Comprehensive evaluation of a tri-band MIMO antenna with SAR and MIMO performance for 5G devices Authors: S Murugeswari, R Dhananjeyan, M Suganthy, S Selvakumar, P Sundaravadivel, D Rajesh Kumar Published: 2026 ABSTRACT AbstractIn this work, a compact two-port MIMO antenna has been proposed for 5G millimeter wave applications. The antenna is fabricated on Rogers RT/Duroid 5880 substrate and occupies a footprint of only 12.5 × 4.8 × 0.65 mm³, which is among the smallest dimensions reported in the literature for comparable designs. Even with such tight physical constraints, the antenna is able to provide a gain ranging from 9.3 to 9.7 dBi across three distinct operating bands, namely 26.47–30.85 GHz, 31.29–32.86 GHz, and 35.8–38.48 GHz. The multiband behavior is mainly attributed to the meandered slot-loaded structure, which effectively increases the electrical length without any corresponding increase in the physical dimensions of the antenna. The MIMO performance parameters are also found to be satisfactory. Port isolation of − 20 dB has been achieved, while the envelope correlation coefficient (ECC) is maintained below 0.01. The channel capacity loss is recorded as 0.33 bps/Hz, which confirms that the mutual coupling between the ports is sufficiently low and that stable multi-stream data transmission can be supported. The measured results for S-parameters, gain, radiation patterns, and surface current distributions are in good agreement with the simulated values, thereby validating the design experimentally. SAR analysis has been carried out using a multi-layered human hand phantom model. The obtained SAR values are found to be within the permissible exposure limits for all three operating bands, which is of particular relevance since the antenna is intended for use in handheld and wearable 5G devices. On comparison with recently published works in the literature, the proposed design demonstrates competitive performance in terms of physical size, gain, bandwidth coverage, and electromagnetic safety compliance. Hence, the design can be considered as a suitable candidate for IoT and wearable 5G communication systems.Subject terms: Energy science and technology, Engineering Introduction mmWave MIMO antenna research has moved quickly over the past few years, producing designs that are increasingly compact and better suited to 5G deployment. Iqbal et al.1 built an EBG-backed wearable MIMO antenna with strong port isolation for body-centric use. Aggarwal et al.2 surveyed mmWave antenna design principles, practical applications, and the key challenges facing next-generation wireless systems. Hussain and Kim3 tackled full spatial coverage by merging microwave and mm-wave MIMO modules into a single architecture, reaching 360° pattern diversity for IoT use. El Halaoui et al.4 put forward a dual-band inverted-F array antenna at 28/38 GHz aimed at 5G mobile handsets. Elalaouy et al.5 cut mutual coupling in a two-port MIMO antenna by etching a defected ground structure into the ground plane. P.S.B. G et al.6 reviewed where planar mmWave MIMO antenna design currently stands and where it is likely headed. Alfakhri7 used a cross-shaped decoupling structure to get dual polarization and better isolation in UWB MIMO antennas. Ali et al.8 built a dual-band 2 × 2 MIMO slot antenna with low mutual coupling, aimed squarely at 5G networks. Elsharkawy et al.9 reported a dual-band MIMO antenna system operating at 28/38 GHz, wherein the footprint was kept compact while maintaining good integration potential. Sabek et al.10 focused on further improvements in isolation and size reduction for a dual-band MIMO configuration. In the work of Megahed et al.11, it was demonstrated that wideband MIMO arrays are capable of maintaining high port isolation across 5G mmWave bands. A different design approach was adopted by Rahman et al.12, where bio-mimetic principles were employed to realize a novel mmWave MIMO antenna geometry. Tariq et al.13 incorporated a metasurface layer into the antenna structure, which resulted in enhancement of both radiation performance and inter-port isolation. Hasan et al.14 proposed a dual-band omnidirectional mmWave antenna providing full azimuthal coverage, which is considered suitable for mobile terminal applications. The use of a defected ground structure (DGS) beneath a 4-port MIMO layout was investigated by Khalid et al.15, through which isolation losses were directly reduced for 5G mmWave scenarios. El-Hassan et al.16 achieved quad-band operation within a single MIMO platform, addressing the requirements of future generation mobile networks. Patel et al.17,18 designed CPW-fed four-port UWB MIMO antennas, with one variant employing an inverted-L geometry to improve impedance matching behaviour. Munir et al.19 simultaneously achieved wide bandwidth and high inter-element isolation in a four-element MIMO system intended for reliable 5G communication links. Esmail and Koziel20 made use of metamaterial loading to enhance antenna gain and to introduce beam-tilting capability into the design. A compact circularly polarized MIMO antenna operating across dual bands at 28 and 38 GHz was presented by Alnemr et al.21, and the design was found to be well suited for portable device applications. Murthy22 explored the integration of metamaterial concepts into dielectric resonator antenna structures, with the objective of improving isolation performance at mmWave frequencies. Singh et al.23 investigated how geometric modifications to MIMO antenna structures can directly contribute to isolation enhancement. Finally, Jayanthi and Kalpana24 developed a six-element MIMO antenna demonstrating strong gain characteristics at 28/38 GHz bands, and the design was experimentally validated for high-throughput wireless communication applications. Ali et al.25 reported a dual-band planar MIMO antenna design for the 27/39 GHz frequency range, where compact physical dimensions and adequate inter-port isolation were achieved together. In the work of Sokunbi et al.26, slot-based decoupling techniques were utilized to obtain self-isolation characteristics in a wideband MIMO configuration for mmWave 5G use. Tadesse et al.27 proposed a four-port planar MIMO antenna with a compact footprint operating at 28/38 GHz, which makes it appropriate for applications where available board space is limited. The problem of simultaneously reducing antenna size and improving bandwidth in UWB MIMO antennas for 5G mmWave systems was taken up by Abbas et al.28. Kumar et al.29 presented a detailed review of the design challenges, fabrication and measurement considerations, and practical aspects involved in the development of mmWave antennas for next-generation 5G networks. Compact wearable antennas for healthcare and body-centric communication have attracted steady research attention, with most work converging on flexible, ultra-wideband designs that clear strict SAR thresholds. Karad et al.30 built a hexagonal-shaped UWB antenna and ran detailed SAR tests for healthcare use. Their next paper31 brought in a Shri Yantra-shaped fractal flexible antenna targeting super-wideband performance on 5G wearable healthcare platforms. From the same group, a flower bud-shaped flexible UWB antenna32 and a foam-based compact flexible wideband antenna33 were both put through body-worn healthcare test scenarios. What these four papers share is a consistent push to combine SAR compliance, flexible substrates, and bio-compatible materials — a combination that directly frames the SAR evaluation carried out here. Tiwari et al.34 fabricated a flexible four-port MIMO antenna on Rogers 3003 substrate at 18 × 8.5 × 0.25 mm³, covering 28/38 GHz with peak gains of 4.15 dBi and 7.73 dBi, SAR within FCC/ICNIRP limits, and bending tests confirming wearable fitness. Tiwari et al.35 built a compact quad-port dual-band MIMO at 24 GHz and 38 GHz inside a 7 × 24 × 0.25 mm³ footprint, using circular patches with modified T-shaped ground slots to reach isolation above 21 dB, peak gain over 5.68 dBi, and ECC below 0.018. Hussain et al.36 got frequency-reconfigurable operation at 24.8 GHz and 28 GHz through simple geometric changes alone, with no switching circuitry and miniaturized dimensions. Hussain et al.37 developed a bendable MWCNT/PDMS composite antenna for WLAN and Wi-Fi 6E wearable use in 5G/6G, running across 5.19–5.6 GHz and 6.68–6.92 GHz with gain above 4.5 dBi, radiation efficiency over 87% under bending, and SAR compliance confirmed on a body phantom. Across these works, the pattern is clear — compact size, flexible substrates, multiband coverage, and SAR validation are the four design priorities now driving wearable mmWave antenna development, and the proposed antenna addresses all four at once. In spite of the significant amount of research work that has been carried out in this area, the existing mmWave MIMO antenna designs are still associated with certain practical limitations that affect their suitability for direct use in 5G deployment scenarios. The issue of antenna size remains a commonly encountered trade-off, as many of the designs reported in the literature are able to meet the desired performance targets only by adopting a relatively larger footprint, which restricts their use in compact or wearable form factors. Another limitation that is frequently observed is that most antennas are designed to operate over only one or two mmWave bands, and this narrow frequency coverage reduces their overall usefulness for 5G systems that span a wider range of operating frequencies. Although isolation enhancement techniques such as defected ground structures and metamaterial loading are now regularly employed, maintaining adequately low mutual coupling between ports in a physically small multi-port configuration is still a challenging task. It is also noted that experimental validation is lacking in a considerable number of published works, and measured MIMO performance parameters as well as SAR analysis results are not reported in most of the studies available in the literature. These observations collectively suggest that there exists a clear requirement for an antenna that combines small physical size, high gain, multiband operation, and good port isolation, while also being supported by measured performance data and electromagnetic safety assessment. The antenna design proposed in this work attempts to address each of the above-mentioned limitations in a systematic manner. With a total footprint of 12.5 × 4.8 × 0.65 mm³, the antenna is sufficiently compact to be accommodated within wearable and portable device platforms without any significant compromise on physical size. While most of the designs reported in the literature are limited to dual-band operation, the proposed antenna is capable of covering three distinct mmWave bands, which provides comparatively broader frequency coverage across the 5G spectrum. The meandered slot-loaded structure employed in this design contributes to both gain enhancement and port isolation improvement, with values reaching up to 9.7 dBi and − 20 dB respectively, and this is achieved without the incorporation of any additional decoupling elements. The ECC is maintained below 0.01 and the channel capacity loss is found to be as low as 0.33 bps/Hz, which are indicative of stable multi-stream data transmission capability. Unlike several works in the existing literature, the performance of the proposed antenna has been verified through prototype measurements, and the reflection coefficients, transmission coefficients, radiation patterns, and gain have all been experimentally confirmed across the three operating bands. SAR analysis conducted on a multilayered human hand phantom further validates the design from an electromagnetic safety standpoint, with the obtained values found to be within permissible exposure limits for all operating bands, which is an aspect that has not been adequately addressed in most comparable reported designs. Antenna design and analysis The geometry of the proposed two-port MIMO antenna has been developed by taking into consideration compact resonant principles that are well suited for millimeter-wave frequency operation. The antenna is realized on a low-loss Rogers RT/Duroid substrate having a thickness of 0.65 mm, which was selected primarily to ensure minimal dielectric losses and to maintain stable electrical performance at higher operating frequencies. The overall antenna structure comprises two identical radiating elements that are arranged in a symmetric configuration, so as to preserve radiation pattern stability and to achieve adequate reduction in mutual coupling between the two ports. Design principle The design of the radiating element is governed by the concept of effective electrical length, which determines the resonance of planar structures. For microstrip and स्लॉट-based configurations, the resonant frequency can be approximated using the guided wavelength , given by: where cis the speed of light in free space, fis the resonant frequency, and is the effective permittivity of the substrate. The effective permittivity accounts for fringing fields and is expressed as1,2: where is the substrate dielectric constant, his the substrate thickness, and Wrepresents the effective width of the radiating structure. For the proposed antenna, the resonant behavior is approximated by considering the effective electrical length of the meandered slot, which satisfies the half-wavelength condition: Unlike conventional straight-line radiators, the use of a meandered-slot configuration increases the current path within a limited physical area. This enables the antenna to resonate at lower frequencies while maintaining a compact footprint, which is critical for mm-Wave integrated devices. Figure 1 (a) displays the top view of the antenna layout, which consists of two identical radiating elements—Antenna 1 and Antenna 2—symmetrically placed on either side of a common substrate, which is a 0.65 mm thick RO3010 material, known for its low loss and stable dielectric properties, ideal for high-frequency operation. The compact geometry aims to reduce the overall footprint while maintaining effective MIMO performance. Figure 1 (b) provides a detailed dimensional breakdown of Antenna 1, highlighting the meandered slot design used to achieve compactness and wideband operation. The meandered and stub-loaded structure enhances the effective electrical length, enabling resonance at lower frequencies without increasing physical size. The precise dimensions such as L1 = 7.3 mm and W1 = 1.9 mm, among others listed in Table 1, suggest meticulous optimization to support desired frequency bands and achieve high isolation, low envelope correlation coefficient (ECC), and good impedance matching. This structure is particularly suited for integration in portable wireless systems demanding multi-band operation with minimal mutual coupling between ports. (a) Structure of proposed MIMO antenna, (b) Detailed dimension of antenna 1. Table 1Various parameters of the proposed MIMO antenna.LWL1L2L3L4L5L6L7L8L9L10W112.54.87.30.340.170.170.260.260.260.260.110.131.9 Various parameters of the proposed MIMO antenna. Figure 2 presents the simulated S-parameters of the proposed MIMO antenna across the frequency range of 24 to 42 GHz, demonstrating its reflection and transmission characteristics. The key parameters—S11 and S22 (reflection coefficients of Antenna 1 and Antenna 2), and S12 and S21 (mutual coupling between the ports)—are plotted to evaluate impedance matching and isolation. The antenna achieves three distinct impedance bandwidths below the − 10 dB threshold: 26.47–30.85 GHz, 31.29–32.86 GHz, and 35.8–38.48 GHz, which are highlighted by the shaded regions. These bands confirm the antenna’s multi-band operation within the millimeter-wave spectrum, suitable for 5G applications. The S11 and S22 curves drop significantly below − 10 dB within these bands, indicating good impedance matching and low reflection. Additionally, the transmission coefficients S12 and S21 remain well below − 15 dB across the operational bands, suggesting excellent port isolation and low mutual coupling—critical for effective MIMO performance. Overall, the antenna exhibits efficient multi-band functionality with strong isolation and minimal signal degradation between ports. Simulated S Parameters of the proposed antenna. Figure 3 demonstrates the design evolution of the proposed MIMO antenna and its corresponding reflection coefficient performance across successive stages. Figure 3 (a) presents the antenna development through five incremental steps, starting from a basic rectangular slot configuration in Step 1 to the final complex geometry in the Proposed design. With each step, additional slotting and structural modifications are introduced to enhance impedance bandwidth and resonance behavior. Figure 3 (b) shows the simulated reflection coefficients (S11) for each step over the 24–42 GHz range. The initial Step 1 design exhibits poor impedance matching across the desired bands, with reflection coefficients mostly above − 10 dB. As the geometry becomes more complex (Steps 2–4), the matching improves, and new resonances begin to appear within targeted mm-wave frequency ranges. The Proposed design achieves significant improvement, presenting three well-defined impedance bandwidths (highlighted in blue), where the reflection coefficient drops below − 10 dB—specifically over 26.47–30.85 GHz, 31.29–32.86 GHz, and 35.8–38.48 GHz. This progression clearly validates the effectiveness of the design modifications in achieving multi-band operation and optimal impedance matching for 5G millimeter-wave applications. (a) Design evolution of the proposed antenna and its, (b) Reflection coefficient. Parametric analysis when altering (a) W1, (b) L1. Figure 4 presents a parametric study of how small variations in the feeding-stub width (W1) and slot length (L1) affect the antenna’s impedance response. Figure 4 (a) compares the reflection coefficient for three values of W1 (1.8 mm, 1.9 mm, and 2.1 mm): narrowing W1 to 1.8 mm shifts the first resonance upward by roughly 0.3 GHz and degrades matching around 37 GHz, while widening W1 to 2.1 mm lowers the lowest‐band resonance by about 0.5 GHz but slightly weakens the mid‐band dip. The optimized 1.9 mm width balances these shifts, yielding all three bands below − 10 dB with robust matching. Figure 4 (b) shows the effect of tuning L1 between 7.2 mm, 7.3 mm, 7.4 mm, and 7.5 mm: increasing L1 lengthens the effective electrical path, pulling the middle resonance from 32.5 GHz down to approximately 31.9 GHz and slightly broadening the upper band, whereas shortening L1 shifts that resonance higher and narrows the lowest band. The chosen 7.3 mm value delivers the most symmetric multi‐band response, confirming that precise control of W1 and L1 is critical for centering all three operational bands and achieving optimal impedance matching. Surface current distribution of antenna 1 at three distinct frequencies. Figures 5 and 6 depict the surface current distribution of the proposed MIMO antenna, providing valuable insights into its resonant behaviour and mutual coupling characteristics at three key frequencies: 28.5 GHz, 32 GHz, and 37 GHz. Figure 5 focuses on Antenna 1 alone, showing that at each resonant frequency, strong current concentrations (highlighted in red) are localized along specific arms and slots of the radiator. At 28.5 GHz, intense surface currents are observed along the vertical stubs near the feed line, indicating dominant radiation from these segments. At 32 GHz, the current is more uniformly distributed across the meandered arms, suggesting an efficient excitation of the second resonant mode. At 37 GHz, higher-frequency modes emerge, and current is concentrated near the shorter horizontal segments, demonstrating higher-order resonance behavior. Figure 6 expands the view to the full MIMO structure, with Port 1 excited. It reveals strong current confinement around Antenna 1 across all three frequencies, while Antenna 2 exhibits minimal induced currents. This suggests excellent port isolation and effective suppression of mutual coupling. The current profiles reaffirm that each resonant mode is well excited by the meandered-slotted geometry and that the structure supports independent multi-band operation. The use of a symmetrical layout and carefully engineered slot placement contributes to the minimized interaction between ports, a critical factor for MIMO performance at millimeter-wave bands. Surface current distribution of proposed MIMO antenna when port 1 is excited. To further validate the physical behaviour of the proposed tri-band MIMO antenna and provide a computationally efficient alternative to full-wave electromagnetic simulation, a lumped-element equivalent circuit model is developed as illustrated in Fig. 7 (a) and analyzed using Keysight ADS. The equivalent circuit topology, presented in Fig. 7(a), consists of two identical antenna element sub-circuits connected through a central mutual coupling T-network, reflecting the symmetric two-port architecture of the proposed design. Each antenna element is modelled by a series feed network comprising a feedline inductance LF = 0.18 nH and a feed resistance RF = 2.10 Ω, which represent the inductive reactance and ohmic loss of the W₁ = 1.9 mm wide microstrip feedline on the Rogers RT/Duroid 5880 substrate, respectively. A shunt substrate capacitance CSUB = 0.042 pF is connected between the feed junction and the ground plane to account for the fringing electric field between the antenna metallization and the low-permittivity substrate (εr = 2.2), which influences the effective resonant frequency of each band. (a) Equivalent circuit of the proposed antenna, (b) Comparison of S parameters between EM full wave analysis and equivalent circuit (c). The three operational resonant bands of each antenna element are individually represented by three parallel series LCR branches connected between the feed junction node and the common ground rail. The first resonant branch (Table 2), comprising L₁ = 0.320 nH, C₁ = 0.098 pF, and R₁ = 4.80 Ω, models the fundamental resonance of the meandered main stub (L1 = 7.3 mm) at 28.5 GHz corresponding to Band 1 (26.47–30.85 GHz). The relatively large inductance L₁ reflects the extended electrical length of the meandered path, while the radiation resistance R₁ = 4.80 Ω is consistent with the 9.3 dBi gain achieved at this band. The second resonant branch with L₂ = 0.210 nH, C₂ = 0.118 pF, and R₂ = 5.50 Ω represents the horizontal slot arm resonance at 32 GHz corresponding to Band 2 (31.29–32.86 GHz). The reduced inductance L₂ compared to L₁ reflects the shorter effective electrical path of the horizontal slot arms, while the slightly higher capacitance C₂ accounts for the stronger capacitive coupling associated with the slot geometry at this frequency. The third resonant branch with L₃ = 0.150 nH, C₃ = 0.121 pF, and R₃ = 6.20 Ω models the higher-order resonance of the short loading stubs L5–L8 at 37.5 GHz corresponding to Band 3 (35.8–38.48 GHz). The progressively decreasing inductance values from L₁ to L₃ are physically consistent with the shorter radiating paths responsible for higher-frequency resonances in the meandered-slot geometry, while the increasing radiation resistance from R₁ to R₃ reflects the proportionally higher radiation efficiency observed at higher operational frequencies. Table 2Extracted lumped-element component values of the equivalent circuit model for the proposed two-port tri-band MIMO antenna.ComponentSymbolValuePhysical meaningFeed inductanceLF0.18 nHInductive reactance of W1 = 1.9 mm microstrip feedlineFeed resistanceRF2.10 ΩOhmic loss of feedline on Rogers RT/Duroid 5880Substrate capacitanceCSUB0.042 pFFringing capacitance between metallization and substrate (εr = 2.2)Branch inductanceL₁0.320 nHEffective inductance of meandered stub (L1 = 7.3 mm)Branch capacitanceC₁0.098 pFSlot capacitance at Band 1 resonanceRadiation resistanceR₁4.80 ΩRadiation loss at Band 1 (related to gain = 9.3 dBi)Branch inductanceL₂0.210 nHEffective inductance of horizontal slot armsBranch capacitanceC₂0.118 pFSlot capacitance at Band 2 resonanceRadiation resistanceR₂5.50 ΩRadiation loss at Band 2Branch inductanceL₃0.150 nHEffective inductance of short stubs (L5–L8)Branch capacitanceC₃0.121 pFSlot capacitance at Band 3 resonanceRadiation resistanceR₃6.20 ΩRadiation loss at Band 3 (related to gain = 9.7 dBi)Mutual inductanceLM0.080 nHMagnetic field coupling through shared Rogers substrateMutual resistanceRM0.060 ΩResistive mutual loss (related to S12 < − 15 dB isolation)Coupling capacitanceCM0.032 pFElectric field coupling between the two antenna elements Extracted lumped-element component values of the equivalent circuit model for the proposed two-port tri-band MIMO antenna. The electromagnetic coupling between the two antenna elements is represented by a central mutual coupling T-network consisting of a series mutual inductance LM = 0.080 nH on each side, a shunt mutual resistance RM = 0.060 Ω, and a shunt coupling capacitance CM = 0.032 pF connected to the ground reference. The small values of LM and RM are physically consistent with the high port isolation (S₁₂ < −15 dB) achieved in the proposed design, indicating that the inter-element magnetic coupling through the shared substrate is effectively suppressed by the symmetric meandered geometry. The coupling capacitance CM = 0.032 pF captures the residual electric field coupling between the two antenna elements, which is inherently limited by the compact spacing and the low permittivity of the Rogers RT/Duroid 5880 substrate. The accuracy of the extracted component values is verified using the standard LC resonance formula: As tabulated in Table 3, the resonant frequencies computed from the extracted L and C values 28.43 GHz, 32.05 GHz, and 37.42 GHz for Bands 1, 2, and 3, respectively deviate from the target center frequencies by less than 1.2%, confirming the physical consistency and extraction accuracy of all component values. The maximum error of 1.14% observed at Band 3 is attributed to the higher-order distributed nature of the short stub resonance, which is inherently less accurately captured by a single lumped LC pair compared to the fundamental and second resonances at lower bands. Table 3Verification of extracted equivalent circuit component values against target resonant frequencies.Bandfr (GHz)L (nH)C (pF)Computed fr (GHz)Error (%)Band 128.50.320.09828.430.25Band 2320.210.11832.050.16Band 3370.150.12137.421.14 Verification of extracted equivalent circuit component values against target resonant frequencies. The S-parameter responses synthesized from the equivalent circuit model in ADS are compared against the full-wave CST EM simulation results in Fig. 7 (b) & (c). Both responses confirm the tri-band operational coverage across the three designated frequency bands, with the equivalent circuit correctly predicting the − 10 dB impedance bandwidth boundaries for all three bands. The observable differences in null depth, resonance center frequency, and fine spectral structure between the two responses are physically expected and are inherent to the lumped-element approximation of distributed electromagnetic behavior. Specifically, the shallower resonance nulls and slightly shifted band centers in the equivalent circuit response are attributed to the inability of the lumped model to capture substrate dispersion, higher-order radiation modes, and edge diffraction effects that are fully resolved in the three-dimensional EM simulation. The transmission coefficient S₁₂ predicted by the equivalent circuit model is approximately 3–5 dB higher than the EM simulation result, reflecting the inherent limitation of the simplified mutual coupling T-network in fully capturing the complex inter-element electromagnetic interaction at millimeter-wave frequencies. Despite these expected differences, the close agreement in band coverage, impedance matching trend, and overall S-parameter behaviour between the equivalent circuit and EM simulation confirms the validity of the proposed equivalent circuit model as an accurate and computationally efficient tool for the physical interpretation and initial design optimization of the proposed two-port tri-band MIMO antenna. Measured results and discussions Figure 8 illustrates a fabricated prototype of the proposed MIMO antenna. Figure 9 presents a comparative analysis of the simulated and measured S-parameters of the proposed MIMO antenna, highlighting both reflection and transmission characteristics. Figure 9 (a) presents the reflection coefficients (S11 and S22) for both ports. The simulated curves (S11 Sim and S22 Sim) demonstrate good impedance matching across three distinct bands—26.47–30.85 GHz, 31.29–32.86 GHz, and 35.8–38.48 GHz—where the reflection coefficients remain below − 10 dB. The measured results (S11 Mea and S22 Mea), although exhibiting minor deviations due to fabrication and connector tolerances, closely follow the simulated trends, confirming the antenna’s effective multi-band operation and design robustness. Figure 9 (b) illustrates the transmission coefficients (S12 and S21), representing the mutual coupling between the two ports. Both simulated and measured results show that the transmission coefficients remain well below − 15 dB over the desired frequency bands, indicating excellent isolation and low mutual coupling. The consistency between simulation and measurement confirms the antenna’s viability for high-performance MIMO applications at millimeter-wave frequencies, ensuring reliable signal separation and minimal interference between elements. Fabricated prototype of the proposed antenna with measurement setups. Simulated and measured S parameters (a) Reflection coefficient, (b) Transmission coefficient. Figure 10 illustrates the simulated and measured far-field radiation patterns of the proposed antenna at three resonant frequencies: 28.5 GHz, 32 GHz, and 37.5 GHz. At all frequencies, both the E-plane (Theta) and H-plane (Phi) components demonstrate consistent radiation behaviour, with good agreement between simulated and measured results. At 28.5 GHz, the radiation pattern exhibits a broad main lobe and symmetrical profile, indicating efficient radiation with minimal side lobes. At 32 GHz, the antenna maintains its directional stability, with the measured patterns closely following the simulated curves, confirming reliable performance in the second band. At 37.5 GHz, the patterns preserve their shape, showing a slightly more omnidirectional behaviour in the H-plane, typical of higher-order resonances. Overall, the consistency between simulation and measurement validates the antenna’s stable radiation characteristics across its operating bands, making it well-suited for millimeter-wave applications that demand predictable far-field behaviour. Simulated and measured far-field pattern of the proposed antenna at (a) 28.5 GHz, (b) 32 GHz, (c) 37.5 GHz. Figure 11 (a–c) illustrates the simulated and measured gain performance of both antenna elements in the proposed MIMO configuration across the 24–42 GHz frequency range. The antenna demonstrates stable and relatively high gain within the three operational bands: 26.47–30.85 GHz, 31.29–32.86 GHz, and 35.8–38.48 GHz. The simulated gain values for Antenna 1 and 2 remain above 8 dBi across the designated bands, peaking at around 10.5 dBi, indicating efficient radiation characteristics. The measured gain results closely follow the simulated trends, with slight variations attributed to fabrication tolerances and measurement setup imperfections. Despite these deviations, the measured gains remain within acceptable limits, confirming the reliability of the antenna design. The consistent gain performance across all bands validates the proposed antenna’s suitability for millimeter-wave MIMO applications requiring stable and directional radiation characteristics. Simulated and measured gain of the proposed two-port tri-band MIMO antenna across the three operational millimeter-wave bands: (a) 26.47–30.85 GHz, (b) 31.29–32.86 GHz, and (c) 35.8–38.48 GHz. MIMO performance Figure 12, along with the accompanying equations, comprehensively evaluates the MIMO performance of the proposed antenna design. The graphical plots illustrate four critical metrics: envelope correlation coefficient (ECC), diversity gain (DG), mean effective gain (MEG), and channel capacity loss (CCL). The ECC, shown in Fig. 12 (a), remains below 0.005 across all operating bands, indicating excellent isolation and minimal signal correlation between antenna elements. This behaviour is mathematically supported by Eq. (1), which calculates ECC based on the far field patterns. MIMO Parameters of the proposed antenna (a) envelope correlation coefficient, (b) diversity gain, (c) mean effective gain, (d) channel capacity loss. Correspondingly, Fig. 12 (b) demonstrates that the diversity gain remains near the ideal value of 10 dB, as expressed by Eq. (2), confirming the antenna’s ability to enhance signal reliability in multipath environments. Figure 12 (c) presents the MEG values for both ports, which remain balanced, signifying consistent radiation efficiency. Finally, Fig. 12 (d) shows that the CCL remains below 0.4 bits/s/Hz, suggesting negligible degradation in system capacity, as defined by Eqs. (3) and (4), which employ the ECC matrix to compute channel capacity constraints. Where, Collectively, the results and equations validate the antenna’s high-performance characteristics, ensuring. its suitability for advanced MIMO applications in the millimeter-wave spectrum. SAR analysis The proposed antenna is designed for integration into handheld and wrist-worn 5G millimeter-wave devices such as smartphones, smartwatches, and IoT wearables. In such devices, the hand and wrist represent the primary body region in closest proximity to the antenna during operation. Therefore, a multi-layered human hand phantom model is selected as the most anatomically representative scenario for near-field electromagnetic exposure assessment. Figure 13 illustrates the Specific Absorption Rate (SAR) analysis for the proposed antenna when placed near a human hand phantom model. This setup mimics realistic wearable or handheld device usage, where electromagnetic exposure to human tissue must be evaluated for safety. Figure 13 includes a multi-layered model of human skin, composed of the epidermis, dermis, hypodermis, and muscle layers, which accurately represents the dielectric and conductive nature of human tissue. The SAR values are measured at three key operating frequencies (27.5 GHz, 32 GHz, and 37.5 GHz) and remain well below the standard safety limit of 1.6 W/kg averaged over 1 gram and 10 g of tissue, as specified by international guidelines (FCC/IEEE). Table 4 records SAR values that range from 0.9546 to 1.3245 W/kg, confirming that the antenna poses minimal health risk while maintaining performance integrity. These results validate the antenna’s suitability for integration into user-centric, body-worn millimeter-wave devices without violating electromagnetic exposure constraints. Table 5 details the dielectric and electrical properties of human skin layers and provides insight into SAR behaviour. The epidermis, being the outermost layer, has moderate permittivity and conductivity, affecting surface-level absorption. The dermis, due to its higher water content, exhibits greater permittivity (25–30) and conductivity (1.5–2.0 S/m), contributing significantly to SAR. The hypodermis, primarily composed of fat, has lower permittivity and conductivity, acting as a partial insulator. Lastly, muscle tissue, being highly conductive and permittivity (up to 55 εr, 5 S/m), absorbs deeper penetrating fields, but its effect depends on how much of the field reaches it. These layer-specific characteristics influence how electromagnetic energy is distributed within the hand phantom and play a critical role in ensuring realistic and safe SAR predictions for mm-wave antenna designs. Placement analysis of proposed antenna on human hand phantom model. Table 4Simulated SAR values of the proposed antenna.Frequency(GHz)SAR (W/Kg)1 g tissue10 g tissue27.51.250.98745321.13281.0137.51.32450.9546 Simulated SAR values of the proposed antenna. Table 5Dielectric and Electrical Properties of Human Skin Layers LayerRelative Permittivity (εr)Conductivity (σ) [S/m]Thickness [mm]Frequency27.5 GHz32 GHz37.5 GHz27.5 GHz32 GHz37.5 GHz27.5 GHz32 GHz37.5 GHzEpidermis87.578.51012~ 0.1~ 0.1~ 0.1Dermis2118.516.5283236~ 1.5~ 1.5~ 1.5Hypodermis (Fat)4.54.23.92.83.23.8~ 2.0~ 2.0~ 2.0Muscle454036354045Variable (> 5 mm)Variable (> 5 mm)Variable (> 5 mm) Dielectric and Electrical Properties of Human Skin Layers Table 6Performance comparison of the proposed antenna with literature.RefsYearSize (mm³)SubstrateThickness(mm)Bands (GHz)PortsGain (dBi)Isolation(dB)ECCCCL(bps/Hz)SAR (W/kg)1 g / 10 gElabd &Al-Gburi38202417.76 × 17.76×1.52RogersTMM41.5227.66–28.38 /37.70–38.2648.9> 68 / >90< 3 × 10⁻⁵< 0.031 g: 0.022510 g: 0.00576(28 GHz, 15 dBm)Haqueet al.39202537.8 × 37.8×0.80Rogers RTDuroid 58800.8027.2–32.2 /36.5–40.7410.14> 29< 0.007< 0.15NotperformedFranciset al.40202431 × 42 × 0.40Rogers RTDuroid 58800.4027.76–28.15 /32.02–32.46 /37.39–38.5945.88–8.66> 30< 0.0010.151 g: 0.507 (28G)1 g: 0.640 (32G)1 g: 0.538 (38G)[10 g: not reported]Alsaab &Shaban41202550 × 50 × 0.787RogersRT-58800.78728 / 35 / 3849.6 / 7.8 /13.734.5 / 22 / 300.000150.039 / 0.203 /0.024NotperformedTiwariet al.42202418 × 8.5 × 0.25Rogers 3003(Flexible)0.2528 / 3844.15 (28G) /7.73 (38G)> 200.03 (28G) /0.01 (38G)0.151 g/10 g:0.11/0.08 (28G)0.05/0.04 (38G)Sharmaet al.43202617 × 22 × 0.787(MIMO alone)RogersRT58800.78726.45–29.27 /37.04–39.1244.80/7.37(w/o FSS)9.96/11.48(with FSS)> 20< 0.18(with FSS)< 0.40 (w/o FSS)< 0.30 (with FSS)1 g: 0.0265 (28G)1 g: 0.0386 (38G)[50 mW, w/o FSS;10 g: not reported]Tiwariet al.44202535 × 35 × 0.254Rogers RT/Duroid 58800.25424–40(wideband)418.2> 44< 0.0005< 0.30NotperformedKaradet al.30202436 × 48 × 2Foam(εr = 1.07)2.01.98–13.57(UWB;3.1–10.6 GHz)1(Single)8.05N/A(Single-port)N/AN/A1 g: 0.016610 g: 0.0250@ 4.9 GHz,1 mm dist.Proposed202412.5 × 4.8 × 0.65Rogers RTDuroid 58800.6526.47–30.85 /31.29–32.86 /35.80–38.4829.3–9.7> 20< 0.0050.331 g/10 g:1.25/0.987 (27.5G)1.13/1.01 (32G)1.32/0.955 (37.5G) Performance comparison of the proposed antenna with literature. Elabd & Al-Gburi 38 17.76 × 17.76 ×1.52 Rogers TMM4 27.66–28.38 / 37.70–38.26 1 g: 0.0225 10 g: 0.00576 (28 GHz, 15 dBm) Haque et al.39 37.8 × 37.8 ×0.80 Rogers RT Duroid 5880 27.2–32.2 / 36.5–40.7 Not performed Francis et al.40 Rogers RT Duroid 5880 27.76–28.15 / 32.02–32.46 / 37.39–38.59 1 g: 0.507 (28G) 1 g: 0.640 (32G) 1 g: 0.538 (38G) [10 g: not reported] Alsaab & Shaban41 Rogers RT-5880 9.6 / 7.8 / 13.7 0.039 / 0.203 / 0.024 Not performed Tiwari et al.42 Rogers 3003 (Flexible) 4.15 (28G) / 7.73 (38G) 0.03 (28G) / 0.01 (38G) 1 g/10 g: 0.11/0.08 (28G) 0.05/0.04 (38G) Sharma et al.43 17 × 22 × 0.787 (MIMO alone) Rogers RT5880 26.45–29.27 / 37.04–39.12 4.80/7.37 (w/o FSS) 9.96/11.48 (with FSS) < 0.18 (with FSS) < 0.40 (w/o FSS) < 0.30 (with FSS) 1 g: 0.0265 (28G) 1 g: 0.0386 (38G) [50 mW, w/o FSS; 10 g: not reported] Tiwari et al.44 Rogers RT/ Duroid 5880 24–40 (wideband) Not performed Karad et al.30 Foam (εr = 1.07) 1.98–13.57 (UWB; 3.1–10.6 GHz) 1 (Single) N/A (Single-port) 1 g: 0.0166 10 g: 0.0250 @ 4.9 GHz, 1 mm dist. Rogers RT Duroid 5880 26.47–30.85 / 31.29–32.86 / 35.80–38.48 1 g/10 g: 1.25/0.987 (27.5G) 1.13/1.01 (32G) 1.32/0.955 (37.5G) The proposed antenna demonstrates clear superiority over the compared works across several critical dimensions. In terms of physical compactness, it achieves the smallest footprint of 12.5 × 4.8 × 0.65 mm³ among all the compared designs. The designs reported in38,40,43, and44 occupy volumes of approximately 480, 520, 293, and 310 mm³ respectively, while39 and41 are substantially larger at 1143 mm³ and 1967 mm³, making the proposed antenna several times more compact than any of these works. Although42 has a comparable footprint of 18 × 8.5 × 0.25 mm³, it only achieves dual-band operation and requires a four-port configuration, whereas the proposed design delivers tri-band coverage with just two ports. This brings to the second major advantage: architectural simplicity. Every competing mmWave MIMO work in the table38 — through44 — employs a four-port configuration, often relying on additional decoupling mechanisms such as EBG structures38, parasitic elements39,40, defected ground structures, or frequency-selective surfaces43 to manage mutual coupling. The proposed antenna achieves comparable MIMO diversity performance using only two ports without any such auxiliary decoupling structure, significantly reducing circuit complexity and fabrication cost. The third distinguishing feature is tri-band operation within a two-port architecture. Among the compared works, only40 (Francis et al.) and41 (Alsaab & Shaban) also cover three millimeter-wave bands, but both demand four ports and occupy substrates 13 and 50 times larger in volume, respectively. No other two-port design in the comparison achieves tri-band mmWave coverage, making the proposed design uniquely positioned in this respect. In terms of gain, the proposed antenna delivers 9.3–9.7 dBi consistently across all three bands, which substantially outperforms42 (4.15/7.73 dBi) and43 without FSS (4.80/7.37 dBi), and remains competitive with38 (8.9 dBi) and41 (9.6/7.8/13.7 dBi). While44 reports a higher gain of 18.2 dBi, this is achieved on a 35 × 35 × 0.254 mm³ substrate using a 1 × 2 array configuration — a fundamentally different and larger structure that cannot be fairly compared on size grounds. Similarly39, achieves 10.14 dBi but on a substrate nearly 30 times the volume of the proposed design. Regarding MIMO diversity metrics, the proposed antenna’s ECC below 0.005 is better than39 (< 0.007)42, (0.03/0.01 per band), and43 (< 0.18 with FSS), and is fully comparable to40 (< 0.001) and41 (0.00015) — both of which require four-port designs. The CCL of 0.33 bps/Hz remains within the acceptable threshold of 0.4 bps/Hz across all operating bands. Finally, electromagnetic safety validation through SAR analysis is performed on a realistic human hand phantom model, yielding values of 1.25/0.987, 1.13/1.01, and 1.32/0.955 W/kg (1 g/10 g) at 27.5, 32, and 37.5 GHz respectively — all well below the FCC/IEEE limit of 1.6 W/kg. Critically, three of the six mmWave MIMO references —39,41, and44 — entirely omit SAR analysis despite being intended for portable 5G devices. Among those that do report SAR, the proposed antenna uniquely combines SAR compliance with tri-band coverage and a two-port compact design, a combination that none of the compared works achieves simultaneously. Conclusion In conclusion, a novel two-port MIMO antenna has been designed and validated for millimeter-wave 5G applications, offering a unique balance between compact size, high gain, multiband operation, and robust MIMO performance. The antenna employs a meandered and stub-loaded geometry fabricated on a low-loss Rogers RT Duroid 5030 substrate, leading to efficient radiation and miniaturization. The structure supports tri-band operation over the frequency ranges of 26.47–30.85 GHz, 31.29–32.86 GHz, and 35.8–38.48 GHz—covering essential mm-Wave bands relevant for future 5G networks. The design ensures excellent reflection characteristics (S11, S22 < − 10 dB) and high isolation (S12, S21 < − 15 dB), along with a low ECC (< 0.01), which translates into minimal signal correlation and enhanced diversity. The antenna achieves peak gain values approaching 9.7 dBi and maintains stable gain across all three bands, as confirmed through simulation and measurement. MIMO performance metrics such as diversity gain (~ 10 dB), mean effective gain (balanced), and low channel capacity loss (< 0.4 bps/Hz) further validate the effectiveness of the design. A notable advantage of the proposed antenna is its ultra-compact form factor, which is significantly smaller than comparable designs in the literature. SAR analysis also confirms compliance with international safety standards when deployed in close proximity to the human body, ensuring suitability for wearable and handheld device integration. The human hand phantom-based SAR analysis, conducted across all three operational bands, directly validates the antenna’s electromagnetic safety for its targeted application in hand-held and wrist-worn 5G mmWave devices. Thus, the proposed antenna is not only a high-performing candidate for 5G systems but also a safe and scalable solution for next-generation wireless applications in dense and miniaturized environments. Author contributions S. Murugeswari - conceptualized the research idea, carried out the antenna design, performed simulations, and prepared the initial manuscript draft.R. Dhananjeyan - contributed to the theoretical analysis, optimization of antenna parameters, and validation of simulation results.Suganthy M. - performed fabrication of the prototype and conducted experimental measurements including S-parameters and gain characterization.S. Selvakumar - carried out MIMO performance analysis, including ECC, diversity gain (DG), mean effective gain (MEG), and channel capacity loss (CCL) evaluations.P. Sundaravadivel - conducted the SAR analysis using the human hand phantom model and validated electromagnetic safety compliance.D. Rajesh Kumar - supervised the overall research work, contributed to result interpretation, critically revised the manuscript for important intellectual content, and approved the final version for submission. Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.