A unified framework for synchronization optimization in directed multiplex networks

1. Problem Statement

Real-world complex systems (e.g., financial markets, neural networks, power grids) are rarely isolated; they consist of interconnected multilayer systems where interactions are often directed (asymmetric) and frustrated (involving phase lags or time delays). While synchronization in undirected multiplex networks has been extensively studied, a unified analytical framework for directed multiplex networks that accounts for both structural directionality and dynamical frustration remains largely unexplored.

The core challenges addressed are:

• How to derive optimal natural frequency distributions that maximize synchronization in directed duplex networks.
• How to achieve perfect synchronization at a specific coupling strength in the presence of phase frustration.
• How to optimize network topology (rewiring) and frequency assignment on fixed structures to enhance collective coherence.

2. Methodology

A. Model Formulation

The authors model the system as a duplex network (two layers) of $N$ phase oscillators.

• Dynamics: They employ a generalized Sakaguchi-Kuramoto model for directed layers. The phase evolution for node $i$ in layer $l$ ($l=1,2$) is governed by:
  $$\frac{d\theta_i^{(l)}}{dt} = \omega_i^{(l)} + \frac{K}{N}\sum_{j=1}^N A_{ij}^{(l)} \sin(\theta_j^{(l)} - \theta_i^{(l)} - \alpha^{(l)}) + K \sin(\theta_i^{(l')} - \theta_i^{(l)})$$
  Where:
  • $\omega_i^{(l)}$ is the natural frequency.
  • $A_{ij}^{(l)}$ is the directed adjacency matrix ($A_{ij}=1$ if link $i \to j$ exists).
  • $\alpha^{(l)}$ is the layer-specific phase-lag (frustration) parameter.
  • The last term represents interlayer coupling between mirror nodes.

B. Theoretical Framework: Multiplex Synchrony Alignment Function (MSAF)

To analyze synchronization, the authors linearize the system around the coherent state (small phase differences) and derive steady-state solutions using the Moore-Penrose pseudoinverse of modified Laplacian matrices.

• They define the Multiplex Synchrony Alignment Function (MSAF), denoted as $J(\tilde{\omega}_m^{(l)}, L_m^{(l)})$.
• The global order parameter $R$ is inversely related to the MSAF: $R \approx 1 - \frac{1}{4K^2} \sum J$.
• Key Insight: Minimizing the MSAF maximizes the synchronization order parameter. The MSAF integrates structural properties (Laplacians of both layers) and dynamical characteristics (frequencies and phase lags).

C. Optimization Strategies

The paper proposes three distinct optimization approaches:

1. Optimal Frequency Derivation: Deriving frequency sets that minimize MSAF to enhance synchronization over a broad range of coupling strengths.
2. Perfect Synchronization Derivation: Deriving frequency sets that force the MSAF to zero at a specific target coupling strength $K_p$, even with non-zero phase lags.
3. Network Design & Rearrangement:
   • Rewiring: An accept-reject algorithm to simultaneously rewire links in both layers to minimize MSAF.
   • Frequency Assignment: A greedy swapping algorithm to optimally assign a fixed set of frequencies to nodes on a fixed topology.

3. Key Contributions

1. Unified Analytical Framework: The introduction of the MSAF specifically for directed multiplex networks with phase frustration. This extends previous monolayer and undirected multiplex theories.
2. Two Classes of Optimal Frequencies:
   • Perfect Synchronization Set: Frequencies linearly dependent on the out-degree and frustration parameters ($\omega \propto K_p s_{out} \sin \alpha$). These guarantee $R=1$ at a prescribed $K_p$.
   • Optimal Broadband Set: Frequencies derived to minimize MSAF generally, providing robust synchronization across a wide range of coupling strengths.
3. Inverse Problem Solutions:
   • Demonstration that simultaneous rewiring of directed layers significantly improves synchronization.
   • Proof that optimal frequency arrangement on fixed networks (via greedy swapping) outperforms random allocation.
4. Discovery of Structural-Dynamical Signatures: Identification of three specific correlations inherent in synchrony-optimized directed networks.

4. Key Results

Numerical Validation

Simulations were performed on three duplex configurations: SF-SF (Scale-Free/Scale-Free), SF-ER (Scale-Free/Erdős-Rényi), and ER-ER (Erdős-Rényi/Erdős-Rényi).

• Non-Frustrated Regime ($\alpha=0$): The analytically derived optimal frequencies achieved synchronization at significantly lower coupling strengths compared to standard Uniform, Normal, and Lorentzian distributions.
• Frustrated Regime ($\alpha \neq 0$):
  • The Perfect Frequency Set successfully achieved global synchronization ($R=1$) exactly at the target coupling strength $K_p$ (tested at $K_p=0.3$ and $0.5$), whereas standard distributions failed to synchronize.
  • The Optimal Frequency Set provided robust synchronization across a broad range of $K$.

Structural-Dynamical Correlations

Analysis of the optimized networks revealed three robust correlations:

1. Frequency-Out-Degree Correlation: A strong positive correlation between a node's natural frequency magnitude and its out-degree. High-frequency oscillators tend to be hubs that influence many others. (Correlation coefficients $\approx 0.95$).
2. Neighbor Frequency Correlation: A consistent negative correlation between an oscillator's frequency and the mean frequency of its neighbors (both in-neighbors and out-neighbors). High-frequency nodes are surrounded by low-frequency nodes.
3. Mirror Node Correlation: A significant negative correlation (anti-correlation) between the frequencies of mirror nodes across the two layers. A high-frequency node in Layer 1 is paired with a low-frequency node in Layer 2.

5. Significance and Implications

• Theoretical Advancement: This work bridges the gap between directed network theory and multiplex synchronization, providing a rigorous mathematical tool (MSAF) to handle asymmetry and frustration simultaneously.
• Practical Applications:
  • Power Grids: The findings suggest that stabilizing power grids (modeled as directed multiplex systems) requires aligning generator frequencies with their out-degree (influence) and ensuring anti-correlated frequencies between mirror nodes (e.g., across different market layers or control zones).
  • Neural Systems: The identified correlations offer a blueprint for understanding how brain regions (directed multiplex layers) organize their oscillatory frequencies to maintain cognitive coherence.
  • Resilient Systems: The proposed rewiring and frequency assignment strategies provide actionable methods to design communication and infrastructure networks that are resilient to desynchronization.
• Future Directions: The framework opens avenues for studying higher-order interactions (simplicial complexes) and exotic states like chimera states in directed multiplex settings.

In summary, the paper provides a comprehensive toolkit for designing and optimizing directed multiplex networks, demonstrating that specific alignments between network topology (directionality) and oscillator dynamics (frequencies) are critical for achieving robust global synchronization.