pith. machine review for the scientific record. sign in

arxiv: 2605.07047 · v1 · submitted 2026-05-07 · ⚛️ physics.plasm-ph

Recognition: no theorem link

Accelerating integrated modeling with surrogate-based optimization: the MAESTRO workflow

A. E. White, A. Ho, A. J. Creely, A. Martin-Sanabria, A. Saltzman, C. Holland, G. Snoep, J. C. Hillesheim, J. Hall, J. Pimentel-Aldaz, K. Yanna, M. Muraca, N. T. Howard, P. B. Snyder, P. de Lara Montoya, P. Rodriguez-Fernandez, T. Body

Authors on Pith no claims yet

Pith reviewed 2026-05-11 00:58 UTC · model grok-4.3

classification ⚛️ physics.plasm-ph
keywords plasma modelingsurrogate optimizationfusion reactor designtransport modelingintegrated simulationsteady-state profilesquasilinear transport
0
0 comments X

The pith

The MAESTRO workflow couples surrogate-based optimization in PORTALS with external solvers to deliver accurate steady-state plasma profile predictions.

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

This paper presents the MAESTRO workflow as a method to accelerate integrated modeling of fusion plasmas by linking the PORTALS framework's surrogate optimization with separate solvers for plasma equilibrium, pedestal physics, divertor constraints, and heating. Improvements to the surrogate models allow handling of discontinuities in transport fluxes that arise from numerical problems or physical instabilities with extreme stiffness. The result is efficient calculation of steady-state plasma profiles using full physics models rather than simplified approximations. Such predictions matter because they are needed to design and optimize future fusion reactors, where exhaustive simulations have historically been too slow for practical use.

Core claim

The MAESTRO workflow enables efficient coupling of the PORTALS surrogate-based transport solver with external equilibrium, pedestal, divertor, and heating models, providing accurate steady-state plasma profile predictions by improving the surrogate modeling of quasilinear transport to handle discontinuities in the fluxes.

What carries the argument

The MAESTRO workflow, which integrates the surrogate-based optimization of the PORTALS framework with external physics solvers for equilibrium and other components.

If this is right

  • Steady-state plasma profiles can be predicted efficiently with complete physics models instead of reduced approximations.
  • Discontinuities in transport fluxes from numerical or physical sources can be managed without loss of accuracy.
  • The approach supports optimization loops needed for fusion reactor design.
  • Coupling becomes practical because the surrogate solver avoids repeated expensive transport calculations.

Where Pith is reading between the lines

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

  • This coupling strategy could enable faster parameter scans over reactor operating points that are currently too expensive to explore.
  • Similar surrogate integration methods might extend to other coupled multi-physics problems beyond steady-state transport.
  • Performance on specific designs such as ITER or DEMO could be tested by comparing run times and profile agreement against conventional workflows.

Load-bearing premise

The surrogate models can accurately capture transport fluxes even when discontinuities arise from numerical issues or physical instabilities with extreme stiffness.

What would settle it

A direct comparison in which MAESTRO predictions deviate substantially from full nonlinear simulations or from experimental plasma profiles in cases with stiff transport or flux discontinuities would falsify the accuracy of the workflow.

read the original abstract

This paper introduces the MAESTRO workflow, that enables the coupling of the PORTALS framework [P. Rodriguez-Fernandez et al, Nucl. Fusion 2024] with external solvers for the plasma equilibrium, pedestal physics, divertor constraints and heating. The surrogate-based optimization nature of the transport solver is ideally suited for external coupling, allowing efficient steady-state predictions of plasma profiles with full physics models. Improvements in the surrogate modeling of quasilinear transport models with PORTALS are presented, which enable the efficient handling of discontinuities in the transport fluxes that can arise from numerical issues or physical instabilities with extreme stiffness. The combination of physics-informed methods and advanced numerical techniques allows the MAESTRO workflow to provide accurate and efficient predictions of steady-state plasma profiles, which are critical for fusion reactor design and optimization.

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.

Referee Report

1 major / 1 minor

Summary. The manuscript introduces the MAESTRO workflow, which couples the existing PORTALS surrogate-based optimization framework for quasilinear transport modeling with external solvers for plasma equilibrium, pedestal physics, divertor constraints, and heating. It describes improvements to the surrogate construction that are intended to handle discontinuities in transport fluxes arising from numerical artifacts or extreme physical stiffness, thereby enabling efficient computation of steady-state plasma profiles for fusion reactor design.

Significance. If the accuracy claims hold under validation, the workflow would represent a practical advance in accelerating integrated plasma modeling by leveraging surrogate optimization for external coupling, which is a recognized bottleneck in fusion simulations. The approach builds directly on the prior PORTALS framework without introducing new free parameters or ad-hoc entities, and the emphasis on physics-informed handling of stiff regimes aligns with needs in reactor optimization.

major comments (1)
  1. [Abstract] Abstract: The central claim that MAESTRO 'provide[s] accurate and efficient predictions of steady-state plasma profiles' rests on the assertion of improved surrogate modeling for discontinuities, yet the text supplies no quantitative error metrics, benchmark comparisons against full-physics reference runs, or specific test cases with known discontinuities or extreme stiffness. This absence is load-bearing because the fidelity of the surrogate under the stiff regimes critical for fusion applications cannot be assessed from the given material.
minor comments (1)
  1. The reference to the PORTALS framework should be expanded to a full bibliographic entry rather than the abbreviated form given.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback on our manuscript describing the MAESTRO workflow. We have addressed the major comment regarding the need for quantitative validation metrics in our point-by-point response below. We believe these revisions will clarify the accuracy and efficiency of the approach for the fusion modeling community.

read point-by-point responses

  1. Referee: The central claim that MAESTRO 'provide[s] accurate and efficient predictions of steady-state plasma profiles' rests on the assertion of improved surrogate modeling for discontinuities, yet the text supplies no quantitative error metrics, benchmark comparisons against full-physics reference runs, or specific test cases with known discontinuities or extreme stiffness. This absence is load-bearing because the fidelity of the surrogate under the stiff regimes critical for fusion applications cannot be assessed from the given material.

    Authors: We thank the referee for highlighting this important point. Although the manuscript provides qualitative demonstrations of the MAESTRO workflow's capabilities through its coupling with external solvers and the handling of discontinuities via improved surrogate techniques, we agree that the absence of explicit quantitative error metrics and specific benchmark comparisons limits the ability to fully evaluate the surrogate's performance in stiff regimes. In the revised version, we will incorporate a dedicated validation subsection that includes quantitative error metrics (e.g., mean absolute percentage errors and maximum relative errors) for the surrogate model on test cases with discontinuities. We will also present benchmark comparisons against full-physics reference solutions for cases involving extreme stiffness, thereby substantiating the accuracy claims made in the abstract. revision: yes

Circularity Check

0 steps flagged

No significant circularity; MAESTRO is a new coupling workflow built on referenced prior components

full rationale

The paper introduces the MAESTRO workflow as a coupling of the existing PORTALS surrogate framework with external solvers for equilibrium, pedestal, divertor, and heating physics. The abstract and description present this coupling and improvements in surrogate handling of discontinuities as the core contribution, without any derivation chain in which a claimed prediction, profile, or result reduces by the paper's own equations or definitions to a quantity defined solely in terms of its inputs or a self-citation. The referenced PORTALS work is a prior publication and the surrogate construction is a standard modeling technique rather than a tautological redefinition. No load-bearing step equates an output to its construction inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the work relies on the previously published PORTALS framework and standard assumptions of quasilinear transport modeling in tokamak plasmas.

pith-pipeline@v0.9.0 · 5520 in / 1084 out tokens · 38675 ms · 2026-05-11T00:58:34.006404+00:00 · methodology

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

40 extracted references · 27 canonical work pages · 1 internal anchor

  1. [1]

    Nuclear Fusion64(7), 076034 (2024)

    Rodriguez-Fernandez, P., Howard, N.T., Saltzman, A., Kantamneni, S., Candy, J., Holland, C., Balandat, M., Ament, S., White, A.E.: Enhancing MAESTRO workflow37 predictive capabilities in fusion burning plasmas through surrogate-based optimization in core transport solvers. Nuclear Fusion64(7), 076034 (2024). https://doi.org/10.1088/1741-4326/ad4b3d. 2024-06-05

    work page doi:10.1088/1741-4326/ad4b3d 2024

  2. [2]

    Tokamak profile prediction using direct gyrokinetic and neoclassical simula- tion

    Candy, J., Holland, C., Waltz, R.E., Fahey, M.R., Belli, E.: Tokamak pro- file prediction using direct gyrokinetic and neoclassical simulation. Physics of Plasmas16(6), 060704 (2009). https://doi.org/10.1063/1.3167820

    work page doi:10.1063/1.3167820 2009

  3. [3]

    Nuclear Fusion63(3), 036003 (2023)

    Siena, A.D., Rodriguez-Fernandez, P., Howard, N.T., Navarro, A.B., Bilato, R., G¨ orler, T., Poli, E., Merlo, G., Wright, J., Greenwald, M., Jenko, F.: Predictions of improved confinement in SPARC via energetic particle turbulence stabilization. Nuclear Fusion63(3), 036003 (2023). https://doi.org/10.1088/1741-4326/acb1c7. 2023-02-01

    work page doi:10.1088/1741-4326/acb1c7 2023

  4. [4]

    IPP-Report (Max-Planck- Institut f¨ ur Plasmaphysik)IPP 5/98(February) (2002)

    Pereverzev, G.V., Yushmanov, P.N.: ASTRA. IPP-Report (Max-Planck- Institut f¨ ur Plasmaphysik)IPP 5/98(February) (2002)

    2002

  5. [5]

    Computer software

    Breslau, J., Gorelenkova, M., Poli, F., Sachdev, J., Pankin, A., Perumpilly, G., Yuan, X., Glant, L.: TRANSP. Computer software. USDOE Office of Science (SC), Fusion Energy Sciences (FES) (2018). https://doi.org/10. 11578/DC.20180627.4

    2018

  6. [6]

    Physics of Plasmas30(9), 092510 (2023)

    Lyons, B.C., McClenaghan, J., Slendebroek, T., Meneghini, O., Neiser, T.F., Smith, S.P., Weisberg, D.B., Belli, E.A., Candy, J., Hanson, J.M., Lao, L.L., Logan, N.C., Saarelma, S., Sauter, O., Snyder, P.B., Stae- bler, G.M., Thome, K.E., Turnbull, A.D.: Flexible, integrated modeling of tokamak stability, transport, equilibrium, and pedestal physics. Physi...

    work page doi:10.1063/5.0156877 2023

  7. [7]

    Meneghini, O., Slendebroek, T., Lyons, B.C., McLaughlin, K., McCle- naghan, J., Stagner, L., Harvey, J., Neiser, T.F., Ghiozzi, A., Dose, G., Guterl, J., Zalzali, A., Cote, T., Shi, N., Weisberg, D., Smith, S.P., Grierson, B.A., Candy, J.: FUSE (Fusion Synthesis Engine): A Next Generation Framework for Integrated Design of Fusion Pilot Plants. arXiv. arXi...

    work page internal anchor Pith review doi:10.48550/arxiv 2024

  8. [8]

    https://github.com/pabloprf/MITIM-fusion

    Rodriguez-Fernandez, P.: MITIM: a toolbox for modeling tasks in plasma physics and fusion energy. https://github.com/pabloprf/MITIM-fusion. Version 1.1 (2024). https://mitim-fusion.readthedocs.io/en/latest/

    2024

  9. [9]

    A theory-based transport model with com- prehensive physics

    Staebler, G.M., Kinsey, J.E., Waltz, R.E.: A theory-based transport model with comprehensive physics. Physics of Plasmas14(5), 055909 (2007). https://doi.org/10.1063/1.2436852

    work page doi:10.1063/1.2436852 2007

  10. [10]

    https://doi.org/10.48550/arXiv.2209.06731

    Mandell, N.R., Dorland, W., Abel, I., Gaur, R., Kim, P., Martin, M., Qian, 38MAESTRO workflow T.: GX: a GPU-native gyrokinetic turbulence code for tokamak and stel- larator design (2022). https://doi.org/10.48550/arXiv.2209.06731. https: //arxiv.org/abs/2209.06731v3 2023-02-28

    work page doi:10.48550/arxiv.2209.06731 2022

  11. [11]

    Journal of Computational Physics324, 73–93 (2016)

    Candy, J., Belli, E.A., Bravenec, R.V.: A high-accuracy Eulerian gyroki- netic solver for collisional plasmas. Journal of Computational Physics324, 73–93 (2016). https://doi.org/10.1016/j.jcp.2016.07.039

    work page doi:10.1016/j.jcp.2016.07.039 2016

  12. [12]

    Physics of Plasmas31(6), 062501 (2024)

    Rodriguez-Fernandez, P., Howard, N.T., Saltzman, A., Shoji, L., Body, T., Battaglia, D.J., Hughes, J.W., Candy, J., Staebler, G.M., Creely, A.J.: Core performance predictions in projected SPARC first-campaign plasmas with nonlinear CGYRO. Physics of Plasmas31(6), 062501 (2024). https: //doi.org/10.1063/5.0209752. 2024-06-03

    work page doi:10.1063/5.0209752 2024

  13. [13]

    Nuclear Fusion62(7), 076036 (2022)

    Rodriguez-Fernandez, P., Howard, N.T., Candy, J.: Nonlinear gyroki- netic predictions of SPARC burning plasma profiles enabled by surrogate modeling. Nuclear Fusion62(7), 076036 (2022). https://doi.org/10.1088/ 1741-4326/AC64B2. 2022-05-16

    2022

  14. [14]

    Astudillo, R., Frazier, P.I.: Bayesian Optimization of Composite Func- tions. arXiv. arXiv:1906.01537 [cs, math, stat] (2019). https://doi.org/10. 48550/arXiv.1906.01537. http://arxiv.org/abs/1906.01537 2024-05-21

    work page arXiv 1906

  15. [15]

    Journal of Plasma Physics (accepted) (2026)

    Howard, N.T., Rodriguez-Fernandez, P., Hall, J., Muraca, M., Saltzman, A., Ho, A., Hillesheim, J.C., Creely, A.: Performance and transport in the ARC tokamak. Journal of Plasma Physics (accepted) (2026)

    2026

  16. [16]

    Gyro-Landau fluid equations for trapped and passing particles

    Staebler, G.M., Kinsey, J.E., Waltz, R.E.: Gyro-Landau fluid equations for trapped and passing particles. Physics of Plasmas12(10), 102508 (2005). https://doi.org/10.1063/1.2044587

    work page doi:10.1063/1.2044587 2005

  17. [17]

    Plasma Physics and Controlled Fusion47(5 A) (2005)

    Bourdelle, C.: Turbulent particle transport in magnetized fusion plasma. Plasma Physics and Controlled Fusion47(5 A) (2005). https://doi.org/ 10.1088/0741-3335/47/5A/023

    work page doi:10.1088/0741-3335/47/5a/023 2005

  18. [18]

    Nuclear Fusion62(9), 096005 (2022)

    Dudding, H.G., Casson, F.J., Dickinson, D., Patel, B.S., Roach, C.M., Belli, E.A., Staebler, G.M.: A new quasilinear saturation rule for tokamak turbulence with application to the isotope scaling of transport. Nuclear Fusion62(9), 096005 (2022). https://doi.org/10.1088/1741-4326/ac7a4d. 2022-07-14

    work page doi:10.1088/1741-4326/ac7a4d 2022

  19. [19]

    Journal of Plasma Physics 86(5), 865860502 (2020)

    Creely, A.J., Greenwald, M.J., Ballinger, S.B., Brunner, D., Canik, J., Doody, J., F¨ ul¨ op, T., Garnier, D.T., Granetz, R., Gray, T.K., Hol- land, C., Howard, N.T., Hughes, J.W., Irby, J.H., Izzo, V.A., Kramer, G.J., Kuang, A.Q., LaBombard, B., Lin, Y., Lipschultz, B., Logan, N.C., Lore, J.D., Marmar, E.S., Montes, K., Mumgaard, R.T., Paz- Soldan, C., R...

    work page doi:10.1017/s0022377820001257 2020

  20. [20]

    Journal of Plasma Physics86(5), 865860503 (2020)

    Rodriguez-Fernandez, P., Howard, N.T., Greenwald, M.J., Creely, A.J., Hughes, J.W., Wright, J.C., Holland, C., Lin, Y., Sciortino, F., Team, t.S.: Predictions of core plasma performance for the SPARC tokamak. Journal of Plasma Physics86(5), 865860503 (2020). https://doi.org/10. 1017/S0022377820001075. 2021-07-12

    2020

  21. [22]

    Physics of Plasmas30(8), 082304 (2023)

    Molina Cabrera, P.A., Rodriguez-Fernandez, P., G¨ orler, T., Bergmann, M., H¨ ofler, K., Denk, S.S., Bielajew, R., Conway, G.D., Yoo, C., White, A.E., ASDEX Upgrade Team: Isotope effects on energy transport in the core of ASDEX-Upgrade tokamak plasmas: Turbulence measurements and model validation. Physics of Plasmas30(8), 082304 (2023). https://doi. org/1...

    work page doi:10.1063/5.0143416 2023

  22. [23]

    Nuclear Fusion62(4), 042003 (2022)

    Rodriguez-Fernandez, P., Creely, A.J., Greenwald, M.J., Brunner, D., Ballinger, S.B., Chrobak, C.P., Garnier, D.T., Granetz, R., Hartwig, Z.S., Howard, N.T., Hughes, J.W., Irby, J.H., Izzo, V.A., Kuang, A.Q., Lin, Y., Marmar, E.S., Mumgaard, R.T., Rea, C., Reinke, M.L., Riccardo, V., Rice, J.E., Scott, S.D., Sorbom, B.N., Stillerman, J.A., Sweeney, R., Ti...

    work page doi:10.1088/1741-4326/ac1654 2022

  23. [24]

    Advances in Neural Information Processing Systems2018-Decem(NeurIPS), 9884–9895 (2018)

    Wilson, J.T., Hutter, F., Deisenroth, M.P.: Maximizing acquisition functions for Bayesian optimization. Advances in Neural Information Processing Systems2018-Decem(NeurIPS), 9884–9895 (2018). arXiv: 1805.10196

    work page arXiv 2018

  24. [25]

    Balandat, M., Karrer, B., Jiang, D.R., Daulton, S., Letham, B., Wil- son, A.G., Bakshy, E.: BoTorch: A Framework for Efficient Monte- Carlo Bayesian Optimization. arXiv. arXiv:1910.06403 [cs, math, stat] (2020). https://doi.org/10.48550/arXiv.1910.06403. http://arxiv. org/abs/1910.06403 2023-12-19 40MAESTRO workflow

    work page doi:10.48550/arxiv.1910.06403 1910

  25. [26]

    Plasma Physics and Controlled Fusion50(9), 095010 (2008)

    Belli, E.A., Candy, J.: Kinetic calculation of neoclassical transport including self-consistent electron and impurity dynamics. Plasma Physics and Controlled Fusion50(9), 095010 (2008). https://doi.org/10.1088/ 0741-3335/50/9/095010

    2008

  26. [27]

    Plasma Phys

    Brambilla, M.: Numerical simulation of ion cyclotron waves in tokamak plasmas. Plasma Phys. Control. Fusion41(1) (1999)

    1999

  27. [28]

    Computer Physics Communications159(3), 157– 184 (2004)

    Pankin, A., McCune, D., Andre, R., Bateman, G., Kritz, A.: The tokamak Monte Carlo fast ion module NUBEAM in the national transport code collaboration library. Computer Physics Communications159(3), 157– 184 (2004). https://doi.org/10.1016/j.cpc.2003.11.002

    work page doi:10.1016/j.cpc.2003.11.002 2004

  28. [29]

    Physics of Plasmas16(5) (2009)

    Snyder, P.B., Groebner, R.J., Leonard, A.W., Osborne, T.H., Wil- son, H.R.: Development and validation of a predictive model for the pedestal height. Physics of Plasmas16(5) (2009). https://doi.org/10. 1063/1.3122146

    2009

  29. [30]

    Nuclear Fusion65(8), 086002 (2025)

    Body, T., Kallenbach, A., Eich, T.: A simple, accurate model for detach- ment access. Nuclear Fusion65(8), 086002 (2025). https://doi.org/10. 1088/1741-4326/ade4d9. 2026-03-21

    2025

  30. [31]

    https: //freegs.readthedocs.io/en/latest/ (2024)

    Dudson, B.: FreeGS: Free boundary Grad-Shafranov solver. https: //freegs.readthedocs.io/en/latest/ (2024). https://freegs.readthedocs.io/ en/latest/

    2024

  31. [32]

    Citrin, J., Goodfellow, I., Raju, A., Chen, J., Degrave, J., Donner, C., Felici, F., Hamel, P., Huber, A., Nikulin, D., Pfau, D., Tracey, B., Riedmiller, M., Kohli, P.: TORAX: A Fast and Differentiable Tokamak Transport Simulator in JAX. arXiv. arXiv:2406.06718 [physics] (2024). https://doi.org/10.48550/arXiv.2406.06718. http://arxiv.org/abs/2406. 06718 2...

    work page doi:10.48550/arxiv.2406.06718 2024

  32. [33]

    Journal of Plasma Physics (accepted) (2026)

    Hillesheim, J.C., Creely, A.J., Eich, T.H., Howard, N.T., Leuthold, N., Sweeney, R., LeViness, A., Nelson, A.O., Nichols, L., Tinguely, R.A., Usoltseva, M., Battaglia, D., Body, T.A.J., Hansen, C., Logan, N.C., Mumgaard, R.T., Rodriguez-Fernandez, P., Snyder, P.B., Sorbom, B.N., Wright, J.C.: Overview of the physics basis for the ARC fusion power plant. J...

    2026

  33. [34]

    Nuclear Fusion66(2), 026005 (2025)

    Saltzman, A., Rodriguez-Fernandez, P., Body, T., Ho, A., Howard, N.T.: Impact of model uncertainty on SPARC operating scenario predictions with empirical modeling. Nuclear Fusion66(2), 026005 (2025). https:// doi.org/10.1088/1741-4326/ae2342. 2025-12-19

    work page doi:10.1088/1741-4326/ae2342 2025

  34. [35]

    Pearlstein: On the Grad-Shafranov equation as an eigenvalue problem, with implications for q solvers

    LoDestro, L.L., L.D. Pearlstein: On the Grad-Shafranov equation as an eigenvalue problem, with implications for q solvers. Physics of Plasmas MAESTRO workflow41 1(1), 90–95 (1994). https://doi.org/10.1063/1.870464

    work page doi:10.1063/1.870464 1994

  35. [36]

    Plasma Physics and Controlled Fusion38(12), 2163–2186 (1996)

    Porcelli, F., Boucher, D., Rosenbluth, M.N.: Model for the sawtooth period and amplitude. Plasma Physics and Controlled Fusion38(12), 2163–2186 (1996). https://doi.org/10.1088/0741-3335/38/12/010

    work page doi:10.1088/0741-3335/38/12/010 1996

  36. [37]

    Exact Solutions of the Sine-Gordon Equation Describing Oscillations in a Long (but Finite) Josephson Junction

    Hager, R., Chang, C.S.: Gyrokinetic neoclassical study of the bootstrap current in the tokamak edge pedestal with fully non-linear Coulomb collisions. Physics of Plasmas23(4) (2016). https://doi.org/10.1063/1. 4945615

    work page doi:10.1063/1 2016

  37. [38]

    PhD thesis, Princeton University (1986)

    Hammett, G.: Fast ion studies of ion cyclotron heating in the PLT tokamak: Phd thesises. PhD thesis, Princeton University (1986)

    1986

  38. [39]

    Plasma Physics and Controlled Fusion63(1), 012001 (2020)

    Arbon, R., Candy, J., Belli, E.A.: Rapidly-convergent flux-surface shape parameterization. Plasma Physics and Controlled Fusion63(1), 012001 (2020). https://doi.org/10.1088/1361-6587/abc63b

    work page doi:10.1088/1361-6587/abc63b 2020

  39. [40]

    Nuclear Fusion64(5), 056003 (2024)

    Kallenbach, A., Dux, R., Henderson, S.S., Tantos, C., Bernert, M., Day, C., McDermott, R.M., Rohde, V., Zito, A., Team, t.A.U.: Divertor enrich- ment of recycling impurity species (He, N2, Ne, Ar, Kr) in ASDEX Upgrade H-modes. Nuclear Fusion64(5), 056003 (2024). https://doi.org/ 10.1088/1741-4326/ad3139. 2026-02-03

    work page doi:10.1088/1741-4326/ad3139 2024

  40. [41]

    Gardner, J.R., Pleiss, G., Bindel, D., Weinberger, K.Q., Wilson, A.G.: GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration. arXiv. arXiv:1809.11165 [cs, stat] (2021). https://doi. org/10.48550/arXiv.1809.11165. http://arxiv.org/abs/1809.11165 2024- 05-21

    work page doi:10.48550/arxiv.1809.11165 2021