1-Bit Hybrid Reconfigurable Intelligent Surface

• 1-Bit HRIS is a reconfigurable metasurface with binary phase control that reflects and simultaneously senses incident electromagnetic waves.
• The architecture employs power splitting between reflection and sensing paths to enable efficient channel estimation and angle-of-arrival localization.
• Electromagnetic pre-randomization and algorithmic optimization mitigate quantization lobes and beamforming limitations inherent to 1-bit control.

Searching arXiv for papers on 1-bit and hybrid RIS/HRIS to ground the article in current literature. A 1-bit Hybrid Reconfigurable Intelligent Surface (HRIS) is a reconfigurable metasurface that simultaneously redirects incident electromagnetic waves and senses a portion of them locally, while restricting the dynamically controlled element state to a binary configuration, typically two reflection-phase states such as $0$ and $\pi$. In the HRIS literature, the hybrid property denotes the coexistence of a controllable reflection path and a sensing or receiving path within the same metasurface aperture, usually realized through power splitting, waveguide coupling, or a small number of RF chains integrated at the surface (Alexandropoulos et al., 2021). The 1-bit specialization is not always explicit in early HRIS papers, but it follows directly by quantizing the reflection control variables of the underlying HRIS models (Alexandropoulos et al., 2021), and it has been instantiated more directly in later designs based on PIN-loaded resonant patches, parallel-plate or substrate-integrated waveguides, binary measurement matrices, and low-RF-chain sensing architectures (Keshmiri et al., 7 Jul 2025, Yang et al., 2022).

1. Concept and defining characteristics

Conventional RISs are nearly passive metasurfaces composed of sub-wavelength meta-atoms that only reflect incident signals in controllable ways. In such systems, each meta-atom implements a tunable reflection coefficient, often controlled by discrete states such as 1-bit or 2-bit phase shifts via PIN diodes or varactors. A conventional RIS does not directly observe the channel and therefore relies on external entities, typically the base station, to estimate the cascaded channels and then transmit control commands for phase-profile configuration (Alexandropoulos et al., 2021).

An HRIS extends this architecture by equipping each meta-atom, or each group of meta-atoms, with a simultaneous reflecting-and-sensing functionality. In the foundational HRIS formulation, each element splits the impinging wave into a reflected part and a sensed part. The reflected part is used for beam steering, focusing, or coverage enhancement, while the sensed part is coupled into a guided structure and routed to a small number of RF chains for baseband processing at or near the surface (Alexandropoulos et al., 2021). This dual operation enables local channel estimation, angle-of-arrival estimation, localization, and potentially self-configuration without requiring a fully external feedback loop (Alexandropoulos et al., 2021).

In a 1-bit HRIS, the reflection coefficient is quantized to two dynamically controllable states. The most common abstraction is

$\Gamma_n \in \{e^{j0}, e^{j\pi}\},$

or equivalently $\{+1,-1\}$ under unit-modulus reflection assumptions (Alexandropoulos et al., 2021, Yang et al., 2022). In some implementations, the active real-time control remains 1-bit even when structural pre-coding or geometry introduces additional fixed phase offsets. A specific example is a PIN-diode HRIS where each element has two diode states, while two different slot geometries generate four overall discrete reflection phases separated by roughly $90^\circ$; the active control is still binary because only the PIN ON/OFF state changes in operation (Keshmiri et al., 7 Jul 2025).

The literature uses “hybrid” in more than one sense. In reflecting-and-sensing HRISs, hybrid refers to simultaneous reflection and reception with a few RF chains (Alexandropoulos et al., 2021, Zhang et al., 2022, Zhang et al., 2022). In group-based over-the-air index modulation, hybrid can refer to active and passive partitions, and in an enhanced variant also absorbing groups (Arslan et al., 2024). In hybrid relay-reflecting surfaces, hybrid denotes a small subset of active amplify-and-forward elements embedded in an otherwise reflective surface (Nguyen et al., 2021, Nguyen et al., 2021). These variants are related but not identical. A plausible implication is that “1-bit HRIS” should be interpreted with care, because the binary constraint may apply to reflection phase, combining phase, group state, or only the controllable part of an otherwise structurally diversified element design.

2. Electromagnetic architecture and 1-bit implementations

The canonical reflecting-and-sensing HRIS meta-atom consists of a resonant scatterer loaded by a tunable element, plus a coupling structure that siphons part of the incident field into a waveguide. In one representative design, the meta-atom uses a resonant patch or ring loaded by a varactor diode to control effective capacitance and reflection phase, while a via couples RF energy into a substrate integrated waveguide (SIW) that carries the sensed signal to a small number $N_r \ll N$ of RF chains (Alexandropoulos et al., 2021). The same via or a separate conductor can also carry the DC bias (Alexandropoulos et al., 2021). HFSS simulations at $19$ GHz show substantial reflection-phase variation, nearly $360^\circ$, while the sensing path remains modest but non-negligible, with $|S_{21}|$ below about $-15$ dB across the operating band (Alexandropoulos et al., 2021).

A more explicit 1-bit reflecting-and-sensing implementation is built from independently tunable resonant patch elements loaded with PIN diodes, where a small rectangular slot in the ground plane couples part of the incident wave into a parallel-plate waveguide (PPWG) (Keshmiri et al., 7 Jul 2025). Two coaxial connectors at the ends of the PPWG collect the multiplexed sensing signal (Keshmiri et al., 7 Jul 2025). The reported unit cell uses a square patch of side length $\pi$0 mm on Rogers 4003 substrate of thickness $\pi$1 mm and relative permittivity $\pi$2, with period $\pi$3 mm and operating frequency around $\pi$4 GHz (Keshmiri et al., 7 Jul 2025). The slot has width $\pi$5 mm and length $\pi$6 mm or $\pi$7 mm, and the PIN diode is modeled in HFSS as a $\pi$8 pF capacitor in the OFF state and a $\pi$9 resistor in the ON state (Keshmiri et al., 7 Jul 2025). The reflection magnitude reduction caused by the sensing slot is reported to be less than $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$0 dB, indicating minimal efficiency loss from sensing (Keshmiri et al., 7 Jul 2025).

The same paper addresses a central difficulty of low-bit reflectarrays: quantization lobes. With $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$1-scale spacing and 1-bit control, regular phase quantization yields strong undesired lobes. To suppress them, the design introduces pre-coded phase randomization by randomly assigning one of two slot lengths to each element, thereby creating fixed geometry-dependent phase offsets while retaining 1-bit ON/OFF control per PIN diode (Keshmiri et al., 7 Jul 2025). For a 19-element 1D array, different random slot distributions generate markedly different far-field patterns, and the selected distribution suppresses the strong quantization lobes usually associated with regular 1-bit phase patterns (Keshmiri et al., 7 Jul 2025). This suggests that low-resolution control can be partially compensated at the electromagnetic design stage rather than only through signal processing.

Other 1-bit RIS hardware abstractions assume each element is a unit-modulus phase-only reflector with two states, physically consistent with PIN-diode realization (Yang et al., 2022). In that formulation there is no explicit sensing path, but the signal model and measurement-matrix design can be transferred to the sensing side of a 1-bit HRIS, especially when the surface uses a single RF chain or a very small number of RF chains (Yang et al., 2022). A plausible implication is that 1-bit HRIS design is most naturally realized as a co-design of binary switching hardware, low-complexity guided-wave sensing networks, and pattern libraries or codebooks optimized jointly for beamforming and estimation.

3. Mathematical models

A standard HRIS power-splitting abstraction assigns to each meta-atom $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$2 a reflection-power coefficient $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$3, where $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$4 is the reflected fraction and $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$5 is the sensed fraction (Alexandropoulos et al., 2021). If the incident complex baseband field is $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$6, then the reflected and sensed outputs are modeled as

$\Gamma_n \in \{e^{j0}, e^{j\pi}\},$7

where $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$8 is the controllable reflection coefficient and $\Gamma_n \in \{e^{j0}, e^{j\pi}\},$9 is the sensing-path phase or amplitude factor (Alexandropoulos et al., 2021). In an idealized interpretation,

$\{+1,-1\}$0

with

$\{+1,-1\}$1

accounting for parasitic losses (Alexandropoulos et al., 2021).

At system level, an uplink multi-user HRIS-assisted MIMO model uses a BS with $\{+1,-1\}$2 antennas, an HRIS with $\{+1,-1\}$3 elements and $\{+1,-1\}$4 RF chains, and $\{+1,-1\}$5 single-antenna users. Let $\{+1,-1\}$6 denote the user-to-HRIS channel, $\{+1,-1\}$7 the HRIS-to-BS channel, and

$\{+1,-1\}$8

Then the cascaded channel is

$\{+1,-1\}$9

the BS pilot observation is

$90^\circ$0

and the HRIS sensing observation is

$90^\circ$1

where $90^\circ$2 denotes the analog combining induced by waveguides and RF chains (Alexandropoulos et al., 2021). This structure is retained under 1-bit quantization; only $90^\circ$3 and possibly $90^\circ$4 become discrete (Alexandropoulos et al., 2021).

A closely related formulation represents the reflecting and sensing operations through matrices $90^\circ$5 and $90^\circ$6. The reflected signal is

$90^\circ$7

with

$90^\circ$8

while the sensed signal is

$90^\circ$9

where $N_r \ll N$0 if element $N_r \ll N$1 feeds RF chain $N_r \ll N$2, and zero otherwise (Zhang et al., 2022, Zhang et al., 2022). This formulation is especially convenient for channel-estimation analysis and makes the 1-bit specialization transparent: $N_r \ll N$3 and, in more aggressive discretizations,

$N_r \ll N$4

(Zhang et al., 2022, Zhang et al., 2022).

In low-cost single-RF-chain 1-bit sensing formulations, the surface operates over $N_r \ll N$5 time slots, each with a binary phase pattern. For a uniform linear array with $N_r \ll N$6 elements and target DOAs $N_r \ll N$7, the element-domain echo field is

$N_r \ll N$8

and the scalar received signal in time slot $N_r \ll N$9 is

$19$0

which stacks into

$19$1

with measurement matrix $19$2 consisting of $19$3 rows (Yang et al., 2022). Although developed for one-bit RIS-assisted DOA estimation rather than HRIS, this measurement-matrix view transfers directly to the sensed side of a 1-bit HRIS (Yang et al., 2022).

4. Channel estimation, sensing, and localization

One of the primary motivations for HRISs is that purely reflective RISs make channel estimation difficult because they only allow external estimation of cascaded channels. HRISs break this limitation by sensing a fraction of the incident field directly (Alexandropoulos et al., 2021, Zhang et al., 2022, Zhang et al., 2022). In the multi-user uplink setting, the HRIS can first estimate the user-to-HRIS channel using its local sensing circuitry and then enable the BS to estimate the HRIS-to-BS channel using the reflected pilots and the HRIS-reported local estimate (Alexandropoulos et al., 2021). In the ideal noise-free case, the HRIS requires

$19$4

pilot symbols to estimate the $19$5 user-to-HRIS channel, while the BS requires

$19$6

to estimate the $19$7 HRIS-to-BS channel (Alexandropoulos et al., 2021). For $19$8, $19$9, and $360^\circ$0, both channels can be recovered with 64 pilots overall due to pilot reuse across the two estimation stages (Alexandropoulos et al., 2021). In the same general scenario, a purely reflective RIS method cited as state of the art requires over 90 pilots to estimate only the cascaded channel (Alexandropoulos et al., 2021).

The noisy analysis yields closed-form MSE expressions for the local user-to-HRIS estimate and the BS-side HRIS-to-BS estimate in terms of the sensing matrix or analog combiner and the reflection configuration (Zhang et al., 2022). A central empirical finding is the trade-off governed by the common splitting ratio $360^\circ$1: increasing $360^\circ$2, meaning more reflection and less sensing, improves the HRIS-to-BS channel estimate at the BS, while degrading the user-to-HRIS estimate at the HRIS (Alexandropoulos et al., 2021). In the reported example with $360^\circ$3, $360^\circ$4, $360^\circ$5, $360^\circ$6, 70 pilots, and transmit SNR 30 dB, increasing $360^\circ$7 up to about $360^\circ$8 improves BS-side estimation with only minor degradation of HRIS-side estimation, whereas pushing $360^\circ$9 further yields little additional BS-side benefit and sharply worsens the HRIS-side estimate (Alexandropoulos et al., 2021). This suggests a balanced operating point around 50/50 power split for many HRIS-assisted channel-estimation tasks (Alexandropoulos et al., 2021).

AoA estimation is another key HRIS sensing function. In the 2021 HRIS study, a square HRIS with $|S_{21}|$0 or $|S_{21}|$1, element spacing about $|S_{21}|$2, a single RF chain, and $|S_{21}|$3 sensed samples can estimate elevation AoA with RMSE approaching the Cramér–Rao lower bound even when only 20% of the incident power is sensed, corresponding to $|S_{21}|$4 (Alexandropoulos et al., 2021). Increasing the sensed fraction improves the AoA estimate, and in that single-RF-chain setting a smaller HRIS with $|S_{21}|$5 can outperform a larger one with $|S_{21}|$6 because the narrower beams of the larger aperture complicate the single-chain estimation problem (Alexandropoulos et al., 2021). A plausible implication for 1-bit HRIS is that temporal design of binary phase masks becomes especially important when continuous analog combining is unavailable, because the spatial information must be encoded through a sequence of coarse patterns.

A direct 1-bit HRIS AoA architecture is described in the 2025 PPWG-based design (Keshmiri et al., 7 Jul 2025). There, $|S_{21}|$7 different random diode masks are applied sequentially, and for each mask the complex voltage difference

$|S_{21}|$8

between the two PPWG ports is measured (Keshmiri et al., 7 Jul 2025). Over a discretized angle grid from $|S_{21}|$9 to $-15$0 in $-15$1 steps, these measurements define a sensing matrix $-15$2, yielding

$-15$3

where $-15$4 is sparse and indicates the active AoA bin (Keshmiri et al., 7 Jul 2025). With CGS-based recovery, $-15$5 masks, and SNR $-15$6 dB, the system achieves AoA estimation within $-15$7 (Keshmiri et al., 7 Jul 2025). Performance degrades when SNR drops to 15 dB or when the number of masks is reduced to 8 (Keshmiri et al., 7 Jul 2025). Using a multi-layer perceptron trained on only 4 masks, the same architecture maintains high classification accuracy across a wide SNR range and achieves robust $-15$8-level angle classification (Keshmiri et al., 7 Jul 2025).

Related low-cost 1-bit RIS sensing work develops a measurement-matrix design based on a modified genetic algorithm and performs DOA recovery using atomic norm minimization (Yang et al., 2022). For $-15$9, $\pi$00 measurements, $\pi$01 targets at $\pi$02 and $\pi$03, and SNR $\pi$04 dB, the method reports RMSE $\pi$05, compared with $\pi$06 for OGSBI and $\pi$07 for $\pi$08-SVD (Yang et al., 2022). Although this work is not explicitly hybrid, its 1-bit measurement-matrix construction and gridless sparse recovery are directly relevant to HRIS sensing subsystems (Yang et al., 2022).

Localization with HRISs can extend beyond user positioning to joint calibration of the surface itself. In downlink multicarrier settings with a single-RF-chain reflecting-and-sensing HRIS, multi-stage estimators have been proposed for joint 3D user localization and 6D HRIS state estimation, including 3D position and 3D rotation matrix (Ghazalian et al., 2024), and for 2D joint user localization and HRIS position-orientation calibration (Ghazalian et al., 2022). These works show that the power-splitting ratio $\pi$09 induces a nontrivial trade-off: too little sensing power undermines BS–HRIS parameter estimation, while too little reflection power undermines the BS–HRIS–UE path, producing an optimal intermediate $\pi$10 for joint state estimation (Ghazalian et al., 2024, Ghazalian et al., 2022). In the reported setups, the optimal point appears near $\pi$11 (Ghazalian et al., 2024). This suggests that 1-bit HRIS localization systems, despite coarse phase control, can still benefit strongly from accurate split-ratio design.

5. Beamforming, pattern design, and quantization effects

The main direct consequence of 1-bit control is reduced beamforming flexibility relative to continuous-phase or multi-bit RISs. A conventional RIS with continuous phase can implement near-perfect phase alignment and achieve array gain approximately $\pi$12, whereas 1-bit phase shifts reduce the achievable gain and increase sidelobes (Alexandropoulos et al., 2021). The 2021 HRIS overview states that for large arrays the typical array-gain reduction under 1-bit quantization is around 1–2 dB (Alexandropoulos et al., 2021). The PPWG-based 1-bit HRIS paper identifies quantization lobes as a dominant issue when element spacing is about $\pi$13 and phase control is binary (Keshmiri et al., 7 Jul 2025).

Two complementary strategies appear in the literature for mitigating low-resolution effects. The first is electromagnetic pre-randomization, as described above through binary slot-length patterns that suppress regular quantization lobes (Keshmiri et al., 7 Jul 2025). The second is algorithmic optimization of the binary pattern sequence. In the one-bit RIS DOA estimation work, each time-slot pattern is a row of a measurement matrix $\pi$14, and the design objective combines beam-pattern matching and low mutual coherence (Yang et al., 2022). The far-field beampattern for a row $\pi$15 is

$\pi$16

and the modified genetic algorithm minimizes

$\pi$17

with a penalty factor

$\pi$18

to enforce sidelobe and pointing constraints (Yang et al., 2022). To support sparse recovery, the measurement vectors are additionally selected to keep their pairwise correlation low, reflecting a practical restricted isometry or low-coherence condition (Yang et al., 2022).

Within HRIS design, a plausible implication is that 1-bit pattern design should be viewed not merely as approximation of a continuous phase profile, but as a codebook-construction problem. The same binary profiles can serve different functions in different subframes: high-gain communication beams, low-coherence sensing patterns, or mixed operation where the reflected portion supports communication and the sensed portion supports estimation (Yang et al., 2022, Alexandropoulos et al., 2021). The early HRIS analysis already notes that even when beamforming gain is reduced by 1-bit reflection, the main structural advantage of HRIS—direct observation of the user-to-HRIS channel—remains intact (Alexandropoulos et al., 2021).

6. Variants and applications

Reflecting-and-sensing communications and ISAC

The reflecting-and-sensing interpretation of HRIS directly supports integrated sensing and communications. A continuous-phase framework for simultaneous downlink communications and sensing considers an XL-MIMO BS, a single-antenna user, and an HRIS that both reflects toward the user and absorbs a fraction of the incident signal for localization of passive targets in an area of interest (Gavras et al., 2024). The user-side received signal is

$\pi$19

while the sensing-side observation is

$\pi$20

where $\pi$21 is now the absorbed fraction and $\pi$22 the reflected fraction (Gavras et al., 2024). Position Error Bound constraints for multiple area-of-interest points are then incorporated into a joint optimization of BS beamforming, HRIS reflection, and analog combining (Gavras et al., 2024). Simulation results show a clear rate–coverage trade-off: increasing $\pi$23 improves localization coverage while reducing downlink rate (Gavras et al., 2024).

A secure ISAC extension uses the same dual-functional HRIS to support secure downlink communication and simultaneous localization of both a legitimate user and an eavesdropper in a bistatic sensing setting (Gavras et al., 29 Apr 2025). There the global splitting coefficient is denoted $\pi$24, with the reflected user channel

$\pi$25

and the sensing signal

$\pi$26

(Gavras et al., 29 Apr 2025). The Fisher information scales with $\pi$27, so larger absorption improves positioning, while larger reflection improves secrecy spectral efficiency (Gavras et al., 29 Apr 2025). The work explicitly frames continuous-phase HRIS as an upper bound for a 1-bit implementation, recommending post-optimization phase quantization to $\pi$28 for both reflection and combining (Gavras et al., 29 Apr 2025).

Group-based active–passive HRIS for over-the-air index modulation

A different HRIS variant partitions the surface into $\pi$29 groups of $\pi$30 elements and uses group states to convey index bits over the air (Arslan et al., 2024). In the OTA-RIS-IM scheme, each group is either passive or active, encoded by $\pi$31, and the received user signal is

$\pi$32

with instantaneous SNR

$\pi$33

(Arslan et al., 2024). The enhanced E-OTA-RIS-IM scheme adds an absorption state per group, increasing the index-modulation alphabet and the spectral efficiency to

$\pi$34

bpcu (Arslan et al., 2024). Simulations show that OTA-RIS-IM achieves the best BER because all groups always contribute, while E-OTA-RIS-IM increases spectral efficiency at the cost of BER degradation due to absorption states (Arslan et al., 2024). The paper assumes continuous phase alignment, but explicitly notes that a 1-bit HRIS can be obtained by constraining $\pi$35, with the same signaling structure preserved (Arslan et al., 2024).

Hybrid relay-reflecting surfaces

Another branch of the literature embeds a small number of active amplify-and-forward elements into a largely passive surface (Nguyen et al., 2021, Nguyen et al., 2021). In this hybrid relay-reflecting intelligent surface, the diagonal coefficient matrix is

$\pi$36

with $\pi$37 for passive elements and $\pi$38 for active elements (Nguyen et al., 2021). The active subset can compensate for performance degradation due to limited-resolution phase shifters (Nguyen et al., 2021). In a representative 8-by-2 MIMO scenario with $\pi$39 elements and 2-bit phase shifters, a dynamic HR-RIS with only four active elements achieves up to 42.8 percent spectral-efficiency improvement and 41.8 percent energy-efficiency improvement relative to a conventional RIS (Nguyen et al., 2021). The earlier 2021 HR-RIS paper also formulates fixed and dynamic architectures, derives alternating-optimization and water-filling style coefficient updates, and emphasizes that a small number of active elements is usually preferable (Nguyen et al., 2021). While these works do not specialize to reflecting-and-sensing HRISs, they are relevant to 1-bit HRIS because they treat low-resolution phase control as a practical impairment that hybridization can mitigate (Nguyen et al., 2021, Nguyen et al., 2021).

7. Performance trends, design trade-offs, and open issues

Across the surveyed literature, several recurring trade-offs define 1-bit HRIS design.

The first is the reflection–sensing trade-off, governed by $\pi$40 or $\pi$41. Larger sensing fraction improves local estimation, localization, or target detection, but weakens the reflected communication path (Alexandropoulos et al., 2021, Gavras et al., 2024, Gavras et al., 29 Apr 2025, Ghazalian et al., 2024, Ghazalian et al., 2022). The converse improves communication or secrecy but degrades sensing. Multiple studies identify an interior operating point—often near a balanced split—as optimal for joint tasks rather than pure communications or pure sensing (Alexandropoulos et al., 2021, Ghazalian et al., 2024, Ghazalian et al., 2022).

The second is the hardware–algorithm trade-off. One-bit control greatly simplifies bias networks, switching circuitry, power consumption, and manufacturability (Yang et al., 2022, Keshmiri et al., 7 Jul 2025). The compensating cost is algorithmic: pattern design, sparse recovery, neural classification, binary optimization, or codebook construction become more important because the hardware no longer provides fine-grained analog control (Yang et al., 2022, Keshmiri et al., 7 Jul 2025). This suggests that 1-bit HRISs are most attractive when offline optimization and moderate digital processing are acceptable in exchange for substantial hardware simplification.

The third is the aperture–resolution trade-off. Larger HRISs improve array gain, channel-estimation conditioning, and spatial resolution (Alexandropoulos et al., 2021, Gavras et al., 2024, Gavras et al., 29 Apr 2025). But larger apertures can also complicate single-RF-chain sensing if the beams become too narrow, as observed in the HRIS AoA study where $\pi$42 outperformed $\pi$43 for one-RF-chain estimation in certain regimes (Alexandropoulos et al., 2021). For 1-bit surfaces this trade-off can be sharper because coarse phase quantization tends to widen main lobes and raise sidelobes while large apertures sharpen them. A plausible implication is that the optimal aperture depends strongly on the sensing architecture, especially the number of RF chains and the temporal coding scheme.

The fourth is the role of RF-chain count $\pi$44. Even a small number of sensing chains can substantially improve channel estimation over purely reflective RISs (Alexandropoulos et al., 2021, Zhang et al., 2022). In the HRIS cascaded-channel comparison, performance gains become pronounced once $\pi$45 (Alexandropoulos et al., 2021). This indicates that the defining benefit of HRIS lies less in phase resolution than in the existence of direct observations at the surface. Accordingly, 1-bit reflection quantization does not erase the structural estimation advantage so long as the sensing network remains sufficiently informative (Alexandropoulos et al., 2021).

Open problems recur throughout the literature. One is the lack of explicit information-theoretic limits for 1-bit HRISs combining quantized reflection, limited RF chains, and power splitting (Alexandropoulos et al., 2021). Another is quantization-aware optimization: most continuous-phase HRIS works can be adapted by quantization, but do not solve the discrete problem natively (Gavras et al., 2024, Gavras et al., 29 Apr 2025). Electromagnetic–algorithmic co-design remains underdeveloped, particularly for mutual coupling, neighboring-element interactions, and the mapping from switch states to actual S-parameters in binary hardware (Alexandropoulos et al., 2021, Keshmiri et al., 7 Jul 2025). Localization and ISAC under multipath, wideband operation, multi-target conditions, and full 2D large-aperture 1-bit HRISs are also active challenges (Keshmiri et al., 7 Jul 2025, Ghazalian et al., 2024). The active–passive hybridization line raises additional questions about optimal placement and number of active elements under stringent power budgets and coarse phase control (Nguyen et al., 2021, Nguyen et al., 2021).

In aggregate, the literature supports a precise characterization of the 1-bit HRIS as a low-complexity dual-functional metasurface whose binary control constrains reflection and sometimes sensing or group state, but whose sensing capability fundamentally alters the estimation and autonomy properties of the surface. The most consistent conclusion is that quantization mainly limits beamforming finesse, whereas the defining benefit of HRIS—the ability to directly observe part of the impinging field—survives coarse control and often becomes more valuable when element hardware is kept deliberately simple (Alexandropoulos et al., 2021, Keshmiri et al., 7 Jul 2025, Yang et al., 2022).