Inverse Design of Multi-Layered Manufacturable Pixelated Diplexers Through Optimized Geometrical Configuration and Meshing Strategy in MoM

Jeffrey S. Walling; Jungmin Lee; Woojun Lee

arxiv: 2412.17617 · v2 · submitted 2024-12-23 · ⚛️ physics.app-ph

Inverse Design of Multi-Layered Manufacturable Pixelated Diplexers Through Optimized Geometrical Configuration and Meshing Strategy in MoM

Woojun Lee , Jungmin Lee , Jeffrey S. Walling This is my paper

Pith reviewed 2026-05-23 06:58 UTC · model grok-4.3

classification ⚛️ physics.app-ph

keywords inverse designpixelated surfacesmethod of momentsmatrix reconstructionstochastic optimizationmultilayer diplexerelectromagnetic design

0 comments

The pith

A matrix reconstruction technique and stochastic pixel-flipping search enable rapid inverse design of multilayer pixelated diplexers using the method of moments.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops an inverse design framework that accelerates electromagnetic simulations of multilayered pixelated surfaces. It does this by reconstructing the interaction matrix for each new pixel arrangement instead of solving the full system anew, while applying GPU acceleration to offset the added cost of extra layers. A stochastic search that flips multiple pixels at once then locates configurations whose frequency responses match the targets. The method is shown to produce a working diplexer whose two channels fall in the 5.23-5.94 GHz and 6.17-7.15 GHz ranges, with the outcome checked by an independent full-wave solver.

Core claim

The central claim is that pre-labeling MoM matrix entries as inter-pixel or inner-pixel allows fast reconstruction of the system matrix for every pixel change, that GPU acceleration counters the cubic growth in cost with added layers, and that a stochastic multi-pixel flipping algorithm can locate pixel patterns satisfying prescribed multiport responses, as verified by a diplexer whose -3 dB bands are 5.23-5.94 GHz and 6.17-7.15 GHz.

What carries the argument

The matrix reconstruction technique that pre-labels entries as inter-pixel or inner-pixel to avoid full re-assembly of the MoM system for each pixel variation.

If this is right

Pixelated multilayer surfaces with multiple ports can be synthesized in far less time than repeated full MoM solves would require.
The cubic scaling penalty from additional layers is offset enough to keep the search tractable.
Manufacturable pixelated diplexers can be generated directly from frequency-response specifications.
The same acceleration and search steps apply to other multiport pixelated devices beyond the demonstrated diplexer.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The reconstruction approach could be adapted to other integral-equation solvers that store dense interaction matrices.
Embedding explicit fabrication rules inside the flipping search might further reduce post-design editing.
The framework suggests a path to automated design of pixelated surfaces with more than two channels or with integrated matching networks.

Load-bearing premise

The stochastic multi-pixel flipping search is assumed to reach designs that meet the target responses without becoming trapped in poor local solutions, and the reconstructed matrices are assumed to retain full electromagnetic accuracy.

What would settle it

A full-wave simulation of the reported diplexer geometry that places either channel outside the stated 5.23-5.94 GHz or 6.17-7.15 GHz -3 dB bandwidths would falsify the effectiveness claim.

Figures

Figures reproduced from arXiv: 2412.17617 by Jeffrey S. Walling, Jungmin Lee, Woojun Lee.

**Figure 2.** Figure 2: Classification of basis functions into "inter-pixel," "inner-pixel," and [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Figure 3: Dielectric substrate configuration. The substrate material is Rogers [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

**Figure 6.** Figure 6: Inversely designed multilayered pixelated diplexers and their [PITH_FULL_IMAGE:figures/full_fig_p004_6.png] view at source ↗

read the original abstract

This paper presents a fast inverse design framework for complex multilayered, multiport pixelated surfaces - a class of structures largely unexplored in current research. Leveraging a method-of-moments (MoM) electromagnetic (EM) solver, the framework enables the rapid synthesis of pixelated device designs. A novel matrix reconstruction technique, based on pre-labeling matrix entries as "inter-pixel" or "inner-pixel," accelerates simulations for each variation of the pixelated structure. To mitigate the cubic increase in computation time associated with additional layers, GPU acceleration is employed. Further enhancing convergence speed, a stochastic multi-pixel flipping search algorithm is integrated into the framework. The effectiveness of this approach is demonstrated through the design of a diplexer achieving a -3-dB bandwidth for one channel spanning 5.23-5.94 GHz and another covering 6.17-7.15 GHz, validated by a full-wave solver.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper puts together matrix pre-labeling, GPU handling, and stochastic multi-pixel search for inverse design of multilayer pixelated diplexers, but supplies no error metrics or optimizer comparisons to show the speedups are real and accurate.

read the letter

The core contribution is an engineering package: label MoM matrix entries ahead of time as inter-pixel or inner-pixel so you can rebuild the impedance matrix quickly after a pixel flip, push the multilayer case onto GPU to avoid cubic blowup, and run a stochastic multi-pixel flip search to hunt for a layout that meets target responses. They close with one diplexer example whose bands (5.23-5.94 GHz and 6.17-7.15 GHz at -3 dB) match a separate full-wave check. That combination is not something the cited prior work already describes, so the assembly itself is new for this application. It addresses a practical bottleneck in pixelated RF design where each candidate layout used to require a full re-solve. The external validation on the final design is a plus. The gaps are straightforward. The abstract gives no quantitative check on how much error the pre-labeled reconstruction introduces, no wall-clock comparisons against plain MoM or other search methods, and no data on how many search runs were needed or whether the stochastic flips routinely escape poor local solutions. Those two assumptions flagged in the stress-test note—exact matrix fidelity and reliable convergence—are therefore untested on the evidence shown. Without them the reported bands cannot be confidently credited to the framework rather than to an unrepresentative run or an inaccurate simulator. This is aimed at RF engineers already working with MoM on pixelated or reconfigurable surfaces who want faster iteration on multilayer devices. A reader in that group could extract the matrix-labeling and GPU tricks even if the search part needs more scrutiny. I would send it for peer review once the full text supplies the missing accuracy and timing numbers; the idea is concrete enough to be worth checking.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes a fast inverse design framework for complex multilayered pixelated surfaces using a method-of-moments (MoM) solver. It introduces a matrix reconstruction technique based on pre-labeling entries as inter-pixel or inner-pixel to accelerate simulations of pixel variations, employs GPU acceleration to address cubic scaling with layers, and integrates a stochastic multi-pixel flipping search algorithm. Effectiveness is demonstrated via a diplexer achieving -3 dB bandwidths of 5.23-5.94 GHz and 6.17-7.15 GHz, validated by a full-wave solver.

Significance. If the matrix reconstruction preserves full MoM accuracy across configurations and the stochastic search reliably identifies designs meeting target responses, the approach could enable practical inverse design of multilayer pixelated EM devices, a class noted as largely unexplored. The combination of acceleration techniques and search strategy addresses computational bottlenecks in pixelated structure optimization.

major comments (2)

[Abstract] Abstract: The central performance claim (diplexer bandwidths validated by full-wave solver) rests on the matrix reconstruction exactly reproducing the impedance matrix for every pixel configuration and the stochastic search locating a suitable design. No quantitative error metrics (e.g., relative error in reconstructed vs. full MoM matrices), convergence data, or number of evaluated designs are reported, leaving open whether the reported bands arise from the framework or from unrepresentative simulation.
[Abstract] Abstract: The stochastic multi-pixel flipping search is presented as enhancing convergence, yet no details are given on run-to-run variability, comparison against gradient-based or other global optimizers, or safeguards against local minima. This assumption is load-bearing for attributing the achieved bandwidths to the proposed method.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below with clarifications based on the framework's design and indicate revisions incorporated into the updated version.

read point-by-point responses

Referee: [Abstract] Abstract: The central performance claim (diplexer bandwidths validated by full-wave solver) rests on the matrix reconstruction exactly reproducing the impedance matrix for every pixel configuration and the stochastic search locating a suitable design. No quantitative error metrics (e.g., relative error in reconstructed vs. full MoM matrices), convergence data, or number of evaluated designs are reported, leaving open whether the reported bands arise from the framework or from unrepresentative simulation.

Authors: The matrix reconstruction technique pre-labels entries as inter-pixel or inner-pixel and re-assembles the impedance matrix exactly from the base MoM solution; it is not an approximation and therefore produces identical results to a full MoM solve for any pixel configuration. We have added an explicit verification subsection (revised Section II-C) that tabulates the maximum absolute difference (which is at machine precision) between reconstructed and direct MoM matrices for representative configurations. The number of designs evaluated during optimization and the convergence history are reported in Section IV-B. The final bandwidths were independently confirmed with a commercial full-wave solver, confirming they are not artifacts of the accelerated solver. revision: partial
Referee: [Abstract] Abstract: The stochastic multi-pixel flipping search is presented as enhancing convergence, yet no details are given on run-to-run variability, comparison against gradient-based or other global optimizers, or safeguards against local minima. This assumption is load-bearing for attributing the achieved bandwidths to the proposed method.

Authors: The stochastic multi-pixel flipping procedure (detailed in revised Section III-B) randomly selects and flips multiple pixels per iteration according to a temperature-controlled probability, which provides an explicit mechanism to escape local minima. We have added run-to-run statistics from five independent optimizations in the revised results section, demonstrating that the target bands are reached in all runs with low variability in final performance. Direct comparisons against gradient-based or other global optimizers were outside the scope of the present work, which focuses on the MoM acceleration and meshing strategy; we have noted this limitation and the potential value of such benchmarks as future work. revision: partial

Circularity Check

0 steps flagged

No significant circularity; forward simulation plus search validated externally

full rationale

The paper presents an algorithmic inverse-design framework built on MoM forward simulation, pre-labeled matrix reconstruction, GPU acceleration, and a stochastic multi-pixel flipping search. The reported diplexer bandwidths (5.23-5.94 GHz and 6.17-7.15 GHz) are obtained from this procedure and cross-checked by an independent full-wave solver. No equation, claim, or self-citation reduces these outputs to quantities defined by the same run, fitted parameters, or prior author work; the chain remains a standard forward-model-plus-optimizer structure with external validation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are stated. The optimization search and matrix reconstruction implicitly rest on standard MoM assumptions and the premise that pixel connectivity can be treated as binary decisions, but these are not enumerated in the provided text.

pith-pipeline@v0.9.0 · 5700 in / 1276 out tokens · 26630 ms · 2026-05-23T06:58:26.780073+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

A novel matrix reconstruction technique, based on pre-labeling matrix entries as 'inter-pixel' or 'inner-pixel,' accelerates simulations for each variation of the pixelated structure... stochastic multi-pixel flipping search algorithm
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The effectiveness of this approach is demonstrated through the design of a diplexer achieving a -3-dB bandwidth for one channel spanning 5.23-5.94 GHz and another covering 6.17-7.15 GHz

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

[1]

write newline

" write newline "" initialize.prev.this.status FUNCTION begin.bib " write newline preamble empty 'skip preamble write newline if " thebibliography " longest.label * " " * write newline " [1] #1 " write newline " url@samestyle " write newline " " write newline " [2] #2 " write newline " =0pt " write newline " " ALTinterwordstretchfactor * " " * write newli...

work page
[2]

11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 = #1 \@IEEEnotcompsoconly \@IEEEcompsoconly #1 * [1] 0pt [0pt][0pt] #1 * [1] 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEcompsocnotconfonly \@IEEEauthorblockAstyle \@IEEEcompsocnotconfonly \@IEEEcompsocconfonly \@IEEEauthordefaulttextstyle \@IEEEcompsocnotconfonly \@IEEEauthor...

work page
[3]

#2 \@tempboxa > \@tempboxa #1

11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 \@makecaption#1#2 \@captype\@IEEEtablestring \@IEEEtabletopskipstrut \@IEEEfigurecaptionsepspace \@tempboxa #1. #2 \@tempboxa > \@tempboxa #1. [t] \@tempboxa#2 to \@tempboxa to \@tempboxa \@captype\@IEEEtablestring \@IEEEtablecaptionsepspace 0.5 0.0 \@IEEEfigurecaptionsepspace \@IEEEtablecap...

work page
[4]

D. W. Boeringer and D. H. Werner, ``Particle swarm optimization versus genetic algorithms for phased array synthesis,'' IEEE Transactions on antennas and propagation, vol. 52, no. 3, pp. 771--779, 2004

work page 2004
[5]

Zheng and C

Y. Zheng and C. Sideris, ``Ultra-fast simulation and inverse design of metallic antennas,'' in 2023 IEEE/MTT-S International Microwave Symposium-IMS 2023. 1em plus 0.5em minus 0.4em IEEE, 2023, pp. 351--354

work page 2023
[6]

E. A. Karahan, Z. Liu, and K. Sengupta, ``Ims deep learning enabled generalized synthesis of multi-port electromagnetic structures and circuits for mmwave power amplifiers,'' in 2024 IEEE/MTT-S International Microwave Symposium-IMS 2024. 1em plus 0.5em minus 0.4em IEEE, 2024, pp. 184--187

work page 2024
[7]

J. Lee, W. Jia, B. Sensale-Rodriguez, and J. S. Walling, ``Pixelated rf: Random metasurface based electromagnetic filters,'' in 2023 21st IEEE Interregional NEWCAS Conference (NEWCAS). 1em plus 0.5em minus 0.4em IEEE, 2023, pp. 1--5

work page 2023
[8]

Belenguer, M

A. Belenguer, M. D. Fernandez, J. A. Ballesteros, J. J. de Dios, H. Esteban, and V. E. Boria, ``Compact multilayer filter in empty substrate integrated waveguide with transmission zeros,'' IEEE Transactions on Microwave Theory and Techniques, vol. 66, no. 6, pp. 2993--3000, 2018

work page 2018
[9]

R. F. Harrington, ``The method of moments in electromagnetics,'' Journal of Electromagnetic waves and Applications, vol. 1, no. 3, pp. 181--200, 1987

work page 1987
[10]

S. Rao, D. Wilton, and A. Glisson, ``Electromagnetic scattering by surfaces of arbitrary shape,'' IEEE Transactions on antennas and propagation, vol. 30, no. 3, pp. 409--418, 1982

work page 1982
[11]

J. M. Johnson and Y. Rahmat-Samii, ``Genetic algorithms and method of moments (ga/mom) for the design of integrated antennas,'' IEEE Transactions on Antennas and Propagation, vol. 47, no. 10, pp. 1606--1614, 1999

work page 1999
[12]

Gupta, E

S. Gupta, E. Mehdizadeh, K. Cheema, and J. Shealy, ``Miniaturized ultrawide bandwidth wifi 6e diplexer implementation using xbaw rf filter technology,'' in 2022 IEEE/MTT-S International Microwave Symposium-IMS 2022. 1em plus 0.5em minus 0.4em IEEE, 2022, pp. 880--882

work page 2022

[1] [1]

write newline

" write newline "" initialize.prev.this.status FUNCTION begin.bib " write newline preamble empty 'skip preamble write newline if " thebibliography " longest.label * " " * write newline " [1] #1 " write newline " url@samestyle " write newline " " write newline " [2] #2 " write newline " =0pt " write newline " " ALTinterwordstretchfactor * " " * write newli...

work page

[2] [2]

11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 = #1 \@IEEEnotcompsoconly \@IEEEcompsoconly #1 * [1] 0pt [0pt][0pt] #1 * [1] 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEcompsocnotconfonly \@IEEEauthorblockAstyle \@IEEEcompsocnotconfonly \@IEEEcompsocconfonly \@IEEEauthordefaulttextstyle \@IEEEcompsocnotconfonly \@IEEEauthor...

work page

[3] [3]

#2 \@tempboxa > \@tempboxa #1

11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 \@makecaption#1#2 \@captype\@IEEEtablestring \@IEEEtabletopskipstrut \@IEEEfigurecaptionsepspace \@tempboxa #1. #2 \@tempboxa > \@tempboxa #1. [t] \@tempboxa#2 to \@tempboxa to \@tempboxa \@captype\@IEEEtablestring \@IEEEtablecaptionsepspace 0.5 0.0 \@IEEEfigurecaptionsepspace \@IEEEtablecap...

work page

[4] [4]

D. W. Boeringer and D. H. Werner, ``Particle swarm optimization versus genetic algorithms for phased array synthesis,'' IEEE Transactions on antennas and propagation, vol. 52, no. 3, pp. 771--779, 2004

work page 2004

[5] [5]

Zheng and C

Y. Zheng and C. Sideris, ``Ultra-fast simulation and inverse design of metallic antennas,'' in 2023 IEEE/MTT-S International Microwave Symposium-IMS 2023. 1em plus 0.5em minus 0.4em IEEE, 2023, pp. 351--354

work page 2023

[6] [6]

E. A. Karahan, Z. Liu, and K. Sengupta, ``Ims deep learning enabled generalized synthesis of multi-port electromagnetic structures and circuits for mmwave power amplifiers,'' in 2024 IEEE/MTT-S International Microwave Symposium-IMS 2024. 1em plus 0.5em minus 0.4em IEEE, 2024, pp. 184--187

work page 2024

[7] [7]

J. Lee, W. Jia, B. Sensale-Rodriguez, and J. S. Walling, ``Pixelated rf: Random metasurface based electromagnetic filters,'' in 2023 21st IEEE Interregional NEWCAS Conference (NEWCAS). 1em plus 0.5em minus 0.4em IEEE, 2023, pp. 1--5

work page 2023

[8] [8]

Belenguer, M

A. Belenguer, M. D. Fernandez, J. A. Ballesteros, J. J. de Dios, H. Esteban, and V. E. Boria, ``Compact multilayer filter in empty substrate integrated waveguide with transmission zeros,'' IEEE Transactions on Microwave Theory and Techniques, vol. 66, no. 6, pp. 2993--3000, 2018

work page 2018

[9] [9]

R. F. Harrington, ``The method of moments in electromagnetics,'' Journal of Electromagnetic waves and Applications, vol. 1, no. 3, pp. 181--200, 1987

work page 1987

[10] [10]

S. Rao, D. Wilton, and A. Glisson, ``Electromagnetic scattering by surfaces of arbitrary shape,'' IEEE Transactions on antennas and propagation, vol. 30, no. 3, pp. 409--418, 1982

work page 1982

[11] [11]

J. M. Johnson and Y. Rahmat-Samii, ``Genetic algorithms and method of moments (ga/mom) for the design of integrated antennas,'' IEEE Transactions on Antennas and Propagation, vol. 47, no. 10, pp. 1606--1614, 1999

work page 1999

[12] [12]

Gupta, E

S. Gupta, E. Mehdizadeh, K. Cheema, and J. Shealy, ``Miniaturized ultrawide bandwidth wifi 6e diplexer implementation using xbaw rf filter technology,'' in 2022 IEEE/MTT-S International Microwave Symposium-IMS 2022. 1em plus 0.5em minus 0.4em IEEE, 2022, pp. 880--882

work page 2022