New Self-Isolated Wideband MIMO Antenna System for 5G mm-Wave Applications Using Slot Characteristics

This paper proposes a simple novel technique for self-isolating a MIMO antenna for mm-wave applications. MIMO antenna elements with inter-element separation of 0.2 mm (<inline-formula> <tex-math notation="LaTeX">$0.023\lambda $ </tex-math></inline-formula> at 35 GHz) and measured high isolation (>50 dB) are presented. By observing the concentration of surface waves on the radiating patch, several rigorously optimized slots of different shapes, positions, and dimensions are etched on the patch to enhance the inter-element isolation and increase the bandwidth within 28-37.5 GHz. The circuit models of the reference and proposed antennas have been presented. The coupling is measured by the level of the differences in the output voltage in both antennas. The novel mm-wave antenna exhibits high impedance bandwidth (>29%), high isolation (>50 dB), high efficiency (>90%), and low envelope correlation coefficient (< 0.005). Two configurations of MIMO antenna (i.e., <inline-formula> <tex-math notation="LaTeX">$1\times 2$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$1\times 4$ </tex-math></inline-formula>) are fabricated and measured to validate the simulation outcomes. The single reference antenna has dimensions of 10x12 mm 2 while the 1x2 array has dimensions of 19x12 mm 2. The presented design is the first to exhibit such wideband isolation improvement without any external decoupling structure at the mm-wave frequency range, to the best of the authors’ knowledge.


I. INTRODUCTION
T HE 5G wireless technology offers high data rate, low latency, low energy and cost, increased spectral efficiency and channel capacity, especially when used in MIMO technology [1], [2]. Therefore, mm-wave antenna arrays are explored to meet the user requirements of high throughput, with a very close distance between antenna elements to enable miniaturization [3]. However, antenna arrays design at mm-wave bands has been a growing concern among antenna engineers because of the inevitable, undesired mutual coupling between the antenna elements [4].
Various techniques have been reported in the literature to mitigate mutual coupling at mm-wave bands [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20]. The defective fractal ground was reported by [5] to achieve 13 dB isolation improvement between two compact patch antenna arrays in the 11.4-11.78 GHz range. Metasurfaces have been used to achieve a maximum of 15 dB isolation improvement within 27-32 GHz [6]. A maximum of 31.5 dB mutual coupling reduction at 60. 4 GHz was achieved in [7] using a metasurface shield. The authors in [8] proposed Complementary Split Ring Resonator (CSRR) as a decoupling element in a MIMO array to achieve 31.8 dB isolation improvement and improved the radiation characteristics within 24.7-25.3 GHz. Defected Ground Surfaces (DGS) [9] were used to enhance the isolation of two integrated antenna systems by 3 dB in the dual-band of 4G (3.8 and 5.5 GHz), and 5 dB in the 5G bands . Frequency Selective Surface (FSS) was reported to increase the isolation of a dielectric resonator (DRA) MIMO antenna by a maximum of 30 dB in the 57-63 GHz bandwidth [10]. In [11], a hybrid isolator was used as an EBG unit cell to improve the isolation of a MIMO dielectric resonator antenna by a maximum of 25 dB within 59.3-64.8 GHz impedance bandwidth. The authors in [12] have used metallic vias as a decoupling mechanism to achieve about 19.8 dB isolation improvement in the E-plane, and 22.7 dB in the H-plane of a 1 × 2 MIMO DRA antenna at 26 GHz. Metal strips were printed on a MIMO DRA to achieve a maximum isolation improvement of 12 dB at 27.5-28. 35 GHz [13]. The authors in [15] used multiple notches structures on the ground planes of a MIMO Vivaldi antenna to achieve a 37.3 dB maximum mutual coupling reduction enhancement within 24.55-28.5 GHz. The feed line structure of a 56.5-64 GHz slot patch antenna was modified to achieve a maximum isolation improvement of 38.6 dB [16]. Several Electromagnetic Band Gap (EBG) surfaces were employed as decoupling structures in [17], [18], [19]. Defective Ground Surface (DGS) was used by [20] to decouple a slot-array patch antenna by 50 dB between 1.2-1.35 GHz. In [21], the authors used near-field resonator placed above 2 and 8element patch antennas to realize a 20 dB coupling reduction within 6.1% bandwidth. Also, the authors in [22] proposed the use of the dielectric block to achieve over 20 dB coupling reduction between very close patch antenna elements in both E-and H-planes around 4.32-5.48 GHz bandwidth.
This paper presents a simple, yet efficient, self-isolated MIMO antenna system in the 5G mm-wave frequency band. Tightly spaced patch antenna MIMO elements with a maximum isolation enhancement of 35 dB over the 28-37.5 GHz mm-wave frequency band is achieved. The impedance bandwidth and isolation improvements are enabled by meticulously introducing U-slot, H-slot, and five circular slots on the patch, which are illustrated by studying the surface current distributions. The authors believe that this is the first time such concept is reported at the mm-wave frequency band of 28 GHz. 1 × 2 and 1 × 4 MIMO antenna arrays are presented to validate the proposed novel isolation mechanism. The proposed antenna exhibits high gain, efficiency, low correlation coefficient, and good radiation properties throughout the working bandwidth. The proposed highly isolated antenna is fabricated and tested.
The manuscript is organized as follows: Section II describes the antenna structure, which includes the evolution (i.e., development stages) of the proposed antenna array, the concept of wideband self-isolation, and the parametric analyses. Section III illustrates a 4-element antenna array to verify the MIMO capability of the proposed antenna, while Section IV concludes the paper.

II. ANTENNA STRUCTURE AND DEVELOPMENT
In this Section, first, the design methodology of the MIMO antenna is discussed. The self-isolation mechanism is then thoroughly explained through the equivalent circuit models and S-parameter results. Rigorous parametric analyses are also presented to optimize the performance of the MIMO antenna. Finally, far-field results are presented and discussed.  Fig. 1 consists of two copper patches, each with 0.035 mm thickness, mounted on Rogers RT5870 substrate of dielectric constant 2.33, loss tangent 0.0012, height 0.508 mm, and a total size of 19x12 mm 2 . The patch dimensions are accurately designed to operate at the mm-wave frequency band of 28 GHz. The two antennas are placed side-by-side along the x-axis (H-plane) with an edge-to-edge distance of 0.2 mm. The two patches are fed  with 50 microstrip lines, each line with optimized dimensions of 1.11 mm x 3.9 mm and located at L 12 = 2.86 mm (See Fig. 2).

A. ANTENNA DESIGN AND EVOLUTION
From Fig. 1, to realize the proposed antenna with a wide bandwidth and improved isolation between the elements, a design evolution is performed where the initial reference antenna is denoted as Ref.  Table 1 shows the dimensions of the proposed antenna. It is worthy of note that the U, H, and circular slots were used as a result of meticulous observation of the surface currents on the patch and the effect of different types of slots on the antenna coupling. As a result, the considered slots' shapes exhibit better coupling reduction compared to others due to further surface current reduction.

B. EQUIVALENT CIRCUIT AND DECOUPLING CONFIGURATION
The equivalent circuit of the reference antenna (Ref. A) is a modified version of the equivalent circuit of an insetfed microstrip antenna [31], [32]. As seen in Fig. 3, the feedline and patch are represented by series step-in-width L-C elements, connected in parallel with a gap capacitor of capacitance (g c ), caused by the ground and the radiating patch. This combination is connected in series to another gap capacitor of capacitance (g c ), caused by the separation between the two antennas. The inductance (L) and capacitance values (C) of the step-in-width feedline and patch are calculated using the step-in-width formulas [32]: While the gap capacitance (g c ) is calculated as follows [33]: where I L is the element length (in mm), Z OL is the element impedance assumed to be 50 , f is the central operating frequency taken as 35 GHz and λ g is the guide wavelength.
Equations (1) and (2) are used to derive the values of L 1 and C 1 seen in Fig. 3, while (3) is used to derive C 2 and C 3 . The parameter g used in (3) is the distance measured in μm between the two antennas (for C 2 ) and the thickness of the substrate (for C 3 ). The element lengths I L of the patch and feedline are equal to L 4 , and L 11 , respectively, as shown in Fig. 2. From Fig. 3, L 1 and C 1 are the equivalent inductance and capacitance of the patch and feedline. The coupling between the two antennas is assumed to be capacitive due to the electric field between both antennas. Using National Instruments (NI) Multisim software, an AC voltage source of 1 V peak is used to excite the circuit in Fig. 3. A voltage 25 nV is obtained as the coupling voltage transferred to the other antenna, measured across C 2 .
Using the same concept, the equivalent circuit model of the proposed self isolated MIMO antenna is developed and depicted in Fig. 4. C 1 -C 15 are derived using Equation (3), while L 1 -L 13 are derived using equation (1). The U-slot consists of three arms, equivalent to three series L-C branches, all connected in parallel. The H-slot consists of five branches, each consisting of series L-C branches. The circular slots consist of five capacitances created by the gaps; and four inductances created by the space between any two slots in the five circular slots. The parameter g used in the calculation is the width of each slot.
NI Multism software and the same 1V-peak AC voltage source are used for the circuit simulation. From Fig. 4, it can be observed that the output voltage, measured across the capacitor C 15 is 8.25 nV, which is about 67% reduction compared to the reference antenna. It can be inferred that the decrease in voltage is a consequence of the additional U-slot, H-slot, and the five circular slots, which have collectively absorbed the coupling field emanating from the antenna. These etched structures are responsible for reducing the surface currents responsible for the mutual coupling between the two antennas when brought very close to each other. The presented equivalent circuits that demonstrate the output voltage reduction verify the concept of self-decoupling.

C. WIDE BAND AND SELF-ISOLATION MECHANISM
Herein, we explain the self-isolation technique employing the surface current distributions. In addition, we demonstrate how the different incorporated slots can simultaneously increase the matching bandwidth as well as increase the isolation in a MIMO array configuration.
It is well known that microstrip patch antennas are very effective because of their simple, robust, and compact structure. However, microstrip antennas are limited by low bandwidth, gain, and surface waves [4]. According to [34], the surface wave is predominant in a microstrip antenna if the normalized substrate electrical thickness h/λ 0 satisfies the condition: where h is the substrate thickness, λ 0 is the wavelength in free space, and r is the dielectric constant of the substrate. Substituting the parameters of our antenna in (4), the left side and right side of the equation are approximately 0.06 and 0.03, respectively. Hence, it is evident that the surface waves effect is a potential cause for mutual coupling in this MIMO antenna. Apart from the surface waves, space waves (i.e., near-fields) also contribute to coupling in microstrip antennas, in the near-field region of the antenna, especially if the distance between the antenna elements is less than half wavelengths. This coupling effect is confined in the region above the substrate of the coupled patch antennas [35]. One way to improve the impedance bandwidth of the antenna is through the use of slots [36]. Slots enable more resonant modes to be excited in the patch by interrupting the uniformity of the patch, thereby, inducing perturbations, counteracting or constraining EM waves. This will disturb the surface current distribution in the patch, resulting in multiple resonances being excited in the patch. The EM wave propagation can be effectively controlled by meticulously optimizing the slots dimensions and locations, which can eventually alter the inductive and capacitive response of the entire patch. Slots' dimensions and locations were chosen by carefully observing the current paths on the surface of the antenna. The disturbed surface current introduced due to the slots is responsible for the increase in the isolation between the antenna elements when the positions and dimensions of the slots are carefully optimized. To ascertain the impedance bandwidth improvement and isolation enhancement capability of the slots, the antenna is arranged in a 1 × 2 configuration, as shown in Fig. 5 with dimension L 15 = 19 mm X L 16 = 12 m, and edge-to-edge separation of d = 0.2 mm, which is approximately 0.023λ 0 at 35 GHz.  To enable the antenna work over a wider bandwidth, five circular slots are meticulously optimized and incorporated into the patch (see proposed in Figs. 1d and 2). It is clear from Fig. 6 that the antenna now resonates from 30-41 GHz with a maximum isolation improvement of 67 dB at 33.2 GHz compared to the reference antenna (Ref. A). It is worthy of note that the major contributor to the coupling reduction and bandwidth enhancement is the carefully placed H-slot in addition to the U-slot and circular slots. Figure 7 shows good agreement between the simulated and measured S 11 , S 22 , and S 21 of the reference 1 × 2 array (Ref. A) and the proposed 1 × 2 self-isolated antenna array.

D. PARAMETRIC ANALYSES
This subsection highlights the results obtained by a rigorous tuning of some key antenna parameters to improve its overall performance. The parameters and dimensions of the antenna in Figs. 2 and 5 are tuned and the results are discussed. Fig. 8a illustrates the effect of changing the edge-to-edge distance on the isolation characteristics while keeping other parameters constant. The isolation increases as the distance between the two antennas increases. However, the chosen smallest distance in this work (d = 0.2 mm) still exhibits very good isolation; less than 20 dB throughout the working bandwidth, fulfilling MIMO isolation requirements. The reflection coefficient is not affected by the change of the inter-element spacing, hence, not shown. Fig. 8b depicts the effect of the number of circular slots on the S-parameters while keeping other parameters constant. The proposed antenna with five slots exhibits the best reflection and transmission characteristics over the 30-41 GHz impedance bandwidth.
The effect of changing the tilting angle "c" of the five circular slots (see Fig. 2) on the reflection coefficient of the proposed 1 × 2 self-isolated MIMO antenna is shown in Fig. 9. The widest impedance bandwidth is observed when c = 5.4 • . Fig. 10a depicts that L 8 = 1.24 mm (see Fig. 2) is the best position to place the H-slot above the U-slot to obtain the desired reflection and transmission characteristics. Furthermore, Fig. 10b shows that the optimum rectangular slot width L 14 to obtain the best reflection and isolation characteristics is 0.25 mm, while other parameters are unaltered. Fig. 11 plots the simulated radiation patterns of the proposed self-isolated 1 × 2 5G mm-wave antenna compared with the reference array (Ref. A) in the E-and H-planes at 31, 35, and 38 GHz. After inserting the slots and as a result of the surface current reduction, the radiation patterns of the array become more directive along the broadside direction over the frequency band of interest. Furthermore, a slight beam deflection is observed at 38 GHz, which is expected because of the high bandwidth of the antenna. A good agreement between the simulated and measured radiation patterns of  the proposed self-isolated mm-wave 1 × 2 antenna at 31, 35, and 38 GHz is depicted in Fig. 12. Fig. 13a shows the fabricated prototypes of the Ref. A, 1 × 2, and 1 × 4 MIMO antennas. Fig. 13b shows the radiation pattern measurement set-up in an anechoic chamber.

III. MIMO ANTENNA CONFIGURATION
To verify the MIMO capability of the proposed antenna, a 4-element array is designed and arranged along the x-axis (H-plane) as shown in Fig. 14 with an edge-to-edge distance of d = 0.2 mm. The simulated surface current distribution of the 1 × 4 antenna when one port is excited, and the others are terminated with a matched load at 33 GHz is shown in Fig. 15. Fig. 15a shows that the surface current diminishes as the distance between the excited antenna (Ant.1) and the other antennas increases. Ant. 4 has almost no surface current when Ant. 1 is excited. Additionally, when Ant. 2 is excited, Ant. 1 and Ant. 3 exhibit nearly the same small surface current distribution, while Ant. 4 reveals a minimal surface current, as shown in Fig. 15b. The same phenomenon is observed when Ant. 3 or Ant. 4 are excited, as shown in Figs. 15c and 15d. This shows the effectiveness of the added slots in mitigating the surface currents. The measured and simulated E-and H-plane radiation patterns of the 1 × 4 MIMO array at 32 GHz are depicted in Fig. 17. The MIMO array exhibits consistent broadside radiation characteristics in both planes with almost no beam deflection.
The gain and efficiency are two important performance indices of an antenna that relates the radiated power to the input power. Coupling affects the effective input impedance of the elements (see Fig. 6a), resulting in a higher reflection than isolated elements, and thus, reduces the radiated power. Therefore, reducing the coupling effect on the element matching enhances the radiated power improving the gain and efficiency [5], [6], [7]. Moreover, considerable amount of power is coupled to the neighbouring antennas in case of poor isolation between the MIMO elements. Fig. 18a depicts that the 1 × 2 antenna array peak gain has been improved compared to the reference antenna (Ref. A) without slots, owing to the isolation improvement over the working bandwidth. It is worth noting that the proposed MIMO antenna has good gain flatness except for the notched band at 34 GHz. This is due to the low isolation improvement observed around 34 GHz, as seen in Fig. 7b.
However, Fig. 18(a-b) shows that the gain and total efficiency at 34 GHz of 7.5 dB and 90%, are still sufficient for efficient radiation. The simulated total efficiency of the 1 × 2 antenna array is shown in Fig. 18b. As expected, the total efficiency has an average value of about 91% compared with 55% of the reference antenna (Ref. A). The efficiency increase is mainly due to the isolation enhancement of the antenna.
The Envelope correlation coefficient (ECC) is an important performance criterion in MIMO antennas. It can be determined using S-parameters or far-field characteristics. The S-parameters method is useful when efficiency is considerably high (i.e., lossless antenna) [37]. The far-fields method is more reliable for lossy antennas as it considers the far-fields characteristics of the two antennas. ECC based on far-fields can be expressed as [37]: where F 1 (θ, φ) is the field pattern when antenna 1 is excited and antenna 2 is terminated by 50 load, and vice-versa for F 2 (θ, φ). The simulated ECC of the reference (Ref. A) and the proposed 1×2 MIMO antennas, based on far field parameters, is depicted in Fig. 19a. The proposed antenna exhibits ECC less than −20 dB over the impedance bandwidth, with a considerable coupling reduction compared with the reference antenna (Ref. A) indicating an isolation improvement. The diversity gain is another important parameter for the evaluation of the MIMO system and can be obtained using [39]: where 10 is the apparent maximum diversity gain. The ECC of the proposed antenna is less than 0.01, thereby, making its diversity gain around 10 dB within the operational bandwidth as depicted in Fig. 19b. This improved diversity gain is a result of the improved isolation between the MIMO elements. Another important performance index for MIMO antenna is the Total Active Reflection Coefficient (TARC). As stated in [40], TARC is defined as the ratio of the squareroot of total reflected power divided by the square-root of total incident power. Unlike the traditional S 11 of a single antenna, TARC considers the effect of the mutual coupling between the MIMO elements and the incident wave phase. For a two-element MIMO antenna, TARC can be calculated as [40] TARC = s 11 + s 12 e jθ 2 + s 21 + s 22 e jθ 2 / √ 2 (7) Fig. 20 shows the TARC of the proposed 1×2 self-isolated MIMO antenna in the presence of mutual coupling. It can  be observed that TARC exhibits similar behaviour as the antenna's traditional reflection coefficients (S 11 and S 22 ). However, the 10-dB TARC bandwidth is slightly affected due to the mutual coupling effect. Table 2 depicts the comparison of the proposed antenna with recent literature. The proposed antenna exhibits the largest impedance bandwidth combined with high efficiency, gain, and maximum isolation improvement with minimal edge-to-edge spacing. The proposed antenna is the only design in Table 2 that simultaneously acquires all these attractive features without compromising any fundamental design objective.

ACKNOWLEDGMENT
The authors acknowledge Mr Sagiru Gaya from Khalifa University for the valuable discussions during this work.

IV. CONCLUSION
A novel, simple, and low-cost self-isolated mm-wave 1 × 2 MIMO antenna with an impedance bandwidth of 28-37.5 GHz, 51 dB maximum isolation, and edge-to-edge spacing of 0.023λ 0 has been presented. No additional or external isolating structures have been used. Various antenna parameters have been investigated to achieve high bandwidth and the smallest possible edge-to-edge spacing. Moreover, a 1×4 MIMO configuration of the antenna has been presented and analyzed. Good agreement between the measured and simulated results has been achieved. Compared to previously published designs, the proposed MIMO antenna exhibits a wide bandwidth greater than 29%, improved efficiency greater than 90%, minimal envelope correlation coefficient less than 0.005, and significant isolation improvement greater than 50 dB with minimal inter-element spacing.