Metamaterials are artificially synthesized structures engineered to obtain effective permittivity and permeability that do not exist in nature. Compared to natural materials, metamaterials possess ground-breaking properties for electromagnetic (EM) [1–3], acoustic [4–6], and thermal [7–9] waves, which give rise to a rich variety of applications such as broadband absorbers [10], negative reflection [11], and thermal emitters [12]. As the two-dimensional (2D) version of metamaterials, metasurfaces have advantages in power consumption, weight, volume, and processing difficulty [13,14]. In addition, metasurfaces are able to tailor EM waves [15] dynamically by designing material structures elaborately. This distinctive property gives metasurfaces considerable potential applications in tuning EM waves, including absorption [16,17], wavefront manipulation [18,19], and polarization conversion [20]. Different from analog metasurfaces [21–24], described by effective medium parameters, it is now possible for digital metasurfaces to modulate EM waves by digital numbers [25–28]. By digitizing specific physical parameters, a bridge between the physical world and digital world has been built. Nevertheless, polarization regulation is rarely involved in digital coding metasurfaces.

Since polarization carries valuable information in signal transmission, it occupies a peculiar position among numerous characteristics of the EM waves. Manipulating polarization is extremely significant for practical applications ranging from communicating [29] to imaging [30]. To control the polarization of EM waves, various kinds of passive [31–33] and active [34–36] reconfigurable polarization converters have been proposed in recent literature. For passive converters, an asymmetric waveguide is adopted to realize polarization transformation. Another effective method uses a geometric phase in a metasurface structure to achieve spatially varying polarization conversion. In active polarization converters, PIN diodes are commonly introduced to accomplish conversion among different polarization states, including linear-to-linear [37] and linear-to-circular [38]. However, most of the aforementioned designs have a drawback in flexibility, for which they can only achieve one or two polarization conversion effects.

In this paper, we propose a programmable metasurface based on single-pole double-throw (SPDT) switches. Different from the aforementioned works, this design employs switches based on silicon-on-insulator (SOI) technology, which offers high performance and ultralow cost. Two SPDT switches in a symmetrical structure form a double-pole double-throw (DPDT) switch to realize flexible selections on the incident and emitted waves. More importantly, since each SPDT switch can be controlled independently by a field programmable gate array (FPGA), the proposed metasurface is able to agilely encode binary polarization information on incident EM waves by a single programmable metasurface. We show that an ASCII code sequence of the characters contained in a 9-bit imaging frame of a near electric field can be completely transmitted. With low cost and easy fabrication, we believe our design extends the study of polarization and shows great potential applications in field information storage, processing, communications, and imaging.

We clearly see that each kind of coding element enables a channel of metasurface for the transmitted wave, such as the x- and y-polarized transmissions or polarization conversions. Once two or more kinds of coding elements are programmed, the relevant channels are activated, as shown in Figs. 4(a)–4(d). The related simulation results are provided in Figs. 4(e)–4(h), in which the transmission states of four channels are clearly observed, showing good consistency with the element performance. The simulation configuration is the same as that illustrated in Fig. 3(a), in which the two horn antennas are placed 500 mm away from the metasurface. In patterns E, F, and G, two kinds of coding elements with the same amount are grouped in the patterns, and the related results are presented, respectively, in Figs. 4(e)–4(g), in which the maximum transmission coefficient (S21) is about −6dB at 2.68 GHz. The other channels in these patterns are closed. Hence, the magnitudes of S21 in these channels are much lower than in the single transmitting channels. In pattern H, all four channels are enabled, and the maximum transmission coefficient S21 further reduces by about 3 dB compared to patterns E–G, because only a quarter of the elements are programmed for one channel. It should be noted that in the simulation of each channel, the polarizations of the transmitting and receiving horn antennas are set for the related polarization of the channel. We should also notice that in patterns A, B, and D, we intend to present the spatial energy conversion between two orthogonal polarizations, which cannot be used for information transmissions. On the contrary, in pattern C, two independent channels are established for low-interference transmissions.

Due to the symmetrical property, the X–X or X–Y polarization transmission can be easily induced by rotating the whole system 90 deg. Therefore, we select four representative patterns for measurements. Due to the slight frequency offset in the measurement results, we extend the frequency range from 2.5 to 3.5 GHz in the data plot. In Figs. 5(c) and 5(d), it can be clearly observed that the gains of the co-polarized transmitted waves (X–X) for pattern A and cross-polarized waves (X–Y) for pattern B are about −5.5dB at 3 GHz, when the metasurface is excited by the x-polarized EM waves. Meanwhile, the other two channels Y–Y and Y–X have good isolations, which remain at about −20dB. The transmission loss mainly results from the loss resistor of the SPDT switch and the structure loss. In addition, the combined patterns E and H are measured and presented in Figs. 5(e) and 5(f), where the activated elements are marked in specific colors. As depicted in Fig. 5(e), the blue curve represents the result of pattern E, in which only the elements 00 (in red) are activated, while the red curve indicates the transmission coefficient (S21) of the metasurface with only half elements 01 (in orange) being activated. Comparing Figs. 5(c) and 5(d), the transmission coefficients of the metasurface in these two cases are slightly lower by about 3 dB. This is because only half of the elements of the metasurface are activated for one channel (X–X), while the other half are activated for the other channel (X–Y). Likewise, in Fig. 5(f), the measured power (in blue) of pattern F for only a quarter of the activated elements 01 is lower than the case when half of the elements 01 are excited. Compared to the simulated results, the power levels of the measured patterns have good agreement, and the slight frequency deviation is due to the error in PCB fabrication and circuit modeling. Besides, when the SPDT switches in each element are in the off states, the transmission coefficients S21 of the four coding patterns are all below −20dB, indicating that the metasurface can achieve prefect reflections.

Take the letter H as an example, whose binary ASCII code is 01001000. Under the control of the FPGA, the state of the second and fourth supercells is set as X-to-X polarization (encoded as 1), and the state of the remaining six supercells is set as X-to-Y polarization (encoded as 0). In Fig. 1(b), when the x-polarized wave is incident, we observe that the energy of the x-polarized transmitted wave is mainly concentrated on the second and fourth supercells, whose corresponding information code is 01001000, indicating the letter H. Thus, the information can be transmitted through the energy levels of the supercells. Under the same setting, the energy distribution of the y-polarized transmitted wave is just opposite to the x-polarized, as shown in Fig. 6(c). In the same way, we can characterize other letters by setting specific supercell states. Figures 6(d) and 6(e) represent the letter I (the two graphs have opposite energy distributions for the same reason). As shown in Figs. 6(d) and 6(e), the energy graphs of both metasurfaces represent the letter I. Consequently, we can recognize the letters represented by Figs. 6(f)–6(l) as S(01010011), U(01010101), E(01000101), P(01010000), and S, E, U, respectively. We remark that the horizontal dimension of the near-field result is slightly compressed; hence, the white boxes indicating 0 and 1 are not uniformly spaced. The slightly inhomogeneous energy distribution and position deviation are mainly due to two reasons: 1) a nonideal wavefront is generated from the horn antenna; and 2) cross coupling exists between adjacent elements.

We demonstrated a novel programmable metasurface integrated with the SPDT switches on the top and bottom metallic layers. By controlling the SPDT switches in each element, we can convert the transmitted EM waves between the linear cross- and co-polarizations. Each element has four polarization states encoded as 00, 01, 10, and 11. We have programmed different element states into various sequences to further manipulate the polarizations of the transmitted EM waves. Besides the switchable polarization conversions, a perfect reflection is achieved when all switches on the metasurface work in the off state. It should be noted that the presented metasurface only supports plane waves at normal incidence. For oblique incidence, the structure and period dimension must be redesigned for higher diffraction orders. More importantly, we can further modulate the information on near-field distributions using the single programmable metasurface. Six characters in binary ASCII code sequences are designed, simulated, and measured. Both simulated and measured results proved good performance and high flexibility, indicating that the proposed design has a great effect on polarization conversions and dynamic field information editing and transmission. We envision that this work will enrich the programmable dimension of information metasurfaces [39,40] and facilitate the next generation of wireless communications and imaging.