Energy efficiency and sum rate tradeoffs for massive MIMO systems with underlaid devicetodevice communications
 Serveh Shalmashi^{1}Email authorView ORCID ID profile,
 Emil Björnson^{2},
 Marios Kountouris^{3},
 Ki Won Sung^{1} and
 Mérouane Debbah^{3}
https://doi.org/10.1186/s1363801606781
© Shalmashi et al. 2016
Received: 30 March 2016
Accepted: 20 July 2016
Published: 29 July 2016
Abstract
In this paper, we investigate the coexistence of two technologies that have been put forward for the fifth generation (5G) of cellular networks, namely, networkassisted devicetodevice (D2D) communications and massive MIMO (multipleinput multipleoutput). Potential benefits of both technologies are known individually, but the tradeoffs resulting from their coexistence have not been adequately addressed. To this end, we assume that D2D users reuse the downlink resources of cellular networks in an underlay fashion. In addition, multiple antennas at the BS are used in order to obtain precoding gains and simultaneously support multiple cellular users using multiuser or massive MIMO technique. Two metrics are considered, namely the average sum rate (ASR) and energy efficiency (EE). We derive tractable and directly computable expressions and study the tradeoffs between the ASR and EE as functions of the number of BS antennas, the number of cellular users and the density of D2D users within a given coverage area. Our results show that both the ASR and EE behave differently in scenarios with low and high density of D2D users, and that coexistence of underlay D2D communications and massive MIMO is mainly beneficial in low densities of D2D users.
Keywords
1 Introduction
The research on future mobile broadband networks, referred to as the fifth generation (5G), has started in the past few years. In particular, stringent key performance indicators (KPIs) and tight requirements have been introduced in order to handle higher mobile data volumes, reduce latency, increase the number of connected devices and at the same time increase the energy efficiency (EE) [1, 2]. The current network and infrastructure cannot cope with 5G requirements—fundamental changes are needed to handle future heterogeneous deployments as well as new trends in user behavior such as high quality video streaming and future applications such as eHealth and virtual reality. 5G technology is supposed to evolve existing networks and at the same time integrate new dedicated solutions to meet the KPIs [2]. The new key concepts for 5G include massive MIMO (multipleinput multipleoutput), ultra dense networks (UDN), devicetodevice (D2D) communications, and huge number of connected devices, known as machinetype communications (MTC). The potential gains and properties of these different solutions have been studied individually, but the realistic gains when they coexist and share network resources are not very clear so far. In this paper, we study the coexistence of two of these key concepts, namely massive MIMO and D2D communication.
Massive MIMO is a type of multiuser MIMO (MUMIMO) technology where the base station (BS) uses an array with hundreds of active antennas to serve tens of users on the same time/frequency resources by coherent transmission processing [3, 4]. Massive MIMO techniques are particularly known to be very spectralefficient, in the sense of delivering high sum rates for a given amount of spectrum [5]. This comes at the price of deploying more transceiver hardware, but the solution is still likely to improve the energy efficiency of networks [6, 7]. On the other hand, in a D2D communication, user devices can communicate directly with each other and the user plane data is not sent through the BS [8]. D2D communication is considered for close proximity applications which have the potential to achieve high data rates with little amount of transmission energy, if interference is wellmanaged. In addition, D2D communications can be used to decrease the load of the core network. D2D users either have their own dedicated time/frequency resources (overlay approach), which in turn leads to elimination of the crosstier interference between the two types of users (i.e., cellular and D2D users), or they transmit simultaneously with cellular users in the same resource (underlay approach). In the 3rd Generation Partnership Project (3GPP) standardization’s document [9], the underlay case for D2D communications is recommended in the uplink direction, while at the same time, downlink reuse in the time devision duplexing (TDD) scenario is considered for future study. Even though the majority of studies in this area have focused on the uplink, authors in [10] and [11] show the importance of downlink reuse in singleantenna settings. In [10], machinetype communication is enabled via D2D communication where fixed rate zerooutage downlink transmission is achieved, and the work in [11] shows that lower outage probability can be achieved in the downlink compared to the uplink over consecutive time slots in a multicell environment.
We consider two network performance metrics in this work: The average sum rate (ASR) in bit/s and the EE which is defined as the number of bits transmitted per Joule of energy consumed by the transmitted signals and the transceiver hardware. It is wellknown that these metrics depend on the network infrastructure, radio interface, and underlying system assumptions [7, 12, 13]. The motivation behind our work is to study how the additional degrees of freedom resulting from the high number of antennas in the BS can affect the ASR and EE of a multitier network where a D2D tier is bypassing the BS, and how a system with massive MIMO is affected by adding a D2D tier. We focus on the downlink since greater part of the payload data and network energy consumption are ascociated to the downlink [12]. We assume that each D2D pair is transmitting simultaneously with the BS in an underlay fashion. In addition, we assume that the communication mode of each user (i.e., D2D or cellular mode) has already been decided by higher layers. We compare the energy efficiency (EE) and average sum rate (ASR) gains that massive MIMO [6] has been claimed to provide with similar EE and ASR gains that D2D communications can provide.
1.1 Related work
The relation between the number of BS antennas, ASR and EE in cellular networks has been studied in [6, 7, 14, 15] among others. The tradeoff between ASR and EE was described in [6] for massive MIMO systems with negligible circuit power consumption. This work was continued in [14] where radiated power and circuit power were considered. In [7], joint downlink and uplink design of a cellular network was studied in order to maximize EE for a given coverage area. The maximal EE was achieved by having a hundred BS antennas and serving tens of users in parallel, which matches well with the massive MIMO concept. Furthermore, the study [15] considered a downlink scenario in which a cellular network has been overlaid by small cells. It was shown that by increasing the number of BS antennas, the array gain allows for decreasing the radiated signal energy while maintaining the same ASR. However, the energy consumed by the transceiver chains increases. Maximizing the EE is thus a complicated problem where several counteracting factors need to be balanced. This stands in contrast to the maximization of the ASR, which is relatively straightforward since the sum capacity is the fundamental upper bound for the ASR.
There are a few works in the D2D communication literature where the base stations have multiple antennas [16–20]. In [16], uplink MUMIMO with one D2D pair was considered. Cellular user equipments (CUEs) were scheduled if they were not in the interferencelimited zone of the D2D user. The study [17] compared different multiantenna transmission schemes. In [18, 19], two power control schemes were proposed for a multicell MIMO network. In [20], the ergodic capacity for a scenario with only one CUE and one D2D user is derived and cases of high and low SNR as well as high number of antennas in the downlink have been studied.
The more relevant works to our setup are [21, 22]. The former investigates the mode selection problem in the uplink of a network with potentially many antennas at the BS. The impact of the number of antennas on the qualityofservice (QoS) and transmit power was studied when users need to decide their mode of operation (i.e., D2D or cellular). The study [22] only employs extra antennas in the network to protect the CUEs from interference of D2D users in the uplink.
There are other related works in the context of massive MIMO and D2D communications which study different angles, such as [23] which uses D2D to enable local CSI exchanges in a frequency devision duplexing (FDD) massive MIMO system and [24] which exploits D2D communications to create virtual MIMO in order to avoid huge feedback overhead for CSI acquisition in downlink of a massive MIMO FDD system. Another study [25] considers the problem of BS precoder design and power allocation in multicasting massive MIMO with underlaid D2D communications. The ASR in D2D communications is mostly studied in the context of interference and radio resource management [26, 27]. There are a few works that consider EE in D2D communications, but only for singleantenna BSs, e.g., [28, 29], and [30], where the first one proposed a coalition formation method, the second one designed a resource allocation scheme, and the third one aimed at prolonging the battery life of user devices.
The spatial degrees of freedom offered by having multiple antennas at BSs are very useful in the design of future mobile networks, because spatial precoding enables dense multiplexing of users while keeping the interuser interference under control. In particular, the performance for cell edge users, which have almost equal signaltonoise ratios (SNRs) to several BSs, can be greatly improved since only the desired signals are amplified by the transmit precoding [31–33]. In order to model the random number of users and random user positions, we use mathematical tools from stochastic geometry [34] which are powerful in analytically quantifying certain metrics in closedform. Some work in the literature of D2D communications in singleantenna systems that exploits these tools can be found in [35–40]. There are certain studies in the context of MUMIMO for single tier network in stochastic geometry considering equal or smaller than number of antennas and users like [41, 42]. In this paper, our analysis holds for any number of antennas, but the simulations investigate mostly “massive MIMO regimes” with many more antennas than users.
1.2 Contributions

A tractable model for underlaid D2D communication in massive MIMO systems: we model a twotier network with two different user types. The firsttier users, i.e., CUEs, are served in the downlink by a BS using massive multiuser MIMO precoding to cancel interference. The secondtier users, i.e., D2D users, exploit their close proximity and transmit simultaneously with the downlink cellular transmissions bypassing the BS. The number of D2D transmitters and their locations are modeled according to a homogeneous Poisson point process (PPP) while a fixed number of CUEs are randomly distributed in the network.

Tractable and directly computable expressions: we derive tightly approximated expressions for the coverage probability of D2D users and CUEs. These expressions are directly used to compute our main performance metrics, namely, the ASR and EE. We verify the tightness of these approximations by Monte Carlo simulations. Furthermore, we provide analytical insights on the behavior of these metrics for both CUEs and D2D users.
To the best of our knowledge, the energy efficiency analysis for underlay D2D communications in a network with large number of BS antennas has not been carried out before.

Performance analysis: based on extensive simulations, we characterize the typical relation between the ASR and EE metrics in terms of the number of BS antennas, the number of CUEs, and the D2D user density for a given coverage area and study the incurred tradeoffs in two different scenarios. The modeling and comparative study is an important contribution of the paper.
2 System model
We consider a singlecell scenario where the BS is located in the center of the cell and its coverage area is a disc of radius R. The BS serves U _{ c } singleantenna CUEs which are uniformly distributed in the coverage area. These are simultaneously served in the downlink using an array of T _{ c } antennas located at the BS. It is assumed that 1≤U _{ c }≤T _{ c } so that the precoding can be used to control the interference caused among the CUEs [43].
Let R _{ k,j } denote the distance between the jth D2D Tx to the kth D2D Rx.
The performance analysis for D2D users is carried out for a typical D2D user, which is denoted by the index 0. The typical D2D user is an arbitrary D2D user located in the cell and its corresponding receiver is positioned in the origin. The results for a typical user show the statistical average performance of the network [34]. Therefore, for any performance metric derivation, the D2D users inside the cell are considered and the ones outside the cell are only taken into account as sources of interference. Note that we neglect potential interference from other BSs and leave the multicell case for future work. This is because a cellular user or a D2D receiver at the cell edge will see much more interference from D2D transmitters than from the BS in a neighboring cell, simply because the D2Ds are much closer (e.g., at the other side of the cell edge) leading to potentially huge proximity gains.
We assume equal power allocation for both CUEs and D2D users. Let P _{ c } denote the total transmit power of the BS, then the transmit power per CUE is \(\frac {P_{c}}{U_{c}}\). The transmit power of the D2D Tx is denoted by P _{ d }.
is the received interference power from all D2D users (normalized by A _{ d }).
3 Performance analysis
In this section, we first introduce the performance metrics that are considered in this paper. Then, we proceed to derive the coverage probability for both CUEs and D2D users which are needed to compute these metrics.
3.1 Performance metrics
In this paper, two main performance metrics for the network are considered: the average sum rate (ASR) and energy efficiency (EE). The metrics used here are aligned to requirements that are demanded in 5G [1, 46, 47]. We investigate this scenario in order to get an understanding of how such coexistence would perform. Another important fact is that to the best of our knowledge no one has compared the EE and ASR performance of D2D communication in a massive MIMO system. Both of these technologies are known to bring high ASR and are likely to be more energyefficient. However, there is no work in literature showing the impact of high number of antennas and cellular users along with the density of D2D users in such a setting.
is the coverage probability when the received SINR is higher than a specified threshold β _{ t } needed for successful reception. Note that SINR_{ t } contains random channel fading and random user locations. Finding the supremum guarantees the best constant rate for the D2D users and the CUEs. If we know the coverage probability (\(\mathrm {P}^{t}_{\text {cov}} (\beta _{t})\)), (10) can easily be computed by using line search for each user type independently. Moreover, (10) is easily achievable in practice since the modulation and coding is performed without requiring that every transmitter knows the interference characteristics at its receiver.
where P _{ c }+λ _{ d } π R ^{2} P _{ d } is the total transmission power averaged over the number of D2D users, η is the amplifier efficiency (0<η≤1), C _{0} is the load independent power consumption at the BS, C _{1} is the power consumption per BS antenna, C _{2} is the power consumption per user device, and U _{ c }+2λ _{ d } π R ^{2} is the average number of active users.
In order to calculate the ASR and EE, we need to derive the coverage probability for both cellular and D2D users. The analytic derivation of these expressions is one of the main contributions of this paper.
3.2 Coverage probability of D2D users
We first derive the expression for the coverage probability of D2D users.
Proposition 1
where \(\kappa \triangleq \frac {\zeta }{P_{d} A_{d} R_{0,0}^{\alpha _{d}}}\) with ζ defined in (5), \(y \triangleq \frac {1}{\kappa \beta _{d} R^{\alpha _{c}} + 1}\), \(\text {sinc}(x) = \frac {\sin (\pi x)}{\pi x}\), \(\bar {\gamma }_{d} = \frac {A_{d} R_{0,0}^{\alpha _{d}}P_{d}}{N_{0}}\) is the average D2D SNR, and \(\mathcal {B}(x;a,b)\) is the incomplete Beta function.
Proof
The proof is given in Appendix 1: Proof of Proposition 1. □
The coverage probability expression in Proposition 1 allows us to compute the average data rate of a typical D2D user in (10). The approximation in this proposition is due to neglecting the spatial interference correlation resulting from the fact that multiple interfering streams are coming from the same location (more details can be found in Appendix 1: Proof of Proposition 1). We note that (14) is actually a tight approximation and its tightness is evaluated in Section 4. From the expression in (14), we make several observations as listed below.
Remark 1
This can also be referred to as the interferencelimited regime.
Remark 2
The coverage probability of a typical D2D user is a decreasing function of the D2D density λ _{ d }. Because higher λ _{ d } results in more interference among D2D users. In particular, it can be seen that \(\mathrm {P}^{d}_{ {\text {cov}}}\) in (14) is a function of λ _{ d } through exp(−C λ _{ d }) with \(C\triangleq \frac {\pi R_{0,0}^{2} \beta _{d}^{2/\alpha _{d}}}{{\text {sinc}}\left (\frac {2}{\alpha _{d}}\right)} > 0\). Thus, if λ _{ d }→∞, \(\mathrm {P}^{d}_{{\text {cov}}} \to 0\). Recall that in our model, the D2D Rx is associated to the D2D Tx which is located at a fixed distance away. However, if we had assumed that the D2D Rx’s association to a D2D Tx is based on, for example, the shortest distance or the maximum SINR, then the \(\mathrm {P}^{d}_{{\text {cov}}}\) would have been unaffected by the D2D density (in the highinterference regime). Similar observation can be found in [ 49 , 50 ].
Now, considering the number of BS antennas or the number of CUEs as variables, we have the following behavior of the D2D coverage probability.
Remark 3
\(\mathrm {P}^{d}_{{\text {cov}}}\) is not affected by the number of BS antennas T _{ c }. The BS antennas are used to cancel out the interference among CUEs and they do not have any impact on D2D users’ performance as long as the number of CUEs U _{ c } is constant and does not vary with the number of BS antennas T _{ c }. The coverage probability of a typical D2D user \(\mathrm {P}^{d}_{{\text {cov}}}\) is a decreasing function of U _{ c }. However, increasing the number of CUEs have a small effect on D2D users’ performance. This is due to the fact that the resulting interference from the BS to D2D users does not change significantly by increasing the number of CUEs as the transmit power of the BS is the same irrespective of the number of users and the precoding is independent of the D2D channels. Thus, a change of U _{ c } will only change the distribution of the interference but not its average.
Next, we comment on how changes in the transmit powers of the BS and D2D Tx as well as the distance between D2D user pairs affect the coverage probability of D2D users.
Remark 4
\(\mathrm {P}^{d}_{{\text {cov}}}\) is a decreasing function of the ratio between the transmit power of the BS and of the D2D users, i.e., \(\frac {P_{c}}{P_{d}}\), which is part of the first term in (14) and corresponds to the interference from the BS. For instance, if we fix P _{ c } and decrease P _{ d }, the coverage probability for D2D users decreases as the interference from the BS would be the dominating factor. At the same time, if we decrease P _{ c }, it would improve the coverage of D2D users.
Remark 5
\(\mathrm {P}^{d}_{{{\text {cov}}}}\) is a decreasing function of the distance between D2D TxRx pairs R _{0,0} and the cell radius R. Increasing the cell radius with the same D2D user density reduces the effect of the interference from the BS. Also by decreasing the distance between D2D TxRx pairs, it is evident that a better performance for D2D users can be obtained.
Using Proposition 1, the following corollary provides the optimal D2D user density that maximizes the D2D ASR, i.e., \(\pi R^{2} \lambda _{d} \bar {R}_{d}\), where \(\bar {R}_{d}\) is given in (10). The optimal density is also evident in our numerical results in Section 4.
Corollary 1
Proof
where \(\mathrm {P}^{d}_{\text {cov}}(\beta _{d})\) is given in (14) and depends on λ _{ d } through an exponential function. Taking the second derivative of (17) with respect to λ _{ d }, we observe that for \(\lambda _{d} < 2 \frac {\text {sinc}\left (\frac {2}{\alpha _{d}}\right)}{\pi R_{0,0}^{2}}\beta _{d}^{2/\alpha _{d}}\), the function is concave. Therefore, setting the first derivative of (17) with respect to λ _{ d } to zero yields the optimal D2D user density \(\lambda _{d}^{*}(\beta _{d})\) given in (16) that maximizes the D2D ASR. □
3.3 Coverage probability of cellular users
Next, we compute the coverage probability for CUEs.
Proposition 2
Proof
The proof is given in Appendix Appendix 2: Proof of Proposition 2. □
This proposition gives an expression for the coverage probability of CUEs in which there is only one random variable left. The expectation in (18) with respect to D _{0,BS} is intractable to derive analytically but can be computed numerically. The analytical results of Proposition 1 and Proposition 2 have been verified by Monte Carlo simulations in Section 4. A main benefit of the analytic expressions (as compared to pure Monte Carlo simulations with respect to all sources of randomness) is that they can be computed much more efficiently, which basically is a prerequisite for the multivariable system analysis carried out in Section 4.
Next, we present some observations from the result in Proposition 2 as follows.
Remark 6
The result obtained in Remark 6 has a lower computational complexity compared to the expression in Proposition 2 and at the same time it is a tight approximation for Proposition 2. This can be observed from the denominator of the (7) where the term \(\frac {N_{0}}{A_{d} }\approx 0\).
Remark 7
The coverage probability of a typical CUE \(\mathrm {P}^{c}_{\text {cov}}(\beta _{c})\) is a decreasing function of the D2D user density λ _{ d }. From Proposition 2, only Υ(λ _{ d },s,i) is a function of λ _{ d } which is composed of an exponential term in λ _{ d } multiplied by a polynomial term in λ _{ d }. Thus, if λ _{ d }→∞, the exponential term which has a negative growth dominates the polynomial term and \(\mathrm {P}^{c}_{{\text {cov}}}(\beta _{c}) \to 0\).
We proceed to analyze the behavior of Proposition 2 by considering a number of special cases. The impact of these special cases is also corroborated in our numerical results in Section 4.
Corollary 2
where \(s = \frac {A_{d}}{\zeta } D_{0,{\text {BS}}}^{\alpha _{c}} \beta _{c}\) and \(C_{d}=\frac {\pi P_{d}^{2/\alpha _{d}}}{{\text {sinc}}\left (\frac {2}{\alpha _{d}}\right)}\).
Corollary 2 applies for any case of MUMIMO and massive MIMO is a form of MUMIMO [52,53]. The important distinction is that MUMIMO has traditionally been considered for the case of equal number of antennas and users, while massive MIMO employs a large number of antennas compared to the number of users [52 , 53]. As a ruleofthumb, T _{ c }>50 and T _{ c }/U _{ c }>2 are required to exploit the favorable propagation of massive MIMO [52]. Next, we consider the case where massive number of antennas are deployed in the BS.
Corollary 3
Proof
Corollary 3 gives an indication that the desired signal can be amplified by adding more antennas. However, note that even if the power gain becomes much stronger than the D2D interference, it will, in practice, eventually become limited by pilot contamination, hardware distortion, and/or finite modulation sizes.
In the results so far, we have discussed the case where there exist some D2D users as underlay to the cellular network, that is, λ _{ d }≠0, However, it is interesting to see what can be achieved without D2D users.
Corollary 4
where Γ(·)is the Gamma function and ζ is defined in (5).
Proof
where (a) follows from the CCDF of \(\mathbf {h}_{0}^{H} \mathbf {v}_{0}^{2}\) with \(2\mathbf {h}_{0}^{H} \mathbf {v}_{0}^{2} \sim {\chi ^{2}_{2}}\) given D _{0,BS} and setting \(l= \frac {N_{0}}{\zeta } \beta _{c}\) and \(z = D_{0,\text {BS}}^{\alpha _{c}}\) with PDF \(f(z)=\frac {2}{\alpha _{c}R^{2}}z^{\frac {2}{\alpha _{c}}1}\). Step (b) follows from taking the expectation with respect to z which is similar to the expression in (35) with the Laplace transform \( \mathcal {L}_{z}(l) = \frac {2}{\alpha _{c} R^{2}}\Gamma \left (\frac {2}{\alpha _{c}}\right)l^{2/\alpha _{c}}\). Simplifying the kth derivative to \(\frac {\mathrm {d}^{k}}{\mathrm {d}l^{k}} ~l^{2/\alpha _{c}} = (1)^{k} l^{\frac {2}{\alpha _{c}}k} \prod _{i=0}^{k1}\left (\frac {2}{\alpha _{c}}+i\right)\) and using the identity \(\frac {1}{k!}\prod _{i=0}^{k1}\left (\frac {2}{\alpha _{c}}+i\right) = {\frac {2}{\alpha _{c}}+k1 \choose k}\), (23) follows. □
The closedform results in Corollary 4 for λ _{ d }=0 depends only on noise rather than interference and perhaps can result in higher ASR for CUEs. The ASR for λ _{ d }>0 also depend on noise but its impact is much smaller. However, we note that this result is obtained for a singlecell scenario. Thus, comparing Proposition 2 and Corollary 4 and evaluating the potential performance gain/loss due to introducing D2D communications would make more sense in a multicell scenario.
Using the results from Proposition 1 and Proposition 2, we proceed to evaluate the network performance in terms of the ASR and EE from (9) and (12), respectively.
4 Numerical results
System and simulation parameters
Description  Parameter  Value 

D2D TX power  P _{ d }  6 dBm 
BS TX power  P _{ c }  30 dBm 
Cell radius  R  500 m 
Bandwidth  B _{ w }  20 MHz 
Thermal noise power  N _{0}  −101 dBm 
Noise figure in UE  F  5 dB 
Carrier frequency  f _{ c }  2 GHz 
D2D pair distance  R _{0,0}  35 m 
Pathloss exponent between devices  α _{ d }  3 
Pathloss exponent between BS–device  α _{ c }  3.67 
Pathloss coefficient between devices  A _{ d }  38.84 dB 
Pathloss coefficient between BS–device  A _{ c }  30.55 dB 
Amplifier efficiency  η  0.3 
Loadindependent power in BS  C _{0}  5 W 
Power per BS antenna  C _{1}  0.5 W 
Power per UE handset  C _{2}  0.1 W 
Monte Carlo runs  MC  5000 
We consider two scenarios corresponding to the number of CUEs U _{ c } in our evaluations. First, we assume that U _{ c } is chosen as a function of the number of BS antennas T _{ c }. Then, we move on to the case where we fix the number of CUEs and study the tradeoffs among other parameters. Both scenarios are relevant in the design of massive MIMO systems. In order to speed up the numerical computations, based on the insight obtained in Remark 6, we neglected the terms that are very small.
4.1 Number of CUEs as a function of the number of BS antennas
In this scenario, we assume that there is a fixed ratio between the number of CUEs U _{ c } and the number of BS antennas T _{ c }. We assume this ratio to be \(\frac {T_{c}}{U_{c}}=5\). Simply put, to serve one additional user, we add five more antennas at the BS since the main gains from massive MIMO come from multiplexing of many users rather than only having many antennas.
As seen in Fig. 4 a, for U _{ c }=1 user and T _{ c }=5 antennas, the rate contributed from the CUEs to the sum rate is low as there is only one CUE. This rate is in a comparable level as the contribution of D2D users sum rate to the total ASR. Adding D2D users to the network (i.e., increasing λ _{ d }), which may cause interference, will nevertheless leads to an increase in the ASR. This increase in the ASR continues until reaching a certain density that gives the maximum ASR. By further increasing λ _{ d }, the interference between D2D users reduces their coverage probability as previously observed in Remark 2. This limits the per link data rate and even a high number of D2D users cannot compensate for the D2D rate loss. At the same time, increasing λ _{ d } tremendously affects the CUEs sum rate (cf. Remark 7). Consequently, as λ _{ d } increases, the ASR decreases.
By increasing the number of CUEs and BS antennas to U _{ c }=14 users and T _{ c }=70 antennas, respectively, in Fig. 4 a, the average rates of the CUEs become higher than the case with U _{ c }=1 user and T _{ c }=5 antennas as expected from Corollary 3 and the multiplexing gain from having many CUEs. However, by introducing a small number of D2D users, there is a substantial probability that the interference from the D2D users reduces the CUEs’ rates per link as observed in Remark 7. The reduction in these rates are not compensated in the ASR by the contribution of the D2D users’ rates. Note that, as we stated in Remark 3, when U _{ c } is scaled with T _{ c }, it impacts the D2D coverage probability, but the decrease in the performance of D2D users is not significant. Furthermore, if we keep increasing λ _{ d }, even though the rate per link decreases for both CUEs and D2D users, there is a local minima after which the aggregate D2D rate over all D2D users becomes higher and the ASR increases again. The second turning point follows from the same reasoning as for the case of U _{ c }=1 user and T _{ c }=5 antennas, i.e., in higher D2D densities, the interference from D2D users are the limiting factor for the ASR. This effect can also be observed in Fig. 4 b where the ASR performance is depicted versus different number of CUEs (and BS antennas) for two D2D densities. At the lower density, the ASR is linearly increasing with U _{ c } (and T _{ c }), however, in the interferencelimited regime (higher λ _{ d }), increasing the number of CUEs and BS antennas do not impact the network ASR performance.
Furthermore, if we plot the EE versus U _{ c }, we see a different behavior for low and high D2D densities. Figure 7 b illustrates that in the low D2D density regime (λ _{ d }=10^{−6}), even though the ASR increases linearly, the EE almost stays the same as the number of CUEs, and correspondingly the number of BS antennas, increases. From (13), we can observe that for a fixed λ _{ d }, only the circuit power is changed by increasing U _{ c } and T _{ c }. At the same time, the circuit power dominates the total power consumption and increases almost linearly leading to an (almost) constant EE. The network performance in terms of the EE is poor with high density of D2D users (λ _{ d }=10^{−4}). This is due to the fact that the sum rate contributed by the CUEs is already degraded by the interference from high number of D2D users, and additionally, increasing U _{ c } (and accordingly T _{ c }) increases the circuit power without any gain in the total ASR. Consequently, the EE decreases. Thus, massive MIMO is only beneficial in terms of EE if the D2D user density is small, as the resulting ASR gain compensates the significant increase in the circuit power consumption due to higher number of U _{ c } and T _{ c }. On the other hand, for high D2D user density, the EE performance degrades with higher number of U _{ c } and T _{ c } since massive MIMO gains cannot compensate the higher circuit power consumption. Therefore, in the latter case dedicated resources or underlaying with fewer BS antennas could be more beneficial.
4.2 Fixed number of CUEs
Figure 8 b illustrates that when the D2D user density is low, the EE benefits from adding extra BS antennas until the sum of the circuit power consumption of all antennas dominates the performance and leads to a gradual decrease in the EE. As the figure implies, there exists an optimal number of BS antennas which is relatively small since the main massive MIMO gains come from multiplexing rather than just having many antennas. However, in a high density D2D scenario, which is an interferencelimited scenario, the EE decreases monotonically with T _{ c }. Increasing the number of BS antennas in this region cannot improve the ASR significantly, as shown in Fig. 8 a; at the same time, the circuit power consumption increases as a result of the higher number of BS antennas, which in turn leads to decreasing network EE.
The conclusion is that the D2D user density has a very high impact on a network that employs the massive MIMO technology. In the downlink, these two technologies can only coexist at low D2D user densities and careful interference coordination. The number of CUEs should be a function of the number of BS antennas in order to benefit from massive MIMO in terms of the ASR and EE. Otherwise, at high D2D user densities, D2D communication should use the overlay approach rather than the underlay, that is, dedicated time/frequency resources should be allocated to the D2D tier. As discussed earlier, adding successive interference cancellation or using any other interference cancellation technique may change the conclusions but at the same time it would increase the complexity and scalability of the problem.
4.3 The effect of other system parameters
So far, we have discussed the results based on constant transmit power P _{ c }, D2D transmit power P _{ d }, and distance between D2D TxRx pairs R _{0,0} given in Table 1. Now we comment on the choice of these parameters and study their effects on the system performance. From Proposition 1, Proposition 2, and Remark 4, it is evident that the coverage probability for both D2D and cellular tiers, and consequently the network ASR and EE, depend on the ratio of P _{ d } and P _{ c }. Therefore, we fix P _{ c } and vary P _{ d }.
Figure 9 b depicts the EE as a function of λ _{ d } under the same two levels of D2D transmit power. It is observed that lower P _{ d } is more beneficial in terms of the EE in both cases of U _{ c }=1 user and U _{ c }=14 users. This is particularly visible in higher density of D2D users (e.g., λ _{ d }=3×10^{−5}) with U _{ c }=1 user and T _{ c }=5 antennas when the interference is the limiting factor. With U _{ c }=14 users and T _{ c }=70 antennas, the CUEs have higher impact on the ASR, and as a consequence, the system benefits from lower transmit power of D2D users in terms of the EE. Therefore, we have chosen P _{ d }=6 dBm in the previous performance evaluation, as it has a better impact on the ASR as well as EE, especially in higher number of BS antennas.
5 Conclusions
We studied the coexistence of two key 5G concepts: devicetodevice (D2D) communication and massive MIMO. We considered two performance metrics, namely, the average sum rate in bit/s and the energy efficiency in bit/Joule. We considered a setup with uniformly distributed cellular users in the cell, while the D2D transmitters are distributed according to a Poisson point process. We derived tractable expressions for the coverage probabilities of both cellular and D2D users which led to computation of the average sum rate and energy efficiency. We then studied the tradeoff between the number of base station antennas, the number of cellular users, and the density of D2D users for a given coverage area in the downlink. Our results showed that both the average sum rate and energy efficiency behave differently in scenarios with low and high density of D2D users. Underlay D2D communications and massive MIMO can only coexist in low densities of D2D users with careful interference coordination, because the massive MIMO gains vanish when the interference from the D2D tier becomes extremely large. The number of cellular users should scale with the number of BS antennas in order to benefit from massive MIMO in terms of the average sum rate and energy efficiency. If there is a high density of D2D users, the D2D communication should use the overlay approach rather than the underlay or the network should only allow a subset of the D2D transmissions to be active at a time.
6 Endnote
^{1} The assumption that the D2D Tx are distributed in the whole \(\mathbb {R}^{2}\) plane removes any concern about the boundary effects and makes the model more mathematically tractable. The boundary effects are local effects in which users at the network boundary experience less interference than the ones closer to the center, because they have fewer neighbors.
7 Appendix 1: Proof of Proposition 1
Step (a) comes from the fact that g _{0,0}^{2}∼ exp(1) and (b) follows since the noise and interference terms are mutually independent. In step (c), the Laplace transform defined as \(\mathcal {L}_{x}(s) = \mathbb {E}_{x}\left [e^{ sx}\right ]\) is identified.
for a,b>0.
where (a) is based on the probability generating functional (PGFL) [48], and (b) follows from the fact that G∼ exp(1) and \(\mathcal {L}\left [e^{t}\right ] = \frac {1}{s+1}\). Step (c) follows by solving the integral in step (b) and using \(\text {sinc}(x) = \frac {\sin (\pi x)}{\pi x}\).
Substituting (26) and (32) in (25) concludes the proof of Proposition 1. □
8 Appendix 2: Proof of Proposition 2
where (a) follows from the CCDF of \(\mathbf {h}_{0}^{H} \mathbf {v}_{0}^{2}\) with \(2\mathbf {h}_{0}^{H} \mathbf {v}_{0}^{2} \sim \chi ^{2}_{2(T_{c}U_{c}+1)}\) given D _{0,BS} and I _{ d,c }. Since the BS employs ZF with perfect CSI under Rayleigh fading channel to cancel out the interference while boosting the desired signal power, (U _{ c }−1) degrees of freedom is used to null the interference created from other cellular users, hence array gain is reduced to T _{ c }−U _{ c }+1. The distribution of the desired signal power \(\mathbf {h}_{0}^{H} \mathbf {v}_{0}^{2}\), as mentioned above, follows from the projection of a random vector v _{0} into an independent (T _{ c }−U _{ c }+1)dimensional space of the channel. A detailed proof can be found in [41 , 42] and references therein.
Substituting (36) in (33) and using the Faà di Bruno’s formula for the ith derivative of a composite function f(g(s)) with f(s)=e ^{ s } and \(g(s) =  \frac {\pi \lambda _{d} P_{d}^{2/\alpha _{d}}}{\text {sinc}\left (\frac {2}{\alpha _{d}}\right)} s^{2/\alpha _{d}}\), Proposition 2 follows. □
Declarations
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 A Osseiran, F Boccardi, V Braun, K Kusume, P Marsch, M Maternia, O Questeh, M Schellmann, H Schotten, H Taoka, H Tulberg, MA Uusitalo, B Timus, M Fallgren, Scenarios for 5G mobile and wireless communications: the vision of the METIS project. IEEE Commun. Mag.52(5), 26–35 (2014).View ArticleGoogle Scholar
 E Björnson, E Jorswieck, M Debbah, B Ottersten, Multiobjective signal processing optimization: The way to balance conflicting metrics in 5G systems. IEEE Signal Process. Mag.31(6), 14–23 (2014).View ArticleGoogle Scholar
 TL Marzetta, Noncooperative cellular wireless with unlimited numbers of base station antennas. IEEE Trans. Wireless Commun.9(11), 3590–3600 (2010).View ArticleGoogle Scholar
 F Rusek, D Persson, LK Buon, EG Larsson, TL Marzetta, O Edfors, F Tufvesson, Scaling up MIMO: opportunities and challenges with very large arrays. IEEE Signal Process. Mag.30(1), 40–60 (2013).View ArticleGoogle Scholar
 E Björnson, EG Larsson, M Debbah, Massive MIMO for maximal spectral efficiency: how many users and pilots should be allocated?IEEE Trans. Wireless Commun.15(2), 1293–1308 (2016).View ArticleGoogle Scholar
 HQ Ngo, EG Larsson, TL Marzetta, Energy and spectral efficiency of very large multiuser MIMO systems. IEEE Trans. Commun.61(4), 1436–1449 (2013).View ArticleGoogle Scholar
 E Björnson, L Sanguinetti, J Hoydis, M Debbah, Optimal design of energyefficient multiuser MIMO systems: is massive MIMO the answer?IEEE Trans. Wireless Commun.14(6), 3059–3075 (2015).View ArticleGoogle Scholar
 K Doppler, M Rinne, C Wijting, CB Ribeiro, K Hugl, Devicetodevice communication as an underlay to LTEadvanced networks. IEEE Commun. Mag.47(12), 42–49 (2009).View ArticleGoogle Scholar
 3rd Generation Partnership Project, study on LTE device to device proximity services; Radio aspects, 3GPP TR 36.843 (2014).Google Scholar
 NK Pratas, P Popovski, Zerooutage cellular downlink with fixedrate D2D underlay. IEEE Trans. Wireless Commun.14(7), 3533–3543 (2015).View ArticleGoogle Scholar
 JF Schmidt, MK Atiq, U Schilcher, C Bettstetter, in Proc. IEEE Int. Symp. on Personal, Indoor, Mobile Radio Commun. (PIMRC). Underlay devicetodevice communications in LTEA: uplink or downlink (Hong Kong, 2015).Google Scholar
 S Tombaz, A Vastberg, J Zander, Energy and costefficient ultrahighcapacity wireless access. IEEE Wireless Commun.18(5), 18–24 (2011).View ArticleGoogle Scholar
 G Auer, V Giannini, C Desset, I Godor, P Skillermark, M Olsson, MA Imran, D Sabella, MJ Gonzalez, O Blume, A Fehske, How much energy is needed to run a wireless network?IEEE Wireless Commun.18(5), 40–49 (2011).View ArticleGoogle Scholar
 H Yang, TL Marzetta, in Proc. IEEE Online Conference on Green Commun. (OnlineGreenCom). Total energy efficiency of cellular large scale antenna system multiple access mobile networks (Piscataway, NJ, 2013).Google Scholar
 E Björnson, M Kountouris, M Debbah, in Proc. IEEE Int. Conf. on Telecommun. (ICT). Massive MIMO and small cells: improving energy efficiency by optimal softcell coordination (Casablanca, Morocco, 2013).Google Scholar
 H Min, J Lee, S Park, D Hong, Capacity enhancement using an interference limited area for devicetodevice uplink underlaying cellular networks. IEEE Trans. Wireless Commun.10(12), 3995–4000 (2011).View ArticleGoogle Scholar
 W Yu, L Liang, SJ H. Zhang, JCF Li, M Lei, in Proc. IEEE Global Telecommun. Conf. (GLOBECOM). Performance enhanced transmission in devicetodevice communications: beamforming or interference cancellation? (Anaheim, CA, 2012).Google Scholar
 N Reider, G Fodor, A distributed power control and mode selection algorithm for D2D communications. EURASIP J. on Wireless Commun. Netw. 2012(1), 1–25 (2012).View ArticleGoogle Scholar
 MGDS Rêgo, TF Maciel, HDHM Barros, FRP Cavalcanti, G Fodor, in Proc. Int. Symp. of Wireless Commun. Syst. (ISWCS). Performance analysis of power control for devicetodevice communication in cellular MIMO systems (Paris, 2012).Google Scholar
 Y Ni, S Jin, W Xu, M Matthaiou, S Shaoand, H Zhu, in Proc. IEEE Int. Conf. on Commun. Workshops. Beamforming and interference cancellation schemes for D2D communications (London, UK, 2015).Google Scholar
 S Shalmashi, E Björnson, SB Slimane, M Debbah, in Proc. IEEE Wireless Commun. Network. Conf. (WCNC). Closedform optimality characterization of networkassisted devicetodevice communications (Istanbul, Turkey, 2014).Google Scholar
 X Lin, RW Heath Jr, JG Andrews, The interplay between massive MIMO and underlaid D2D networking. IEEE Trans. Wireless Commun.14(6), 3337–3351 (2015).View ArticleGoogle Scholar
 H Yin, L Cottatellucci, D Gesbert, in Proc. IEEE Asilomar Conf. Signals, Systems, and Computers. Enabling massive MIMO systems in the FDD mode thanks to D2D communications (Pacific Grove, CA, 2014).Google Scholar
 SAM Ghanem, L Cottatellucci, D2D cooperation to avoid instantaneous feedback in nonreciprocal massive MIMO systems. IEEE COMSOC MMTC Electronics. Lett.10(1), 39–42 (2015). Special issue on Novel approaches and technologies for 5G.Google Scholar
 BS Amin, YR Ramadan, AS Ibrahim, MH Ismail, in Proc. IEEE Wireless Commun. Network. Conf. (WCNC). Power allocation for devicetodevice communication underlaying massive MIMO multicasting networks (New Orleans, LA, 2015).Google Scholar
 S Shalmashi, G Miao, SB Slimane, in Proc. IEEE Int. Symp. on Personal, Indoor, Mobile Radio Commun. (PIMRC). Interference management for multiple devicetodevice communications underlaying cellular networks (London, UK, 2013).Google Scholar
 M Zulhasnine, S Changcheng, A Srinivasan, in Proc. IEEE Int. Conf. on Wireless and Mobile Computing, Networking, and Commun. (WiMob). Efficient resource allocation for devicetodevice communication underlaying LTE network (Niagara Falls, ON, Canada, 2010).Google Scholar
 E Yaacoub, O Kubbar, in Proc. IEEE Global Telecommun. Conf. Workshops. Energyefficient devicetodevice communications in LTE public safety networks (Anaheim, CA, 2012).Google Scholar
 S Mumtaz, KMS Huq, A Radwan, J Rodriguez, RL Aguiar, in Proc. IEEE Int. Conf. on Commun. (ICC). Energy efficient interferenceaware resource allocation in LTED2D communication (Sydney, Australia, 2014).Google Scholar
 F Wang, C Xu, L Song, Q Zhao, X Wang, Z Han, in Proc. IEEE Int. Conf. on Commun. (ICC). Energyaware resource allocation for devicetodevice underlay communication (Budapest, Hungary, 2013).Google Scholar
 R Baldemair, E Dahlman, G Fodor, G Mildh, S Parkvall, Y Selen, H Tullberg, K Balachandran, Evolving wireless communications: addressing the challenges and expectations of the future. IEEE Veh. Technol. Mag.8(1), 24–30 (2013).View ArticleGoogle Scholar
 Bjo, Ë,rnson, E Jorswieck, Optimal resource allocation in coordinated multicell systems. Foundations Trends Commun. Inform. Theory. 9(2–3), 113–381 (2013).View ArticleMATHGoogle Scholar
 D Gesbert, M Kountouris, RW Heath, CB Chae, T Salzer, Shifting the MIMO paradigm. IEEE Signal Process. Mag.24(5), 36–46 (2007).View ArticleGoogle Scholar
 M Haenggi, Stochastic Geometry for Wireless Networks (Cambridge University Press, Cambridge, UK, 2013).MATHGoogle Scholar
 X Lin, J Andrews, A Ghosh, Spectrum sharing for devicetodevice communication in cellular networks. IEEE Trans. Wireless Commun.13(12), 6727–6740 (2014).View ArticleGoogle Scholar
 N Lee, X Lin, JG Andrews, RW Heath Jr, Power control for D2D underlaid cellular networks: modeling, algorithms and analysis. IEEE J. Sel. Areas Commun.33(1), 1–13 (2015).View ArticleGoogle Scholar
 MC Erturk, S Mukherjee, H Ishii, H Arslan, Distributions of transmit power and SINR in devicetodevice networks. IEEE Commun. Lett.17(2), 273–276 (2013).View ArticleGoogle Scholar
 H ElSawy, E Hossain, MS Alouini, Analytical modeling of mode selection and power control for underlay D2D communication in cellular networks. IEEE Trans. Wireless Commun.62(11), 4147–4161 (2014).View ArticleGoogle Scholar
 Z Chen, M Kountouris, in Proc. IEEE Int. Conf. on Commun. Workshops. Guard zone based D2D underlaid cellular networks with twotier dependence (London, UK, 2015).Google Scholar
 G George, RK Mungara, A Lozano, An analytical framework for devicetodevice communication in cellular networks. IEEE Trans. Wireless Commun.14(11), 6297–6310 (2015).View ArticleGoogle Scholar
 HS Dhillon, M Kountouris, JG Andrews, Downlink MIMO HetNets: Modelling, ordering results and performance analysis. IEEE Trans. Wireless Commun.12(10), 5208–5222 (2013).View ArticleGoogle Scholar
 V Chandrasekhar, M Kountouris, JG Andrews, Coverage in multiantenna twotier networks. IEEE Trans. Wireless Commun.8(10), 5314–5327 (2009).View ArticleGoogle Scholar
 E Björnson, M Bengtsson, B Ottersten, Optimal multiuser transmit beamforming: a difficult problem with a simple solution structure [lecture notes]. IEEE Signal Process. Mag.31(4), 142–148 (2014).View ArticleGoogle Scholar
 H ShokriGhadikolaei, C Fischione, P Popovski, M Zorzi, Design aspects of short range millimeter wave networks: a MAC layer perspective. IEEE Network. 30(3), 88–96 (2016).View ArticleGoogle Scholar
 M Kountouris, JG Andrews, Downlink SDMA with limited feedback in interferencelimited wireless networks. IEEE Trans. Wireless Commun.11(8), 2730–2741 (2012).Google Scholar
 JG Andrews, S Buzzi, W Choi, SV Hanly, A Lozano, ACK Soong, JC Zhang, What will 5G be?IEEE J. Sel. Areas Commun.32(6), 1065–1082 (2014).View ArticleGoogle Scholar
 P Popovski, V Braun, HP Mayer, P Fertl, et al, ICT317669METIS/D1.1 scenarios, requirements and KPIs for 5G mobile and wireless system. Technical report (2013). http://www.metis2020.com/. Accessed 30 Apr 2013.
 S Weber, JG Andrews, Transmission capacity of wireless networks. Foundations Trends Netw. 5(2–3), 109–281 (2012).MATHGoogle Scholar
 J HanShin, JS Young, X Ping, JG Andrews, Heterogeneous cellular networks with flexible cell association: a comprehensive downlink SINR analysis. IEEE Trans. Wireless Commun.11(10), 3484–3495 (2012).View ArticleGoogle Scholar
 JG Andrews, F Baccelli, RK Ganti, A tractable approach to coverage and rate in cellular networks. IEEE Trans. Commun.59(11), 3122–3134 (2011).View ArticleGoogle Scholar
 N Jindal, MIMO broadcast channels with finiterate feedback. IEEE Trans. Inf. Theory. 52(11), 5045–5060 (2006).MathSciNetView ArticleMATHGoogle Scholar
 E Björnson, EG Larsson, TL Marzetta, Massive MIMO: Ten myths and one critical question. IEEE Commun. Mag. 54(2), 114–123 (2016).View ArticleGoogle Scholar
 EG Larsson, F Tufvesson, O Edfors, TL Marzetta, Massive MIMO for next generation wireless systems. IEEE Commun. Mag. 52(2), 186–195 (2014).View ArticleGoogle Scholar
 H Xing, S Hakola, in Proc. IEEE Int. Symp. on Personal, Indoor, Mobile Radio Commun. (PIMRC). The investigation of power control schemes for a devicetodevice communication integrated into OFDMA cellular system (Istanbul, Turkey, 2010), pp. 1775–1780.Google Scholar
 S Shalmashi, E Björnson, M Kountouris, KW Sung, M Debbah, in Proc. IEEE Int. Conf. on Commun. Workshops. Energy efficiency and sum rate when massive MIMO meets devicetodevice communication (London, UK, 2015).Google Scholar