Parameter retrieval
Below is a practical, pragmatic, and logical process diagram/workflow diagram for parameter retrieval based on the comprehensive guide provided. The diagram is structured to cover various methods and cases discussed in the lecture transcript, addressing key concepts like scattering parameters, branching, degrees of freedom, and different retrieval techniques. Since this is a text-based response, I’ll describe the workflow in a structured, hierarchical format that could be easily visualized as a flowchart or diagram. Each section corresponds to a method or decision point from the lecture.
Parameter Retrieval Workflow Diagram
Start
- Input: Engineered material (metamaterial, photonic crystal, mixture, etc.) with thickness ( D ).
- Goal: Determine effective electromagnetic properties (e.g., permittivity (\epsilon), permeability (\mu), refractive index (n), impedance (Z)).
- Data Available: Scattering parameters ((S_{11}, S_{21}, S_{12}, S_{22})) from simulation or lab measurement.
Step 1: Define the Problem
- Question: How to replace the engineered material with a homogeneous slab of thickness (D) that yields the same scattering parameters?
- Challenges:
- Branching (multiple correct solutions due to wave periodicity).
- Degrees of freedom (number of measurements vs. unknowns).
- Decision Point: Choose a parameter retrieval method based on material properties and available data.
Branching Decision: Select Retrieval Method
- S-Parameter Retrieval
- Field Averaging
- Transfer Matrix (T-Matrix) / S-Matrix
- Optimization
- Plane Wave Expansion Method (PWEM)
- Single-Pass Phase Slope
1. S-Parameter Retrieval Workflow
- Input: Scattering parameters ((S_{11}, S_{21})) from simulation or measurement.
- Process:
- Calculate Intermediate Parameters:
- Local reflection coefficient ((r)): From interface properties.
- Single-pass transmission coefficient ((T = e^{j k_0 n D})): Phase accumulated across slab.
- Formulate Scattering Equations:
- (S_{11} = r \frac{1 - T^2}{1 - r^2 T^2})
- (S_{21} = T \frac{1 - r^2}{1 - r^2 T^2})
- Invert Equations:
- Solve for impedance ((Z)): (Z = \sqrt{\frac{(1 + S_{11})^2 - S_{21}^2}{(1 - S_{11})^2 - S_{21}^2}})
- Resolve sign ambiguity: Assume passive material ((\text{Re}(Z) \geq 0)).
- Solve for refractive index ((n)): (n = \frac{1}{k_0 D} \ln(T) + \frac{2\pi m}{k_0 D}), where (m) is branch index.
- Handle Branching:
- If first frequency point:
- Test multiple (m) values.
- Check passivity condition: (\text{Im}(\mu) \geq 0), (\text{Im}(\epsilon) \geq 0).
- Select (m) satisfying all frequencies consistently.
- If subsequent frequency:
- Use previous frequency point ((f_n)) and Taylor series fit.
- Pick root with closest imaginary part to known (n'').
- Select (m) minimizing deviation in real part ((n')).
- If first frequency point:
- Output: (n = n' + j n''), (Z), then (\mu = n Z), (\epsilon = n / Z).
- Calculate Intermediate Parameters:
- Issues: Branching requires additional frequency data or assumptions.
2. Field Averaging Workflow
- Input: Simulation of unit cell with periodic boundaries.
- Process:
- Simulate Unit Cell:
- Apply Bloch wave vector ((k)).
- Compute zero-order Bloch mode fields ((E, H, D, B)) throughout unit cell.
- Average Fields:
- Use Stokes’ Theorem on Maxwell’s equations.
- Compute line integrals around unit cell perimeter for (E) and (H).
- Compute surface integrals for (D) and (B).
- Position fields on Yee grid (staggered).
- Calculate Average Properties:
- Solve constitutive relations: (D = \epsilon E + \kappa H), (B = \mu H + \kappa E).
- Obtain (\epsilon, \mu, \kappa) tensors.
- Correct Numerical Dispersion:
- Compute dispersion factor ((\gamma)): (\gamma = \frac{v_{\text{grid}}}{c}), where (v_{\text{grid}} < c).
- Adjust: (\epsilon_{\text{eff}} = \gamma \epsilon_{\text{avg}}), (\mu_{\text{eff}} = \gamma \mu_{\text{avg}}).
- Output: Effective (\epsilon), (\mu), (\kappa).
- Simulate Unit Cell:
- Advantages: No branching.
- Drawbacks: Computationally intensive, simulation-only.
3. Transfer Matrix (T-Matrix) / S-Matrix Workflow
- Input: Unit cell scattering parameters or layered structure.
- Process:
- T-Matrix Approach:
- Divide unit cell into layers.
- Compute transfer matrix relating fields: (F_{\text{right}} = T F_{\text{left}}).
- Solve eigenvalue problem: Eigenvalues (e^{j k_z D}).
- Extract (n = k_z / k_0), handle branching as in S-parameter method.
- S-Matrix Approach (Stable Alternative):
- Compute scattering matrix for each layer.
- Assemble overall (S)-matrix for unit cell.
- Rearrange into generalized eigenvalue problem: (S A_{\text{in}} = e^{j k_z D} A_{\text{out}}).
- Solve for (k_z), then (n = k_z / k_0).
- Handle branching similarly to S-parameter method.
- Output: Effective (n), then (\mu), (\epsilon) via impedance relations.
- T-Matrix Approach:
- Note: S-Matrix is numerically stable; T-Matrix is simpler but unstable.
4. Optimization Workflow
- Input: Target scattering parameters ((S_{\text{target}})) from simulation or measurement.
- Process:
- Initial Guess: (\mu_0), (\epsilon_0).
- Simulate Homogeneous Slab:
- Use transfer matrix or similar to compute (S_{\text{calc}}).
- Compare:
- Error = (|S_{\text{calc}} - S_{\text{target}}|).
- If error < tolerance, stop.
- Refine Guess:
- Use optimization algorithm (e.g., gradient descent, genetic algorithm).
- Update (\mu), (\epsilon).
- Repeat steps 2-4 until convergence.
- Alternative (Lorentz-Drude):
- Optimize Lorentz-Drude parameters instead of (\mu), (\epsilon).
- Fit frequency sweep to match (S_{\text{target}}).
- Output: Effective (\mu), (\epsilon).
- Advantages: Avoids branching explicitly.
- Drawbacks: Iterative, computationally costly.
5. Plane Wave Expansion Method (PWEM) Workflow
- Input: Anisotropic unit cell, subwavelength regime.
- Process:
- Define Unit Cell: 2D grid for uniaxial structure (e.g., extruded in (z)).
- Set Bloch Wave Vector ((k)):
- Small magnitude ((k \approx 0.01 \cdot 2\pi/a)) for long wavelength limit.
- Direction arbitrary.
- Solve Eigenvalue Problem:
- Call PWEM solver with unit cell and (k).
- Get two low-order bands (E-mode, H-mode).
- Extract Refractive Indices:
- E-mode (E-field out-of-plane): (n_{\text{extraordinary}} = |k| / k_0).
- H-mode (H-field out-of-plane, E in-plane): (n_{\text{ordinary}} = |k| / k_0).
- Form Tensor:
- (\epsilon = \text{diag}(n_{\text{ordinary}}^2, n_{\text{ordinary}}^2, n_{\text{extraordinary}}^2)).
- Output: Effective (\epsilon) tensor.
- Advantages: No branching in subwavelength limit.
- Limitations: Only for subwavelength, non-resonant cases.
6. Single-Pass Phase Slope Workflow
- Input: Transmission phase ((\phi)) over frequency or wavelength sweep.
- Process:
- Measure Phase:
- Obtain (\phi(f)) or (\phi(\lambda)) from (S_{21}).
- Unwrap phase to remove (2\pi) jumps.
- Calculate Slope:
- Frequency: (\frac{d\phi}{df}) via finite difference.
- Wavelength: (\frac{d\phi}{d\lambda}).
- Extract Refractive Index:
- (n = -\frac{c}{2\pi D} \frac{d\phi}{df}) (frequency domain).
- (n = -\frac{\lambda^2}{2\pi D} \frac{d\phi}{d\lambda}) (wavelength domain).
- Output: Effective (n).
- Measure Phase:
- Advantages: Avoids branching by using slope (derivative eliminates (m)).
- Requirements: Multi-point data for slope calculation.
Step 2: Validate Results
- Check Passivity: (\text{Im}(n) \geq 0), (\text{Im}(\mu) \geq 0), (\text{Im}(\epsilon) \geq 0).
- Compare with Bounds: Wiener bounds for non-resonant cases; expect deviations for resonant structures.
- Consistency: Test across frequencies or thicknesses if applicable.
End
- Output: Effective material properties ((\epsilon, \mu, n, Z)).
- Application: Use in simulations or design with simpler homogeneous models.
Notes for Visualization
- Flowchart Structure:
- Start with a single "Input" box.
- Branch into six parallel paths (one for each method).
- Each path has sequential steps (boxes) with decision points (diamonds) for branching or iteration.
- Converge to a "Validate" box, then "Output."
- Color Coding: Use different colors for each method to distinguish them.
- Annotations: Add notes for advantages/drawbacks at each method’s branch.
This workflow captures the lecture’s key methods and decision points pragmatically, addressing branching, computational feasibility, and data requirements. Let me know if you’d like a specific section refined further!