Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link

arXiv:2604.09335 · eess.SP · Submitted 2026-04-10 · 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: "Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link".

Dev: Here's the gist of the paper. Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link is introduced to provide a simple yet powerful approach by maximizing the absolute value of the determinant of an equivalent MIMO channel, which achieves the same determinant value as the optimal unitary BD-RIS.

Rosa: This low-rank solution admits a closed-form symmetric unitary scattering matrix whose rank is exactly twice the channel’s degrees of freedom (2r).

Dev: Furthermore, this Max-Det solution is orders of magnitude faster to compute than existing iterative methods while achieving near-optimal rates in practical scenarios.

Rosa: Problem Formulation and Proxy Metric The work addresses the problem of maximizing the absolute value of the determinant as a proxy for rate, showing that this is asymptotically optimal when the direct Tx-Rx link is obstructed.

Paper summary: Rosa: Looking at this paper, "Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link," it really boils down to finding a computationally efficient way to design these reconfigurable intelligent surfaces, or BD-RISs <ref:2604.09335#pg1>.

Dev: Yeah, the authors show that instead of wrestling with complex iterative algorithms, you can maximize the absolute value of the equivalent MIMO channel's determinant <ref:2604.09335#pg2>. This leads to a closed-form symmetric unitary scattering matrix whose rank is exactly twice the degrees of freedom <ref:2604.09335#pg1>.

Taro: What that means for us on the ground is that we get a solution with fewer connection impedances when we use a q-stem topology, specifically with q equals two r minus one stems <ref:2604.09335#pg2>.

Rosa: And they show this low-rank approach achieves the same determinant value as the optimal unitary BD-RIS configuration <ref:2604.09335#pg1>, which is a strong claim since that's what we usually aim for with these designs.

Dev: They also prove that this rate loss relative to the optimal unitary solution vanishes at high signal-to-noise ratio or when the number of BD-RIS elements becomes large <ref:2604.09335#pg1>.

Taro: So, in simple terms, they found a mathematically simpler way to get near-optimal performance that is also much faster to compute than what we currently use for these kinds of optimization problems <ref:2604.09335#pg1>.

Rosa: It seems the authors are saying that for systems where the direct link between the transmitter and receiver might be blocked, maximizing this determinant is an excellent proxy to maximize the rate <ref:2604.09335#pg2>.

Dev: And they conclude that this solution is computationally efficient, which makes high-performance BD-RIS deployment much more feasible even with large surfaces and limited computational resources <ref:2604.09335#pg1>.

Taro: The main implication for us is that if you're working on a system where the direct channel might be obstructed, this method gives you a viable, low-cost path to near-optimal performance <ref:2604.09335#pg2>.

Conclusion: Rosa: So, this paper is about finding the best way to set up these reconfigurable surfaces, these BD-RISs, so they maximize that determinant value of the signal link between your transmitter and receiver <ref:2604.09335#pg1>.

Dev: Yeah, they introduce a low-rank solution for this problem. It’s not trying to solve the whole massive optimization headache; it’s finding a simpler matrix structure that gives you the same determinant value as the most ideal unitary BD-RIS configuration <ref:2604.09335#pg1>.

Taro: That means they found a low-rank scattering matrix, its rank is twice the degrees of freedom of your system, and that’s a very specific structure you can use <ref:2604.09335#pg1>.

Rosa: Exactly. And they show this solution is super fast to calculate compared to those slow iterative methods we usually have to run <ref:2604.09335#pg1>.

Dev: That speed comes from a closed-form solution, Algorithm one shows you how it works by doing some compact singular value decompositions on two matrices <ref:2604.09335#pg1>. The complexity scales linearly with the number of RIS elements and quadratically with the degrees of freedom <ref:2604.09335#pg1>.

Taro: And they connect this back to real hardware, showing that you can implement this low-rank solution efficiently using a q-stem topology with only q equals two r minus one stems <ref:2604.09335#pg2>. That’s a big deal for physical deployment.

Rosa: It really suggests that when the direct channel is blocked, maximizing this determinant is a very good stand-in for actually maximizing the actual achievable data rate <ref:2604.09335#pg2>.

Dev: The analysis using log majorization proves that as you get more signal power or more RIS elements, the difference between this low-rank solution and the absolute optimal unitary one gets smaller and smaller <ref:2604.09335#pg1>.

Taro: So, even if you aren't at the absolute theoretical maximum yet, this method gets you very close without needing those heavy iterative solvers <ref:2604.09335#pg1>.

Rosa: It makes high-performance BD-RIS deployment much more practical because the computation cost drops by orders of magnitude <ref:2604.09335#pg1>.

Dev: We still need to keep an eye on those scenarios where the direct link is unblocked, though they suggest a different approach for that case <ref:2604.09335#pg1>.

Taro: The next thing to look at is whether this determinant-based optimization works for other metrics, like the mean square error or the Frobenius norm instead of just the determinant <ref:2604.09335#pg1>.

Department of Communications Engineering, Universidad de Cantabria · Signal and System Theory Group, Universität Paderborn · IHP - Leibniz-Institut für Innovative Mikroelektronik, Frankfurt (Oder) · Institute for Communications Technology, Technische Universität Braunschweig

eess.SP

Submitted: 2026-04-10

Updated: 2026-11-05

Comments: 12 pages, 5 figures

Code: https://github.com/IgnacioSantamaria/Code-MaxDet

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link is introduced to provide a simple yet powerful approach by maximizing the absolute value of the determinant…

Key concepts

Determinant Maximization Proxy
Instead of directly maximizing signal rate, the paper uses the absolute value of the determinant as a proxy metric. This is shown to be asymptotically optimal for maximizing the actual communication rate when there is no direct link between the transmitter and receiver, providing a simpler optimization target.
Low-Rank Solution (Rank 2r)
The optimal scattering matrix derived has a rank exactly twice the degrees of freedom (DoF) of the MIMO system. This low-rank structure simplifies computation significantly compared to complex iterative methods and allows for efficient physical implementation using a minimal number of impedances.
q-Stem Topology Implementation
The paper demonstrates that this optimal low-rank solution can be perfectly realized using a q-stem topology with only q = 2r - 1 stems. This means the required number of reconfigurable circuits is minimized, offering a practical advantage over fully connected implementations.
Log-Majorization Theory
This mathematical tool is used to prove that the rate loss between this low-rank Max-Det solution and other optimal solutions (like the unitary one) becomes negligible as the signal strength increases or more RIS elements are added. It establishes that maximizing the determinant closely aligns with maximizing achievable communication rates in practical scenarios.

Terminology

Summary

Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link is introduced to provide a simple yet powerful approach by maximizing the absolute value of the determinant of an equivalent MIMO channel, which achieves the same determinant value as the optimal unitary BD-RIS. p. 0

Where the paper says it

1 Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link Ignacio Santamaria, Senior Member, IEEE, Mohammad Soleymani, Senior Member, IEEE, Jesús Gutiérrez, Member, IEEE, Eduard Jorswieck, Fellow, IEEE Abstract—Beyond-diagonal reconfigurable intelligent surfaces (BD-RISs) significantly improve wireless performance by allowing tunable interconnections among elements, but their design in multiple-input multiple-output (MIMO) systems has so far relied on complex iterative algorithms or suboptimal approximations. This work introduces a simple yet powerful approach: instead of directly maximizing the achievable rate, we maximize the absolute value of the determinant of the equivalent MIMO channel.

This low-rank solution admits a closed-form symmetric unitary scattering matrix whose rank is exactly twice the channel’s degrees of freedom (2r). p. 0
Where the paper says it

We derive a closed-form symmetric unitary scattering matrix whose rank is exactly twice the channel’s degrees of freedom (2r). Remarkably, this low-rank solution achieves the same determinant value as the optimal unitary BD-RIS.

Furthermore, this Max-Det solution is orders of magnitude faster to compute than existing iterative methods while achieving near-optimal rates in practical scenarios. p. 0
Where the paper says it

Moreover, the proposed solution can be perfectly implemented using a q-stem BD-RIS architecture with only q = 2r − 1 stems, requiring a minimum number of reconfigurable circuits. The resulting Max-Det solution is orders of magnitude faster to compute than existing iterative methods while achieving near-optimal rates in practical scenarios.

Problem Formulation and Proxy Metric

The work addresses the problem of maximizing the absolute value of the determinant as a proxy for rate, showing that this is asymptotically optimal when the direct Tx-Rx link is obstructed. The objective is formulated as minimizing a function involving log-determinant and channel power terms, where s represents the rank of the scattering matrix. p. 0

Where the paper says it

This work introduces a simple yet powerful approach: instead of directly maximizing the achievable rate, we maximize the absolute value of the determinant of the equivalent MIMO channel. We derive a closed-form symmetric unitary scattering matrix whose rank is exactly twice the channel’s degrees of freedom (2r).

A key motivation is that when SNR increases, the solution maximizing the SNR-independent term, log det H(Θ)H(Θ)H, converges to a solution that maximizes the rate. p. 2
Where the paper says it

, σhr > 0), the last term of (5) vanishes. Consequently, the solution that maximizes the SNR-independent term, log det H(Θ)H(Θ) H , converges to the solution that maximizes the rate as the SNR increases.

Closed-Form Solution Derivation

Theorem 1 presents a closed-form solution for the Max-Det problem (P2), showing that the optimal rank of the resulting scattering matrix is twice the DoF of the MIMO system, i.e., rank = 2r in P2. p. 0

Where the paper says it

1 Optimal symmetric low-rank BD-RIS configuration maximizing the determinant of a MIMO link Ignacio Santamaria, Senior Member, IEEE, Mohammad Soleymani, Senior Member, IEEE, Jesús Gutiérrez, Member, IEEE, Eduard Jorswieck, Fellow, IEEE Abstract—Beyond-diagonal reconfigurable intelligent surfaces (BD-RISs) significantly improve wireless performance by allowing tunable interconnections among elements, but their design in multiple-input multiple-output (MIMO) systems has so far relied on complex iterative algorithms or suboptimal approximations. This work introduces a simple yet powerful approach: instead of directly maximizing the achievable rate, we maximize the absolute value of the determinant of the equivalent MIMO channel.

Algorithm 1 provides a pseudocode to obtain this optimal Max-Det BD-RIS by performing compact SVDs on F and G, and then on matrix A = VF1, V∗G1. p. 3
Where the paper says it

Theorem 1 provides a closed-form solution for the optimal BD-RIS that maximizes det H(Θ)H(Θ) H , which coincides with the maximum rate solution as the SNR or the number of BD-RIS elements grows. A pseudocode to obtain the optimal Max-Det BD-RIS is shown3 in Algorithm 1.

The complexity of this solution scales linearly with the number of RIS elements and quadratically with the number of DoF of the MIMO channel. p. 3
Where the paper says it

Therefore, the complexity is O(6Mr2 ), which scales linearly with the number of RIS elements and quadratically with the number of DoF of the MIMO channel. Rate penalty of the Max-Det solution In this subsection, we bound the rate gap between the Max-Det solution and the symmetric, but non-unitary, BDRIS that maximizes the rate (cf.

Rate Gap Analysis

An analysis based on log-majorization results is exploited to bound the rate gap between the Max-Det solution and the optimal unitary but not symmetric BD-RIS solution. Lemma 1 shows that if det(H(Θ)) = det(H(Θ')), then the singular values of H(Θ) log-majorize those of H(Θ'). This implies that the optimal solution in terms of rate is the one whose singular values log-majorize those of any other solution with equal determinant. p. 4

Where the paper says it

Lemma 1 shows that, of all equivalent channels with the same determinant, the optimal one in terms of rate is the one whose singular values log-majorize those of any other. This optimal solution is σ ↓ H = σ ↓ F σ ↓ G, obtained by the BD-RIS Θ = VF1VH G1 , which achieves an optimal matching between the eigenmodes of F and those of G.

Lemma 2 bounds the rate gap between the optimal unitary BD-RIS and the optimal symmetric BD-RIS, showing that this gap tends to zero as either SNR or the number of BDRIS elements grows unbounded. p. 4
Where the paper says it

The following result bounds the rate penalty between the optimal symmetric and unitary solutions, showing that it disappears when the SNR or the number of elements of the BD-RIS grows unbounded. [Rate gap].

Implementation via q-Stem Topology

The low-rank Max-Det solution can be implemented efficiently using a q-stem topology with just q = 2r − 1 stems. Lemma 3 proves that there exists a B with the q-stem structure such that det(FΘBGH) = det(FΘlrGH), meaning this solution can be implemented with the minimum possible number of impedances. p. 6

Where the paper says it

If q = 2r−1, there exists a B with corresponding scattering matrix ΘB = (IM + jZ0B) −1 (IM − jZ0B) such that | det(FΘBGH)| = | det(FΘlrGH)|. Therefore, ΘB is a Max-Det solution that can be implemented with the q-stem topology.

Simulation Validation

Numerical simulations demonstrate that when the direct channel is blocked, the rate achieved by the Max-Det solution is very close to that obtained with iterative algorithms that maximize rate, while exhibiting a computational cost orders of magnitude lower. In scenarios with an unblocked direct channel, the Max-Det solution with phase correction shows improvements compared to the standard Max-Det, but its determinant is not a good proxy for rate in certain regimes. p. 7

Where the paper says it

Here, the phase-corrected Max-Det solution shows some improvement with respect to the standard Max-Det, but there is still a noticeable difference with the MO solution. In this regime, the determinant of the BD-RIS channel is not a good proxy for rate, which is to be expected.

The q-stem topology perfectly recovers the optimal Theta without any penalty with respect to the fully connected implementation when q = 2r − 1. p. 8
Where the paper says it

We observe in Fig. 5 that, with exactly q = 2r − 1 = 7 stems, the q-stem topology perfectly recovers the optimal Θ without any penalty with respect to the fullyconnected implementation.

Conclusions and Future Work

The work concludes that for a blocked direct channel, maximizing the determinant is an excellent proxy to maximize rate, and the resulting solution is computationally efficient. The rank of the optimal scattering matrix being twice the number of DoFs translates into fewer connection impedances when using q-stem topology. p. 8

Where the paper says it

An analysis of the solution allows us to bound the gap between the Max-Det solution and the solution that maximizes rate, showing that the gap decreases as the SNR or the number of BD-RIS elements increases. The rank of the optimal scattering matrix is twice the number of DoFs of the MIMO link, a fact that translates into fewer connection impedances when BDRIS is implemented in the q-stem topology.

Future research is suggested to study whether closed-form BD-RIS solutions can be obtained for other cost functions besides the determinant, such as the Frobenius norm or mean squared error. p. 8
Where the paper says it

The solution can be extended to the multiple access channel, although a detailed analysis of its performance in multi-user scenarios is left for future work. It is also of interest for future research to study whether it is possible to obtain closed-form BD-RIS solutions for other cost functions besides the determinant, such as the Frobenius norm of the equivalent channel or the mean squared error.

Pruning redundant reasoning tokens

The paper introduces a simple yet powerful approach by maximizing the absolute value of the determinant of an equivalent MIMO channel. This low-rank solution achieves the same determinant value as the optimal unitary BD-RIS. p. 0

Where the paper says it

We derive a closed-form symmetric unitary scattering matrix whose rank is exactly twice the channel’s degrees of freedom (2r). Remarkably, this low-rank solution achieves the same determinant value as the optimal unitary BD-RIS.

Using log-majorization theory, it is proven that the rate loss relative to the optimal unitary BD-RIS vanishes at high signal-to-noise ratio (SNR) or when the number of BD-RIS elements becomes large. p. 0
Where the paper says it

Remarkably, this low-rank solution achieves the same determinant value as the optimal unitary BD-RIS. Using log-majorization theory, we prove that the rate loss relative to the optimal unitary BD-RIS vanishes at high signal-to-noise ratio (SNR) or when the number of BD-RIS elements becomes large.

Moreover, the proposed solution can be perfectly implemented using a q-stem BD-RIS architecture with only q = 2r − 1 stems, requiring a minimum number of reconfigurable circuits. p. 0
Where the paper says it

Using log-majorization theory, we prove that the rate loss relative to the optimal unitary BD-RIS vanishes at high signal-to-noise ratio (SNR) or when the number of BD-RIS elements becomes large. Moreover, the proposed solution can be perfectly implemented using a q-stem BD-RIS architecture with only q = 2r − 1 stems, requiring a minimum number of reconfigurable circuits.

How the training signal is built

The equivalent channel is modeled as H(Θ) = FΘGH where G ∈ C Nt×M is the channel from the Tx to the BDRIS, F ∈ C Nr×M is the channel from the BD-RIS to the Rx, and Θ is the M × M BD-RIS matrix. For a symmetric MIMO system with Nt = Nr = r, this channel can be expressed using SVD factorizations of F and G. p. 8

Where the paper says it

The generalization to asymmetric MIMO systems is straightforward. Substituting in (1) the SVD factorizations of F and G given in (2) and (3), the equivalent channel can be expressed as H(Θ) = UF ΣF 0r×(M−r) VH F ΘVG ΣG 0(M−r)×r UH G .

Where the gains come from

The gains in computational efficiency stem from the closed-form solution derived in Theorem 1, which is orders of magnitude faster to compute than existing iterative methods. The implementation advantage comes from the inherent low-rank structure of the scattering matrix, enabling efficient realization using a minimal number of impedance elements when implemented under the q-stem topology. p. 1

Where the paper says it

Interestingly, the optimal scattering matrix is shown to have a rank exactly equal to twice the number of degrees of freedom (DoF) of the MIMO channel. From an implementation standpoint, this inherent low-rank structure enables efficient realization using a minimal number of impedance elements when the BD-RIS is implemented under the q-stem topology [15]–[17].

The rank of the optimal scattering matrix is twice the number of degrees of freedom (DoF) of the MIMO channel (i.e., rank = 2r with r = min(Nt, Nr)). p. 1

Where the paper says it

• We show that the problem of maximizing the determinant with a symmetric BD-RIS has a closed-form solution whose rank is twice the number of degrees of freedom (DoF) of the channel (i. , rank = 2r with r = min(Nt, Nr)).

Compared to the optimal unitary BDRIS, a unitary and symmetric solution simply doubles the rank of the solution. p. 1
Where the paper says it

, rank = 2r with r = min(Nt, Nr)). Compared to the optimal unitary BDRIS, a unitary and symmetric solution simply doubles the rank of the solution.

Both solutions yield equivalent MIMO channels, whose determinants have the same absolute value. p. 1
Where the paper says it

Compared to the optimal unitary BDRIS, a unitary and symmetric solution simply doubles the rank of the solution. Both solutions yield equivalent MIMO channels, whose determinants have the same absolute value.

The optimal scattering matrix is shown to have a rank exactly equal to twice the number of degrees of freedom (DoF) of the MIMO channel. This fact translates into fewer connection impedances when BDRIS is implemented in the q-stem topology. p. 8

Where the paper says it

An analysis of the solution allows us to bound the gap between the Max-Det solution and the solution that maximizes rate, showing that the gap decreases as the SNR or the number of BD-RIS elements increases. The rank of the optimal scattering matrix is twice the number of DoFs of the MIMO link, a fact that translates into fewer connection impedances when BDRIS is implemented in the q-stem topology.

The low-rank Max-Det solution can be implemented optimally with a q-stem topology with just q = 2r − 1 stems. This corroborates Corollary 2 in [18]. p. 4

Where the paper says it

, (11) where ρ = P/(Ntσ 2 ) is the per-antenna SNR, and σf1 (σfr ) and σg1 (σgr ) are the max (min) singular values of F and G, respectively Furthermore, ∆R → 0 as ρ → ∞ or M → ∞. A corollary that follows from Lemma 1 is that, among all equivalent channels with maximum determinant, the worstcase in terms of rate is the channel whose singular values are log-majorized by any other solution with equal determinant.

The analysis based on log-majorization results is exploited to bound the rate gap between the Max-Det solution and the optimal unitary but not symmetric BDRIS solution. This gap tends to zero either when the SNR or the number of BD-RIS elements increases. p. 1

Where the paper says it

• An analysis based on log-majorization results is exploited to bound the rate gap between the Max-Det solution and the optimal unitary but not symmetric solution. This gap tends to zero either when the SNR or when the number of BD-RIS elements tends to infinity.

Therefore, in these asymptotic regimes, the proposed solution maximizes the achievable rate. p. 1
Where the paper says it

This gap tends to zero either when the SNR or when the number of BD-RIS elements tends to infinity. Therefore, in these asymptotic regimes, the proposed solution maximizes the achievable rate.

The optimal symmetric BD-RIS in Theorem 1 gives an equivalent channel whose singular values are log-majorized by those of the optimal unitary solution with optimal eigenmode matching. This means that among all rank-r matrices satisfying ΘHΘ ⪯ IM, Θ = VF1VH G1 maximizes the rate and also maximizes the determinant (and, incidentally, also the Frobenius norm) of the equivalent channel. p. 4

Where the paper says it

This optimal solution is σ ↓ H = σ ↓ F σ ↓ G, obtained by the BD-RIS Θ = VF1VH G1 , which achieves an optimal matching between the eigenmodes of F and those of G. Therefore, among all rank-r matrices satisfying ΘHΘ ⪯ IM, Θ = VF1VH G1 maximizes the rate and also maximizes the determinant (and, incidentally, also the Frobenius norm) of the equivalent channel.

p. 4
Where the paper says it

This optimal solution is σ ↓ H = σ ↓ F σ ↓ G, obtained by the BD-RIS Θ = VF1VH G1 , which achieves an optimal matching between the eigenmodes of F and those of G. Therefore, among all rank-r matrices satisfying ΘHΘ ⪯ IM, Θ = VF1VH G1 maximizes the rate and also maximizes the determinant (and, incidentally, also the Frobenius norm) of the equivalent channel.

The penalty in rate between the optimal unitary solution and the symmetric solution can be bounded by Lemma 2, showing that it disappears when SNR or M increases unbounded. The numerical simulations confirm that maximizing the determinant is an excellent proxy to maximize rate in a scenario without a direct channel. p. 7

Where the paper says it

The results in Fig. 3 show that maximizing the determinant is an excellent proxy to maximize rate in a scenario without a direct channel.

The solution in Theorem 1 is only optimal if the direct channel is blocked. When a direct channel between Tx and Rx, Hd, exists, a closed-form solution for the BD-RIS that maximizes the determinant (let alone the rate) is unknown and seems unlikely. p. 5

Where the paper says it

A suboptimal solution for unblocked direct channels The solution in Theorem 1 is only optimal if the direct channel is blocked. When a direct channel between Tx and Rx, Hd, exists, the equivalent channel is H(Θ) = Hd + FΘGH, and a closed-form solution for the BD-RIS that maximizes the determinant (let alone the rate) is unknown and seems unlikely.

Therefore, to maximize rate with an unblocked direct link, it is still necessary to apply optimization algorithms such as those proposed in [9], [10]. p. 5
Where the paper says it

When a direct channel between Tx and Rx, Hd, exists, the equivalent channel is H(Θ) = Hd + FΘGH, and a closed-form solution for the BD-RIS that maximizes the determinant (let alone the rate) is unknown and seems unlikely. Therefore, to maximize rate with an unblocked direct link, it is still necessary to apply optimization algorithms such as those proposed in [9], [10].

To maximize rate with an unblocked direct link, it is possible to propose a suboptimal solution to the maximum rate problem with an unblocked direct link, which offers a competitive solution. This involves choosing a phase term that maximizes capacity by solving Problem P3. p. 7

Where the paper says it

To find the phase term that maximizes capacity (cf. Problem P3 in (13)), we applied numerical optimization.

The low-cost method [8] degrades somewhat in the region where the direct channel is very weak, but not significantly. The BD-RIS link determines the performance in this region. p. 7

Where the paper says it

The low-cost method [8] degrades somewhat in this region, but not significantly. Clearly, the BD-RIS link determines the performance in this region.

p. 7
Where the paper says it

The low-cost method [8] degrades somewhat in this region, but not significantly. Clearly, the BD-RIS link determines the performance in this region.

p. 7
Where the paper says it

The low-cost method [8] degrades somewhat in this region, but not significantly. Clearly, the BD-RIS link determines the performance in this region.

The q-stem topology perfectly recovers the optimal Theta without any penalty with respect to the fullyconnected implementation when q = 2r − 1. p. 8
Where the paper says it

We observe in Fig. 5 that, with exactly q = 2r − 1 = 7 stems, the q-stem topology perfectly recovers the optimal Θ without any penalty with respect to the fullyconnected implementation.

p. 8
Where the paper says it

We observe in Fig. 5 that, with exactly q = 2r − 1 = 7 stems, the q-stem topology perfectly recovers the optimal Θ without any penalty with respect to the fullyconnected implementation.

p. 8
Where the paper says it

We observe in Fig. 5 that, with exactly q = 2r − 1 = 7 stems, the q-stem topology perfectly recovers the optimal Θ without any penalty with respect to the fullyconnected implementation.

p. 8
Where the paper says it

We observe in Fig. 5 that, with exactly q = 2r − 1 = 7 stems, the q-stem topology perfectly recovers the optimal Θ without any penalty with respect to the fullyconnected implementation.

The rank-r unitary solution is not strictly unitary, but a matrix with orthogonal columns and with r singular values equal to 1 and the remaining singular values equal to 0. We use this term loosely, understanding it to be a matrix with orthogonal columns and with r singular values equal to 1 and the remaining singular values equal to 0. p. 3

Where the paper says it

4Strictly speaking, a rank-r matrix is not unitary. We use this term loosely, understanding it to be a matrix with orthogonal columns and with r singular values equal to 1 and the remaining singular values equal to 0.

The optimal symmetric BD-RIS in Theorem 1 gives an equivalent channel whose determinant is Dmax, where Dmax = det(ΣF) det(ΣG). The solution of Algorithm 1 generates an equivalent channel whose determinant is Dmax, and the rank-r unitary solution also attains Dmax. p. 3

Where the paper says it

≥ σg[r] , which we shall also denote as σ ↓ F and σ ↓ G. The solution of Algorithm 1 generates an equivalent channel whose determinant is Dmax, and the rank-r unitary4 , but not symmetric, solution Θ = VF1VH G1 also attains Dmax.

The resulting Max-Det solution exhibits orders of magnitude lower computational complexity than existing iterative methods while achieving near-optimal rates in practical scenarios. This makes high-performance BDRIS deployment feasible even with large surfaces and limited computational resources. p. 0

Where the paper says it

The resulting Max-Det solution is orders of magnitude faster to compute than existing iterative methods while achieving near-optimal rates in practical scenarios. This makes high-performance BDRIS deployment feasible even with large surfaces and limited computational resources.

The results shown are the average of 200 independent simulations (channel realizations) keeping the Tx, Rx and BD-RIS positions fixed. The objective of the first experiment is to compare the solution that maximizes the rate and the proposed solution that maximizes the absolute value of the determinant of an equivalent channel. p. 7

Where the paper says it

(18) The results shown are the average of 200 independent simulations (channel realizations) keeping the Tx, Rx and BD-RIS positions fixed. Perfomance assessment The objective of the first experiment is to compare the solution that maximizes the rate and the proposed solution that maximizes the absolute value of the determinant of the equivalent channel (Algorithm 1).

The comparison in Fig. 3 shows that maximizing the determinant is an excellent proxy to maximize rate in a scenario without a direct channel. p. 7

Where the paper says it

The results in Fig. 3 show that maximizing the determinant is an excellent proxy to maximize rate in a scenario without a direct channel.

The differences with the optimal unitary BD-RIS are minimal, and at high SNRs and when the number of BD-RIS elements increases, the differences are negligible. p. 7
Where the paper says it

As can be seen, the differences with the optimal unitary BD-RIS are minimal. Moreover, at high SNRs and when the number of BD-RIS elements increases, the differences are negligible, thus validating the results of Remark 2.

The low-cost method [8] provides practically identical results to those of the MO algorithm in a region where the direct channel is very weak. The BD-RIS link determines the performance in this region. The q-stem topology perfectly recovers the optimal Theta without any penalty with respect to the fullyconnected implementation when q = 2r − 1.

The equivalent channel is H(Θ) = aHd + FΘGH, and we consider scenarios with and without a direct channel. The SNR at the receiver depends on the equivalent MIMO channel and, consequently, on the BD-RIS. p. 7

Where the paper says it

Problem P3 in (13)), we applied numerical optimization. 7Note that the SNR at the Rx depends on the equivalent MIMO channel and, consequently, on the BD-RIS.

The simulation results compare three methods in a 4×4 MIMO scenario without a direct channel. The comparison highlights that the Max-Det solutions provide rates very close to those obtained with iterative algorithms that maximize rate, with a computational cost orders of magnitude lower. p. 8

Where the paper says it

The rank of the optimal scattering matrix is twice the number of DoFs of the MIMO link, a fact that translates into fewer connection impedances when BDRIS is implemented in the q-stem topology. The numerical simulations show that, when the direct channel is blocked, the rate achieved by the Max-Det solution is very close to that obtained with iterative algorithms that maximize rate, with a computational cost orders of magnitude lower.

The low-cost method [8] degrades somewhat in this region, but not significantly. p. 7

Where the paper says it

In the first region (green shaded region), where the direct channel is very weak, the Max-Det solutions (with or without phase correction) provide practically identical results to those of the MO algorithm, which is consistent with the results presented in Fig. The low-cost method [8] degrades somewhat in this region, but not significantly.

The BD-RIS link determines the performance in this region. The q-stem topology perfectly recovers the optimal Theta without any penalty with respect to the fullyconnected implementation when q = 2r − 1.

The structure of the susceptance matrix B in the q-stem topology is B = B11 BT12 Γ, where B11 = BT11 is a q × q symmetric and real matrix, B12 is a (M − q) × q block and Γ = diag(bqq+1,... bM)) is a (M − q) × (M − q) diagonal matrix. p. 6

Where the paper says it

The susceptance matrix of the q-stem topology is structured as B = B11 BT 12 B12 Γ , (15) where B11 = BT 11 is a q × q symmetric and real matrix, B12 is a (M − q) × q block and Γ = diag(bqq+1, . bM)) is a (M − q) × (M − q) diagonal matrix.

Theorem 1 and Corollary 2 in [15] prove that the 2r low-rank Max-Det solution can be implemented by a q-stem topology with q = 2r−1 stems. This means that det(FΘBGH) = det(FΘlrGH). p. 6

Where the paper says it

Let B be a real and symmetric M × M matrix with the q-stem structure as in (15), and let Θlr = QQT be the rank-2r MaxDet solution. If q = 2r−1, there exists a B with corresponding scattering matrix ΘB = (IM + jZ0B) −1 (IM − jZ0B) such that | det(FΘBGH)| = | det(FΘlrGH)|.

The linear system in (17) has 2Mr equations and ν unknowns, and when q = 2r − 1, the number of unknowns is larger than the number of equations, thus ensuring a solution exists because W has full column rank generically. The complexity is O(6Mr2), which scales linearly with the number of RIS elements and quadratically with the number of DoF. p. 3

Where the paper says it

Algorithm 1 involves two compact SVDs of matrices F and G, with a computational cost of O(Mr2 ) each, and a compact SVD of matrix A of dimensions M × 2r, with a computational cost of O(M(2r) 2 ). Therefore, the complexity is O(6Mr2 ), which scales linearly with the number of RIS elements and quadratically with the number of DoF of the MIMO channel.

The singular values of the Gram matrix AHA in (21) are 1 ± cos(θk) for k = 1,..., r, where θk are the principal angles between the subspaces spanned by VF1 and V∗G1. p. 9

Where the paper says it

The angles 0 ≤ θk ≤ π/2, k = 1, . , r are the principal angles between the subspaces spanned by VF1 and V∗ G1 [41, Chapter 9].

The columns of the unitary matrix V in (21) correspond to singular values 1 ± cos(θk) for k = 1,. p. 9
Where the paper says it

The singular values of the Gram matrix AHA in (21) are 1 ± cos(θk) for k = 1, . The columns of the unitary matrix V in (21) are vk = 1 √ 2 pk rk , corresponding to singular values 1 + cos(θk) for k = 1, .

.., r, and corresponding to singular values 1 − cos(θk) for k = r + 1,. p. 9
Where the paper says it

The singular values of the Gram matrix AHA in (21) are 1 ± cos(θk) for k = 1, . The columns of the unitary matrix V in (21) are vk = 1 √ 2 pk rk , corresponding to singular values 1 + cos(θk) for k = 1, .

., 2r.

The analysis of the Gram matrix AHA shows that the contribution from each VH F1 u1,ku T1,kVG1 is cos(θk/2)2pkrT k. Similarly, there is a negative contribution from each VH F1 u2,ku T2,kVG1 which is - sin(θk/2)2pkrT k. p. 9

Where the paper says it

The fact that ∥u1,k∥ = 1 can easily be checked. Similarly, the k-th vector of U2, which corresponds to the singular value σk = p 1 − cos(θk), can be expressed as u2,k = 1 p 2(1 − cos(θk)) VF1 pk − V∗ G1 rk .

The analysis shows that the optimal symmetric BD-RIS in Theorem 1 gives an equivalent channel whose singular values are log-majorized by those of the optimal unitary solution with optimal eigenmode matching. This means that among all rank-r matrices satisfying ΘHΘ ⪯ IM, Θ = VF1VH G1 maximizes the rate and also maximizes the determinant (and, incidentally, also the Frobenius norm) of the equivalent channel.

The analysis shows that among all equivalent channels with maximum determinant, the worst-case in terms of rate is the channel whose singular values are log-majorized by any other solution with equal determinant. This implies that the optimal symmetric BD-RIS in Theorem 1 gives an equivalent channel whose singular values are log-majorized by those of the optimal unitary solution with optimal eigenmode matching. p. 4

Where the paper says it

Therefore, among all rank-r matrices satisfying ΘHΘ ⪯ IM, Θ = VF1VH G1 maximizes the rate and also maximizes the determinant (and, incidentally, also the Frobenius norm) of the equivalent channel. The optimal symmetric BD-RIS in Theorem 1 gives an equivalent channel whose singular values are log-majorized by those of the optimal unitary solution with optimal eigenmode matching.

The rate gap between the optimal unitary BD-RIS and the optimal symmetric BD-RIS can be bounded as ∆R ≤ r log (1 + ρσ2 fr σgr)σ2 f1 σg1 (1 + ρσ2 f1 σg1)σ2 fr σgr! This gap goes to zero when ρ → ∞ or M → ∞. p. 5

Where the paper says it

From this result, the inequality (11) follows. It is easy to check that ∆R goes to zero when ρ → ∞ or M → ∞ (since all singular values concentrate around √ M).

The bound provided in Lemma 2 suggests that the penalty in rate is generally negligible, which will be confirmed in Sec. V. The numerical simulations confirm this, showing that at high SNRs and when the number of BD-RIS elements increases, the differences are negligible. p. 7

Where the paper says it

As can be seen, the differences with the optimal unitary BD-RIS are minimal. Moreover, at high SNRs and when the number of BD-RIS elements increases, the differences are negligible, thus validating the results of Remark 2.

The low-cost method [8] degrades somewhat in this region, but not significantly. The q-stem topology perfectly recovers the optimal Theta without any penalty with respect to the fullyconnected implementation when q = 2r − 1.

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. Closed-form optimization for Beyond-Diagonal Reconfigurable Intelligent Surfaces (BDRIS) can be implemented much faster than existing iterative methods while achieving near-optimal rates in practical scenarios because the proposed solution admits a closed form solution.

  2. The system can utilize a q-stem BD-RIS architecture with only q equals two r minus one stems, which requiring a minimum number of reconfigurable circuits, reducing hardware complexity.

  3. The improved AI system can maximize the absolute value of the determinant of the equivalent MIMO channel as a proxy for rate, and this approach is shown to be asymptotically optimal in certain cases when SNR or the number of BD-RIS elements grows.

  4. The system can perform real-time channel optimization by utilizing Algorithm 1, which requires only two compact SVDs of F and G and one compact SVD of matrix A, resulting in a computational cost scaling linearly with the number of RIS elements.

  5. For blocked direct channels, the system can propose a suboptimal but competitive solution by maximizing the Frobenius norm of the equivalent channel using Algorithm 1 or related methods, which offers a competitive solution.

Abstract

Beyond-diagonal reconfigurable intelligent surfaces (BD-RISs) significantly improve wireless performance by allowing tunable interconnections among elements, but their design in multiple-input multiple-output (MIMO) systems has so far relied on complex iterative algorithms or suboptimal approximations. This work introduces a simple yet powerful approach: instead of directly maximizing the achievable rate, we maximize the absolute value of the determinant of the equivalent MIMO channel. We derive a closed-form symmetric unitary scattering matrix whose rank is exactly twice the channel's degrees of freedom (2r). Remarkably, this low-rank solution achieves the same determinant value as the optimal unitary BD-RIS. Using log-majorization theory, we prove that the rate loss relative to the optimal unitary BD-RIS vanishes at high signal-to-noise ratio (SNR) or when the number of BD-RIS elements becomes large. Moreover, the proposed solution can be perfectly implemented using a q-stem BD-RIS architecture with only q=2r-1 stems, requiring a minimum number of reconfigurable circuits. The resulting Max-Det solution is orders of magnitude faster to compute than existing iterative methods while achieving near-optimal rates in practical scenarios. This makes high-performance BD-RIS deployment feasible even with large surfaces and limited computational resources.

Sources

Related papers