Ferkans — Interactive Telecom Tutor

Breaking the Half-Space Barrier

Passive RIS reflects. Its physics is simple: an element sits in front of a ground plane, and incident waves can only go back to the half-space from which they came. This is a fundamental constraint: a user behind the RIS sees nothing. If the BS and UE are on opposite sides of an RIS panel, the classical architecture fails — the RIS shields rather than serves.

STAR-RIS (Simultaneously Transmitting and Reflecting RIS) removes this constraint by replacing the ground plane with a partially-transparent layer. Each element produces two waves: one reflected back to the incident half-space, and one transmitted through to the other side. The RIS now covers the full space — every direction a wave can reach. Multi-user deployments on both sides of the surface become natural.

The golden thread: the RIS still programs propagation, but now with two diagonals $\boldsymbol{\Phi}^r, \boldsymbol{\Phi}^t$ instead of one. Energy conservation ties them together: $|r_n|^2 + |t_n|^2 \leq 1$ for passive hardware.

STAR-RIS

Simultaneously Transmitting and Reflecting Reconfigurable Intelligent Surface. Each element radiates waves in both half-spaces simultaneously (reflect + transmit), with coefficients $r_n, t_n$ satisfying $|r_n|^2 + |t_n|^2 \leq 1$ . Enables full-space coverage from a single panel.

Energy Splitting (ES)

A STAR-RIS protocol where each element continuously divides its incident energy between the reflection and transmission paths with per-element amplitudes $(a_n^r, a_n^t)$ satisfying $(a_n^r)^2 + (a_n^t)^2 = 1$ . Best-performing protocol; requires continuous amplitude control hardware.

Definition:
STAR-RIS System Model

A STAR-RIS element has two outputs: a reflected wave (to the incident-side half-space) and a transmitted wave (to the opposite half-space). Each is characterized by a complex coefficient:

$r_n = a_n^r e^{j\theta_n^r}, \quad t_n = a_n^t e^{j\theta_n^t}, \qquad n = 1, \ldots, N.$

Stacking into diagonal matrices, $\boldsymbol{\Phi}^r = \text{diag}(r_n)$ and $\boldsymbol{\Phi}^t = \text{diag}(t_n)$ . For a user $k \in \mathcal{U}^r$ on the reflection side, the effective channel uses $\boldsymbol{\Phi}^r$ :

$\mathbf{h}_{k,\text{eff}}^{r,H} = \mathbf{h}_{k,d}^H + \mathbf{h}_{k,2}^{r,H} \boldsymbol{\Phi}^r \mathbf{H}_1.$

For a user $k \in \mathcal{U}^t$ on the transmission side:

$\mathbf{h}_{k,\text{eff}}^{t,H} = \mathbf{h}_{k,d}^H + \mathbf{h}_{k,2}^{t,H} \boldsymbol{\Phi}^t \mathbf{H}_1.$

The two half-spaces share the same $\mathbf{H}_1$ (BS-to-RIS) but different $\mathbf{h}_{k,2}^r$ / $\mathbf{h}_{k,2}^t$ channels.

Note that passive RIS is the special case $|t_n| = 0$ (no transmission) with $|r_n| = 1$ . STAR-RIS reduces to passive RIS when all elements are set to "reflect-only" mode.

,

Theorem: Energy Conservation in Passive STAR-RIS

For any passive STAR-RIS element,

$|r_n|^2 + |t_n|^2 \leq 1, \qquad n = 1, \ldots, N,$

with equality for a perfectly lossless element. The phases $\theta_n^r, \theta_n^t$ can be independent only within certain hardware-specific relations — for instance, a common physical implementation (electric and magnetic currents) imposes

$\theta_n^t - \theta_n^r \in \{\pi/2, -\pi/2\}$

at each element. Whether the phases are fully independent or coupled defines the hardware model (symmetric, asymmetric, or fully coupled STAR-RIS).

A lossless passive element cannot create power. The incident wave splits into reflected + transmitted; the sum of their powers equals the incident power. Phases are free; amplitudes must respect conservation.

Proof

Passivity

A passive element has no power source. Total reflected power + transmitted power $\leq$ incident power.

Per-element coefficient bound

For a unit-power incident wave, reflected amplitude is $|r_n|$ and transmitted is $|t_n|$ . Squared-magnitude conservation gives $|r_n|^2 + |t_n|^2 \leq 1$ .

Phase coupling from physics

In the electric-magnetic current model (Liu et al. 2021), the electric-current induced reflection and magnetic-current-induced transmission are naturally $\pi/2$ out of phase. Other implementations (purely dielectric, purely metallic) have different coupling. $\blacksquare$

Two Hardware Models: Coupled vs. Independent Phases

Two hardware implementations dominate the STAR-RIS literature:

Coupled-phase model: $\theta_n^t = \theta_n^r \pm \pi/2$ . One phase controls both; easier to manufacture; slightly less flexible for optimization.
Independent-phase model: $\theta_n^t, \theta_n^r$ are independent. Requires more complex hardware (both electric and magnetic control); fully flexible for optimization.

The coupled-phase model is physically realistic for most passive implementations; the independent-phase model is an idealization used in algorithmic studies. Some engineering realizations (active layers, meta-surfaces with separate E- and H-currents) approach the independent-phase ideal. We mainly work with the simpler independent-phase model below and flag where the coupled model would differ.

STAR-RIS: Full-Space Coverage Geometry

A STAR-RIS panel deployed between a BS (top) and two user sets:

\mathcal{U}^r

on the reflection side (same as the BS) and

\mathcal{U}^t

on the transmission side (behind the panel). Each element radiates two beams — one in each half-space — programmed via

(r_n, t_n)

.

Key Takeaway

STAR-RIS extends passive RIS to full-space coverage. Each element has two independent (or coupled- $\pi/2$ ) complex coefficients: $r_n$ for reflection and $t_n$ for transmission. Energy conservation constrains $|r_n|^2 + |t_n|^2 \leq 1$ . The extra flexibility enables simultaneous multi-user coverage of both half-spaces, at the cost of more optimization variables and a more complex hardware design.

Example: Indoor-Outdoor Coverage from a Single Panel

A STAR-RIS panel is mounted on a building window. The BS is outside (serving outdoor pedestrian UEs via the reflection side), and indoor IoT devices are on the transmission side. How does STAR-RIS serve both user groups with one panel?

Solution

Reflection side

For outdoor UEs, the RIS appears as a conventional reflecting RIS: $\boldsymbol{\Phi}^r$ forms coherent beams toward each outdoor user. Received signal uses the $r_n$ coefficients.

Transmission side

For indoor UEs, the RIS acts as a window-transparent beamformer: $\boldsymbol{\Phi}^t$ shapes the transmitted beam. The BS signal passes through the RIS with controllable phase and amplitude, reaching indoor users who would otherwise have no outdoor-RIS path.

Simultaneous operation

Energy-splitting protocol (see Section 10.2): each element reflects a fraction $|r_n|^2$ of energy to outdoor UEs and transmits $|t_n|^2$ to indoor UEs. Optimal split per element depends on user priorities and channel strengths — solved jointly with $\mathbf{W}$ .

Comparison to passive

With passive RIS: outdoor-only coverage. Indoor UEs would need a second RIS panel mounted inside. STAR-RIS saves one deployment point and halves the hardware cost for this common geometry. $\blacksquare$

Historical Note: STAR-RIS: From Academic Curiosity to Industry Standard

2021–present

The idea of a transparent reconfigurable surface predates the RIS literature by decades — frequency-selective surfaces (FSS) of the 1990s could block some frequencies and pass others. The reconfigurable version, where transmission/reflection coefficients can be adjusted in real time, emerged around 2020 in Mu et al. (2022) and Xu et al. (2021), introducing the "STAR-RIS" name and the three-protocol framework.

Early prototypes used metamaterial metasurfaces with electric and magnetic resonators; commercial implementations have emerged in the 2024–2025 window. ETSI has begun drafting STAR-RIS standards as part of the broader RIS specification. What started as a theoretical curiosity has become the architectural default for any deployment where coverage through a barrier (window, wall with low attenuation) is required.

🔧Engineering Note

STAR-RIS Physical Realizations

Three hardware approaches to STAR-RIS:

Chiral metasurface: orthogonal electric and magnetic layers enable independent $r_n, t_n$ control. Complex fabrication; research-grade at mmWave.
Hybrid dielectric+metallic: dielectric transmission layer with metallic reflective patches. Coupled phases; easier to manufacture but less flexible.
Amplifier-based (active STAR-RIS): each element has dual active amplifiers for r/t paths. Most flexible, most expensive, not covered here (see Ch. 9 for active RIS foundations).

Commercial implementations (ETSI 2024 draft) typically use option 2 with coupled- $\pi/2$ phases. Performance-focused research deployments use option 1.

Practical Constraints

•
Typical reflection amplitude $a_n^r$ : $0.6$ - $0.9$ (rest transmitted or lost to material).
•
Typical transmission amplitude $a_n^t$ : $0.4$ - $0.7$ .
•
Combined loss (energy not reflected or transmitted): $\leq 10\%$ in good designs.
•
Bandwidth: typically narrower than passive RIS (resonant structure more sensitive).

Common Mistake: Coupled vs. Independent Phases: Check Hardware

Mistake:

"Optimize $\theta_n^r$ and $\theta_n^t$ as two independent variables."

Correction:

Most passive STAR-RIS hardware imposes a coupling $\theta_n^t - \theta_n^r = \pm \pi/2$ . Ignoring the coupling overestimates the achievable rate by $0.5$ - $1\,\text{dB}$ in typical scenarios. For independent-phase STAR-RIS (active or chiral-metasurface), the coupling can be dropped. Always verify the hardware model before running the optimization.

The STAR-RIS Concept