Pith. sign in

REVIEW 1 major objections 7 minor 106 references

Algorithms design magnonic devices; magnonic devices compute

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

A perspective article surveys inverse-design magnonics and proposes the term 'AI magnonics' for the convergence of machine-learning-based design tools and magnonic neuromorphic hardware.

T0 review reviewed 2026-07-09 challenge →

load-bearing objection Useful survey of inverse-design magnonics; the 'AI magnonics' label oversells a convergence where one pillar is undemonstrated. the 1 major comments →

arxiv 2607.07324 v1 pith:5S4GCLAI submitted 2026-07-08 cond-mat.mes-hall

Perspectives on inverse design for AI magnonics

Franz Vilsmeier , Florian Bruckner , Claas Abert , Dieter Suess , Andrii V. Chumak This is my paper
classification cond-mat.mes-hall
keywords designmagnonicmagnonicsfielddevicesinverseparadigmalgorithm
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

This paper argues that inverse design—specifying a desired device function and letting an optimization algorithm find the structure that achieves it—represents a paradigm shift for magnonics, the field that uses spin waves (collective excitations of magnetic materials) to carry and process information. The authors survey the field along two axes: what can be optimized (topology, material parameters, and magnetic field landscape) and how it can be optimized (gradient-free, gradient-based, and neural-network-based methods, enabled by differentiable micromagnetic solvers). They then identify open frontiers including robust design against fabrication tolerances, input shaping and transducer optimization, explicit incorporation of nonlinear spin-wave effects as a design resource, spatially structured amplification, and self-adapting media. The paper's central conceptual claim is that magnonics and artificial intelligence are converging from two directions—machine-learning tools that design magnonic devices, and magnonic devices that serve as neuromorphic hardware—and proposes the term 'AI magnonics' for this convergence, culminating in the vision of a universal, reconfigurable magnonic platform that can be software-reprogrammed for any functionality.

Core claim

The paper is a perspective rather than an experimental result, so its central contribution is organizational and conceptual. The core claim is that inverse design, already mature in photonics, transfers to magnonics with a richer design space—saturation magnetization, anisotropy, damping, and a reconfigurable external field landscape have no photonic analogues—and that this transfer, combined with the development of differentiable micromagnetic solvers sharing infrastructure with neural network training, creates a natural bridge to machine-learning-based design. The authors document that the field has moved from its founding works in 2021 to experimental demonstrations of inverse-designed RF

What carries the argument

Inverse design loop (objective function plus optimizer plus differentiable micromagnetic forward solver); five design variables (topology, material parameters, field landscape, input shaping, nonlinear effects); three algorithm classes (gradient-free, gradient-based, neural-network-based); differentiable micromagnetic solvers (SpinTorch, magnum.np, NeuralMag) that share automatic-differentiation infrastructure with deep learning frameworks; the Landau-Lifshitz-Gilbert equation as the governing physics; the concept of AI magnonics as the convergence of AI-designed magnonics with magnonic hardware for AI.

Load-bearing premise

The paper assumes that inverse design, which has proven productive in photonics, will transfer to magnonics as a tractable paradigm despite the magnonic design landscape being more nonlinear, more non-convex, and more computationally expensive to evaluate per simulation. The optimism about scaling to a universal reconfigurable platform depends on this tractability holding without a rigorous argument for why it should.

What would settle it

If systematic sensitivity studies show that inverse-designed magnonic devices cannot tolerate realistic fabrication deviations, or if the computational cost of micromagnetic simulation prevents optimization from scaling beyond toy problems to multi-bit circuits, the paradigm-shift claim would be undermined.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If inverse design proves tractable at scale, magnonic device development would shift from intuition-guided iterative refinement to automated specification-driven synthesis, potentially reaching device configurations that manual design could never discover.
  • The shared computational infrastructure between differentiable physics solvers and neural network training means that surrogate models and generative design tools could accelerate magnonic design-space exploration by orders of magnitude, once sufficient training data exists.
  • A universal reconfigurable magnonic platform—trained once for a suite of functionalities and reprogrammed on nanosecond timescales via software-defined field landscapes—would change magnonics from a field of individual device demonstrations into a systematic engineering discipline.
  • Explicit incorporation of nonlinear spin-wave dynamics into the optimization loop would unlock classes of functionality inaccessible to linear interference, including Boolean logic and multilayer neuromorphic mappings, with the excitation amplitude itself becoming a design parameter.
  • Robust optimization incorporating fabrication tolerances into the design loop is a prerequisite for any topology-optimized magnonic device to leave simulation and be realized experimentally; the field's first such demonstration remains an open milestone.

Where Pith is reading between the lines

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

  • The paper draws an explicit analogy to the scaling of large language models, suggesting that complex magnonic functionalities might 'emerge' from a general inverse-design infrastructure once scaled. This analogy is suggestive but untested: whether the magnonic design landscape exhibits the kind of scaling laws that underpin emergence in language models is an open empirical question that the paper
  • The convergence of AI-designed magnonics and magnonic hardware for AI could, if realized, create a feedback loop where magnonic neuromorphic hardware is itself designed by AI tools, potentially optimizing the physical substrate for the computation it performs—a form of hardware-software co-design that goes beyond what either field achieves alone.
  • The richness of the magnonic design space (five design variables versus photonics' primarily geometric optimization) cuts both ways: it offers more degrees of freedom but also a more rugged, high-dimensional optimization landscape whose tractability for gradient-based methods in the strongly nonlinear regime is not guaranteed.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

1 major / 7 minor

Summary. This perspective article surveys the emerging field of inverse-design magnonics, organizing the literature along two axes: design variables (topology, material parameters, field landscape) and algorithmic tools (gradient-free, gradient-based, neural-network-based). The authors provide a clear overview of differentiable micromagnetic solvers (SpinTorch, magnum.np, NeuralMag) and identify several open frontiers including robust design, input shaping, nonlinear effects, amplification, and self-adapting systems. The article culminates in the proposal of 'AI magnonics' as a term for the convergence of ML-based design tools with magnonic neuromorphic hardware, and a long-term vision of a universal reconfigurable magnonic platform. The paper is well-structured, the literature coverage is broad, and the identification of open problems is thoughtful and grounded in the actual state of the field.

Significance. The manuscript provides a timely and useful organizational framework for a rapidly growing subfield. Its strengths include a comprehensive Table 1 of demonstrated inverse-design magnonic work, a clear taxonomy of design variables (Fig. 2) and algorithmic classes (Fig. 3), and honest acknowledgment of fundamental unknowns (Sec. 4.6). The discussion of nonlinearity as a design resource (Sec. 4.3) and the distinction between deterministic and stochastic nonlinear effects are particularly well-framed. The in-situ experimental approach of Zenbaa et al. is correctly highlighted as a significant advance for bridging the simulation-to-experiment gap. The perspective is appropriately forward-looking without overclaiming demonstrated results.

major comments (1)
  1. Abstract and Sec. 4.5/4.7: The central organizational claim is that 'magnonics and artificial intelligence are converging from two directions—machine-learning tools for designing magnonic devices, and magnonic devices as hardware for neuromorphic computation.' However, by the paper's own accounting (Sec. 3.1.3, Sec. 4.5.2, Table 1), the first direction—neural-network-based magnonic design—has zero demonstrated instances. The paper itself distinguishes neural-network-based approaches (Fig. 3c) from gradient-based methods that merely share computational infrastructure (autodiff, PyTorch) with ML. The 'AI' in 'AI magnonics' thus refers on the design side to shared software infrastructure rather than to any actual use of AI/ML for design. This is acknowledged honestly in Sec. 3.1.3 and Sec. 4.5.2, but the abstract and Sec. 4.7 frame the convergence as an 'emerging paradigm' without clearly信号
minor comments (7)
  1. Sec. 2.1: The statement that 'all topology-optimised magnonic devices remain at the simulation stage; no experimental realisation has yet been demonstrated' is correct but could be stated more prominently, e.g., in the abstract or Table 1.
  2. Table 1: The entry for Abert et al. [30] lists the design variable as '—' and the objective as 'General-purpose.' While this is accurate for a solver paper, it is visually inconsistent with the other rows that report specific devices. A footnote or separate category would clarify.
  3. Sec. 4.6: The discussion of parameter counts (from 5 to ~10^5) and configuration spaces (~10^162 states) is useful but the distinction between 'independent parameters' and 'reachable configurations' could be made sharper with a brief formal definition.
  4. Sec. 3.2: The discussion of LLM-based code generation for micromagnetic solvers is interesting but somewhat tangential to the core topic. If retained, it would benefit from a concrete example or reference specific to magnonics rather than general PDE solvers.
  5. Fig. 1: The check marks and question marks are described in the caption but their meaning could be made more explicit in the figure itself for readers scanning the article.
  6. Sec. 4.5.1: The energy efficiency comparison (25 aJ per operation vs. 7-nm CMOS) cites [6] which is a directional coupler paper, not an inverse-design result. The context of this comparison should be clarified to avoid implying it is an inverse-design achievement.
  7. The self-citation rate is notable (refs [24,27,30,32,34] and several others are by the present authors), but this is understandable given that the authors are among the pioneers of the field. The bibliography otherwise appears to represent the field adequately.

Circularity Check

0 steps flagged

No circularity found: perspective/survey paper with no derivation chain to reduce to its inputs

full rationale

This is a perspective/survey paper, not a derivation. It proposes the term 'AI magnonics' as an organizational label for a convergence of two research directions (ML-based magnonic design and magnonic neuromorphic hardware) and surveys the existing literature. There is no chain of equations, no fitted parameter presented as a prediction, and no self-citation that load-bears a mathematical claim. The paper's central claim is conceptual and aspirational; it is supported by citing external literature (Table 1, Table 2) and by the authors' own prior experimental and simulation work (e.g., [24,27,30,32,34]), but these citations serve as evidence of the field's state, not as premises that define the conclusion. The fact that several cited works are co-authored by the present authors is normal for a group active in the area; it does not create circularity because the claims being made (that inverse design is a growing paradigm, that ML-based design is not yet demonstrated in magnonics, that neuromorphic magnonic hardware exists) are externally verifiable and not defined in terms of the present paper's conclusions. No step reduces to its inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 2 invented entities

The paper is a perspective with no free parameters, no new axioms in the mathematical sense, and no physical entities introduced beyond naming conventions. The 'axioms' listed are domain assumptions underlying the perspective's optimism. The 'invented entities' are conceptual labels and visions, not postulated physical objects.

axioms (3)
  • domain assumption Inverse design, as a general computational paradigm, transfers productively from photonics and structural mechanics to magnonics despite the latter's nonlinearity and non-convex landscape.
    This is the foundational premise of the entire perspective, invoked implicitly throughout Sections 1–4. The paper argues for it by analogy and by citing early demonstrations, but it is not proven.
  • domain assumption The LLG equation, solved by differentiable micromagnetic solvers, provides a fully predictive forward model for deterministic nonlinear spin-wave dynamics.
    Stated in Sec. 4.3: 'For deterministic nonlinearities, the forward model remains fully predictive and gradient-based optimisation applies.' This is standard in micromagnetics but is a modeling assumption underlying all gradient-based inverse design discussed.
  • domain assumption Spin-wave circuits can operate at attojoule-scale energies (approximately 25 aJ per operation for a magnonic half-adder).
    Cited in Sec. 4.5.1 from reference [6]. Used to motivate the energy-efficiency argument for neuromorphic magnonics. The estimate's assumptions (e.g., transducer losses, fan-out) are not critically examined.
invented entities (2)
  • AI magnonics no independent evidence
    purpose: A label for the convergence of machine-learning-based magnonic design tools and magnonic neuromorphic hardware.
    The term is coined in this paper as a naming act. The two phenomena it describes are independently documented in the cited literature (Papp et al. 2021 [25] for both directions). The label itself has no falsifiable handle outside the paper.
  • Universal magnonic device no independent evidence
    purpose: A long-term vision of a single reconfigurable magnonic platform reprogrammable for any functionality via software-defined field landscapes.
    Described in Sec. 4.7 as a 'long-range vision.' The closest existing realization is the 7×7 current-loop platform of Zenbaa et al. [32,34], which demonstrated multiple functionalities but not universality. This is a speculative concept, not a demonstrated entity.

reviewed 2026-07-09 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Perspectives on inverse design for AI magnonics." pith.science (2026). https://pith.science/paper/5S4GCLAI

@misc{pith2026260707324,
  author       = {Pith},
  title        = {Pith review of: Perspectives on inverse design for AI magnonics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5S4GCLAI}},
  note         = {Machine review of arXiv:2607.07324}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Inverse design - specifying a desired functionality and letting a computational algorithm find the optimal structure - has emerged as a powerful paradigm for magnonic device engineering. In this article, we survey the rapidly growing field of inverse-design magnonics, organising it along two axes: the design variables (topology, material parameters, and magnetic field landscape) and the algorithmic toolbox (gradient-free, gradient-based, and neural-network-based methods) together with the differentiable micromagnetic solvers that enable them. We then identify open frontiers that we consider most promising for the next phase of the field: sensitivity analysis and robust design to bridge the gap between simulation and experiment; input shaping and transducer optimisation; the incorporation of nonlinear spin-wave effects as an explicit design resource; spatially structured amplification; self-adapting media and machine-learning-based design; and the long-term vision of a universal, reconfigurable magnonic platform. We argue that magnonics and artificial intelligence are converging from two directions - machine-learning tools for designing magnonic devices, and magnonic devices as hardware for neuromorphic computation - and propose the term AI magnonics to describe this emerging paradigm.

Figures

Figures reproduced from arXiv: 2607.07324 by Andrii V. Chumak, Claas Abert, Dieter Suess, Florian Bruckner, Franz Vilsmeier.

Figure 1
Figure 1. Figure 1: Inverse design as a paradigm shift in magnonics. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Design variables in inverse-design magnonics. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The algorithmic toolbox of inverse-design magnonics. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: AI magnonics: the convergence of inverse design and neuromorphic magnonics. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

106 extracted references · 106 canonical work pages · 4 internal anchors

  1. [1]

    Magnon spintronics,

    A. V. Chumak, V. I. Vasyuchka, A. A. Serga, and B. Hillebrands, “Magnon spintronics,”Nature Physics, vol. 11, pp. 453–461, 2015

    work page 2015

  2. [2]

    The 2021 magnonics roadmap,

    A. Barman, G. Gubbiotti, S. Ladak, A. O. Adeyeye, M. Krawczyk, J. Gräfe, C. Adelmann, S. Cotofana, A. Naeemi, V. I. Vasyuchka, B. Hillebrands, S. A. Nikitov, H. Yu, D. Grundler, A. V. Sadovnikov, A. A. Grachev, S. E. Sheshukova, J.-Y. Duquesne, M. Marangolo, G. Csaba, W. Porod, V. E. Demidov, S. O. Demokritov, H. Schultheiss,et al., “The 2021 magnonics ro...

    work page 2021

  3. [3]

    Advances in magnetics roadmap on spin-wave computing,

    A. V. Chumak, P. Kabos, M. Wu, C. Abert, C. Adelmann, A. O. Adeyeye, J. Akerman, F. G. Alber, G. E. W. Bauer, M. Becherer,et al., “Advances in magnetics roadmap on spin-wave computing,”IEEE Transactions on Magnetics, vol. 58, no. 6, p. 0800172, 2022

    work page 2022

  4. [4]

    The 2024 magnonics roadmap,

    B. Flebus and A. in Magnonics Collaboration, “The 2024 magnonics roadmap,”Journal of Physics: Condensed Matter, vol. 36, no. 36, p. 363501, 2024. 34 co-authors; includes section on inverse design

    work page 2024

  5. [5]

    Reconfigurable nanoscale spin-wave directional coupler,

    Q. Wang, P. Pirro, R. Verba, A. Slavin, B. Hillebrands, and A. V. Chumak, “Reconfigurable nanoscale spin-wave directional coupler,”Science Advances, vol. 4, no. 1, p. e1701517, 2018

    work page 2018

  6. [6]

    A magnonic directional coupler for integrated magnonic half- adders,

    Q. Wang, M. Kewenig, M. Schneider, R. Verba, F. Kohl, B. Heinz, M. Geilen, M. Mohseni, B. Lägel, F. Ciubotaru, C. Adelmann, C. Dubs, S. D. Cotofana, O. V. Dobrovolskiy, T. Brächer, P. Pirro, and A. V. Chumak, “A magnonic directional coupler for integrated magnonic half- adders,”Nature Electronics, vol. 3, pp. 765–774, 2020

    work page 2020

  7. [7]

    Reconfigurable submicrom- eter spin-wave majority gate with electrical transducers,

    G. Talmelli, T. Devolder, N. Trager, J. Forster, S. Wintz, M. Weigand, H. Stoll, M. Heyns, G. Schütz, I. P. Radu, J. Gräfe, F. Ciubotaru, and C. Adelmann, “Reconfigurable submicrom- eter spin-wave majority gate with electrical transducers,”Science Advances, vol. 6, no. 51, p. eabb4042, 2020

    work page 2020

  8. [8]

    All-magnonic repeater based on bistability,

    Q.Wang, R.Verba, K.Davídková, B.Heinz, S.Tian, Y.Rao, M.Guo, X.Guo, C.Dubs, P.Pirro, and A. V. Chumak, “All-magnonic repeater based on bistability,”Nature Communications, vol. 15, p. 7577, 2024

    work page 2024

  9. [9]

    Integrated magnonic neural circuits based on nonlinear wave neurons

    M. Guo, X. Jing, K. Davídková, R. Verba, Z. Zhou, X. Guo, C. Dubs, C. Gao, Y. Rao, K. Cai, J. Li, P. Pirro, A. V. Chumak, and Q. Wang, “Integrated magnonic neural circuits based on nonlinear wave neurons,” 2026. arXiv:2606.11703

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  10. [10]

    An all-magnonic neuron with tunable fading memory,

    D. Breitbach, M. Bechberger, H. Mortada, B. Heinz, R. Verba, Q. Wang, C. Dubs, M. Car- pentieri, G. Finocchio, D. Rodrigues, A. A. Hamadeh, and P. Pirro, “An all-magnonic neuron with tunable fading memory,” 2026. arXiv:2509.18321

    work page arXiv 2026

  11. [11]

    Pattern recognition in reciprocal space with a magnon-scattering reservoir,

    L. Körber, C. Heins, T. Hula, J.-V. Kim, S. Thlang, H. Schultheiss, K. Schultheiss, and J. Fassbender, “Pattern recognition in reciprocal space with a magnon-scattering reservoir,” Nature Communications, vol. 14, p. 3954, 2023. 15

    work page 2023

  12. [12]

    Tarantola,Inverse Problem Theory and Methods for Model Parameter Estimation

    A. Tarantola,Inverse Problem Theory and Methods for Model Parameter Estimation. SIAM, 2005

    work page 2005

  13. [13]

    Generating optimal topologies in structural design using a homogenization method,

    M. P. Bendsøe and N. Kikuchi, “Generating optimal topologies in structural design using a homogenization method,”Computer Methods in Applied Mechanics and Engineering, vol. 71, no. 2, pp. 197–224, 1988

    work page 1988

  14. [14]

    Topology optimization approaches: a comparative review,

    O. Sigmund and K. Maute, “Topology optimization approaches: a comparative review,” Structural and Multidisciplinary Optimization, vol. 48, no. 6, pp. 1031–1055, 2013

    work page 2013

  15. [15]

    Giga-voxel computational morpho- genesis for structural design,

    N. Aage, E. Andreassen, B. S. Lazarov, and O. Sigmund, “Giga-voxel computational morpho- genesis for structural design,”Nature, vol. 550, pp. 84–86, 2017

    work page 2017

  16. [16]

    Topology optimization for nano-photonics,

    J. S. Jensen and O. Sigmund, “Topology optimization for nano-photonics,”Laser & Photonics Reviews, vol. 5, no. 2, pp. 308–321, 2011

    work page 2011

  17. [17]

    Adjoint method and inverse design for nonlinear nanophotonic devices,

    T. W. Hughes, M. Minkov, I. A. D. Williamson, and S. Fan, “Adjoint method and inverse design for nonlinear nanophotonic devices,”ACS Photonics, vol. 5, pp. 4781–4787, Dec. 2018

    work page 2018

  18. [18]

    Inverse design and demonstration of a compact and broadband on-chip wavelength demultiplexer,

    A. Y. Piggott, J. Lu, K. G. Lagoudakis, J. Petykiewicz, T. M. Babinec, and J. Vučković, “Inverse design and demonstration of a compact and broadband on-chip wavelength demultiplexer,” Nature Photonics, vol. 9, pp. 374–377, 2015

    work page 2015

  19. [19]

    Inverse design in nanophotonics,

    S. Molesky, Z. Lin, A. Y. Piggott, W. Jin, J. Vučković, and A. W. Rodriguez, “Inverse design in nanophotonics,”Nature Photonics, vol. 12, pp. 659–670, Oct. 2018

    work page 2018

  20. [20]

    Systematic design of phononic band-gap materials and structures by topology optimization,

    O. Sigmund and J. S. Jensen, “Systematic design of phononic band-gap materials and structures by topology optimization,”Philosophical Transactions of the Royal Society of London. Series A, vol. 361, no. 1806, pp. 1001–1019, 2003

    work page 2003

  21. [21]

    Inverse design of phononic meta-structured materials,

    H. W. Dong, C. Shen, Z. Liu, S. D. Zhao, Z. Ren, C. X. Liu, X. He, S. A. Cummer, Y. S. Wang, D. Fang, and L. Cheng, “Inverse design of phononic meta-structured materials,”Materials Today, vol. 80, pp. 824–855, 2024

    work page 2024

  22. [22]

    Solving large-scale inverse magnetostatic problems using the adjoint method,

    F. Bruckner, C. Abert, G. Wautischer, C. Huber, C. Vogler, M. Hinze, and D. Suess, “Solving large-scale inverse magnetostatic problems using the adjoint method,”Scientific Reports, vol. 7, Jan. 2017

    work page 2017

  23. [23]

    Topology optimization for electromagnetics: A survey,

    F. Lucchini, R. Torchio, V. Cirimele, P. Alotto, and P. Bettini, “Topology optimization for electromagnetics: A survey,”IEEE Access, vol. 10, pp. 98593–98611, 2022

    work page 2022

  24. [24]

    Inverse-design magnonic devices,

    Q. Wang, A. V. Chumak, and P. Pirro, “Inverse-design magnonic devices,”Nature Communi- cations, vol. 12, May 2021

    work page 2021

  25. [25]

    Nanoscale neural network using non-linear spin-wave interference,

    Á. Papp, W. Porod, and G. Csaba, “Nanoscale neural network using non-linear spin-wave interference,”Nature Communications, vol. 12, Nov. 2021

    work page 2021

  26. [26]

    Inverse design of magnonic filter,

    Z. R. Yan, Y. W. Xing, and X. F. Han, “Inverse design of magnonic filter,”Journal of Magnetism and Magnetic Materials, vol. 563, p. 169976, Dec. 2022

    work page 2022

  27. [27]

    Inverse- design topology optimization of magnonic devices using level-set method,

    A. A. Voronov, M. Cuervo Santos, F. Bruckner, D. Suess, A. V. Chumak, and C. Abert, “Inverse- design topology optimization of magnonic devices using level-set method,”npj Spintronics, vol. 3, May 2025. 16

    work page 2025

  28. [28]

    Generation of spin-wave pulses by inverse design,

    S. Casulleras, S. Knauer, Q. Wang, O. Romero-Isart, A. V. Chumak, and C. Gonzalez-Ballestero, “Generation of spin-wave pulses by inverse design,”Physical Review Applied, vol. 19, June 2023

    work page 2023

  29. [29]

    Voltage- controlled half adder via magnonic inverse design,

    Z. Chen, G. J. Lim, C. C. I. Ang, T. Jin, F. Tan, B. W. H. Cheng, and W. S. Lew, “Voltage- controlled half adder via magnonic inverse design,”Applied Physics Letters, vol. 126, p. 132406, 2025

    work page 2025

  30. [30]

    Neuralmag: an open-source nodal finite-difference code for inverse micromagnetics,

    C. Abert, F. Bruckner, A. Voronov, M. Lang, S. A. Pathak, S. Holt, R. Kraft, R. Allayarov, P. Flauger, S. Koraltan, T. Schrefl, A. Chumak, H. Fangohr, and D. Suess, “Neuralmag: an open-source nodal finite-difference code for inverse micromagnetics,”npj Computational Materials, vol. 11, June 2025

    work page 2025

  31. [31]

    Experimental demonstration of a spin-wave lens designed with machine learning,

    M. Kiechle, L. Maucha, V. Ahrens, C. Dubs, W. Porod, G. Csaba, M. Becherer, and Á. Papp, “Experimental demonstration of a spin-wave lens designed with machine learning,”IEEE Magnetics Letters, vol. 13, p. 1–5, 2022

    work page 2022

  32. [32]

    A universal inverse-design magnonic device,

    N. Zenbaa, C. Abert, F. Majcen, M. Kerber, R. O. Serha, S. Knauer, Q. Wang, T. Schrefl, D. Suess, and A. V. Chumak, “A universal inverse-design magnonic device,”Nature Electronics, Jan. 2025

    work page 2025

  33. [33]

    Fast magnonic device development with inverse design,

    M. Wu, “Fast magnonic device development with inverse design,”Nature Electronics, Feb. 2025

    work page 2025

  34. [34]

    Realization of inverse-design magnonic logic gates,

    N. Zenbaa, F. Majcen, C. Abert, F. Bruckner, N. J. Mauser, T. Schrefl, Q. Wang, D. Suess, and A. V. Chumak, “Realization of inverse-design magnonic logic gates,”Science Advances, vol. 11, May 2025

    work page 2025

  35. [35]

    Controlling and patterning the effective magnetization in y3fe5o12 thin films using ion irradiation,

    W. T. Ruane, S. P. White, J. T. Brangham, K. Y. Meng, D. V. Pelekhov, F. Y. Yang, and P. C. Hammel, “Controlling and patterning the effective magnetization in y3fe5o12 thin films using ion irradiation,”AIP Advances, vol. 8, Dec. 2017

    work page 2017

  36. [36]

    Spin-wave optics in YIG realized by ion-beam irradiation,

    M. Kiechle, Á. Papp, S. Mendisch, V. Ahrens, M. Golibrzuch, G. H. Bernstein, W. Porod, G. Csaba, and M. Becherer, “Spin-wave optics in YIG realized by ion-beam irradiation,”Small, vol. 19, Feb. 2023

    work page 2023

  37. [37]

    Dispersion-tunable low-loss implanted spin-wave waveguides for large magnonic networks,

    J. Bensmann, R. Schmidt, K. O. Nikolaev, D. Raskhodchikov, S. Choudhary, R. Bhardwaj, S. Taheriniya, A. Varri, S. Niehues, A. El Kadri, J. Kern, W. H. P. Pernice, S. O. Demokritov, V. E. Demidov, S. Michaelis de Vasconcellos, and R. Bratschitsch, “Dispersion-tunable low-loss implanted spin-wave waveguides for large magnonic networks,”Nature Materials, vol...

    work page 1920

  38. [38]

    The effect of ga-ion irradiation on sub-micron-wavelength spin waves in yttrium-iron-garnet films,

    J. Greil, M. Kiechle, Á. Papp, P. Neumann, Z. Kovács, J. Volk, F. Schulz, S. Wintz, M. Weigand, G. Csaba, and M. Becherer, “The effect of ga-ion irradiation on sub-micron-wavelength spin waves in yttrium-iron-garnet films,”Nanotechnology, vol. 36, p. 135301, Feb. 2025

    work page 2025

  39. [39]

    Estab- lishing the magnetoelastic origin of spin-wave routing through focused ion beam patterning,

    F. Naunheimer, J. Greil, V. Ahrens, L. Maucha, Á. Papp, G. Csaba, and M. Becherer, “Estab- lishing the magnetoelastic origin of spin-wave routing through focused ion beam patterning,”

    work page

  40. [40]

    Nanopatterning reconfigurable magnetic landscapes via thermally assisted scanning probe lithography,

    E. Albisetti, D. Petti, M. Pancaldi, M. Madami, S. Tacchi, J. Curtis, W. P. King, Á. Papp, G. Csaba, W. Porod, P. Vavassori, E. Riedo, and R. Bertacco, “Nanopatterning reconfigurable magnetic landscapes via thermally assisted scanning probe lithography,”Nature Nanotechnology, vol. 11, pp. 545–551, Mar. 2016. 17

    work page 2016

  41. [41]

    Phase nanoengineering via thermal scanning probe lithography and direct laser writing,

    V. Levati, D. Girardi, N. Pellizzi, M. Panzeri, M. Vitali, D. Petti, and E. Albisetti, “Phase nanoengineering via thermal scanning probe lithography and direct laser writing,”Advanced Materials Technologies, vol. 8, no. 16, p. 2300166, 2023

    work page 2023

  42. [42]

    Patterning magnonic structures via laser induced crystallization of yittrium iron garnet,

    A. Del Giacco, F. Maspero, V. Levati, M. Vitali, E. Albisetti, D. Petti, L. Brambilla, V. Polewczyk, G. Vinai, G. Panaccione, R. Silvani, M. Madami, S. Tacchi, R. Dreyer, S. R. Lake, G. Woltersdorf, G. Schmidt, and R. Bertacco, “Patterning magnonic structures via laser induced crystallization of yittrium iron garnet,”Advanced Functional Materials, vol. 34...

    work page 2024

  43. [43]

    Programmable Integrated Magnonic Meshes

    P. Florio, M. Vitali, V. Levati, R. M. Ishola, L. Ciaccarini Mavilla, N. Lecis, C. Dubs, R. Bertacco, M. Madami, S. Tacchi, D. Petti, and E. Albisetti, “Programmable integrated magnonic meshes,” 2026. arXiv:2605.00290

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  44. [44]

    Three-dimensional nanoscale control of magnetism in crystalline yttrium iron garnet,

    V. Levati, M. Vitali, A. Del Giacco, N. Pellizzi, R. Silvani, L. Ciaccarini Mavilla, M. Madami, I.Biancardi, D. Girardi, M.Panzeri, P.Florio, M. Cocconcelli, D.Breitbach, P.Pirro, L.Rovatti, N. Lecis, F. Maspero, R. Bertacco, G. Corrielli, R. Osellame, V. Russo, A. Li Bassi, S. Tacchi, D. Petti, and E. Albisetti, “Three-dimensional nanoscale control of ma...

    work page 2025

  45. [45]

    Two-dimensional gradients in magnetic properties created with direct-write laser annealing,

    L. J. Riddiford, J. A. Brock, K. Murawska, J. Wisser, X. Huang, N. A. Shepelin, H. T. Nembach, A. Hrabec, and L. J. Heyderman, “Two-dimensional gradients in magnetic properties created with direct-write laser annealing,”Nature Communications, vol. 16, Dec. 2025

    work page 2025

  46. [46]

    Non-volatile clocked spin wave interconnect for beyond-cmos nanomagnet pipelines,

    S. Dutta, S.-C. Chang, N. Kani, D. E. Nikonov, S. Manipatruni, I. A. Young, and A. Naeemi, “Non-volatile clocked spin wave interconnect for beyond-cmos nanomagnet pipelines,”Scientific Reports, vol. 5, May 2015

    work page 2015

  47. [47]

    Towards magnonic devices based on voltage-controlled magnetic anisotropy,

    B. Rana and Y. Otani, “Towards magnonic devices based on voltage-controlled magnetic anisotropy,”Communications Physics, vol. 2, Aug. 2019

    work page 2019

  48. [48]

    Optically reconfigurable magnetic materials,

    M. Vogel, A. V. Chumak, E. H. Waller, T. Langner, V. I. Vasyuchka, B. Hillebrands, and G. von Freymann, “Optically reconfigurable magnetic materials,”Nature Physics, vol. 11, p. 487–491, May 2015

    work page 2015

  49. [49]

    Optical control of spin waves in hybrid magnonic-plasmonic structures,

    N. Kuznetsov, H. Qin, L. Flajsman, and S. van Dijken, “Optical control of spin waves in hybrid magnonic-plasmonic structures,”Science Advances, vol. 11, no. 2, p. eads2420, 2025

    work page 2025

  50. [50]

    Dynamic magnonic crystals based on spatiotemporal plasmon excitation,

    N. Kuznetsov, H. Qin, L. Flajsman, and S. van Dijken, “Dynamic magnonic crystals based on spatiotemporal plasmon excitation,”Advanced Materials, vol. 37, no. 33, p. 2502474, 2025

    work page 2025

  51. [51]

    Low-loss nanoscopic spin-wave guiding in continuous yttrium iron garnet films,

    H. Qin, R. B. Hollaender, L. Flajsman, and S. van Dijken, “Low-loss nanoscopic spin-wave guiding in continuous yttrium iron garnet films,”Nano Letters, vol. 22, pp. 5294–5300, 2022

    work page 2022

  52. [52]

    Writable spin wave nanochannels in an artificial-spin-ice-mediated ferromagnetic thin film,

    J. Li, W.-B. Xu, W.-C. Yue, Z. Yuan, T. Gao, T.-T. Wang, Z.-L. Xiao, Y.-Y. Lyu, C. Li, C. Wang, F. Ma, S. Dong, Y. Dong, H. Wang, P. Wu, W.-K. Kwok, and Y.-L. Wang, “Writable spin wave nanochannels in an artificial-spin-ice-mediated ferromagnetic thin film,”Applied Physics Letters, vol. 120, no. 13, p. 132404, 2022

    work page 2022

  53. [53]

    Ice sculpting: An artificial spin ice tutorial on controlling microstate and geometry for magnonics and neuromorphic computing,

    R. Sultana, A. K. Mondal, V. S. Bhat, K. Stenning, Y. Li, D. M. Arroo, A. Vasdev, M. R. McCarter, L. E. De Long, J. T. Hastings, J. C. Gartside, and M. B. Jungfleisch, “Ice sculpting: An artificial spin ice tutorial on controlling microstate and geometry for magnonics and neuromorphic computing,”Journal of Applied Physics, vol. 138, p. 061101, Aug. 2025. 18

    work page 2025

  54. [54]

    Reversal of nanomagnets by propagating magnons in ferri- magnetic yttrium iron garnet enabling nonvolatile magnon memory,

    K. Baumgaertl and D. Grundler, “Reversal of nanomagnets by propagating magnons in ferri- magnetic yttrium iron garnet enabling nonvolatile magnon memory,”Nature Communications, vol. 14, Mar. 2023

    work page 2023

  55. [55]

    Perspective on nonvolatile magnon-signal storage and in-memory computation for low-power consuming magnonics,

    A. Nizet, M. Xu, S. S. Joglekar, A. Mucchietto, and D. Grundler, “Perspective on nonvolatile magnon-signal storage and in-memory computation for low-power consuming magnonics,” Applied Physics Letters, vol. 126, Apr. 2025

    work page 2025

  56. [56]

    Magnetic domain walls as reconfigurable spin-wave nanochannels,

    K. Wagner, A. Kakay, K. Schultheiss, A. Henschke, T. Sebastian, and H. Schultheiss, “Magnetic domain walls as reconfigurable spin-wave nanochannels,”Nature Nanotechnology, vol. 11, pp. 432–436, 2016

    work page 2016

  57. [57]

    Skyrmion-based dynamic magnonic crystal,

    F. Ma, Y. Zhou, H. B. Braun, and W. S. Lew, “Skyrmion-based dynamic magnonic crystal,” Nano Letters, vol. 15, pp. 4029–4036, 2015

    work page 2015

  58. [58]

    Tuning magnonic devices with on-chip permanent micromagnets,

    M. Cocconcelli, S. Tacchi, R. Erdélyi, F. Maspero, A. Del Giacco, A. Plaza, O. Koplak, A. Cattoni, R. Silvani, M. Madami, Á. Papp, G. Csaba, F. Kohl, B. Heinz, P. Pirro, and R. Bertacco, “Tuning magnonic devices with on-chip permanent micromagnets,”Physical Review Applied, vol. 22, p. 064063, 2024

    work page 2024

  59. [59]

    Standalone integrated magnonic devices,

    M. Cocconcelli, F. Maspero, A. Micelli, A. Toniato, A. Del Giacco, N. Pellizzi, A. E. Plaza, A. Cattoni, M. Madami, R. Silvani, C. Adelmann, A. A. Hamadeh, P. Pirro, S. Tacchi, F. Ciubotaru, and R. Bertacco, “Standalone integrated magnonic devices,”Advanced Materials, vol. 37, no. 40, p. 2503493, 2025

    work page 2025

  60. [60]

    Monolithic piezo-magnonic-mems for efficient modulation of rf signals,

    M. Cocconcelli, A. Angotti, P. Florio, N. Pellizzi, F. Maspero, and R. Bertacco, “Monolithic piezo-magnonic-mems for efficient modulation of rf signals,” 2026. arXiv:2603.28411

    work page arXiv 2026

  61. [61]

    Nanophotonic particle simulation and inverse design using artificial neural networks,

    J. Peurifoy, Y. Shen, L. Jing, Y. Yang, F. Cano-Renteria, B. G. DeLacy, J. D. Joannopoulos, M. Tegmark, and M. Soljačić, “Nanophotonic particle simulation and inverse design using artificial neural networks,”Science Advances, vol. 4, no. 6, p. eaar4206, 2018

    work page 2018

  62. [62]

    Neural-adjoint method for the inverse design of all-dielectric metasurfaces,

    Y. Deng, S. Ren, K. Fan, J. M. Malof, and W. J. Padilla, “Neural-adjoint method for the inverse design of all-dielectric metasurfaces,”Optics Express, vol. 29, no. 5, pp. 7526–7534, 2021

    work page 2021

  63. [63]

    Training deep neural networks for the inverse design of nanophotonic structures,

    D. Liu, Y. Tan, E. Khoram, and Z. Yu, “Training deep neural networks for the inverse design of nanophotonic structures,”ACS Photonics, vol. 5, no. 4, pp. 1365–1369, 2018

    work page 2018

  64. [64]

    DeepAdjoint: An all-in-one photonic inverse design framework integrating data-driven machine learning with optimization algorithms,

    C. Yeung, B. Pham, Z. Tsai, K. T. Fountaine, and A. P. Raman, “DeepAdjoint: An all-in-one photonic inverse design framework integrating data-driven machine learning with optimization algorithms,”ACS Photonics, vol. 10, no. 4, pp. 884–891, 2023

    work page 2023

  65. [65]

    Parameter extraction and inverse design of semiconductor lasers based on the deep learning and particle swarm optimization method,

    Z. Ma and Y. Li, “Parameter extraction and inverse design of semiconductor lasers based on the deep learning and particle swarm optimization method,”Optics Express, vol. 28, no. 15, pp. 21971–21981, 2020

    work page 2020

  66. [66]

    Photonics inverse design: Pairing deep neural networks with evolutionary algorithms,

    R. S. Hegde, “Photonics inverse design: Pairing deep neural networks with evolutionary algorithms,”IEEE Journal of Selected Topics in Quantum Electronics, vol. 26, no. 1, pp. 1–8, 2020

    work page 2020

  67. [67]

    Global optimization of dielectric metasurfaces using a physics-driven neural network,

    J. Jiang and J. A. Fan, “Global optimization of dielectric metasurfaces using a physics-driven neural network,”Nano Letters, vol. 19, no. 8, pp. 5366–5372, 2019. 19

    work page 2019

  68. [68]

    Optimisation of colour generation from dielectric nanos- tructures using reinforcement learning,

    I. Sajedian, T. Badloe, and J. Rho, “Optimisation of colour generation from dielectric nanos- tructures using reinforcement learning,”Optics Express, vol. 27, no. 4, pp. 5874–5883, 2019

    work page 2019

  69. [69]

    Deep learning for the design of photonic structures,

    W. Ma, Z. Liu, Z. A. Kudyshev, A. Boltasseva, W. Cai, and Y. Liu, “Deep learning for the design of photonic structures,”Nature Photonics, vol. 15, pp. 77–90, 2021

    work page 2021

  70. [70]

    SpinTorch: a PyTorch-based differentiable micromagnetic framework for spin-wave neural networks,

    Á. Papp, W. Porod, and G. Csaba, “SpinTorch: a PyTorch-based differentiable micromagnetic framework for spin-wave neural networks,” 2021. Software available athttps://github.com/ adamopapp/spintorch

    work page 2021

  71. [71]

    magnum.np: aPyTorchbasedGPUenhanced finite difference micromagnetic simulation framework for inverse magnetics,

    F.Bruckner, S.Koraltan, C.Abert, andD.Suess, “magnum.np: aPyTorchbasedGPUenhanced finite difference micromagnetic simulation framework for inverse magnetics,”Scientific Reports, vol. 13, p. 12054, 2023

    work page 2023

  72. [72]

    Mathematical discoveries from program search with large language models,

    B. Romera-Paredes, M. Barekatain, A. Novikov, M. Balog, M. P. Kumar, E. Dupont, F. J. R. Ruiz, J. S. Ellenberg, P. Wang, O. Fawzi, P. Kohli, and A. Fawzi, “Mathematical discoveries from program search with large language models,”Nature, vol. 625, no. 7995, pp. 468–475, 2024

    work page 2024

  73. [73]

    AlphaEvolve: A coding agent for scientific and algorithmic discovery

    A. Novikov, N. V˜ u, M. Eisenberger, E. Dupont, P.-S. Huang, A. Z. Wagner, S. Shirobokov, B. Kozlovskii, F. J. R. Ruiz, A. Mehrabian, M. P. Kumar, A. See, S. Chaudhuri, G. Holland, A. Davies, S. Nowozin, P. Kohli, and M. Balog, “AlphaEvolve: A coding agent for scientific and algorithmic discovery,” 2025. arXiv:2506.13131

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  74. [74]

    CodePDE: An inference framework for LLM-driven PDE solver generation,

    S. Li, T. Marwah, J. Shen, W. Sun, A. Risteski, Y. Yang, and A. Talwalkar, “CodePDE: An inference framework for LLM-driven PDE solver generation,” 2025. arXiv:2505.08783

    work page arXiv 2025

  75. [75]

    Towards end-to-end automation of AI research,

    C. Lu, C. Lu, R. T. Lange, Y. Yamada, S. Hu, J. Foerster, D. Ha, and J. Clune, “Towards end-to-end automation of AI research,”Nature, vol. 651, no. 8107, pp. 914–919, 2026

    work page 2026

  76. [76]

    Threats to scientific software from over-reliance on AI code assistants,

    G. O’Brien, “Threats to scientific software from over-reliance on AI code assistants,”Nature Computational Science, vol. 5, no. 9, pp. 701–703, 2025

    work page 2025

  77. [77]

    Robustness of inverse-designed magnonic devices,

    A. A. Voronovet al., “Robustness of inverse-designed magnonic devices,” 2026. in preparation

    work page 2026

  78. [78]

    Robust design of topology-optimized metasur- faces,

    E. W. Wang, D. Sell, T. Phan, and J. A. Fan, “Robust design of topology-optimized metasur- faces,”Optical Materials Express, vol. 9, p. 469, Jan. 2019

    work page 2019

  79. [79]

    Fabrication- constrained nanophotonic inverse design,

    A. Y. Piggott, J. Petkov, J. Lu, T. M. Babinec, H. Mabuchi, and J. Vučković, “Fabrication- constrained nanophotonic inverse design,”Scientific Reports, vol. 7, p. 1786, 2017

    work page 2017

  80. [80]

    Inverse design of photonic devices with strict foundry fabrication constraints,

    M. F. Schubert, A. K. C. Cheung, I. A. D. Williamson, A. Spyra, and D. H. Alexander, “Inverse design of photonic devices with strict foundry fabrication constraints,”ACS Photonics, vol. 9, no. 7, pp. 2327–2336, 2022

    work page 2022

Showing first 80 references.

This paper was first reviewed by glm-5.2 on July 9, 2026.