Microstrip antennas are widely studied and employed owing to their compact size, low profile, and easy fabrication [1, 2]. However, they suffer from narrow bandwidths [3, 4]. To improve impedance bandwidth, methods such as parallel feeding [5], loading U-shaped slots in the patch [6], and stacking patches [7] have been proposed. Nonetheless, these methods either slightly enhance the bandwidth or significantly increase the profile. To improve antenna performance, scholars have recently proposed using metasurface antennas, which offer the advantages of microstrip antennas while also enabling wide impedance bandwidths and low profiles [8–13]. For instance, the L-probe-fed metasurface antenna reported in [14] achieved a 34.5% impedance bandwidth by exciting the transverse electric and magnetic wave modes of the metasurface maintaining a low profile of only 0.06 λ₀.

Characteristic mode analysis (CMA) is often used to guide metasurface antenna design [15, 16]. For example, the metasurface antennas featuring diamond-shaped patches in [9] and slotted metal square rings in [17] both achieved impedance bandwidths exceeding 31% by integrating slot and metasurface radiation modes using CMA. However, the aperture-coupled feeding in these antennas resulted in strong backward radiation. Meanwhile, multilayer [15] and dipole-based [18] metasurface antennas exhibit front-to-back ratios greater than 20 dB but are limited by their narrow bandwidths. These observations imply that simultaneously achieving a wide bandwidth and a high front-to-back ratio remains a challenging task.

In this paper, a low-profile broadband metasurface antenna is proposed. CMA is conducted to identify two radiation modes that can be excited by the metasurface and the horizontal patch of an L-shaped probe. Adjusting the partial patch sizes and implementing segmentation enhanced the bandwidth and suppressed the sidelobes. The antenna achieved a bandwidth of 40.8% (4.94–7.47 GHz) along with a high front-to-back ratio and stable gain. Moreover, the proposed antenna not only has a simple structure and a low profile but is also low in cost and easy to manufacture, making it suitable for mobile communication system applications.

Fig. 1 presents the designed metasurface antenna, with dimensions of 55 mm × 55 mm × 3.5 mm. The antenna employs two F4B dielectric substrates, where the upper substrate thicknesses h₁ = 2 mm and the lower one h₂ = 1.5 mm. Both substrates have a dielectric constant of 2.65 and a loss tangent of 0.01. Meanwhile, Fig. 1(b) shows that the L-shaped probe structure is composed of a metal probe and a horizontal patch. This structure is placed on the upper surface of the lower substrate, whose lower surface is a metal layer. Furthermore, the metasurface structure is located on the upper surface of the upper substrate.

To analyze the antenna’s radiation mechanism, CMA was conducted. Characteristic modes refer to the surface currents or radiation fields determined by the metal’s shape. Each characteristic mode is independent of all other modes and is not influenced by external excitation sources. According to characteristic mode theory [11, 19], modal significance (MS) is expressed as follows:

Here, λ_n is the eigenvalue of the nth mode. Generally, a mode is considered resonant when MS > 0.707. In this study, CMA simulation was conducted using CST software, with the boundary for the substrate’s lower surface set to infinitely large ground plane.

Fig. 2 illustrates the evolution of the proposed metasurface, with Antenna III as the final design. As shown in Fig. 3(a), Metasurface I comprises a traditional 3×3 array of square patches of length W_p and two rectangular parasitic patches on its left and right. Notably, the unit length was designed to one-sixth of the central frequency’s wavelength, enabling the periodic composite structure to function as a subwavelength element of the metasurface. The two rectangular parasitic patches serve to increase the antenna’s radiation aperture, thereby enhancing antenna gain [9, 20].

Fig. 3 presents the MS curves of Metasurface I, as well as the modal currents and modal radiation patterns for two linearly polarized modes. Fig. 3(c) and 3(d) depict the current direction on the metasurface (denoted using thick black arrows) for Modes 2 and 7, exhibiting orthogonal current and modal radiation patterns similar to those for a traditional 3×3 metasurface [21]. The current directions on each patch are roughly consistent, resulting in a good main beam in the radiation pattern. In Mode 2, the modal current on the entire metasurface is consistent in the x-direction, with no reverse currents. In contrast, Mode 7 exhibits partial reverse currents on the parasitic patches. As a result, Mode 2 was chosen as the target excitation mode over Mode 7. The other modes lack a single-direction modal current and are not considered desirable. Furthermore, Fig. 3(c) shows that the rectangular parasitic patches extend the current path in the x-direction, thereby enhancing the x-direction current intensity and increasing the antenna gain.

Figs. 4 and 5 depict the impact of the L-shaped probe’s horizontal patch on the antenna’s radiation modes. The current direction on the metasurface and on the L-shaped probe’s horizontal patch are indicated by the thick black arrows and long red arrows, respectively. Notably, since the loading of the horizontal patch changed the mode order, the original Mode 2 was renamed Mode 8. The patch introduces a new x-direction linearly polarized radiation mode near 5 GHz, referred to as the new Mode 2.

In Fig. 5, modal currents are the strongest on the horizontal patch at 5 GHz for Mode 2 and at 7 GHz for Mode 8, both in the x-direction. This implies that by adding a probe to the horizontal patch to form an L-shaped probe, these two modes can be successfully excited [14]. Other modes with operating frequencies close to Modes 2 and 8 exhibited either weaker modal currents on the probe’s horizontal patch or low MS values, so they were not excitable and thus not discussed in this paper. Only the modes with strong single-direction current intensities on the L-shaped probe’s horizontal patch, matching the feeding current direction of the actual probe, were considered excitable.

Fig. 5(b) and 5(c) show that, within the potential bandwidth of MS > 0.707 for Mode 8, y-direction currents on the edge patches strengthen as the frequency increases, thereby weakening the currents on the horizontal patch. This hinders the excitation of Mode 8 at high frequencies. However, if only Mode 2 exists in the high-frequency band without Mode 8, the high-frequency bandwidth would decrease significantly. So, this modal current distribution trend needs to be suppressed, and the modal current on the horizontal patch at 8 GHz must be enhanced. Therefore, the structure of Metasurface I is further modified, thereby leading to the creation of Metasurface II.

Metasurface II, depicted in Fig. 6(a), differs from Metasurface I as the three central patches are extended and the six edge patches are shortened in the y-direction. Compared to Fig. 5(c), the modal current on the L-shaped probe’s horizontal patch for Mode 8 at 8 GHz in Fig. 6(e) exhibits a marked improvement, flowing mainly in the x-direction rather than toward the edge patches in the y-direction, thus potentially increasing the linear polarization bandwidth. However, at 8 GHz, the current on the edge and central patches flow oppositely, leading to two sidelobes. This indicates that Metasurface II needs further modification to suppress the sidelobes.

For simplicity, the x-direction current on Metasurface II is divided into four sections in Fig. 7 by its structural symmetry, resembling a four-element antenna array. Notably, the array factor is expressed as follows [13]:

where α and β denote the current amplitudes, while p₁ and p₂ refer to distances from the current elements to the metasurface center. As indicated in Eq. (2), sidelobes can be suppressed by reducing the reverse current amplitude β on the edge patches.

In Metasurface III, as shown in Fig. 8(a), Metasurface II’s six edge patches are uniformly divided into 24 smaller ones to suppress the reverse edge current, thereby reducing sidelobes and enhancing gain. However, this division shifted Mode 8’s resonant point (MS = 1) higher by 0.3 GHz. Considering this MS curve shift, CMA simulations were performed at three frequency points of Mode 8: 7 GHz, 8 GHz, and 8.5 GHz. As shown in Fig. 8, the x-direction modal currents on Metasurface III mainly concentrate on the three central patches, along with suppressed currents on the 24 edge patches. Furthermore, compared to Fig. 6(e), the sidelobes of Mode 8 at 8 GHz and 8.5 GHz in Fig. 8(e) and 8(f) are significantly reduced, confirming the effectiveness of edge patch division.

Since the L-shaped probe is a capacitive coupling feeding structure, it should be fed at the location of the maximum electric field intensity to maximally excite the desired modes. Fig. 9(a) shows the probe positioned at the point of maximum electric field intensity for Modes 2 and 8 to ensure effective excitation [11]. Fig. 9(b) indicates that the impedance bandwidth increases and then decreases as the feeding position d_f increases. Since the impedance bandwidth is widest at d_f = 1.9 mm, this value was chosen for broadband performance.

Fig. 10 compares the simulated |S₁₁| and gain of metasurface antennas using the three metasurfaces. It shows that Antenna III has two resonant points in the passband, corresponding to the resonant frequencies of linearly polarized modes. Furthermore, the impedance bandwidth of | S₁₁| < −10 dB ranges from 4.89 GHz to 7.58 GHz, indicating a 43.1% relative bandwidth, while the gain varies from 8.0 dBi to 10.5 dBi. Compared to Antenna I, Antennas II and III expanded the high-frequency impedance bandwidth by about 1 GHz. Specifically, Antenna III’s gain is enhanced by over 3 dBi at high frequencies compared to Antennas I and II. Furthermore, Fig. 11 confirms that Antenna III’s high-frequency sidelobes are notable suppressed compared to those of Antenna II, consistent with the CMA results.

Fig. 12 presents the fabricated antenna and its measuring environment, while Fig. 13 highlights the measured impedance bandwidth of | S₁₁| < −10 dB is 40.8% (4.94–7.47 GHz), with a gain of 7.6–9.6 dBi. Fig. 14 shows that the measured cross-polarization levels in the E-plane and H-plane are below −29 dB and −18 dB, respectively, with a front-to-back ratio of 18–33 dB in the bandwidth, confirming good directivity. Notably, all the measurements are consistent with the simulated results.

Table 1 presents a comparison of the proposed antenna with recently reported low-profile linearly polarized antennas. It is evident that the proposed design offers a wider bandwidth and higher peak gain than those in [4], [8], [11], and [12], as well as a higher front-to-back ratio than those in [4], [11], [12], and [13]. Overall, it promises excellent performance in terms of bandwidth, gain, and front-to-back ratio.

In this paper, a low-profile broadband metasurface antenna fed by an L-shaped probe was designed using CMA and then experimentally validated. The operating bandwidth was increased by adjusting the partial patch sizes, while segmenting the edge patches enhanced the gain and suppressed the sidelobes. The proposed antenna achieved a wide bandwidth of 40.8%, and is characterized by a low profile, high front-to-back ratio, and stable gain, making it suitable for application in 5G Wi-Fi and 6G Wi-Fi communications.