Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design

arXiv:2611.07548 · eess.SP · Submitted 2026-11-05 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Electrical Engineering. The day in electrical engineering: signals, systems and the hardware that carries them.

Rosa: We're looking at the paper: "Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design".

Dev: Here's the gist of the paper. Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design The minimum desired power gives a tight rate guarantee, set by the interference at the chosen phase, while the weighted sum gives none.

Rosa: Problem Setup and Objectives The paper characterizes when desired-power optimization guarantees the minimum user rate (MUR) in discrete reconfigurable intelligent surface systems with fixed precoding.

Dev: The analysis compares two power objectives: the weighted sum of desired powers, denoted as Pω(θ), and the minimum desired power, D(θ).

Rosa: Both objectives are noted to ignore interference.

Title and authors: Rosa: To summarize the core contribution of this paper, "Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design," they are showing that the minimum desired power objective provides a mathematically provable and sharp lower bound on the user rate when you are using discrete reconfigurable intelligent surface systems with fixed precoding.

Dev: So, they’re contrasting this with the weighted sum objective, which we saw earlier doesn't give any guarantee because it can lead to one user having zero rate while interference stays bounded. That contrast is key to understanding the two different optimization paths available in these systems.

Taro: The authors are essentially arguing that instead of chasing a complex power metric like the weighted sum, designers should focus on selecting a phase based on minimizing the minimum desired power, because that selection gives you a rate guarantee that's actually sharp and depends directly on the interference level at that point.

Rosa: They formalize this with theorem one, establishing that if you pick a phase whose minimum desired power is greater than some fraction of the maximum possible minimum desired power, say alpha times R star, then the resulting rate will be at least proportional to R star.

Dev: The specific bound they derive is linear in R star, with the constant factor dependent on alpha and that interference level kappa hat you mentioned earlier. It’s a very concrete statement about performance reliability.

Taro: I found the numerical results quite compelling, especially how the omitted interference accounts for most of the rate loss—ninety to ninety-four percent of it—which really emphasizes why focusing on that term is so important for practical deployment.

The paper's summary: Rosa: One major improvement they point out is moving away from relying solely on the weighted power sum objective because we know it fails to provide any positive guarantee in certain channel setups.

Dev: That’s a good design constraint for us because it means if we use a weighted sum approach, we have to accept that there are scenarios where the rate might be zero for one user, which is bad for fairness and service level agreements.

Taro: The second major contribution is providing that dual certificate via proposition two; this allows you to certify the retained fraction of optimal rate for any phase without having to enumerate all possible phases, which is a huge win for scalability.

Rosa: So, they’re suggesting that instead of trying to read fairness from the power objective function itself, the system design should focus on certifying the rate associated with the selected phase using this dual certificate. It shifts the focus from power maximization to rate certification based on interference constraints.

Dev: From a controls perspective, this means we can use this certificate during runtime to verify that our current phase selection is meeting a minimum acceptable performance level, even if we don't know the absolute optimal settings for every possible channel realization.

Taro: They also show robustness in their numerical tests; the shortfall in performance only depends on the deployment through that interference kappa hat at the selected phase, and this gap stays well below the optimum even when we change things like precoders or correlation.

The paper's improvements: Rosa: So to wrap up, this paper "Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design" concludes that minimum desired power gives a sharp rate guarantee determined by the interference at the chosen phase, while the weighted sum objective offers no such guarantee.

Dev: The main implication for us is that if you are designing a discrete RIS system with fixed precoding, you should certify your phase selection based on its minimum desired power and the resulting interference level, rather than just trying to maximize some arbitrary weighted sum of powers.

Taro: I think the practical implication is that we need to incorporate this understanding into our deployment strategy, focusing on certifying what the chosen phase actually delivers in terms of rate reliability, not just raw power.

Rosa: Exactly. The paper also notes that as you increase the number of antennas or users, or when you look at finer resolutions at a fixed antenna count like six, the gap between achieved performance and the optimum widens, which means we need to be more careful as our system scales up.

Dev: That makes sense for real-time systems; higher dimensionality means more complex interactions that these simple power objectives don't capture well anymore. So, while this guarantee is sharp now, it gets tighter constraints as the system grows bigger.

Taro: It’s a solid piece of theoretical work showing exactly where the limitations are, and it points us toward focusing on interference control as the key to guaranteeing performance in these discrete systems.

Conclusion: Rosa: So we’re wrapping up on "Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design." The main thing is that they proved that using the minimum desired power objective gives you a solid, sharp rate guarantee, but using the weighted sum objective doesn't give you anything at all.

Dev: That’s right. They showed that when you use the weighted sum of powers, there are specific channel setups where one user gets zero rate while interference stays low, so no positive guarantee exists there.

Taro: And for minimum desired power, they established a tight bound because that guarantee is set directly by how much interference is present at the phase you pick.

Rosa: Exactly. They used this semidefinite dual thing to give us a certificate, meaning we can certify the retained rate for any chosen phase without having to check every single possible phase configuration.

Dev: It means we can use this certificate in our real-time loops to check if our current phase selection is meeting a certain performance floor based on that interference level.

Taro: I just think the numerical validation was really telling; they showed that most of the rate loss comes from that omitted interference, about ninety to ninety-four percent of it.

Rosa: Yeah, and even when we change things like the precoders or correlations, the shortfall stays well below what you need for optimal performance. That robustness is important for deploying anything outside a perfect lab setting.

Dev: It means our control loops can actually rely on this guarantee staying within bounds even if the world misbehaves a bit. The safety case gets much stronger when you have this kind of provable bound tied to the interference at the selected phase.

Taro: I think it really shows that focusing on controlling that interference at the specific phase is more important than just chasing a high weighted sum number.

Rosa: It’s about choosing a metric that actually provides a hard performance guarantee, not one that's just theoretically cleaner looking. That's what this paper on the minimum desired power objective proves.

Dev: Yeah, it sets a clear direction for how we should be designing these discrete RIS systems going forward. Next up, we’ll look at some of those sensing-assisted OFDM waveforms and how they handle localization versus communication tradeoffs.

Eindhoven University of Technology

eess.SP

Submitted: 2026-11-05

Updated: 2026-11-05

Comments: in the submission process to WCL

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 92/100

The gist: Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design The minimum desired power gives a tight rate guarantee, set by the interference at the chosen phase…

Key concepts

Weighted Power Objective Pω(θ)
This objective calculates a score by taking the weighted sum of the desired powers for different users. The paper shows this method is unreliable because, in certain channel conditions, maximizing this sum can cause one user's rate to drop to zero while interference remains controlled.
Minimum Desired Power D(θ)
This objective focuses on finding the phase setting that maximizes the lowest desired power among all users. Unlike the weighted sum, this approach provides a strong mathematical guarantee that ensures the resulting rate is at least a specific fraction of the maximum possible rate.
Semidefinite Dual Certificate
This is a mathematical tool used to create a certificate. Instead of checking every possible phase setting, this dual method gives an inequality that proves how much of the optimal rate is retained for any chosen phase, making it efficient for selection.

Terminology

Summary

Sharp Minimum-Rate Guarantees for Power-Based Discrete RIS Design The minimum desired power gives a tight rate guarantee, set by the interference at the chosen phase, while the weighted sum gives none. p. 2

Problem Setup and Objectives

The paper characterizes when desired-power optimization guarantees the minimum user rate (MUR) in discrete reconfigurable intelligent surface systems with fixed precoding. The analysis compares two power objectives: the weighted sum of desired powers, denoted as Pω(θ), and the minimum desired power, D(θ). p. 0 Both objectives are noted to ignore interference. p. 1 The objective function for the weighted sum is described as a quadratic unconstrained binary optimisation (QUBO). p. 1

Weighted Power Limitations

The analysis shows that the weighted sum of desired powers provides no guarantee. Specifically, on a bounded single-element channel, the exact maximiser drives one user’s rate to zero while the interference stays bounded. p. 0 This implies that no universal factor can be guaranteed by capping interference when using the weighted power objective. p. 1

Minimum Desired Power Guarantee

For the minimum desired power objective, a tight guarantee is established. Theorem 1 states that for any candidate phase θˆ, if D(θˆ) ≥ αD⋆, then R(θˆ) ≥ log2(1 + c(2R⋆ − 1)) ≥ cR⋆. p. 1 The constant c is defined as α/(1 + ˆκ), where κˆ is the maximum interference at the selected phase. p. 1 This bound is linear in R⋆ and its factor c can be replaced by a larger one for each admissible α and κˆ. p. 1

Certification via Semidefinite Dual

A semidefinite dual provides a certificate that bounds the retained fraction for a given phase without enumerating phases. Proposition 2 presents this dual rate certificate, stating that any feasible candidate satisfies R(θˆ)/R⋆ ≥ C(θˆ):= R(θˆ) log2(1 + U). p. 1 The exact diagnostic identity shows that the term for omitted interference is the larger of the two terms in the gap analysis. p. 1

Numerical Validation and Robustness

Numerical evaluations confirm these theoretical guarantees across various settings. The analysis shows that omitted interference causes most of the loss, with 90% to 94% of the total gap attributed to this omission. p. 2 Robustness is demonstrated because the shortfall depends only on the deployment through the interference κˆ at the selected phase. p. 2 The retained fraction stays well below the optimum under every single change, including changes in precoders and correlation. p. 2

Conclusion

The minimum desired power gives a tight rate guarantee, set by the interference at the chosen phase, while the weighted sum gives none. A RIS design should therefore certify the selected phase’s rate rather than read fairness from its power. p. 3 The gap persists across precoders and correlation, widening at higher SNR and (at N=6) finer resolution, and persisting below 0. p. 39 up to 512 antennas and 20 users. p. 2

APPENDIX A

PROOF OF PROPOSITION 1

The proof constructs a bounded instance of the channel model with two users where the phase that maximizes the weighted sum drives one user’s rate to zero as epsilon approaches zero. This construction shows no constant c>0 gives R(θˆω) ≥ cR⋆ for all channels. p. 1

APPENDIX B

PROOF OF THEOREM 1

The proof derives the linear bound cR⋆ from the inequality D(θˆ) ≥ αD⋆ ≥ αg⋆ and uses concavity to establish this bound. Sharpness is shown by a two-element channel construction where the bound holds with equality. p. 1

APPENDIX C

PROOF OF PROPOSITION 2

The proof lifts the problem to a real quadratic form and uses weak duality to establish that any feasible candidate satisfies R(θˆ)/R⋆≥C. The chain of inequalities then yields the result R(θˆ) ≥ log2(1 + qU) ≥ q log2(1 + U). p. 3

REFERENCES

[1] M. Di Renzo, A. p. 3 Zappone, M. p. 3 Debbah, M.-S. Alouini, C. p. 3 Yuen, J. de Rosny, and S. p. 3 Tretyakov, “Smart radio environments empowered by reconfigurable intelligent surfaces: How it works, state of research, and road ahead,” IEEE J. p. 3 Sel. Areas Commun. p. 3 p. 3, vol. 38, no. 11, pp. 2450–2525, Nov. p. 3 2020<ref:2611. p. 007548#pg2>.

[8] M. Schubert and H. p. 3 Boche, “Solution of the multiuser downlink beamforming problem with individual SINR constraints,” IEEE Trans. p. 3 Veh. Technol., vol. 53, no. 1, pp. 18–28, 2004<ref:2611. p. 007548#pg2>.

[9] Q. Wu and R. Zhang, “Beamforming optimization for wireless network aided by intelligent reflecting surface with discrete phase shifts,” IEEE Trans. p. 3 Commun., vol. 68, no. 3, pp. 1838–1851, 2020<ref:2611. p. 307548#pg2>.

[10] B. Di, H. Zhang, L. Song, Y. Li, Z. Han, and H. p. 3 V. Poor, “Hybrid beamforming for reconfigurable intelligent surface based multi-user communications: Achievable rates with limited discrete phase shifts,” IEEE J. p. 3 Sel. Areas Commun., vol. 38, no. 8, pp. 1809–1822, 2020<ref:2611. p. 307548#pg2>.

[11] H. Xie, J. Xu, and Y.-F. Liu, “Max-min fairness in IRS-aided multi-cell MISO systems with joint transmit and reflective beamforming,” IEEE Trans. p. 3 Wireless Commun. p. 3 p. 3, vol. 20, no. 2, pp. 1379–1393, 2021<ref:2611. p. 307548#pg2>.

[12] Q.-U.-A. Nadeem, A. Kammoun, A. p. 3 Chaaban, M. p. 3 Debbah, and M. p. 3-S. Alouini, “Asymptotic max-min SINR analysis of reconfigurable intelligent surface assisted MISO systems,” IEEE Trans. p. 3 Wireless Commun., vol. 19, no. 12, pp. 7748–7764, 2020<ref:2611. p. 307548#pg2>.

[13] S.-N. Jin, D.-W. Yue, and H. p. 3 H. Nguyen, “RIS-aided cell-free massive MIMO systems: Joint design of transmit beamforming and phase shifts,” IEEE Syst. p. 3 J., vol. 17, no. 2, pp. 3093–3104, 2023<ref:2611. p. 307548#pg2>.

[14] G. Yan, L. Zhu, and R. p. 3 Zhang, “Passive reflection optimization for IRSaided multicast beamforming with discrete phase shifts,” IEEE Wireless Commun. p. 3 Lett., vol. 12, no. 8, pp. 1424–1428, 2023<ref:2611. p. 307548#pg2>.

[15] N. D. Sidiropoulos, T. p. 3 N. Davidson, and Z. p. 3-Q. Luo, “Transmit beamforming for physical-layer multicasting,” IEEE Trans. p. 3 Signal Process. p. 3, vol. 54, no. 6, pp. 2239–2251, 2006<ref:2611. p. 307548#pg2>.

[17] H. X. Córdova, “Max-min fairness via a deep-cut ellipsoid algorithm for massive MIMO downlink channels,” IEEE Commun. p. 3 Lett., 2026, to appear<ref:2611. p. 307548#pg2>.

[18] D. Bertsimas, V. p. 3 F. Farias, and N. p. 3 Trichakis, “The price of fairness,” Oper. p. 3 Res., vol. 59, no. 1, pp. 17–31, 2011<ref:2611. p. 007548#pg2>.

[19] M. X. Goemans and D. p. 3 P. Williamson, “Improved approximation algorithms for maximum cut and satisfiability problems using semidefinite programming,” J. p. 3 ACM, vol. 42, no. 6, pp. 1115–1145, Nov. p. 3 1995<ref:2611. p. 007548#pg2>.

[20] Z.-q. Luo, W.-k. Ma, A. M.-c. So, Y. Ye, and S. Zhang, “Semidefinite relaxation of quadratic optimization problems,” IEEE Signal Process. p. 3 Mag., vol. 27, no. 3, pp. 20–34, 2010<ref:2611. p. 007548#pg2>.

[21] P. J. Goulart and Y. p. 3 Chen, “Clarabel: An interior-point conic solver,” 2024, arXiv:2405. p. 312762<ref:2611. p. 007548#pg2>.

--- Page 1 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 1

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

Hernan X. Córdova

Abstract—We characterise when desired-power optimisation guarantees the minimum user rate (MUR) in discrete reconfigurable intelligent surface systems with fixed precoding. For the weighted sum of desired powers, a bounded channel exists whose exact maximiser starves a user: no positive guarantee exists. p. 0 For the minimum desired power, the retained fraction of the optimal MUR is at least a sharp constant set by the phase’s power quality and residual interference. p. 0 A semidefinite dual certifies it without enumerating phases. p. 0 Minimum-power phases keep 88% of the optimum under weak reflection and about half under strong reflection; omitted interference causes most of the loss. p. 0

--- Page 2 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 2

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

Index Terms—Reconfigurable intelligent surface, one-bit phases, max–min fairness, approximation guarantee, dual certificate.

--- Page 3 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 3

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

I. INTRODUCTION

RECONFIGURABLE intelligent surfaces (RIS) steer a multiuser downlink by setting the phase of each passive element [1]–[3]<ref:2611.07548#pg2>. A common and convenient choice is to pick the phases that maximise a desired-power objective, which is quadratic in the phases and suits exact, approximate, and quantum solvers<ref:2611. p. 007548#pg2>. With independent streams, the weakest user’s rate depends not only on its desired power but also on the interference the same phases create (Fig. p. 0 1), which a desired-power objective ignores<ref:2611. p. 007548#pg2>. Power-directed designs keep only desired power. p. 0

--- Page 4 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 4

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

Tx: M ant., precoder W RIS: N one-bit θ user k G HH r Hd sk ik desired power sk (own beam wk) interference ik (other beams wj6=k) Fig. p. 0 1. Signal model. p. 0

--- Page 5 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 4

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

bounded, and one such channel rules out any universal factor (Prop. 1, App. A).

--- Page 6 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 4

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

II. MODEL AND OBJECTIVES

Consider M transmit antennas, K single-antenna users, and N passive RIS elements with one-bit phases θ ∈ Θ = (−1, +1)N. The composite channel is H(θ) = Hd + HHr diag(θ)G. p. 0 The noise-normalised desired and interference powers of user k are given by sk = hHk wk2σ2k, ik = Pj6=khHk wj2σ2k. p. 1 The minimum SINR and the MUR are g(θ) = min k sk(θ)/(1 + ik(θ)), R(θ) = log2 (1 + g(θ)) and R⋆ = maxΘ R. p. 1

--- Page 7 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 4

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

We compare two power objectives: Pω(θ) = Xk ωksk(θ), D(θ) = min k sk(θ). Both ignore interference. p. 1 We assume R⋆ > 0 (hence g⋆ > 0 and D⋆ > 0). p. 1

--- Page 8 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 4

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

Problem Objective. We bound R(θˆ)/R⋆ for a phase θˆ chosen by Pω or D, and identify when no positive bound exists. p. 1

--- Page 9 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 4

Sharp Minimum-Rate Guarantees for

Power-Based Discrete RIS Design

III. POWER-TO-RATE GUARANTEES

A. Weighted Power Has No Guarantee

Proposition 1 (Weighted power can starve a user) Fix weights ωk≥0 with Pk ωk=1, and let θˆω ∈ arg maxθ∈Θ Pω(θ) be any exact weighted-power maximiser. There is a oneparameter family of channels of the form (1), with K=2 users, M=2 antennas, N=1 element, bounded coefficients and interference-to-noise ratios ik≤1, indexed by ǫ ∈ (0, 1/2], on which R(θˆω)/R⋆ → 0 as ǫ ↓ 0. p. 1 p. 1

B. A Sharp Guarantee for Minimum Desired Power

Theorem 1 (Sharp power-to-rate guarantee). For any candidate θˆ ∈ Θ, suppose D(θˆ) ≥ αD⋆, where 0 ≤ α ≤ 1. p. 1 Define κˆ = maxk ik(θˆ) and c = α/(1 + ˆκ). p. 1 Then R(θˆ) ≥ log2(1 + c(2R⋆ − 1))≥ cR⋆. p. 1

C. A Computable Certificate

A semidefinite dual turns the guarantee into a certificate that bounds the retained fraction for a given phase without enumerating phases. Proposition 2 presents this dual rate certificate, stating that any feasible candidate satisfies R(θˆ)R⋆ ≥ C(θˆ):= R(θˆ) log2 (1 + U). p. 1

--- Page 10 ---

PREPRINT, TO BE SUBMITTED TO IEEE WIRELESS COMMUNICATIONS LETTERS 4

III. POWER-TO-RATE GUARANTEES

A. Weighted Power Has No Guarantee

Proposition 1 (Weighted power can starve a user) Fix weights ωk≥0 with Pk ωk=1, and let θˆω ∈ arg maxθ∈Θ Pω(θ) be any exact weighted-power maximiser. There is a oneparameter family of channels of the form (1), with K=2 users, M=2 antennas, N=1 element, bounded coefficients and interference-to-noise ratios ik≤1, indexed by ǫ ∈ (0, 1/2], on which R(θˆω)/R⋆ → 0 as ǫ ↓ 0.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Weighted power objective limitation analysis: The system can be designed to explicitly avoid scenarios where the weighted sum of desired powers gives no guarantee, ensuring that rate optimization does not rely on a non-guaranteed objective.

  2. Minimum desired power guarantee implementation: An AI system can select phases based on the minimum desired power (MDP) to retain at least a sharp constant set by the phase’s power quality and residual interference.

  3. Dual certificate utilization for phase selection: The system can employ a semidefinite dual to certify the retained fraction of optimal MUR for any chosen phase without needing to enumerate all possible phases.

Sources

Related papers