Due to the diverse demands of various application scenarios, extensive explorations have been conducted on the potent electromagnetic control potential of metasurfaces, which possess commanding advantages in profile, weight, integration, etc., over traditional 3D structures, leading to the design of numerous remarkable functionalities such as negative refraction [1–3], anomalous reflection [4–7], meta-holographic imaging [8–11], perfect meta-lenses [12–14], anti-reflective stealth [15–20], and many others [21–36]. Traditional metasurfaces are mostly operating in passive manners that manipulate electromagnetic waves (EWs) by elaborately designing specific geometric structures of the meta-atoms. Although passive metasurfaces are almost omnipotent in manipulating EWs at a lower cost, their functionalities are fixed once their prototypes are manufactured. In pursuit of greater freedom for EW modulations and achieving aperture multiplexing in practical applications, reconfigurable metasurfaces with active devices or media have received widespread attention.

For reconfigurable metasurfaces, essentially, multiple intriguing EM responses are achieved by manipulating the inherent physical parameters of active media (such as liquid crystal [37], graphene [38], or other phase-changing materials [39,40]), or by incorporating active devices [41–43] to introduce discrete components instead of relying solely on structural variations in passive metasurfaces. Especially for those who enable programming capability [44–52], powerful digital EW modulation is achieved by the external field programmable gate array (FPGA) working in real time to physically change the numerically coded sequence of meta-atoms. There is no doubt that the introduction of active devices greatly increases the freedom of EM modulation. However, the accompanying costs including complex wiring, high costs of control systems, and energy consumption still hinder the application of reconfigurable metasurfaces. Besides, although the functionality of reconfigurable metasurfaces also depends on the design of passive structures from a physical perspective, it does not necessarily mean that the passive structures can be easily utilized as a new degree of freedom, as the tight coupling between active and passive modules needs to be considered, which is explained in part of Fig. 1. And from the perspective of designing traditional reconfigurable metasurfaces, the design of passive structures is usually just to better assist the active module in playing its tunable role, rather than placing passive modules and active modules in an equal position. Nevertheless, several research works have made valuable attempts in both the optical [53,54] and microwave [55] bands. However, despite utilizing the freedom of both active and passive modules, these works have not addressed the entanglement caused by their coupling, which may lead to inefficiency or high design complexity in the system.

In order to fully release the freedom of controlling EWs by both active and passive modules, we have constructed a low-complexity prototype composed of weakly coupled passive and active modules to overcome the accompanying complexity increase in meta-atom design. When operating in passive mode, the local transmission phase of EW can be modulated efficiently by adjusting the size of specific metal patterns to realize specified wavefront engineering, while in active mode, the transmission and reflection states of the weakly coupled system can be effectively switched by changing the state of PIN diodes regardless of the non-uniform distribution effect of the passive modules. We have found several works that look similar to our research work on reconfigurable metasurfaces [42,43], which are also controlled by diodes to achieve switching of transmission, reflection, or absorption functions. Although these research works have inspired us, they are still essentially traditional reconfigurable metasurfaces, as the degrees of freedom of passive structures have not been utilized. Therefore, our work was proposed to emphasize the collaborative work of active and passive factors in reconfigurable metasurfaces by introducing weak coupling between the passive and active modules.

Next, we will analyze how to construct the expected weakly coupled reconfigurable metasurface with high efficiency, which can independently control both active and passive modules in the same system. To reduce the verification complexity of the weakly coupled system without losing generality, only a single-layer structure with one PIN diode was initially set up to design the active module in this work. Intuitively, based on the transmission responses of active and passive modules, there are four possible combinations for the two modules fusing in a meta-atom, as schematically shown in Fig. 1. First, if both modules were narrowband responses [depicted in Fig. 1(a)], the phase modulation of the passive module will be limited, because the phase shifts achieved by changing the dimensions of meta-structure are usually accompanied by frequency shifts in the operating band. Obviously, this will make it difficult for the two integrated modules to work together. In a most ideal situation, if both modules possessed broadband transmission capabilities as shown in Fig. 1(b), the integrated cooperation would be much easier. However, under our preliminary assumptions about the active module above, it is almost impossible for a single-layer structure to achieve broadband transmission response with high efficiency. Therefore, as depicted in Fig. 1(c), the assumption that the active module exhibits a narrowband response and the passive module possesses a broadband response would be most likely to be realized under these constraints. Nevertheless, variations in size within the meta-structures of the passive module often lead to frequency shifts in both modules due to coupling effects as shown in Fig. 1(e). Based on the above analysis, it becomes evident that realizing the desired independent control system poses significant challenges. Here, we put forward a hypothesis that if there were only a weak coupling relationship between the two modules [depicted in Fig. 1(d)] when the frequency drift caused by changing the sizes of meta-structures in the passive module occurs, the operating frequency of the active module could almost stay stable within the overlapping range formed by the drifts, as shown in Fig. 1(f); then the envisioned weakly coupled reconfigurable metasurface could be realized.

For the passive module, a high-efficiency phase modulation covering a complete 2π is expected. It is well known that altering the size of the meta-structure to achieve phase shift often results in frequency drift while achieving high-efficiency transmission in a transmission-type system necessitates meeting specific resonance conditions at designated frequencies. Therefore, it is imperative to have sufficient broadband response within the same system to simultaneously fulfill both requirements for realizing the expected passive module, which is also consistent with the above analysis. It should be noted that the mentioned broadband here does not only refer to the bandwidth that EWs can efficiently pass through for a metasurface but also to the premise of being able to freely modulate the transmission phase. One approach to expand bandwidth with high efficiency involves multimode coupling, which typically requires meta-atoms of adequate thicknesses. However, this method will lead to bulky structures for transmission-type metasurfaces and is seldom employed in microwave applications, but finds widespread use in dielectric metasurfaces operating in optical bands [56,57]. Another intelligent way is through polarization conversion [58,59] without relying on bulky thickness, which provides inspiration for us to design wideband transmission-type passive modules.

According to the above theoretical analysis, we choose the appropriate structures to realize our hypothesis. As depicted in Fig. 2(a), the active layer is constructed by a simple dipole-shaped metal patch with two bias lines, and a PIN diode (MADP000907-14020) is inserted in the middle gap. According to Ref. [52], under reverse bias, the PIN diode can be modeled as a series L-C circuit (L=30pH, C=28fF), while with forward bias (V=1.35V, I=10mA), the diode behaves as a series L-R circuit (R=7.8W, L=30pH). Below the active layer, an I-shaped slit, which is the common part of the two modules in our final design [shown in Fig. 3(a)], serves as a filtering barrier for the incident EWs leaked through the non-metal region of the active layer including the vertical polarization part. When the PIN diode is ON, the two metal patches are connected together, reflecting almost all incident EWs. When the diode is OFF, the active layer can be equivalent to a frequency selective surface (FSS), which only allows EWs with specific frequencies to pass through, and after our verification, the active module belongs to the narrowband category. The simulated diagram of the active function under the control of the PIN diode is given in Fig. 2(c). We can see that under two working states, the difference of the transmission coefficient at the center frequency exceeds 0.8, which means that the switching function of the active module performed well. For the passive phase modulation module, we choose a three-layer metal structure consisting of two layers of orthogonal strip gratings with an H-shaped ring sandwiched between them that has been proven to have broadband transmission performance, as shown in Fig. 2(b). The 45°-placed H-shaped metal ring will transfer the transverse electric (TE) polarized EW into transverse magnetic (TM) mode; additionally, the transmission phase of the meta-atom can be modulated flexibly by varying the opening angle θ of the H-shaped ring. With the change of the opening angle θ, the simulation results of the transmission coefficient are given in Fig. 2(d). As we expected, it has broadband properties, and there is always an overlapping frequency range that maintains good transmission as the phase modulation size changes. In summary, we get the decomposition model of the two modules with the expected transmission responses shown in Fig. 1(f).

Considering the coupling between the two modules originating from the overlapping of the local fields, we can achieve it by reasonably designing the thickness of the substrate between them. The uniform arrays of meta-atoms can be regarded as a grating with subwavelength periods, and the scattering process of the incident EW still satisfies the conservation of momentum, which can be expressed by the following formula [22]: kiy±mG=kty,(1)where kiy and kty are the y-direction components of propagation vectors of the incident wave and transmitted wave, respectively. m is the order of the scattering field. G=2πP (P is the period of the unit). Due to the subwavelength size of P, the zeroth-order scattering field only exists in the free space, according to Eq. (1), when the EW is normally incident on a uniform metasurface, indicating that all of the high-order modes are evanescent waves localized in the meta-atoms. The evanescent fields can be expressed as Em=E0me−σ·z·e−ikty·y,(2)where E0m is the amplitude of the mth-order evanescent field, and σ=kty2−(nk0)2 represents the decay factor, where n=εr is the refractive index of the environment, εr is the permittivity of the environment, and k0 is the wave number in free space. kty=mG is the propagation constant of the mth-order evanescent wave. The decay length lem is defined as the distance required for the amplitude of the mth-order evanescent field to decay to 1/e perpendicular to the propagation direction, which can be calculated as lem=1σ=1(mG)2−εrk02.(3)

According to Eq. (3), the decay length decreases as the order m of the evanescent field increases, indicating a weaker coupling between the mode and the external environment. Therefore, we just need to calculate the decay length of the first-order mode (m=1) to represent the effective distance of the module on the external environment. In this case, with the meta-atom size P of 6 mm, the environmental permittivity εr of 3, and the working frequency set at 12 GHz, substituting these parameters into Eq. (3) we can calculate le1=1.05mm. The numerical result shows that when the distance between the effective modulation layers of the two modules is more than 2le1, their coupling effect will be sufficiently weak. Reasonable proof is given by the detailed analysis in Fig. 3(b), which proves that the two modules have a good isolation effect when integrated into the same unit.

As the implementation of our concept, here we introduce the idiographic design of the weakly coupled meta-atom including the active module and the passive module. The schematic model is depicted in Fig. 3(a), which has a multilayered structure consisting of four metal layers and three substrate layers (εr=3.0, tanδ=0.0015). The dipole-shaped active layer is designed with several slots for easy installation of the PIN diode and better impedance matching. In the passive module, the first layer of a strip grating [as shown in Fig. 2(b)] is replaced by the I-shaped slit, since both the I slit and the grating are capable of filtering out quadrature-polarized EWs. Additionally, as explained in the previous section, incorporating the I-shaped slit can enhance the switching function of the active layer by giving a re-reflection of the incoming EWs. It is worth mentioning here that the thickness of the substrate between the active layer [numbered I in Fig. 3(a)] and passive layer [numbered III in Fig. 3(a)] is designed as 3 mm (h1+h2) in the meta-atom, which exceeds the safety distance 2le1 required for the weak coupling we calculated above.

To demonstrate the superior performance of the weakly coupled atom, the proposed meta-atom is studied by using a finite-difference time-domain (FDTD) simulator with periodic boundary conditions. Figure 3(b) shows the switching function under the active mode, as we can see that the weakly coupled atom exhibits low-loss transmission at the OFF state while showing a near-zero transmittance at the ON state. Importantly, we can also see that the effective switching function can still maintain stability with the variation of the opening angle θ of the H-shaped ring, which also demonstrates the weak coupling effect between the active module and the passive module in the same system. Furthermore, the bandwidth performance of the weakly coupled reconfigurable meta-atom is also characterized. Figures 3(c) and 3(e) depict the simulated polarization conversion ratio (PCR) (described by |t|2) under two working states of the PIN diode. It can be observed that in addition to the obvious contrast of the transmission channel, in the OFF mode, the meta-atom exhibits a relatively wide transmission bandwidth of 10.5–13.5 GHz with a good transmittance (|t|^2≥80%). Finally, the corresponding transmission phase Φ [described by arg(t)] under passive mode is shown in Fig. 3(d). As observed, the transmission phase can cover 180° with the change of the opening angle θ within the operating band. Moreover, given the symmetry of the structure, when the “H” is rotated 90° in the xoy plane, we can obtain an additional modulation of more than 180°. In summary, we can obtain a total of 360° of phase-modulation freedom, which means that we can achieve arbitrary desired wavefronts in passive mode.

As a proof of our proposal, based on the weakly coupled reconfigurable meta-atom in Fig. 3, several different functions are implemented, including active function switching and passive wavefront manipulation, as illustrated in Fig. 4(a). First, a simple wavefront manipulation of a 45° anomalous deflection is demonstrated by FDTD simulation. Figure 4(b) depicts the angular distribution intensity of the scattered electric field at three represented frequencies, 11 GHz, 12 GHz, and 13 GHz, under the passive mode of the weakly coupled screen. As we can see, the simulation results are basically consistent with our presuppositions; the deviation angle of the main beam is in good agreement with the theoretical prediction at the typical working frequency, which verifies the wavefront control function under passive mode.

To further display the passive function of the weakly coupled reconfigurable metasurface, more complex wavefront modulation has been demonstrated by calculation and experiment. Similarly, based on the phase data in Fig. 3(d), we have designed a reconfigurable array with a focusing effect on the incident EWs under passive mode. Figure 5(a) shows the fabricated prototype of the designed meta-array with 35×35 reconfigurable elements. Both the active and passive functions of the array are verified by FDTD simulation and microwave anechoic chamber experiments. Figures 4(c) and 4(d) depict the normalized 3D simulated scattering far-field distribution represented by |Ex|2 within xoz and yoz planes at 12 GHz. When diodes are OFF, the weakly coupled reconfigurable array works in the passive mode, and an obvious focal spot can be observed in Fig. 4(c). In addition, we can see that the position of the focus is about 145 mm away from the array front, which is basically consistent with our design. When the diodes work in the ON state, the far-field scattering diagram of the electric field is shown in Fig. 4(d). Clearly, there is no obvious focusing effect, and only a weak electric field can be observed as most of the incoming EWs could not pass through the meta-atoms. The obvious contrast between the two operating states in Figs. 4(c) and 4(d) gives a good interpretation of the function-switching capability of the proposed weakly coupled reconfigurable metasurface under active mode.

Furthermore, the weakly coupled reconfigurable metasurface is verified by experiments. The experimental setup for the scattering far-field distribution measurement of the reconfigurable array is shown in Fig. 5(b). A horn antenna is installed on the robot arm B, and the distance between the horn center and the sample is controlled by a computer to meet the far-field measurement conditions of generating quasi-plane waves on the sample surface. On the other side of the sample, we adopt a probe antenna on the robot arm A to detect the far-field distribution. It is worth mentioning here that we use a three-point calibration method to unify the geometric center of the sample and the antennas, which provides us with a high degree of accuracy in our measurements. By controlling the motion trajectory of the robot arm A, an area of 115mm×115mm with the center located at 50–300 mm above the weakly coupled reconfigurable array is scanned (here we choose the yoz plane for measurement in consideration of the test safety). In addition, it should be noted that the proposed metasurface has a polarization conversion function for incident EWs, so the polarization direction of the transmitting horn antenna should be perpendicular to the receiving probe. The transmitting antenna and the receiving probe were connected to the two ports of a vector network analyzer, model Keysight N5225B. Here, we use a DC power supply (Keysight E36300) to control the working state of the weakly coupled reconfigurable metasurface.

Figure 5(c) shows the FDTD simulated distribution of the electric field intensity under the vertical polarization of the yoz plane within the frequency band of 11–13 GHz under OFF state, which is in good agreement with the corresponding measured 2D scattering Ex field pattern depicted in Fig. 5(e). Similarly, the good matching between the simulation and test under the ON state of the weakly coupled reconfigurable array can also be observed in Figs. 5(d) and 5(f). Figure 5(g) depicts the field intensity distribution at the focal point under different frequencies. Significant amplitude differences can be observed when diodes operate in different states. Specifically, the electric field intensity of diodes being ON is 11.8 dB higher than that of the diodes being OFF at 12 GHz under measurement, which is 3.6 dB lower than the simulation. However, the differences in focusing performance between test and simulation are reasonable, which may be caused by machining inaccuracies or the experimental environment. From the above simulation and measured results, it can be proved that the proposed weakly coupled reconfigurable array has an effective passive wavefront control function and active mode switching capabilities.

To summarize, we have presented a transmissive reconfigurable metasurface, which can fully unleash the freedom of controlling electromagnetic waves by both active and passive modules through weak coupling. The constructive weakly coupled reconfigurable metasurface can realize independent control of active function switching and passive wavefront manipulation within the same aperture. To demonstrate our proposal, a meta-array consisting of 35×35 atoms is designed and verified by accurate experimental testing. The array used to implement the scheme enables switchable anomalous reflection and focusing functions in the microwave regime. The good matching between simulation and testing under ON and OFF states indicates excellent performance of the weakly coupled reconfigurable metasurface. Our exploration lies in the utilization of the physical structure, which is always neglected in reconfigurable metasurface design, emphasizing the collaborative work of active and passive modules in metasurfaces, which provides a more flexible design approach for reconfigurable metasurfaces. Furthermore, by using active elements and structural variations as two degrees of freedom, our method is not only applicable for simple ON/OFF active functions but can also be extended to the cases where each unit is independently controlled for more complicated EM manipulation. In this study, the switching function of the active module is merely one option of active functions. Given the cost and complexity, the proposed weakly coupled reconfigurable metasurface is only a preliminary implementation of the long-neglected degree of freedom in developing reconfigurable metasurfaces, and for per-atom independent control, which is ongoing research and also an extension of this study.