5G is the fifth generation of wireless communication systems that provide higher data speeds, increased communication capacity, and wider coverage than previous systems [1–4]. This is made possible by the wider available bandwidth in the frequency spectrum, which includes the millimeter-wave (mmWave) frequency band [5, 6]. However, transmission in this high-frequency mmWave band is characterized by greater free-space path loss and significant blockage effects caused by obstacles, such as buildings, automobiles, and human bodies [7–10]. To address these issues, mmWave-band 5G antennas often employ array structures in massive-input massive-output (MIMO) systems [11–14]. Notably, since dual-polarization antennas generate two independent polarizations within a single radiating element, they offer an effective solution for antenna deployment in MIMO systems with limited physical space [15–20].

Microstrip patch antennas offer a number of advantages, including low profile and high design flexibility, but they struggle to achieve a wide bandwidth and high port-to-port isolation in dual-polarization applications [20–25]. Various approaches have been proposed to address these limitations in the context of dual-polarization applications. For instance, a via-fed patch with a thick substrate and parasitic elements was proposed, enabling dual-band operation but also exhibiting poor isolation due to mutual coupling [26]. Meanwhile, stacked patches were found to improve both bandwidth and isolation, but they also increased design complexity [27, 28]. Capacitive L- and T-shaped probe-fed patch antennas offer good performance, but require thick substrates and exhibit reduced gain or isolation [29, 30]. Furthermore, the use of coupled-aperture feeding with stacked patches enhanced the bandwidth, but its isolation performance was not reported [31]. Ring-shaped stacked patches achieved an isolation below −20 dB, but they require multiple metal layers, increasing both bulk and complexity [32]. Meanwhile, substrate-integrated waveguide (SIW)-based aperture-fed structures provide high gain and isolation, but offer limited bandwidth due to mode constraints [33]. The observations from these studies highlight the prevalence of tradeoffs among bandwidth, isolation, and complexity, emphasizing the need for an optimized design that balances all these factors.

This paper proposes an aperture-coupled dual-polarization patch antenna with enhanced isolation characteristics achieved using isolation vias. The proposed antenna incorporates a U-shaped aperture to maintain cross-polarization characteristics in a dual-polarization structure with aperture feeding. The port-to-port isolation level is improved using only two isolation vias. Moreover, the proposed approach requires only four metal layers, indicating a relatively simple design. Additionally, the use of a low-profile and low-loss air-cavity antenna substrate enhances the ability to achieve a high gain.

The remainder of this paper is organized as follows. In Section II, we verify the optimal antenna efficiency achieved through 2-port simulations. We then analyze the surface current resulting from the aperture-coupled structure and the isolation vias applied to improve port-to-port isolation performance, and compare the isolation performance between ports. In Section III, the proposed model is fabricated and measured to verify its performance. The simulation and measurement results are presented and compared. Finally, Section IV presents the conclusions.

A cross-sectional view of the proposed wideband U-shaped aperture dual-polarization patch antenna is shown in Fig. 1, clearly depicting its four layers. Aperture-coupled feeding was implemented in this stacked configuration, as shown in Fig. 2, using a stripline. The proposed structure includes three substrates and an air cavity. Substrate 1, realized using Taconic’s TLY-5A (ɛ_r = 2.17, tanδ = 0.0009) with a thickness of 0.127 mm, is implemented as a lid substrate to utilize the air cavity as the patch antenna dielectric. For Substrates 2 and 3, modified polyimide with a dielectric constant of 2.5, dielectric loss of 0.0025, and thickness of 0.05 mm is used. This aperture-coupled feeding method allows for the independent optimization of the radiating element and the feeding line [34].

The size of the patch, which is the primary radiating element, was first evaluated to establish an efficient design approach. The patch size (W_p) in Layer 1 was optimized using the formula for determining the width of a microstrip patch antenna, accounting for the effects of the applied substrate and air-cavity structure [35], expressed as follows:

where W_p represents the width of the patch, v₀ refers to the speed of light in free space, ɛ_r is the dielectric constant of air, and f_r signifies the resonant frequency.

A square patch was adopted to facilitate dual-polarization operations. As depicted in Fig. 3, to identify the appropriate patch size for the aperture-coupled feeding structure, a parametric simulation was performed based on the proposed configuration. As shown in Fig. 4, the resonant frequency and radiation efficiency both change depending on the patch size. Based on the radiation efficiency results, a patch size of 4.1 mm was determined to be most suitable for the target frequency of 28 GHz.

Layer 2, as illustrated in Fig. 2, features two U-shaped apertures for dual polarization, positioned on the ground plane. A 0.5-mm-thick air cavity is introduced between Layers 1 and 2 to minimize transition losses from the aperture to the patch. Meanwhile, Layer 3 includes stripline transmission lines and vias, such as a cavity via and an isolation via. The stripline structure, owing to the ground planes located above and below the signal line, significantly reduced the radiation losses arising from the transmission line. Additionally, the cavity vias were optimized to enhance the coupling between the aperture-coupled slot and the patch located in Layer 1 [36].

As depicted in Fig. 3, the feeding network in Layer 3 operates as a cavity composed of vias with a diameter (L_v) of 0.20 mm and spacing (d_v) of 0.20 mm. The signal transmitted by the stripline is coupled through the U-shaped aperture in Layer 2, facilitating dual polarization. Ports 1 and 2, corresponding to the vertical and horizontal polarizations, respectively, are implemented in an x–y symmetric structure.

Extensive simulations and optimizations were conducted to determine the parameters for radiation efficiency and impedance matching. The results are detailed in Table 1. The width of the U-shaped aperture (W_a) was set to 1.53 mm, which is approximately one-quarter of the guided wavelength (λ_g). With W_a fixed, simulation was performed to optimize the radiation efficiency according to the length of the aperture (L_a). The simulation results, as shown in Fig. 5, indicate that L_a is a critical factor affecting radiation efficiency. We set L_a to 1.05 mm, since it achieved the highest radiation efficiency of 89.8% at the center frequency of 28 GHz. It is also observed that the radiation efficiency exceeds 80% from 26.5 to 29.5 GHz. Hence, the total length of the aperture was optimized to be 0.34λ₀.

A stripline-to-aperture transformer was applied between the feeding line and the radiator to achieve impedance matching. The length of the stripline-to-aperture transformer (L_t) was 0.27λ_g. The simulation results for the reflection coefficient with respect to the width of the stripline-to-aperture transformer (W_t) are presented in Fig. 6, showing that W_t significantly affects not only the antenna’s central frequency but also its reflection coefficient. At a W_t of 0.17 mm, the antenna achieved a reflection coefficient below −10 dB from 26.3 to 29.7 GHz, centered at 28 GHz. Notably, at the end of the design process, the dimensions of the proposed antenna were 10.7 mm × 10.7 mm × 0.78 mm (λ₀ × λ₀ × 0.07λ₀).

Fig. 7 illustrates the simulated current distribution induced in Port 1 at 28 GHz with regard to various designs of the feeding network. As shown in Fig. 7(a), the air-cavity dual-polarized antenna with a straight aperture exhibits severe interference, since the apertures of Ports 1 and 2 are adjacent. In Fig. 7(b), U-shaped apertures are implemented to reduce interference between the apertures, and the striplines are designed with right-angle bends to ensure adequate spacing between them. However, leakage currents within the cavity still result in the current being transferred to Port 2. To address this issue, as depicted in Fig. 7(c), an isolation via is placed at the center of the cavity, and a second isolation via (Isolation-via 2) is positioned 0.6 mm away from the center in the -x and -y directions.

As depicted in Fig. 8, simulations were conducted to verify the isolation between ports based on the design of the feeding network. The isolation level of the air-cavity dual-polarized antenna designed with a straight aperture is observed to be −11.4 dB at 28 GHz. The occurrence of interference between the apertures was also verified. Furthermore, the design with a U-shaped aperture achieved an isolation level of −31.5 dB at 28 GHz, but did not satisfy the requirement for isolation of more than −30 dB within the operating frequency band of 26.5–29.5 GHz. Meanwhile, the air-cavity dual-polarized antenna with isolation vias achieved port-to-port isolation of more than −30 dB within the 24 to 32 GHz band, and was also verified to reduce interference between the two apertures and leakage current within the cavity.

Fig. 9 illustrates the configuration of the proposed antenna, including the structure used for validation. To verify its dual-polarization characteristics, the antenna was fed through two stripline K-type connectors. For optimal measurement, the feeding line was bent to link up with the connector. Furthermore, to address the discontinuity between the connector and the feeding line, impedance matching using a transformer was implemented. Additionally, to maintain a 0.5-mm gap between the aperture and the patch antenna on Layer 1, a 3D-printed structure was employed. For precise alignment of the aperture in the feed section with the patch antenna, two types of hole structures were incorporated into the 3D-printed structure, allowing for relatively accurate alignment of Substrate 1, the 3D-printed structure, and Substrates 2–4, each of which was implemented separately.

Fig. 10 presents photos of the prototype of the proposed antenna. The aforementioned alignment holes were circular in shape, ensuring that the antenna substrate, the 3D-printed structure, and the feeding substrate were well aligned. The overall dimension of the proposed antenna, including the connectors and the 3D-printed structure, was 38 mm × 21 mm × 0.78 mm (3.55λ₀ × 1.96λ₀ × 0.07λ₀).

Fig. 11 presents a comparison of the S-parameter results obtained for the fabricated antenna and the simulation. The measured reflection coefficient results for the fabricated antenna indicate that Port 1 achieved values below −10 dB from 26.4 to 29.3 GHz, while Port 2 attained values below −10 dB from 26.4 to 29.2 GHz. The simulation results for the reflection coefficient of the proposed antenna indicate that Port 1 has values below −10 dB from 26.4 to 29.6 GHz and Port 2 attained values below −10 dB from 26.5 to 29.7 GHz, suggesting similar performance as the measurement results. The simulated isolation results are below −30 dB across the 26–30 GHz band, with a peak value of −31.6 dB. The measured isolation results also remain below −30 dB within the same frequency range, with the peak value being −33.0 dB. Furthermore, the simulation and measurement results for the reflection coefficient agree well, with a slight difference observed in the 29.5-GHz band, likely due to process variations in the 3D-printed structure used to implement the air cavity. Despite these variations, the fabricated antenna exhibited a fractional bandwidth greater than 10%, indicating proper operation at the target frequency.

Fig. 12 presents the measurement results for the co-polarization and cross-polarization radiation patterns of the fabricated antenna. The detailed results are presented in Table 2, which shows that the cross-polarization discrimination (XPD) between the copolarization and cross-polarization in the main lobes of V-pol and H-pol is >20 dB in the target frequency range. Meanwhile, the measured half-power beamwidth (HPBW) for V-pol and H-pol in the target frequency range varies from 60° to 69°. These findings confirm that the V-pol and H-pol characteristics of the proposed antenna are similar.

Fig. 13 presents the gain results for the proposed antenna. In the frequency range of 26.5–29.5 GHz, the V-pol simulation results for gain span 7.2–8.6 dB, while the measurement results range from 6.8 to 8.4 dB. Furthermore, the H-pol simulation results for gain range from 7.1 to 8.4 dB, while the measurement results span 6.7–8.7 dB. Overall, the measured gain results in the target frequency range were similar for the V-pol and H-pol. Moreover, the simulation results closely matched the measurement results. Thus, it is established that the proposed dual-polarization antenna can operate successfully in the target frequency range.

Table 3 presents a performance comparison of previously proposed mmWave-band dual-polarization patch antennas. Since this work focuses on operation in the 5G mmWave band, we compared dual-polarization patch antennas designed for operation in the 5G mmWave (28 GHz) band.

The method proposed in [26] involves a dual-polarization dual-band structure that uses a bridging mechanism based on parasitic elements on a single patch. While this approach follows relatively simple design methods, it exhibited poor port-to-port isolation performance. In [27], a stacked patch structure was employed to achieve a wide bandwidth and isolation performance of −20 dB. However, the design required an antenna substrate with a thickness of 0.89 mm, while the inclusion of parasitic patch elements necessitated a complex design process. Furthermore, in [28], a feed structure similar to that proposed in [27] was used, resulting in >7-dBi gain using a thinner substrate. However, this was accompanied by a 7% reduction in bandwidth. Notably, this result indicates a direct correlation between substrate thickness and bandwidth. In [29], broadband characteristics across the 5G mmWave band were attained using L-probe feeding and parasitic patches, but this process required a thick (0.8 mm) substrate and achieved insufficient peak isolation (−20 dB). In [30], excitation was achieved using a capacitive T-shaped probe situated on the same layer as the square patch, and a defected ground structure was employed to enhance the isolation performance. However, while this configuration reduced the antenna ground plane area, it still necessitated a substrate thickness close to 0.8 mm. In [31], an air-cavity structure was employed to achieve a bandwidth exceeding 10%.

However, the feeding transmission lines on the different layers, which were employed to achieve dual polarization, resulted in increased loss. Moreover, the ground vias situated between the antennas primarily increased the isolation between antenna elements rather than significantly improving port-to-port isolation. In [32], stacked patches were utilized to implement a dual-polarization dual-band antenna, yielding a wide bandwidth of 19.5% in the low band and 7.8% in the high band. However, the use of eight layers in the design significantly increased its complexity. Finally, in [33], an SIW cavity-based aperture feed was used to excite a corner-truncated patch antenna. This design achieved a gain of >7 dBi using only three metal layers, but the inherent characteristics of the SIW cavity resulted in a narrow antenna bandwidth.

In this work, we propose a patch antenna that offers enhanced isolation performance in the 5G mmWave band using a U-shaped aperture feed structure and isolation vias. The proposed antenna achieved −33 dB isolation, ensuring superior port-to-port isolation performance while also permitting the independent operation of each port. Additionally, it attained a high peak gain of 8.6 dBi using a single element. The use of only four substrates simplified the design and implementation process compared to previous works. Notably, the bandwidth was slightly narrower than that reported for previously reported antennas, likely due to the thickness of the antenna substrate [37]. Nevertheless, it exceeded 10%, indicating the applicability of the proposed antenna in the 5G mmWave band.

This paper presents the design method for an aperture-coupled dual-polarization patch antenna with improved isolation characteristics achieved using isolation vias. The proposed antenna exhibits a fractional bandwidth greater than 10% while maintaining a high port-to-port isolation performance. Its key design features include the integration of an air cavity using a 3D-printed structure, a modified U-shaped aperture, and isolation vias to enhance isolation characteristics. The fabrication and measurement results verify that the antenna is capable of achieving gains exceeding 6.8 dBi for V-pol and 6.7 dBi for H-pol within the target frequency band while also maintaining an isolation level below −30 dB. These results establish that the proposed antenna ensures a wide bandwidth and high isolation, making it suitable for 5G mmWave applications.