See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269416921

Impact of User Pairing on 5G Non-Orthogonal Multiple Access

Article  in  IEEE Transactions on Vehicular Technology · December 2014

DOI: 10.1109/TVT.2015.2480766 · Source: arXiv

CITATIONS READS
821 1,167

3 authors:

Zhiguo Ding Pingzhi Fan

The University of Manchester Southwest Jiaotong University
468 PUBLICATIONS   21,103 CITATIONS    526 PUBLICATIONS   12,893 CITATIONS   

SEE PROFILE SEE PROFILE

H. Vincent Poor
Princeton University
2,441 PUBLICATIONS   90,207 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Machine Learning View project

UAV communications using NOMA View project

All content following this page was uploaded by H. Vincent Poor on 14 March 2015.

The user has requested enhancement of the downloaded file.


Impact of User Pairing on 5G Non-Orthogonal Multiple Access
Zhiguo Ding, Member, IEEE, Pingzhi Fan, Fellow, IEEE, and H. Vincent Poor, Fellow, IEEE
Abstract—Non-orthogonal multiple access (NOMA) represents
a paradigm shift from conventional orthogonal multiple access
(MA) concepts, and has been recognized as one of the key
enabling technologies for 5G systems. In this paper, the impact
of user pairing on the performance of two NOMA systems,
NOMA with fixed power allocation (F-NOMA) and cognitive
radio inspired NOMA (CR-NOMA), is characterized. For F-
NOMA, both analytical and numerical results are provided to
demonstrate that F-NOMA can offer a larger sum rate than
arXiv:1412.2799v1 [cs.IT] 8 Dec 2014
orthogonal MA, and the performance gain of F-NOMA over
conventional MA can be further enlarged by selecting users
whose channel conditions are more distinctive. For CR-NOMA,
the quality of service (QoS) for users with the poorer channel
condition can be guaranteed since the transmit power allocated to
other users is constrained following the concept of cognitive radio
networks. Because of this constraint, CR-NOMA has different
behavior compared to F-NOMA. For example, for the user with
the best channel condition, CR-NOMA prefers to pair it with the
user with the second best channel condition, whereas the user
with the worst channel condition is preferred by F-NOMA.
I. I NTRODUCTION
Multiple access in 5G mobile networks is an emerging
research topic, since it is key for the next generation network
to keep pace with the exponential growth of mobile data and
multimedia traffic [1] and [2]. Non-orthogonal multiple access
(NOMA) has recently received considerable attention as a
promising candidate for 5G multiple access [3]–[6]. Partic-
ularly, NOMA uses the power domain for multiple access,
where different users are served at different power levels. The
users with better channel conditions employ successive inter-
ference cancellation (SIC) to remove the messages intended for
other users before decoding their own [7]. The benefit of using
NOMA can be illustrated by the following example. Consider
that there is a user close to the edge of its cell, denoted by
A, whose channel condition is very poor. For conventional
MA, an orthogonal bandwidth channel, e.g., a time slot, will
be allocated to this user, and the other users cannot use this
time slot. The key idea of NOMA is to squeeze another user
with better channel condition, denoted by B, into this time
slot. Since A’s channel condition is very poor, the interference
from B will not cause much performance degradation to A, but
the overall system throughput can be significantly improved
since additional information can be delivered between the
base station (BS) and B. The design of NOMA for uplink
transmissions has been proposed in [4], and the performance
of NOMA with randomly deployed mobile stations has been
characterized in [5]. The combination of cooperative diversity
with NOMA has been considered in [8].
Z. Ding and H. V. Poor are with the Department of Electrical Engineering,
Princeton University, Princeton, NJ 08544, USA. Z. Ding is also with the
School of Computing and Communications, Lancaster University, LA1 4WA,
UK. Pingzhi Fan is with the Institute of Mobile Communications, Southwest
Jiaotong University, Chengdu, China.
1
Since multiple users are admitted at the same time, fre-
quency and spreading code, co-channel interference will be
strong in NOMA systems, i.e., a NOMA system is interference
limited. As a result, it may not be realistic to ask all the
users in the system to perform NOMA jointly. A promising
alternative is to build a hybrid MA system, in which NOMA
is combined with conventional MA. In particular, the users in
the system can be divided into multiple groups, where NOMA
is implemented within each group and different groups are
allocated with orthogonal bandwidth resources. Obviously the
performance of this hybrid MA scheme is very dependent on
which users are grouped together, and the aim of this paper
is to investigate the effect of this grouping. Particularly, tn
this paper, we focus on a downlink communication scenario
with one BS and multiple users, where the users are ordered
according to their connections to the BS, i.e., the m-th user has
the m-th worst connection to the BS. Consider that two users,
the m-th user and the n-th user, are selected for performing
NOMA jointly, where m < n. The impact of user pairing on
the performance of NOMA will be characterized in this paper,
where two types of NOMA will be considered. One is based
on fixed power allocation, termed F-NOMA, and the other is
cognitive radio inspired NOMA, termed CR-NOMA.
For the F-NOMA scheme, the probability that F-NOMA can
achieve a larger sum rate than conventional MA is first studied,
where an exact expression for this probability as well as its
high signal-to-noise ratio (SNR) approximation are obtained.
These developed analytical results demonstrate that it is almost
certain for F-NOMA to outperform conventional MA, and the
channel quality of the n-th user is critical to this probability.
In addition, the gap between the sum rates achieved by F-
NOMA and conventional MA is also studied, and it is shown
that this gap is determined by how different the two users’
channel conditions are, as initially reported in [8]. For example,
if n = M , it is preferable to choose m = 1, i.e., pairing the
user with the best channel condition with the user with the
worst channel condition. The reason for this phenomenon can
be explained as follows. When m is small, the m-th user’s
channel condition is poor, and the data rate supported by this
user’s channel is also small. Therefore the spectral efficiency of
conventional MA is low, since the bandwidth allocated to this
user cannot be accessed by other users. The use of F-NOMA
ensures that the n-th user will have access to the resource
allocated to the m-th user. If (n − m) is small, the n-th user’s
channel quality is similar to the m-th user’s, and the benefit to
use NOMA is limited. But if n >> m, the n-th user can use
the bandwidth resource much more efficiently than the m-th
user, i.e., a larger (n − m) will result in a larger performance
gap between F-NOMA and conventional MA.
The key idea of CR-NOMA is to opportunistically serve
the n-th user on the condition that the m-th user’s quality of
service (QoS) is guaranteed. Particularly the transmit power
 2

allocated to the n-th user is constrained by the m-th user’s
signal-to-interference-noise ratio (SINR), whereas F-NOMA
uses a fixed set of power allocation coefficients. Since the
m-th user’s QoS can be guaranteed, we mainly focus on
the performance of the n-th user offered by CR-NOMA. An
exact expression for the outage probability achieved by CR-
NOMA is obtained first, and then used for the study of the
diversity order. In particular, we show that the diversity order
experienced by the n-th user is m, which means that the
m-th user’s channel quality is critical to the performance of
CR-NOMA. This is mainly because of the imposed SINR
constraint, where the n-th user can be admitted into the
bandwidth channel occupied by the m-th user, only if the m-
th user’s SINR is guaranteed. As a result, with a fixed m,
increasing n does not bring much improvement to the n-th
user’s outage probability, which is different from F-NOMA. If
A. Impact of user pairing on the sum rate
In this subsection, we focus on how user pairing affects
the probability that NOMA achieves a lower sum rate than
conventional MA schemes, which is given by
P(Rm + Rn < R̄m + R̄n ). (4)
The following theorem provides an exact expression for the
above probability as well as its high SNR approximation.
Theorem 1. Suppose that the m-th and n-th ordered users
are paired to perform NOMA. The probability that F-NOMA
achieves a lower sum rate than conventional MA is given by
P(Rm + Rn < R̄m + R̄n ) = (5)
n−1−m   Z ̟
X n − 1 − m (−1)i ̟1 2

the ergodic rate is used as the criterion, a similar difference 1− f (y)(F (y))n−1−m−i
i=0
i m + i ̟4
between F-NOMA and CR-NOMA can be observed. Again   m+i !
take the scenario described in the last paragraph as an example. M−n m+i ̟2 − y
If n = M , in order to yield a large gain over conventional MA, × (1 − F (y)) [F (y)] − F dy
1+y
F-NOMA prefers the choice of m = 1, but CR-NOMA prefers
n−1  
the choice of m = M − 1 , i.e., pairing the user with the best ̟3 X n − 1 ρ (M −n+j+1)̟2

channel condition with the user with the second best channel − (−1)j e− ρ ,
ρ j=0 j M −n+j +1
condition.
1 −x x
II. NOMA W ITH F IXED P OWER A LLOCATION where f (x) = ρe
ρ, F (x) = 1 − e − ρ , ̟1 =
Consider a downlink communication scenario with one BS M! 1−2a2n M!
(m−1)!(n−1−m)!(M−n)! , ̟2 = ̟3 =
a4n , (n−1)!(M−n)!
and M mobile users. Without loss of generality, assume that √
and ̟4 = 1 + ̟2 − 1. At high SNR, this probability can
the users’ channels have been ordered as |h1 |2 ≤ · · · ≤ |hM |2 , be approximated as follows:
where hm denotes the Rayleigh fading channel gain between
 
the BS and the ordered m-th user. Consider that the m-th user 1 ̟3 ̟2n
and the n-th user, m < n, are paired to perform NOMA. P(Rm + Rn < R̄m + R̄n ) ≈ n − ̟1 ̟ , (6)
ρ n
In this section, we focus on F-NOMA, where the BS
allocates a fixed amount of transmit power to each user. Pn−1−m n−1−m
(−1)i R ̟2 n−1−m−i
where ̟ = i=0 i m+i ̟4 y
In particular, denote am and an as the power allocation  h im+i 
coefficients for the two users, where these coefficients are fixed ̟ −y
× y m+i − (1+y)
2
dy, i.e., ̟ is a constant and
and a2m + a2n = 1. According to the principle of NOMA,
am ≥ an since |hm |2 ≤ |hn |2 . The rates achievable to the not a function of ρ.
two users are given by Proof: See the appendix.
!
|hm |2 a2m Theorem 1 demonstrates that it is almost certain for F-
Rm = log 1 + , (1) NOMA to outperform conventional MA, particularly at high
|hm |2 a2n + ρ1 SNR. Furthermore, the decay rate of the probability P(Rm +
and Rn < R̄m + R̄n ) is approximately ρ1n , i.e., the quality of the n-

th user’s channel determines the decay rate of this probability.
Rn = log 1 + ρa2n |hn |2 , (2)
respectively, where ρ denotes the transmit SNR. Note that the
n-th user can decode the message intended for the m-th user B. Asymptotic studies of the sum rate achieved by NOMA
 and Rn is2 always
successfully  achievable at the n-th user, since In addition to the probability P(Rm + Rn < R̄m + R̄n ), it
|h | a2m
Rm ≤ log 1 + |hn |n2 a2 + 1 . is also of interest to study how large of a performance gain
n ρ
On the other hand, an orthogonal MA scheme, such as time- F-NOMA offers over conventional MA, i.e.,
division multiple-access (TDMA), can support the following
data rate: P(Rm + Rn − R̄m − R̄n < R),
1

where R is a targeted performance gain. The probability
R̄i = log 1 + ρ|hi |2 , (3)
2 studied in the previous subsection can be viewed as a special
where i ∈ {m, n}. In the following subsections, the impact case by setting R = 0. An interesting observation for the
of user pairing on the sum rate and the individual user rates cases with R > 0 is that there will be an error floor for
achieved by F-NOMA is investigated. P(Rm +Rn − R̄m − R̄n < R), regardless of how large the SNR
 3

is. This can be shown by studying the following asymptotic C. Impact of user pairing on individual user rates
expression of the sum rate gap: Careful user pairing not only improves the sum rate, but
also has the potential to improve the individual user rates, as
Rm + Rn − R̄m − R̄n (7) shown in this section. We first focus on the probability that
 
1
F-NOMA can achieve a larger rate than orthogonal MA for
→ log + log ρa2n |hn |2 − log (ρ|hm ||hn |) the m-th user which is given by
ρ→∞ a2n
= log |hn | − log |hm |, P(Rm > R̄m ) (12)
 !2 
2 2
which is not a function of SNR. Hence the probability can be |h m | a
expressed asymptotically as follows: =P  1 + m
> (1 + ρ|hm |2 ) .
|hm |2 a2n + ρ1


P Rm + Rn − R̄m − R̄n < R (8) After some algebraic manipulations, the above probability can
→ P (log |hn | − log |hm | < R) . be further rewritten as follows:
ρ→∞  
1 − 2a2n

P(Rm > R̄m ) = P |hm |2 < (13)
When R = 0, P Rm + Rn − R̄m − R̄n < R → 0, which is ρa4n
consistent with Theorem 1, since Z 1−2a4 2n  m−1
ρan ̟5 − (M −m+1)y y

1 = e ρ 1 − e− ρ dy
P Rm + Rn < R̄m + R̄n ∼ n → 0. 0 ρ
ρ ρ→∞ X m − 1
m−1 (1−2a2
!
(−1)i ̟5 − n )(M −m+i+1)
4
= 1−e ρan
,
When R 6= 0, (8) implies
that the probability i M −m+i+1
i=0
P Rm + Rn − R̄m − R̄n < R can be expressed asymptot-
M!
ically as follows: where ̟5 = (m−1)!(M−m)! .
  By applying a series expansion, the above probability can
|hn |2 2R be rewritten as follows:
P (log |hn | − log |hm | < R) → P <2 . (9)
|hm |2 m−1
X m − 1
P(Rm > R̄m ) = (−1)i+1 ̟5 (14)
Directly applying the joint probability density function (pdf) i=0
i
of the users’ channels shown in (31), the probability can be ∞
X (1 − 2a2n )k (M − m + i + 1)k−1
rewritten as follows: × (−1)k .
k!ρk a4k
n
k=1
P (log |hn | − log |hm | < R) (10)
Z ∞Z y Again applying the results in (41) and (42), the above equation
= ̟1 f (x)f (y)[F (x)]m−1 can be approximated as follows:
0 2−2R y
(1 − 2a2n )m
× (F (y) − F (x))
n−1−m M−n
(1 − F (y)) dxdy, P(Rm > R̄m ) ≈ ̟5 , (15)
mρm a4mn

which is quite complicated to evaluate. In [9], a simpler pdf for which means that P(Rm > R̄m ) decays at a rate of ρ1m .
the ratio of two order statistics has been provided as follows: On the other hand, the probability that the n-th user can
experience better performance in a NOMA system than in
m−1
X n−m−1
M! X orthogonal MA systems is given by
f |hm |2 (z) =  
|hn |2 (m − 1)!(n − m − 1)!(M − n)! j =0 j =0 2 2

1 2
 1 2 P(Rn > R̄n ) = P log 1 + ρan |hn | > log(1 + ρ|hn | .

m − 1

n − m − 1 2
(−1)j1 +j2 (τ2 + τ1 z)−2 ,
j1 j2 Following similar steps as previously, we obtain the following:
 
where τ1 = j1 − j2 + n − m and τ2 = M − n + 1 + j2 . By 2 1 − 2a2n
P(Rn > R̄n ) = P |hn | > . (16)
using this pdf, the addressed probability can be calculated as ρa4n
follows: Interestingly P(Rn > R̄n ) in (16) is very much similar to
P(Rm > R̄m ) in (13), which yields the following:
P (log |hn | − log |hm | < R) (11)
m−1
X n−m−1 X n − 1 (−1)i ̟3
n−1
M! X
P(Rn > R̄n ) = 1 − (17)
→ i M −n+i+1
(m − 1)!(n − m − 1)!(M − n)! j =0 j =0 i=0
!
1 2
    2(1−2an )(M −n+i+1)
(−1)j1 +j2 m − 1 n − m − 1 1 1 × 1−e
− ρa4
n ,
− .
τ1 j1 j2 τ2 + 2−2R τ1 τ2 + τ1
 4

R
and its high SNR approximation is given by 1 − e−y , ǫ1 = 2 ρ−1 , b = Iρ , a = 1 + I and b ≤ aǫ1 . The
diversity order achieved by CR-NOMA is given by
(1 − 2a2n )n
P(Rn > R̄n ) ≈ 1 − ̟3 . (18) log Pon
nρn a4n
n lim − = m.
ρ→∞ log ρ
As can be seen from (15) and (18), the two users will have
totally different experience in NOMA systems. Particularly, a Proof: See the appendix.
user with a better channel condition is more willing to perform Theorem 2 demonstrates an interesting phenomenon that, in
NOMA since P(Rn > R̄n ) → 1, which is not true for a user CR-NOMA, the diversity order experienced by the n-th user
with a poor channel condition. Furthermore, it is preferable is determined by how good the m-th user’s channel quality is.
to pair two users whose channel conditions are significantly This is because the n-th user can be admitted to the channel
distinct, since (15) and (18) implies that m should be as small occupied by the m-th user only if the m-th user’s QoS is
as possible and n should be as large as possible. met. For example, if the m-th user’s channel is poor and its
targeted SINR is high, it is very likely that the BS allocates
all the power to the m-th user, and the n-th user might not
III. C OGNITIVE R ADIO I NSPIRED NOMA even get served.
NOMA can be also viewed as a special case of cognitive Recall from the previous section that F-NOMA can achieve
radio systems [10] and [11], in which a user with a strong a diversity gain of n for the n-th user, and therefore the
channel condition, viewed as a secondary user, is squeezed diversity order achieved by CR-NOMA could be much smaller
into the spectrum occupied by a user with a poor channel than that achieved by F-NOMA, particularly if n >> m. This
condition, viewed as a primary user. Following the concept performance difference is again due to the imposed power
of cognitive radio networks, a variation of NOMA, termed as constraint shown in (21).
CR-NOMA, can be designed as follows. Suppose that the BS It is important to point out that CR-NOMA can strictly guar-
needs to serve the m-th user, i.e., a user a with poor channel antee the m-th user’s QoS, and therefore achieve better fairness
condition, due to either the high priority of this user’s messages compared to F-NOMA. In particular, the use of CR-NOMA
or user fairness, e.g., this user has not been served for a long can ensure that a diversity order of m is achievable to the n-th
time. This user can be viewed as a primary user in a cognitive user, and admitting the n-th user into the same channel as the
radio system. The n-th user can be admitted into this channel m-th user will not cause too much performance degradation to
on the condition that the n-th user will not cause too much the m-th user. Particularly the SINR experienced by the m-th
performance degradation to the m-th user. user is strictly maintained at the predetermined level I.
Consider that the targeted SINR at the m-th user is I, which
means that the choices of the power allocation coefficients, am Sum rate achieved by CR-NOMA
and an , need to satisfy the following constraint: Without sharing the spectrum with the n-th user, i.e, all the
|hm |2 a2m bandwidth resource is allocated to the m-th user, the following
1 ≥ I. (20) rate is achievable:
|hm |2 a2n + ρ

R̃m = log 1 + ρ|hm |2 . (23)
This means that the maximal transmit power that can be
allocated to the n-th user is given by It is easy to show that the use of CR-NOMA always achieves
( ) a larger sum rate since
2
|hm |2 − Iρ
an = max 0, , (21) Rm + Rn − R̃m (24)
|hm |2 (1 + I) !
|hm |2 a2m

which means that an = 0 if |hm |2 < ρI . Note that the choice of = log 1 + 2 2 1 + log 1 + ρa2n |hn |2
|hm | an + ρ
an in (21) is a function of the channel coefficient hm , unlike

the constant choice of an used by F-NOMA in the previous − log 1 + ρ|hm |2
section. 1 + ρa2n |hn |2
Since the m-th user’s QoS can be guaranteed due to (20), = log ≥ 0.
1 + ρa2n |hm |2
we only need to study the performance experienced by the n-
th user. Particularly the outage performance of the n-th user This superior performance gain is not surprising, since the key
is defined as follows: idea of CR-NOMA is to serve a user with a strong channel

condition, without causing too much performance degradation
Pno , P log(1 + a2n ρ|hn |2 ) < R , (22) to the user with a poor channel condition.
In addition, it is of interest to study how much the averaged
and the following theorem provides an exact expression for rate gain CR-NOMA can yield, i.e., E {Rn }. This averaged
the above outage probability as well as its approximation. rate gain can be calculated as follows:
Theorem 2. Suppose that the transmit power allocated to the Z ∞Z ∞  
x−b
n-th user can satisfy the predetermined SINR threshold, I, as E {Rn } = log 1 + ρy (25)
b x xa
shown in (21). The n-th user’s outage probability achieved
by CR-NOMA is given by (19), where g(y) = e−y , G(y) = × f|hm |2 ,|hn |2 (x, y)dydx.
 5

M−n
X   n−1−m
X n − 1 − m Z aǫ1
M −n [G(b)]m+i M−n
Pon = ̟5 (−1)i + (−1)i g(y) (1 − G(y)) G(y)n−1−m−i ̟1 (19)
i=0
i m + i i=0
i b

n−1−m

G(y)m+i − G(b)m+i X  n − 1 − m Z b+aǫ1
G(y)m+i − G(b)m+i
× dy + (−1)i (1 − G(y))M−n G(y)n−1−m−i
m+i i=0
i aǫ1 m+i
 m+i !

n−1−m
G 1− baǫ1 − G(b)m+i
X n − 1 − m Z ∞
M−n
|hn |2
× ̟1 g(y)dy + (−1)i g(y) (1 − G(y)) G(y)n−1−m−i ̟1 dy.
i=0
i b+aǫ 1
m+i

In general, the evaluation of the above equation is difficult, and

in the following we provide a case study when n − m = 1.
Particularly, the joint pdf of the channels for this special case 0
10
can be simplified and the averaged rate gain can calculated as
follows: −1
10
Z ∞ Z ∞  
m−1 x−b
E {Rn } = ̟1 f (x)[F (x)] log 1 + ρy −2
b x xa 10 n=3,m=1

P(Rm + Rn < R̄m + R̄n )

M−n
× f (y) (1 − F (y)) dydx (26) −3
Z ∞ 10
−̟1
= f (x)[F (x)]m−1
M −n+1 b −4
10 n=5, m=1
Z ∞  
x−b
× log 1 + ρy d (1 − F (y))M−n+1 dx.
x xa −5
10

After some algebraic manipulations, the above equation can −6

10
be rewritten as follows: Simulation results
Analytical results − Exact expression
Z ∞ Analytical results − High SNR approximation
̟1 −7
10
E {Rn } = f (x)[F (x)]m−1 5 10 15
SNR in dB
20 25 30
M −n+1 b
   (a) m = 1
x−b M−n+1
× log 1 + ρ (1 − F (x))
a
Z ∞ ! 0
10
x−b
1 xa ρ
Simulation results
M−n+1
+ (1 − F (y)) dy dx. Analytical results − Exact expression
ln 2 x 1 + x−b
xa ρy −1
Analytical results − High SNR approximation
10

Now applying Eq. (3.352.2) in [12], the average rate gain can
be expressed as follows: −2
10
P(Rm + Rn < R̄m + R̄n )

Z ∞
̟1
E {Rn } = f (x)[F (x)]m−1 (27) −3
10
n=3,m=2
M −n+1 b
 2
  x a
x−b M−n+1 e ρ(x−b) −4
10
× log 1 +
 ρ (1 − F (x)) −
a ln 2 n=5,m=2
  −5
(M − n + 1)xa 10
×Ei −(M − n + 1)x − dx,
ρ(x − b)
−6
10
5 10 15 20 25 30
where Ei(·) denotes the exponential integral. SNR in dB

(b) m = 2
IV. N UMERICAL S TUDIES
Fig. 1. The probability that F-NOMA realizes a lower sum rate than
In this section, computer simulations are used to evaluate the conventional MA. M = 5. The analytical results are based on Theorem 1.
performance of two NOMA schemes as well as the accuracy
of the developed analytical results.
 6

A. NOMA with fixed power allocation have been developed to show that both P(Rm > R̄m ) and
P(Rn < R̄n ) are decreasing with increasing SNR, which is
In Fig. 1, the probability that F-NOMA realizes a lower
confirmed by the simulation results in Fig. 3. The reason that
sum rate than conventional MA, i.e., P(Rm + Rn < R̄m +
P(Rm > R̄m ) is reduced at a higher SNR is that the m-
R̄n ), is shown as a function of SNR. a2m = 45 and a2n = 15 .
th user’s rate in an F-NOMA system becomes a constant,
As can be seen from both figures, F-NOMA almost always |hm |2 a2m
 
a2m

outperforms conventional MA, particularly at high SNR. The i.e., log 1 + |hm |2 a2 + 1 → log 1 + a2 , which is much
n ρ ρ→∞ n
simulation results in Fig. 1 also demonstrate the accuracy of smaller than R̄m , at high SNR. On the other hand, it is more
the analytical results provided in Theorem 1. For example, likely for Rn to be larger than R̄n since there is a factor of 21
the exact expression of P(Rm + Rn < R̄m + R̄n ) shown outside of the logarithm of R̄n .
in Theorem 1 matches perfectly with the simulation results,
whereas the developed approximation results become accurate 0
10
at high SNR.
Another important observation from Fig. 1 is that increasing
n, i.e., scheduling a user with a better channel condition, will −1
10
make the probability decrease at a faster rate. This observation
is consistent to the high SNR approximation results provided
in Theorem 1 which show that the slope of the curve for the
probability P(Rm +Rn < R̄m + R̄n ) is a function of n. In Fig. −2
10

2, the probability P(Rm + Rn − R̄m − R̄n < R) is shown with

different choices of R. Comparing Fig. 1 to Fig. 2, one can
observe that P(Rm + Rn − R̄m − R̄n < R) never approaches −3
10
zero, regardless of how large the SNR is. This observation Simulation results, m = 2, n = 3, P(Rm > R̄m )
confirms the analytical results developed in (11) which show Simulation results, m = 2, n = 3, P(Rn < R̄n )
Analytical results, m = 2, n = 3, P(Rm > R̄m )
that the probability P(Rm + Rn − R̄m − R̄n < R) is no −4
Analytical results, m = 2, n = 3, P(Rn < R̄n )
10
longer a function of SNR, when ρ → 0. It is interesting to Simulation results, m = 1, n = M , P(Rm > R̄m )
observe that the choice of a smaller m is preferable to reduce Simulation results, m = 1, n = M , P(Rn < R̄n )
Analytical results, m = 1, n = M , P(Rm > R̄m )
P(Rm + Rn − R̄m − R̄n < R), a phenomenon previously Analytical results, m = 1, n = M , P(Rn < R̄n )
reported in [8]. 5 10 15 20 25 30
SNR in dB

0
Fig. 3. The behavior of individual data rates achieved by F-NOMA, P(Rn <
10
Simulation results, m=1 R̄n ) and P(Rm > R̄m ). M = 5. The analytical results are based on (14)
Asymptotic results, m=1 and (17).
Simulation results, m=2
Asymptotic results, m=2

−1
10
P(Rm + Rn − R̄m − R̄n < R)

B. Cognitive radio inspired NOMA

In Fig. 4 the n-th user’s outage probability achieved by
−2
10 CR-NOMA is shown as a function of SNR. As can be seen
from the figure, the exact expression for the outage probability
−Solid lines for the case with R=0.5 BPCU
Pon , P(Rn < R) developed in Theorem 2 matches the
simulation results perfectly. Recall from Theorem 2 that the
−3 −Dashed lines for the case with R=0.1 BPCU
10 diversity order achievable for the n-th user is m. Or in other
words, the slope of the outage probability is determined by
the channel quality of the m-th user, which is also confirmed
−4
by Fig. 4. For example, when increasing m from 1 to 2, the
10
5 10 15 20 25 30 35 40 outage probability is significantly reduced, and its slope is also
SNR in dB
increased. To clearly demonstrate the diversity order, we have
Fig. 2. The probability that the sum rate gap between F-NOMA and
provided an auxiliary curve in the figure which is proportional
conventional MA is larger than R. M = 5 and n = M . The analytical to ρ1m . As can be observed in the figure, this auxiliary curve
results are based on (11). is parallel to the one for P(Rn < R), which confirms that the
diversity order achieved by CR-NOMA is m.
In Fig. 3, two different but related probabilities are shown Since Theorem 2 states that the diversity order of P(Rn <
together. One is P(Rm > R̄m ), i.e., the probability that R) is not a function of n, an interesting question is whether
it is beneficial for the user with a poor channel condition a different choice of n matters. Fig. 5 is provided to answer
to perform F-NOMA, and the other is P(Rn < R̄n ), i.e., this question. While the use of a larger n does bring some
the probability that the user with a strong channel condition reduction of P(Rn < R), the performance gain of increasing
prefers conventional MA. In Section II.C, analytical results n is negligible, particularly at high SNR. This is because the
 7

9
0
Simulation results, m=1, n=m+1
10 Analytical results, m=1, n=m+1
8 Simulation results, m=2, n=m+1
Analytical results, m=2, n=m+1
Simulation results, m=3, n=m+1
7
Analytical results, m=3, n=m+1
−1
10 Simulation results, m=4, n=m+1
6 Analytical results, m=4, n=m+1
m=1
5
P(Rn < R)

E(Rn )
−2
10
4

Simulation results 3
Analytical results
−3 1
10 An auxiliary line with ρm
2

m=2
1
−4
10
5 10 15 20 25 30 0
SNR in dB 5 10 15 20 25 30
SNR in dB

Fig. 4. The outage probability for the n-th user achieved by CR-NOMA, (a) n = m + 1
when n = M . M = 5, R = 1 bit per channel use (BPCU) and I = 5. The
9
analytical results are Theorem 2. m=1, n=M
m=1, n=M−2
8 m=2, n=M
m=2, n=M−2
0 m=M−1, n=M
10 7

I=5 6

5
E(Rn )

4
−1
P(Rn < R)

10
3

2
n=M
n=M−1
n=M−2 I=1 1
n=M−3

−2 0
10 5 10 15 20 25 30
SNR in dB

(b) General Cases

5 10 15 20 25 30
SNR in dB

Fig. 6. The ergodic data rate for the n-th user achieved by CR-NOMA.
Fig. 5. The outage probability for the n-th user achieved by CR-NOMA. M = 5 and I = 5. Analytical results are based on (27).
m = 1, M = 5, and R = 1 BPCU.

channel condition can yield a gain of more than 1 bit per

channel quality of the m-th user becomes a bottleneck for
admitting the n-th user into the same channel.
In Fig. 6 the performance of CR-NOMA is evaluated by
using the ergodic data rate as the criterion. Due to the use
of (21), the m-th user’s QoS can be satisfied, and therefore
we only focus on the n-th user’s data rate, which is the
performance gain of CR-NOMA over conventional MA. Fig.
6 demonstrates that, by fixing (n − m), it is beneficial to
select two users with better channel conditions. While Fig.
5 shows that changing n with a fixed m does not affect the
outage probability, Fig. 6 demonstrates that user pairing has
a significant impact on the ergodic rate. Specifically, when
fixing the choice of m, pairing it with a user with a better
channel use (BPCU) at 30dB. Another interesting observation
from Fig. 6 is that with a fixed n, increasing m will improve
the performance of CR-NOMA, which is different from F-
NOMA. For example, when n = M , Fig. 2 shows that the
user with the worst channel condition, m = 1, is the best
partner, whereas Fig. 6 shows that the user with the second
best channel condition, i.e., m = M − 1, is the best choice.
V. C ONCLUSIONS
In this paper the impact of user pairing on the performance
of two NOMA systems, NOMA with fixed power allocation
(F-NOMA) and cognitive radio inspired NOMA (CR-NOMA),
 8

has been studied. For F-NOMA, both analytical and numerical A PPENDIX
results have been provided to demonstrate that F-NOMA
can offer a larger sum rate than orthogonal MA, and the Proof for Theorem 1: Observe that the sum rate achieved by
performance gain of F-NOMA over conventional MA can be NOMA can be expressed as follows:
further enlarged by selecting users whose channel conditions 
1 + ρ|hm |2



are more distinctive. For CR-NOMA, the channel quality of the Rm + Rn = log 2 2
1 + ρa2n |hn |2 .
user with a poor channel condition is critical, since the transmit ρ|hm | an + 1
power allocated to the other user is constrained following the On the other hand the sum rate achieved by conventional MA
concept of cognitive radio networks. One promising future is given by
direction of this paper is that the analytical results can be used
as criteria designing distributed approaches for dynamic user
12
21
R̄m + R̄n = log 1 + ρ|hm |2 1 + ρ|hn |2 . (28)
pairing/grouping.
Now the addressed probability can be written as follows:

P(Rm + Rn > R̄m + R̄n ) (29)

 2

R EFERENCES 1 + ρ|hm | 2 2


1
2 2
=P 1 + ρa n |h n | > 1 + ρ|h m |
ρ|hm |2 a2n + 1

1 
[1] Q. Li, H. Niu, A. Papathanassiou, and G. Wu, “5G network capacity: × 1 + ρ|hn |2 2
Key elements and technologies,” IEEE Veh. Technol. Mag., vol. 9, no. 1, !
pp. 71–78, Mar. 2014. 1 + ρ|hm |2 1 + ρ|hn |2
=P > 2 .
[2] “5G: A technology vision,” Huawei Technologies Co., Ltd., Shenzhen, (1 + ρa2n |hm |2 )2 (1 + ρa2n |hn |2 )
China, Whitepaper Nov. 2013.
[3] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, “System level After some algebraic manipulations, this probability can be
performance evaluation of downlink non-orthogonal multiple access rewritten as follows:
(NOMA),” in Proc. IEEE Annual Symposium on Personal, Indoor and
Mobile Radio Communications (PIMRC), London, UK, Sept. 2013.
P(Rm + Rn > R̄m + R̄n ) (30)
[4] M. Al-Imari, P. Xiao, M. A. Imran, and R. Tafazolli, “Uplink non-  2

orthogonal multiple access for 5g wireless networks,” in Proc. of the 1 − 2an
= P ρ(|hm |2 + |hn |2 ) + ρ2 |hm |2 |hn |2 > .
11th International Symposium on Wireless Communications Systems a4n
(ISWCS), Barcelona, Spain, Aug 2014, pp. 781–785.
[5] Z. Ding, Z. Yang, P. Fan, and H. V. Poor, “On the performance of The right-hand side of the above inequality is non-negative
non-orthogonal multiple access in 5G systems with randomly deployed since the n-th user will get less power than the m-th user, i.e.,
users,” IEEE Signal Process. Letters, vol. 21, no. 12, pp. 1501–1505, a2n ≤ 12 . Note that the joint pdf of ρ|hm |2 and ρ|hn |2 is given
Dec 2014.
by [13]
[6] J. Choi, “Non-orthogonal multiple access in downlink coordinated two-
point systems,” IEEE Commun. Letters, vol. 18, no. 2, pp. 313–316, M−n
Feb. 2014.
f|hm |2 ,|hn |2 (x, y) = ̟1 f (x)f (y)[F (x)]m−1 (1 − F (y))
[7] T. Cover and J. Thomas, Elements of Information Theory, 6th ed. Wiley × (F (y) − F (x))n−1−m . (31)
and Sons, New York, 1991.
[8] Z. Ding and H. V. Poor, “Cooperative non-orthogonal multiple access In addition, the marginal pdf of |hn |2 is given by
in 5G systems,” IEEE Signal Process. Letters, (submitted) Available
n−1 M−n
on-line at arXiv:1410.5846. f|hn |2 (y) = ̟3 f (y) (F (y)) (1 − F (y)) . (32)
[9] K. Subranhmaniam, “On some applications of Mellin transforms to
statistics: Dependent random variables,” SIAM Journal on Applied By applying the above density functions, the addressed
Mathematics, vol. 19, no. 4, pp. 658–662, Dec. 1970. probability can be expressed as follows:
[10] A. Goldsmith, S. A. Jafar, I. Maric, and S. Srinivasa, “Breaking Z ∞
spectrum gridlock with cognitive radios: An information theoretic
perspective,” Proceedings of the IEEE, vol. 97, no. 5, pp. 894–914,
P(Rm + Rn > R̄m + R̄n ) = f|hn |2 (y)dy (33)
̟2
May 2009. | {z }
[11] G. Zheng, S. Ma, K.-K. Wong, and T.-S. Ng, “Robust beamforming Q2
Z Z
in cognitive radio,” IEEE Trans. Wireless Commun., vol. 9, no. 2, pp.
570–576, Feb. 2010. + f|hm |2 ,|hn |2 (x, y)dxdy .
[12] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and (x+y)+xy>̟2 ,x<y<̟2
Products, 6th ed. New York: Academic Press, 2000. | {z }
Q1
[13] H. A. David and H. N. Nagaraja, Order Statistics. John Wiley, New
York, 3rd ed., 2003. 2 −y
Note that the integral range for x in Q1 is ̟1+y < x <
[14] L. Zheng and D. N. C. Tse, “Diversity and multiplexing : A fundamental ̟2 −y
tradeoff in multiple antenna channels,” IEEE Trans. Inform. Theory, y, and this range implies that 1+y < y, which causes an
√
vol. 49, pp. 1073–1096, May 2003. additional constraint on y, i.e., y > 1 + ̟2 − 1. By applying
 9

the binomial expansion, the joint pdf can be further written as Now the probability Q1 can be approximated as follows:
follows:
X 
n−1−m  Z
n − 1 − m (−1)i ̟2 1 y
 n−1−m−i
Q 1 ≈ ̟1
X 
n−1−m
n−1−m

i=0
i m + i ̟4 ρ ρ
f|hm |2 ,|hn |2 (x, y) = ̟1 (−1)i f (x)  m+i  m+i !
i=0
i y ̟2 − y
M−n
× − dy
× f (y)[F (x)]m−1+i (1 − F (y)) (F (y))n−1−m−i . ρ ρ(1 + y)
n−1−m   Z
̟1 X n − 1 − m (−1)i ̟2 n−1−m−i
Therefore the probability Q1 can now be evaluated as follows: ≈ n y
ρ i=0
i m + i ̟4
 m+i !
n−1−m
! ̟ 2 − y
n − 1 − m (−1)i ̟2 × y m+i −
Z
Q 1 = ̟1
X
f (y)(F (y))n−1−m−i dy. (37)
i m + i ̟4 (1 + y)
i=0
  m+i !
M −n m+i ̟2 − y The high SNR approximation for Q2 is more complicated.
× (1 − F (y)) [F (y)] − F dy.
1+y After applying the series expansion of the exponential func-
(34) tions in (36), we have
i−1
∞ n−1   i+j (M−n+j+1) ̟2i
X ̟3 X n − 1 (−1) ρi−1
On the other hand, Q2 can be calculated as follows: Q2 = (38)
i=0
ρ j=0
j i!
Z ∞ n−1  
∞ X (−1)i ̟3 ̟2i X n − 1
Q2 = ̟3 f (y) (F (y)) n−1
(1 − F (y))M−n
dy (35) = i
(−1)j (M − n + j + 1)i−1 .
i!ρ j
Z̟∞
2
 n−1
i=0 j=0
1 (M −n+1)y y
= ̟3 e − ρ 1 − e− ρ dy. Consider Q2 as a function of ̟2 , and Q2 = 1 is true for
̟2 ρ
̟2 = 0, as can be seen from the definition of Q2 in (33).
On the other hand by letting ̟2 = 0 in (36), we obtain the
By applying the binomial expansion, Q2 can be written as following equality:
follows:
n−1
X 
n−1 1
n−1 ̟3 (−1)j = 1. (39)
̟3 X n − 1 Z ∞
(M −n+j+1)y j M −n+j+1
Q2 = (−1)j
e− ρ dy (36) j=0
ρ j=0
j ̟2
Consequently Q2 can be rewritten as follows:
n−1
X n − 1
̟3 ρ (M −n+j+1)̟2
= (−1)j e− ρ . X∞ n−1 
(−1)i ̟3 ̟2i X n − 1

ρ j=0
j M −n+j +1 Q2 = 1 + (−1)j (40)
i!ρ i j
i=1 j=0
i−1  
Combining (34) with (36), the first part of the theorem is X i−1
× (M − n + 1)i−1−l j l .
proved. l
l=0
To find high SNR approximations for Q1 and Q2 , first
observe that the integral in (34) is calculated for the range Recall the following sums of the binomial coefficients (Eq.
of 0 ≤ y < ̟2 . At high SNR, the two functionsy f (y) and (0.154.3) in [12]):
F (y) can be approximated as follows: f (y) = 1ρ e− ρ ≈ ρ1 and n−1
y X n − 1
F (y) = 1 − e− ρ ≈ yρ , since 0 ≤ y ≤ ̟2 and ρ → ∞. (−1)j j l = 0, (41)
̟2 −y j=0
j
Define u(y) = 1+y . It is straightforward to show
for n − 2 ≥ l ≥ 1 and
0 ≤ u(y) ≤ ̟2 , n−1
X n − 1
(−1)j j n−1 = (−1)n−1 (n − 1)!. (42)
j
for 0 ≤ y ≤ ̟2 , since dg(y)
dy < 0. Therefore at high SNR, we
j=0

can have the following approximation:

Therefore all the components in (40) containing j l , l < (n−1),
  can be removed, since they are equal to zero by using (41).
̟2 − y ̟2 −y ̟2 − y Furthermore, all the components containing j l , l > (n−1) can
F = 1 − e− ρ(1+y) ≈ .
1+y ρ(1 + y) also be ignored, since the one with j = n − 1 is the dominant
 10

b
factor. With these steps, the probability can be approximated b) If |hn |2 < b + aǫ1 , we have aǫ
1− |h 1|2
> |hn |2 ,
as follows: and Q3 can be written as follows:
n

(−1)n ̟3 ̟2n
Q2 ≈ 1 + (−1)n−1 (n − 1)! (43)

Q3 = P aǫ1 < |hn |2 < b + aǫ1 , b < |hm |2 < |hn |2 ,
n!ρn
̟3 ̟2n which is again conditioned on b < aǫ1 .
=1− .
nρn Therefore, the probability Q3 can be written as follows:
Combining (37) and (43), the second part of the theorem is

also proved.  Q3 = P b ≤ |hn |2 < aǫ1 , |hn |2 > |hm |2 > b (48)
2 2 2


+ P |hn | < b + aǫ1 , b < |hm | < |hn |
Proof for Theorem 2: !
Recall that the outage performance of the n-th user is given 2 2 b
+ P |hn | > b + aǫ1 , b < |hm | < .
by 1 − |haǫn1|2


P log(1 + a2n ρ|hn |2 ) < R (44) The first probability in (48) can be calculated by applying (32)
! !
2 I
|hm | − ρ I as follows:
=P log 1 + ρ|hn |2 < R, |hm |2 >

2
|hm | (1 + I) ρ P b ≤ |hn |2 < aǫ1 , |hn |2 > |hm |2 > b
| {z } n−1−m
X n − 1 − m Z aǫ1
Q3 i M−n
  = (−1) g(y) (1 − G(y))
2 I i=0
i b
+ P |hm | < . Z y
ρ
| {z } × G(y)n−1−m−i ̟1 g(x)[G(x)]m−1+i dxdy
Q4 b
The first factor in the above equation can be calculated as X 
n−1−m
n−1−m
 Z aǫ1
follows: = (−1) i
g(y) (1 − G(y))M−n
! i=0
i b
|hm |2 − ρI I m+i


Q3 =P 2 2
|hn | < ǫ1 , |hm | > . (45) n−1−m−i G(y) − G(b)m+i
|hm |2 (1 + I) ρ × G(y) ̟1 dy.
m+i
Recall that the users’ channels are ordered, i.e., |hm |2 < |hn |2 , Following similar steps, the second probability in (48) can be
which brings additional constraints to the integral range in the expressed as
above equation. The constraints can be written as follows:

( ) P aǫ1 < |hn |2 < b + aǫ1 , b < |hm |2 < |hn |2 (49)
b n−1−m   Z
b < |hm |2 < min |hn |2 , . (46) X n−1−m b+aǫ1
M−n
1 − |haǫn1|2 = (−1)i g(y) (1 − G(y))
i=0
i aǫ1
The outage events due to these constraints can be classified as m+i


follows: n−1−m−i G(y) − G(b)m+i
× G(y) ̟1 dy.
1) If |hn |2 < aǫ1 , we have the following: m+i
! The third probability in (48) can be calculated as follows:
|hm |2 − Iρ
P |hn |2 < ǫ1 (47) !
|hm |2 (1 + I) 2 2 b
P |hn | > b + aǫ1 , b < |hm | < (50)
1 − |haǫn1|2


= P |hm |2 (|hn |2 − ǫ1 a) < b|hn |2 = 1.
n−1−m
X n − 1 − m Z ∞
Therefore the probability Q3 can be expressed as fol- i M−n
lows: 1 = (−1) g(y) (1 − G(y))

i=0
i b+aǫ1
Q3 = P b ≤ |hn |2 < aǫ1 , |hn |2 > |hm |2 > b . Z b
aǫ1
1−

2) If |hn |2 > aǫ1 , there are two possible events: × G(y)n−1−m−i |hn |2
̟1 g(x)[G(x)]m−1+i dxdy
b
a) If |hn |2 > b + aǫ1 , we have 1− baǫ1 < |hn |2 , n−1−m
X   Z ∞
|hn |2 n−1−m M−n
and Q3 can be written as follows: = (−1)i g(y) (1 − G(y))
! i=0
i b+aǫ1
b  m+i !
2 2
Q3 = P |hn | > b + aǫ1 , b < |hm | < . b
1− aǫ1
|hn |2
G 1− aǫ1 − G(b)m+i
|hn |2

1 It
× G(y)n−1−m−i ̟1 dy.
is assumed that b ≤ aǫ1 here. For the case of b > aǫ1 , the outage m+i
probability can be calculated in a straightforward way, since there will be
fewer events to analyze. Note that the same diversity order will be obtained Note that Q4 can be obtained easily by applying (32) and the
regardless of the choice of b and aǫ1 . first part of the theorem is proved.
 11

Recall that the first probability in (48) can be expressed as Again applying the binomial expansion, the above probability
follows: can be further expanded as follows:
 
|hm |2 aǫ1


P b ≤ |hn |2 < aǫ1 , |hn |2 > |hm |2 > b 2
P b < |hm | < b + aǫ1 , b + aǫ1 < |hn | <2
n−1−m
X  n − 1 − m Z aǫ1 |hm |2 − b
i
= ̟1 (−1) g(y) n−1−m
X 
n−1−m
 Z b+aǫ1
i=0
i b = ̟1 (−1)i g(x)[G(x)]m−1+i
m+i

i b
M−n n−1−m−i G(y) − G(b)m+i i=0
× (1 − G(y)) G(y) dy, M−n
X M − n Z G( xaǫ1
)
m+i x−b
n−1−m−i+j
× (−1)j [G(y)] dG(y)dx
where the integral range is 0 ≤ y ≤ (aǫ1 ). Note that when j=0
j G(b+aǫ1 )
ρ → ∞, ǫ1 approaches zero, which means y → 0, g(y) ≈ 1 n−1−m   Z b+aǫ1
and G(y) ≈ 1 − y. Therefore the above probability can be
X n−1−m
= ̟1 (−1)i g(x)[G(x)]m−1+i
approximated as follows: i=0
i b

M−n
P b ≤ |hn |2 < aǫ1 , |hn |2 > |hm |2 > b (51) X M − n (−1) j

n−1−m   × (53)
X n−1−m j=0
j n−1−m−i+j
≈ ̟1 (−1)i !
i=0
i  
xaǫ1
n−m−i+j
n−m−i+j
Z aǫ1 m+i

× G − [G(b + aǫ1 )] dx.
n−1−m−i y − bm+i x−b
× y dy
b m+i Compared to (50), the above equation is more complicated;
n−1−m
X n − 1 − m however, this expression is more suitable for asymptotic stud-
≈ ̟1 (−1)i ies, as explained in the following.
i
 i=0
  Recall that the integral range in (53) is b < x < b + aǫ1 .
(aǫ1 )n −bn m+i (aǫ1 )
n−m−i
−bn−m−i When ρ → 0, we have b → 0 and b + aǫ1 → 0, which
m+i+1 − b n−m−i
× → ρ−n . implies x → 0. Therefore the following approximation can be
m+i obtained:
 
Following similar steps, the second probability in (48) can 2 2 |hm |2 aǫ1
be approximated as follows: P b < |hm | < b + aǫ1 , b + aǫ1 < |hn | <
|hm |2 − b


P aǫ1 < |hn |2 < b + aǫ1 , b < |hm |2 < |hn |2 → ρ−n . (52) (54)
n−1−m  
The exact diversity order of the third probability in (48) is X n−1−m
difficult to obtain. Particularly the expression in (50) is difficult ≈ ̟1 (−1)i
i=0
i
to use for asymptotic studies, since the range of y is not limited
M−n
X M − n Z b+aǫ1
and those manipulations related to high SNR approximations (−1)j
cannot be applied here. We first rewrite (50) in an alternative × xm−1+i
j=0
j n − 1 − m − i + j b
form as follows: !
!   n−m−i+j
xaǫ1 n−m−i+j
b × G − [b + aǫ1 ] dx.
P |hn |2 > b + aǫ1 , b < |hm |2 < x−b
1 − |haǫn1|2

|hm |2 aǫ1
 First focus on the following integral which is from the above
= P b < |hm |2 < b + aǫ1 , b + aǫ1 < |hn |2 < . equation:
|hm |2 − b Z b+aǫ1   n−m−i+j
m−1+i xaǫ1
|hm |2 aǫ1 x G dx (55)
Note that b + aǫ1 < |hm |2 −b always holds since |hm |2 < b x−b
b Z aǫ1
aǫ1 .
h i
abǫ1 n−m−i+j
1− |hn |2 ≈ (b + z)m−1+i 1 − e− z dz.
Now applying the joint pdf of the two channel coefficients, 0
we obtain the following expression: We can find the following bounds for the above integral:
 
|hm |2 aǫ1
Z b+aǫ1
P b < |hm |2 < b + aǫ1 , b + aǫ1 < |hn |2 < xm−1+i dx (56)
|hm |2 − b b
n−1−m   Z b+aǫ1 Z b+aǫ1   n−m−i+j
X n−1−m xaǫ1
= ̟1 (−1)i g(x)[G(x)]m−1+i ≥ xm−1+i G dx
i b b x −b
i=0 " #
Z G( xaǫ Z aǫ1 n−m−i+j
x−b )
1
m−1+i 1
× [G(y)]n−1−m−i (1 − G(y))M−n dG(y)dx. ≥ (b + z) 1− dz,
G(b+aǫ1 ) 0 1 + abǫ
z
1
 12

where the lower bound is obtained due to the inequality

abǫ1 1
e− z ≤ ,
1 + abǫ
z
1

when 0 ≤ z ≤ aǫ1 . The upper bound in (56) can be

approximated at high SNR as follows:
Z b+aǫ1
xm−1+i dx (57)
b
(b + aǫ1 )m+i − bm+i 1
= → m+i .
m+i ρ
On the other hand, the lower bound in (56) can be approxi-
mated as follows:
Z aǫ1 " #n−m−i+j
m−1+i 1
(b + z) 1− dz (58)
0 1 + abǫ
z
1

Z aǫ1 +abǫ1
n−m−i+j (w + b − abǫ1 )m−1+i
= (abǫ1 ) dz
abǫ1 wn−m−i+j
m−1+i
X
= (abǫ1 )n−m−i+j (b − abǫ1 )m−1+i−k
k=0
Z aǫ1 +abǫ1 m−1+i
X
× wk−(n−m−i+j) dz , ξk .
abǫ1 k=0

At high SNR, we can show that

 −(n+j)
ρ ln ρ, for k + 1 = n − m − i + j
ξk → . (59)
ρ−(n+j) , otherwise
Since logloglogρ ρ → 0 for ρ → ∞, the lower bound in (56) can
be approximated as follows:
Z aǫ1  n−m−i+j
abǫ1
(b + z)m−1+i dz → ρ−(n+j) . (60)
0 z + abǫ1
Based on the upper and lower bounds in (57) and (60)
and after some algebraic manipulation, we have the following
inequality:
!
−n ˙ 2 2 b ˙ −m , (61)
ρ ≤P |hn | > b + aǫ1 , b < |hm | < aǫ1 ≤ρ
1− |hn |2
   
˙ denotes − log a ≤ − log b when ρ → ∞ [14].
where a≤b log ρ log ρ
Combining (51), (52) and (61), we can obtain the following
asymptotic bounds:
ρ−n ≤Q ˙ −m .
˙ 2 ≤ρ (62)
Following similar steps as above, we can also find that
.
Q3 = ρ−m , which is dominant in Pno , and the proof for the
second part of the theorem is completed. 

View publication stats