A.5)
(n + 1) n · · · (n + 1 − k)
k=0
have been considered insightfully. The closed-form expressions
of outage probability for two users have been derived. Due to After some manipulations, J12 is calculated as follows:
 ∞ ∞
the influence of residual loop interference, the diversity orders 1 − y − x τ
J21 = e ΩLI Ω1 dxdy = χe− Ω1 ,
achieved by two user were zero for FD NOMA. Therefore, ΩLI Ω1 0 τ (yρ+1)
the direct link between the BS and the far user was utilized to (A.6)
convey the information and one diversity order was obtained for Ω1
where χ = Ω1 +τ ρΩLI .
the far user. Based on the analytical results, it was shown that
Similarly, J13 is given by
FD NOMA was superior to HD NOMA at low SNR region  τ
 τ

rather than at high SNR region. Furthermore, the superior J13 = 1 − e− Ω0 1 − χe− Ω1 . (A.7)
of FD NOMA was not apparent with the loop interference
value increasing. Additionally, the delay-limited transmission Combining (A.5), (A.6) and (A.7), we can obtain (26). The
throughput was analyzed. proof is completed.
R EFERENCES
A PPENDIX A: P ROOF OF T HEOREM 3 [1] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, C. L. I, and H. V. Poor,
“Application of non-orthogonal multiple access in LTE and 5G networks,”
The outage probability of D2 can be further expressed as: IEEE Commun. Mag., vol. 55, no. 2, pp. 185–191, Feb. 2017.
[2] Z. Ding, P. Fan, and H. V. Poor, “Impact of user pairing on 5G non-
 M RC    orthogonal multiple-access downlink transmissions,” IEEE Trans. Veh.
FD
PD 2 ,dir
= Pr γD FD
< γth FD
Pr γD2 →D1 > γth2
 2
 2
   Technol., vol. 65, no. 8, pp. 6010–6023, Aug. 2016.
J J
[3] F. Gao, T. Cui, and A. Nallanathan, “On channel estimation and optimal
 11 12
 training design for amplify and forward relay networks,” IEEE Trans.
+ Pr γ1,D2 < γth
FD
, γ D2 →D1 < FD
γth , (A.1) Wireless Commun., vol. 7, no. 5, pp. 1907–1916, May 2008.
 2
 2
 [4] J. B. Kim and I. H. Lee, “Non-orthogonal multiple access in coordinated
J13 direct and relay transmission,” IEEE Commun. Lett., vol. 19, no. 11, pp.
2037–2040, 2015.
where  = 1. J11 and J12 can be calculated as follows: [5] J. Men, J. Ge, and C. Zhang, “Performance analysis of non-orthogonal
 multiple access for relaying networks over nakagami-m fading channels,”
FD 2
2 γth |h 0 | a 2 2 IEEE Trans. Veh. Technol., vol. PP, no. 99, pp. 1–1, 2016.
J11 = Pr |h2 | < 2
− 2 , |h0 | < τ [6] Z. Ding, M. Peng, and V. Poor, H, “Cooperative non-orthogonal multiple
ρ |h0 | a1 ρ + 1 access in 5G systems,” IEEE Commun. Lett., vol. 19, no. 8, pp. 1462–
 
 FD
γth ya2
1465, 2014.
τ
τ
1 − Ωy − Ω1 ρ
2 − ya [7] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor, “Cooperative non-
1 ρ+1
= 1 − e− Ω0 −
2
e 0e dy . (A.2) orthogonal multiple access with simultaneous wireless information and
Ω0
0   power transfer,” IEEE J. Sel. Areas Commun., vol. 34, no. 4, pp. 938–953,
April 2016.
Θ1
[8] Y. Liu, Z. Ding, M. Elkashlan, and J. Yuan, “Non-orthogonal multiple
Based on (A.2), using x = yρa1 + 1, Θ1 can be calculated as access in large-scale underlay cognitive radio networks,” IEEE Trans.
Veh. Technol., vol. 65, no. 12, pp. 10 152–10 157, Dec. 2016.
follows: [9] H. Ju, E. Oh, and D. Hong, “Improving efficiency of resource usage
F D  τ ρa +1 in two-hop full duplex relay systems based on resource sharing and
1 γth
2
1 x−1 a2 (x−1)
Θ1 = e− ρΩ2 e− ρa1 Ω0 e a1 ρΩ2 x dx interference cancellation,” IEEE Trans. Wireless Commun., vol. 8, no. 8,
Ω0 1
pp. 3933–3938, August 2009.
∞  τ ρa1 +1 [10] J. Men and J. Ge, “Full deplex techniques for 5G networks: self-
eϕ  (−1)
n
x interference cancellation, protocol design, and relay selection,” IEEE
= n xn e− a1 ρΩ2 x dx, (A.3) Commun. Mag., vol. 53, no. 5, pp. 128–137, 2015.
φ2 n=0 n!(Ω0 ρa1 ) 1
   [11] Z. Zhang, Z. Ma, M. Xiao, Z. Ding, and P. Fan, “Full-duplex device-to-
Θ2 device aided cooperative non-orthogonal multiple access,” IEEE Trans.
Veh. Technol., vol. PP, no. 99, pp. 1–1, 2016.
FD
[12] Z. Ding, Z. Yang, P. Fan, and H. V. Poor, “On the performance of non-
where ϕ = ρa11Ω0 − ρΩ2 − φ1 , φ1 = a−a
γth2 2
1 ρΩ2
and φ2 = a1 ρΩ0 . orthogonal multiple access in 5G systems with randomly deployed users,”
Note that (A.3) is obtained by using Binomial theorem. IEEE Signal Process. Lett., vol. 21, no. 12, pp. 1501–1505, Dec 2014.
Furthermore, using z = x1 , Θ2 can be given by as follows: [13] T. M. Cover and J. A. Thomas, Elements of information theory, 6th ed.,
Wiley and Sons, New York, 1991.
 a2 z
2n+1 [14] T. Riihonen, S. Werner, and R. Wichman, “Hybrid full-duplex/half-duplex
e− a1 ρΩ2
1 n+1
(−1) φ1 relaying with transmit power adaptation,” IEEE Trans. Wireless Commun.,
Θ2 = dz = (Ei (ψ) −Ei (φ1 ))
1
τ ρa1 +1
z n+2 (n + 1)! vol. 10, no. 9, pp. 3074–3085, September 2011.
[15] D. P. M. Osorio, E. E. B. Olivo, H. Alves, J. C. S. S. Filho, and

n
(1 + a1 ρτ )
n+1 ψ k
e ψ − e φ φk M. Latva-aho, “Exploiting the direct link in full-duplex amplify-and-
1
+ , (A.4) forward relaying networks,” IEEE Signal Process. Lett., vol. 22, no. 10,
(n + 1) n · · · (n + 1 − k) pp. 1766–1770, Oct 2015.
k=0
[16] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products,
where (A.4) can be obtained by using [16, Eq. (3.351.4)]. 6th ed. New York, NY, USA: Academic Press, 2000.