To realize the above concept, we first tactfully design an ASTCM by integrating SCCs, whose meta-atom structure is shown in Fig. 2(a). The meta-atom is a dual-layer structure containing the phase modulation layer and the energy harvesting layer. As depicted in Fig. 2(b), the phase modulation layer is constructed by etching a rectangular patch and two bias lines on the F4BM220 substrate (dielectric constant ε_r = 2.2, loss tangent tanσ = 0.001, thickness h₁ = 1.5 mm), and two low-power single-pole single-throw (SPST) switches are loaded at two gaps between the rectangular patch and bias lines. A metallized via is introduced to connect the rectangular patch and the metallized ground on the back of the substrate, which not only provides bias ground for SPST switches, but more importantly, it breaks the geometric symmetry of the phase modulation layer and therefore ensures that the meta-atom achieves 2-bit reflection phase modulation by tuning the two SPST switches. As shown in Fig. 2(c), the substrate of the energy harvesting layer is F4BM265 (dielectric constant ε_r = 2.65, loss tangent tanσ = 0.0015, thickness h₂ = 1.5 mm). A rectangular patch with a direct current (DC) line is etched on the top of the substrate, and another DC line is etched on the back. To enable ambient solar energy harvesting via a shared aperture, a triple-junction gallium arsenide (GaAs) SCC is directly integrated in the energy harvesting layer on each meta-atom, and its cross-sectional structure is shown in Fig. 2(d). This triple-junction GaAs SCC features a multi-layer stacked structure that harvest different spectra of sunlight through different layers, thereby achieving high-efficiency wide-spectrum photoelectric conversion. More details of the triple-junction GaAs SCC are provided in Supplementary information of Section 1. The anode of the triple-junction GaAs SCC is attached to the top rectangular patch, while its cathode is connected to the bottom DC line through a SCC-specific aluminum welding stripe and a metallized via. To promise simultaneous reflection microwave modulation and ambient solar energy harvesting, the energy harvesting layer is placed in front of the phase modulation layer with a distance of h_air = 2 mm. When the energy harvesting layer is illuminated by sunlight, the integrated SCCs directly convert the sunlight into DC energy and transmit it through DC lines to the back-end SPMC for storage. In addition, the integrated SCCs and metal patches together act as parasitic patches of the phase modulation layer, ensuring that the incident microwave is effectively coupled to the phase modulation layer and vice versa, also ensuring that the reflection microwave can be effectively coupled into the free space. It is noteworthy that the influence of the energy harvesting layer is already well considered in the whole meta-atom design, ensuring that desired reflection microwave modulation performance can be maintained after integrating SCC (see Supplementary information of Section 2 for more details).

To evaluate the reflection microwave modulation performance of the proposed meta-atom, full-wave simulations are conducted with commercial software CST Microwave Studio. As depicted in Fig. 2(e, f), by independently controlling the ON/OFF states of the two SPST switches, 2-bit phase modulation can be achieved at 9 GHz, and the reflection amplitudes corresponding to the four phase coding states are all greater than 0.8. Furthermore, we adopt the time-coding method to further improve the microwave modulation capability. When the metasurface is illuminated by a monochromatic microwave incidence with the frequency of f_c and controlled by a periodic time-coding sequence (with the time period of T₀ = t₀ × L, where t₀ is the time slot, L is the sequence length), it will generate a series of reflection microwave harmonics with frequencies of f_c + mf₀ (m = 0, ±1, ±2, ±3, ... , f₀ = 1 / T₀). Specifically, the equivalent reflection amplitude a^m and phase φ^m of the m^th harmonic can be expressed aswhere a_l and φ_l are the basic reflection amplitude and phase in the l^th time slot, respectively. It can be observed from Eq. (1) that the equivalent reflection coefficient of the m^th harmonic is the superposition of the basic reflection coefficients in a time period. Therefore, here we develop a time-coding strategy based on the particle swarm optimization algorithm, aiming to achieve efficient generation and full-phase synthesis of any desired reflection microwave harmonic by optimizing the time-coding sequence. Without loss of generality, here we assume that the metasurface generates p (p = 1, 2, ... , P) distinct reflection microwave harmonics and independently modulate their phases, and the fitness function can be expressed aswhere m_p is the order of the p^th harmonic, $ a_{\mathrm{d}}^{{m}_{p}} $ are $ \varphi _{\mathrm{d}}^{{m}_{p}} $ are the desired amplitude and phase of the $ m_{p}^{\mathrm{th}} $ harmonic, $ a_{\mathrm{c}}^{{m}_{p}} $ and $ \varphi _{\mathrm{c}}^{{m}_{p}} $ are the current amplitude and phase of the $ m_{p}^{\mathrm{th}} $ harmonic, w₁ and w₂ are weight factors (see Supplementary information of Section 3 for more details on the optimization process of the time-coding sequence). As a verification, two groups of 16 different time-coding sequences are first optimized to achieve equivalent 4-bit phase modulation at the fundamental frequency f_c (i.e., m = 0) and the −1^st harmonic f_c–f₀, respectively. As shown in Fig. 2(g, h), all synthesized equivalent phases match well with the desired phase values (with the phase deviation of less than 4.1°), and the equivalent amplitudes of the fundamental and −1^st harmonic frequencies are greater than 0.7 and 0.83, respectively. These results well demonstrate the ability to efficiently generate any desired reflection microwave harmonic as well as accurately modulate its phase. In addition, independent and simultaneous equivalent 4-bit phase modulation at the −1^st and +2^nd harmonics is further realized. As displayed in Fig. 2(i), by applying 16 different optimized time-coding sequences, equivalent 4-bit phases can be synthesized at both −1^st and +2^nd harmonics (with the deviation from the desired phase values of less than 10°). The phase gradients at the two harmonics show increasing and decreasing trends respectively, indicating that simultaneous and independent phase modulation at multiple frequencies can be achieved. In addition, the equivalent amplitudes of the two harmonics remain at around 0.5, proving the capability of efficient generation of multiple harmonics.

As shown in Fig. 3(a), a metasurface array prototype consisting of 16 × 8 meta-atoms is fabricated using the standard printed circuit board technology. To facilitate the integration of SCCs, four metasurface subarrays consisting of 4×8 meta-atoms are first fabricated. After integrating SCCs on each subarray, the four subarrays are then spliced together to form a complete metasurface array. Photos of the phase modulation layer and the energy harvesting laye of each meta-atom are illustrated in Fig. 3(b, c). Then, a SPMC and a microcontroller-based low-power control circuits (MLCC) are further designed and integrated with the metasurface array to construct the final ASTCM, as shown in Fig. 3(d). Notably, meta-atoms in the same column (along the y-direction) of the phase modulation layer share two independent bias voltages to achieve independent control of the two columns of SPST switches. Therefore, the metasurface array enables one-dimensional phase modulation (along the x-direction) in the space domain. In addition, by designing the DC line routing of the energy harvesting layer, the SCCs on each subarray are connected in parallel and the SCCs between different subarrays are connected in series. This configuration ensures that the photovoltage output by the metasurface array matches the input voltage range of the SPMC, thereby allowing the efficient storage of the harvested solar energy. To improve the solar energy harvesting efficiency, the SPMC is designed based on the maximum power point tracking (MPPT) technology, which helps the ASTCM to dynamically adapt to the fluctuations of the ambient solar irradiance and therefore ensuring the real-time maximization of the harversted energy. Meantime, low-power design strategies are adopted to improve the energy self-sufficiency peformance, including the usage of low-power SPST switches (Archiwave, ARW3172) in the metasurface array design as well as low-power MCU (STMicroelectronics, STM32H743) in the MLCC design. To clearly demonstrate the working mechanism of the ASTCM for solar energy harvesting and self-powing, a simplified equvient circuit schematic is shown in Fig. 3(e). It can be observed that the solar energy harvested by the energy harvesting layer is first stored in the SPMC and then output as a stable driving voltage to drive the MLCC. Finally, the MLCC outputs multiple independent voltage signals to tune the SPST switches loaded on the phase modulation layer.

To further evaluate the realistic solar energy harvesting performance of the ASTCM, we utilize a self-made solar simulator to provide artificial sunlight with controllable irradiance (see Materials and methods for more details). Figure 3(f) shows the measured current-voltage (I-V) and power-voltage (P-V) responses of the ASTCM under different sunlight irradiances. It can be observed that as the sunlight irradiance increases, the solar energy harvested by the ASTCM gradually increases. Under the nominal sunlight irradiance of 100 mW/cm², the solar energy harvested per unit area of the ASTCM can reach 12 mW/cm². Moreover, we measured the power consumption of the ASTCM to evaluate its energy self-sufficiency performance. When all SPST switches are in the ON state, the metasurface array reaches the maximum power consumption, which is measured to be 12.8 mW. We also measured the dynamic power consumption of the MLCC when its 32 channels simultaneously output 1 MHz square wave signals, which is about 177 mW. Hence, the maximum power consumption per unit area of the entire ASTCM can be calculated as 0.66 mW/cm², which is much lower than the solar energy it can harvest. Through straightforward estimation, it can be obtained that the ASTCM can harvest enough ambient solar energy to ensure it uninterrupted operation for 24 hours under the nominal sunlight irradiation for only about 1.3 hours. It is also worth mentioning that benefit from the 3145 mWh lithium battery integrated in the SPMC, the ASTCM enables continuous operation for approximately 16.6 hours with its stored energy, ensuring robustness in practical scenarios without sunlight input, such as at night and in severe weather (such as rain, snow, fog, etc.).

Since the realized ASTCM can be programmed in both the frequency and space domains, we further develop a STC strategy to achieve multi-beamforming at any desired frequency. According to the superposition theorem, the multi-beam farfield pattern can be synthesized by superpositing multiple single-beam farfield patterns. When ASTCM is programmable in one dimension, we assume that it is a linear array with N meta-atoms to generate a total of I main beams. Then the single-beam farfield pattern $ {f}^{i}\left(\theta \right) $ corresponding to the i^th (i = 1, 2, ... , I) main beam can be expressed as,where $ {f}^{u}\left(\theta \right) $ is the farfield pattern of the meta-atom, $ a_{n}^{i} $ and $ \varphi _{n}^{i} $ are the amplitude and phase of the n^th meta-atom respectively, and k is the wave number. When the i^th beam points to $ \theta _{\rm{r}}^{i} $, we can obtain $ \varphi _{n}^{i}=nkd\sin \theta _{\rm{r}}^{i} $. Then, the multi-beam farfield pattern can be synthesized as,where $ \varphi _{0}^{i} $ is the initial phase of the i^th beam. For simplicity, we assume that ASTCM has the ideal reflection amplitude distribution (i.e., $ a_{n}^{i}=1 $). According to Eq. (2), the desired phase $ {\phi }_{n} $ of the n^th meta-atom can be expressed as,

Then by combining Eqs. (2) and (5), the time-coding sequence for each meta-atom can be respectively optimized, and finally, the required STC matrix can be obtained by combining these optimized time-coding sequences (see Supplementary information of Section 4 for more details on the design process of the STC matrix). In addition, by substituting the equivalent amplitude and phase distributions of the ASTCM after applying the STC matrix into Eqs. (3) and (4), the multi-beam farfield pattern generated on the desired harmonic can be calculated.

To further verify the effectiveness of the proposed STC strategy, we performed several dual-beamforming experiments. In these experiments, the configuration parameters are set as follows: f_c = 9 GHz, t₀ =1 μs, L = 16 and f₀ = 62.5 kHz. The first experimental scenario is to generate two main beams pointing to −18° and 24° at the fundamental frequency. Figure 4(a) shows the corresponding optimized STC matrix, and Fig. 4(b, c) record the equivalent amplitude and phase distributions of ASTCM from −5^th to 5^th harmonics after applying the STC matrix. We note that the equivalent amplitude at the fundamental frequency exceeds 0.7, while the equivalent amplitudes at other harmonics are suppressed below 0.15. An equivalent phase distribution with 4-bit resolution is achieved at the fundamental frequency, and it matches the expected phase distribution well (with the maximum phase deviation of less than 4.2°). In addition, both the calculated and measured farfield patterns are illustrated in Fig. 4(d). The measured results are in good agreement with the calculated ones, demonstrating the capability of ASTCM for multi-beamforming at the fundamental frequency. The second experimental scenario is to generate two main beams pointing to −24° and +34° at the +1^st harmonic (f₊₁ = f_c + f₀ = 9.0000625 GHz), and the corresponding optimized STC matrix is shown in Fig. 4(e). After applying this STC matrix, as depicted in Fig. 4(f), ASTCM can efficiently reflect the +1^st microwave harmonic with the equivalent amplitude greater than 0.87. In addition, the equivalent phase distribution is completely consistent with the desired phase distribution, as shown in Fig. 4(g). The corresponding calculated and measured farfield patterns are also in good agreement, as illustrated in Fig. 4(h), incidating that ASTCM can achieve multi-beamforming at harmonic frequencies. The third experimental scenario is to perform independent dual-beamforming at both the −1^st harmonic (f₋₁ = f_c − f₀ = 8.9999375 GHz) and the +2^nd harmonic (f₊₂ = f_c + 2f₀ = 9.000125 GHz). The corresponding optimized STC matrix, the equivalent amplitude and phase distributions of ASTCM are presented in Fig. 4(i–k), respectively. It can be observed that the equivalent amplitudes of the −1^st and +2^nd harmonics are significantly higher than that of the other harmonics, while the equivalent phase distributions at the two harmonics are distinct and thus result in different dual-beamforming effects. As shown in Fig. 4(l), the two main beams at the −1^st harmonic point to ±24°, while the two main beams at the +2^nd harmonic point to −24° and 0°, respectively. Calculated results at both harmonics are in good agreement with the measured results, demonstrating the ability of ASTCM to perform independent multi-beamforming at multiple frequencies.

We can generate multiple beams with different frequencies and spatial propagation directions based on ASTCM to construct multiple independent information transmission channels. Without loss of generality, here we illustrate the process of constructing four independent information transmission channels by performing independent dual-beamforming on two different frequencies. Specifically, we use ASTCM to simultaneously generate dual beams pointing to ±24° at the −2^nd and +3^rd harmonics. Each of the four main beams supports an independent information transmission channel, and different channels are distinguished by different frequencies or spatial propagation directions. For clarity, we name the four channels as CH1 (f₋₂ = f_c−2f₀, $ \theta _{\mathrm{r}}^{1}=-24{^{\circ}} $), CH2 (f₊₃ = f_c + 3f₀, $ \theta _{\mathrm{r}}^{2}=-24{^{\circ}} $), CH3 (f₋₂ = f_c − 2f₀, $ \theta _{\mathrm{r}}^{3}=24{^{\circ}} $) and CH4 (f₊₃ = f_c + 3f₀, $ \theta _{\mathrm{r}}^{4}=24{^{\circ}} $). It is notable that each main beam has an independent initial phase, as can be observed from Eq. (4), which lays the foundation for independent phase modulation on each channel. Considering that each single-beam farfield pattern has low amplitude at other main beam directions, modulating the initial phase of any main beam has little effect on both the amplitude and phase of other main beams in the synthesized multi-beam farfield pattern. Therefore, it can be approximately considered that the phase modulation of each main beam does not interfere with each other. Based on this method, each information transmission channel can simultaneously transmit distinct digital information through quaternary phase shift keying (QPSK) modulation scheme by performing independent 2-bit phase modulation on each main beam.

A series of STC matrices M_ijpq are optimized to realize independent 2-bit phase modulations of the four channels, where i, j, p, q represent the indexs of CH1, CH2, CH3, and CH4, respectively. These indexs take values of 1, 2, 3, and 4, which refer to the 2-bit phase coding states of "00", "01", "10", and "11" respectively. For instance, we can implement 2-bit phase modulation only on CH1 by applying four optimized STC matrices M₁₁₁₁, M₂₁₁₁, M₃₁₁₁ and M₄₁₁₁. Figure 5(a, c) illustrate the farfield patterns generated by the ASTCM at −2^nd and +3^rd harmonics after applying these four STC matrices. It is evident that main beams pointing to ±24° are generated at both two harmonics. The gain fluctuation of all main beams does not exceed 1 dB, and the beam pointing fluctuation does not exceed 1°. As further shown in Fig. 5(b, d), only the main beam of CH1 produces four different phases of 26.3°, 114.1°, 204.8°, and 292.4°, which correspond to "00", "01", "10" and "11" phase coding states of CH1. The phase difference between adjacent phase states is 90°±2.4°, indicating that 2-bit phase modulation is well achieved. At the same time, the main beam phases of CH2, CH3, and CH4 are almost unchanged, indicating that other channels are not affected by the phase modulation of CH1. Here, we regard the current phases of CH2, CH3, and CH4 as their respective ''00'' phase coding states. In addition, we can optimize more STC matrices to achieve 2-bit phase modulation only on CH2, CH3, and CH4, respectively. When four optimized STC matrices M₂₁₂₂, M₂₂₂₂, M₂₃₂₂ and M₂₄₂₂ are applied, only 2-bit phase modulation can be obtained on CH2, while the main beam phases of CH1, CH3 and CH4 are all in their respective "01" phase coding states, as shown in Fig. 5(e–h). Similarly, as depicted in Fig. 5(i–l), when we apply four optimized STC matrices M₃₃₁₃, M₃₃₂₃, M₃₃₃₃ and M₃₃₄₃, only CH3 achieves 2-bit phase modulation, and the phase coding states of CH1, CH2 and CH4 are all "10". Furthermore, when we apply four optimized STC matrices M₄₄₄₁, M₄₄₄₂, M₄₄₄₃ and M₄₄₄₄, 2-bit phase modulation is only achieved on CH4, and the phase coding states of CH1, CH2 and CH3 are maintained at "11", as illustrated in Fig. 5(m–p). These results well verify the capability of ASTCM to efficiently construct four independent information transmission channels by only multiplexing two frequencies and two spatial propagation directions of the refection EM waves, and details of all optimized STC matrices mentioned above can be found in Supplementary information of Section 5. It is worth mentioning that more information transmission channels and diverse modulation schemes can be achieved based on the proposed ASTCM. As a demonstration, we further provide a six-channel case by utilizing three different harmonics (−1^st, 1^st, 2^nd) and two different beam directions (±24°), and we also discuss the challenges and potential solutions for further increasing the channel number (see Supplementary information of Section 6 for more details). In addition, another four-channel case is also provided to implement BPSK, QPSK and 8PSK modulation schemes respectively at different channels (see Supplementary Section 7 for more details).

To vaildate the capability of ASTCM to transmit multiple digital information, we use it to construct a four-channel wireless communication system, whose working mechanism is shown in Fig. 6(a). First, four different original information (named Info.1, Info.2, Info.3 and Info.4) are mapped into four digital bit streams respectively. Then, these digital bit streams are divided into 2-bit forms and subsequently mapped into 2-bit main beam phase streams. According to the above four-channel independent phase modulation method, these four main beam phase streams are interleaved and further converted into STC matrix streams, and output to ASTCM as control signals through MLCC. Incident by a monochromatic carrier wave, ASTCM reflects modulated microwaves carrying the four digital information for space-frequency division multiplexing transmissions. Finally, the four digital information are respectively received by four distinct user terminals (named User1, User2, User3 and User4) operating at different frequencies and located at different spatial positions. For user terminals in the same spatial position (e.g., User1 and User2), independent information transmission is ensured based on frequency division multiplexing; while for user terminals in different spatial positions but with the same operating frequency (e.g., User1 and User3), independent information transmission is achieved based on space division multiplexing. More importantly, ASTCM in such system demonstrates an autonomous and efficient capacity for harvesting the ambient solar energy, coupled with the feature of low power consumption. Such attributes endow the system to be flexibly deployed in real outdoor scenarios, enabling multi-channel information transmission through autonomous ambient solar energy harvesting even in the absence of power coverage.

As a proof of concept, a realistic four-channel wireless communication system is further built in the indoor environment, as illustrated in Fig. 6(b). The transmitting part of the system is composed of the realized ASTCM (including the SPMC and the MLCC), a transmitting (T_x) horn antenna and a microwave carrier generator (not shown in the figure). The T_x antenna connects the output of the microwave carrier generator and illuminate the unmodulated microwave carrier wave to ASTCM. The receiving system consists of four receiving (R_x) horn antennas, two dual-channel software-defined radio platforms (SDRPs), and two post-processing computers (PPCs). The four R_x antennas act as four user terminals respectively, each of which is connected to a demodulation channel of one SDRP, and the received digital information is processed and displayed through the PPC. In the experiment, we transmitted four black and white images (representing calligraphy of the Chinese characters “东”, “南”, “大” and “学”) to the four user terminals, respectively. The received QPSK constellation diagrams and the recovered images are shown in the insets of Fig. 6(b). It can be observed that all the received constellation diagrams have four clear constellation points, and the corresponding error vector magnitudes (EVMs) are calculated as 9.55%, 11.86%, 12.65% and 12.06%, respectively, incidating good communication quality of all four channels (see Supplementary information of Section 8 for more details of the measured communication performance). The four recovered images are all error-free, demonstrating that the system has low inerchannel interference characteristics (see Supplementary information of Movie 1 for more details of the four-channel image transmission experiment). Furthermore, we also evaluated the energy efficiency of the system. Considering that the four channels in the current experiment all adopt the QPSK modulation scheme, the information that can be transmitted in one symbol time is 8 bit. In addition, we measured the power consumption of the system during the four-channel image transmission experiment to be 139 mW. Therefore, the energy consumption per bit of the system is calculated as about 17.4 mW/bit. According to the aforementioned ambient energy harvesting performance verification results, it can be inferred that the system can be easily and solely driven by the ambient solar energy harvested by ASTCM. As shown in Table 1, we also compare our system with a series of state-of-the-art PM-based wireless communication systems and find that our system has the smallest energy consumption per bit to date, demonstrating the high energy efficiency of the system. Moreover, although notable progress has been achieved in the realms of dynamic beamforming, direct information modulation as well as energy harvesting and energy self-sufficiency, respectively, our ASTCM-based system pioneers to implement all of the above functionalities on a single platform, reveling its promising potential for advancing next-generation wireless communications (see Supplementary information of Section 9 for more details of the performance comparision).