Pith. sign in

REVIEW 4 major objections 4 minor 34 references

Machine learning magnetic parameters from spin configurations

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A CNN trained on simulated spin configurations can estimate three magnetic Hamiltonian parameters from a single experimental Lorentz TEM image.

desk verdict A useful simulation-to-experiment protocol idea with solid in-simulation tests, but the experimental validation contains a load-bearing field inconsistency that needs to be fixed before the transfer claim can be taken seriously. read the letter →

arxiv 1908.05829 v4 pith:VDZLUTEN submitted 2019-08-16 cond-mat.dis-nn cond-mat.mtrl-sciphysics.comp-ph

Dingchen Wang , Songrui Wei , Anran Yuan , Fanghua Tian , Kaiyan Cao , Qizhong Zhao , Dezhen Xue , Sen Yang This is my paper · ORCID
classification cond-mat.dis-nncond-mat.mtrl-sciphysics.comp-ph
keywords HamiltonianparameterestimationconvolutionalneuralnetworkspinconfigurationsmicromagneticsimulationLorentzTEMdataaugmentationDzyaloshinskii-Moriyainteractionexchangestiffness
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that a convolutional neural network can estimate three magnetic Hamiltonian parameters — the exchange stiffness $A_{ex}$, the Dzyaloshinskii-Moriya interaction strength $DMI$, and the saturation magnetization $M_{sat}$ — from a single image of a spin configuration. The network is trained not on many experimental images but on a small set of micromagnetic simulations generated under the same temperature and magnetic field as the experiment, with a sliding-window step that multiplies the number of training patches. If the claim holds, one Lorentz TEM micrograph would be enough to recover quantitative Hamiltonian parameters, reproduce the observed configuration, and predict material properties such as coercive and saturation fields. The paper demonstrates the protocol on FeGe and FeGe$_{0.5}$Si$_{0.5}$ skyrmion lattices and reports agreement with independently known values and hysteresis measurements.

What carries the argument

The load-bearing object is a convolutional neural network whose final layer is a sigmoid estimator rather than a classifier. A sliding window of size 32 with step 8 cuts each simulated spin configuration into overlapping patches, magnifying a small training set while preserving the physical meaning of the orientation map; scaling and rotation are rejected as augmentation because they change the spin configuration. The CNN's feature maps automatically learn descriptors of the local spin texture, and the three output neurons produce continuous estimates. The accompanying micromagnetic simulations, run under the same temperature, field, and geometry as the experiment, supply the labeled training distribution; the equal-information-per-patch property of spin configurations is what makes the sliding window physically valid.

What would settle it

Take the same experimental image used in the paper and independently measure $A_{ex}$, $DMI$, and $M_{sat}$ by microwave absorption or neutron scattering; if the CNN estimates disagree by more than the reported scatter, the transfer from simulations to experiment fails. A cheaper test: apply the trained network to synthetically blurred or noisy simulation images with known parameters and observe whether predicted values drift by more than the reported error bars.

Watch

Extended reading notes

Core claim

The central discovery is that the mapping from spin configuration to Hamiltonian parameters can be inverted by a CNN trained purely on simulated data, provided the simulations use the experimental observation conditions. Using 125 simulated images covering a grid of parameter values at a fixed temperature and field, the network learns to output continuous values of $A_{ex}$, $DMI$, and $M_{sat}$ through a sigmoid estimator layer. The sliding-window augmentation works because parameter information is distributed evenly across the image, so each $32\times32$ patch carries the same labels. Tested on simulated images with new random seeds and on parameter combinations absent from training, the estimates lie close to the diagonal; on experimental Lorentz TEM images of FeGe$_{0.5}$Si$_{0.5}$ and FeGe, the estimated parameters reproduce similar spin configurations, fall near the theoretical values from microwave absorption spectroscopy, and predict coercive and saturation fields in agreement with measured hysteresis.

Load-bearing premise

The approach assumes experimental Lorentz TEM spin-configuration images and the micromagnetic simulation images used as training data look statistically similar to the CNN despite differences in resolution, noise, and reconstruction artifacts; if that image distribution differs, the estimated parameters are unreliable.

Editorial extensions

If this is right

  • A single experimental image of a spin texture can replace several conventional measurements, such as ferromagnetic resonance, Brillouin light scattering, or neutron scattering, when estimating the three key magnetic parameters.
  • For a new observation condition, only a handful of simulations at that temperature and field are needed to build a working estimator, since sliding windows expand the training data substantially.
  • The estimated parameters plug back into micromagnetic simulation to reproduce the observed configuration and predict macroscopic behavior, including coercive field, saturation field, and sample volume.
  • Because the final layer is a continuous estimator, the same architecture can be retargeted by retraining on labeled simulations for other Hamiltonian parameters in other condensed-matter systems.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A test the authors leave implicit: train the network on simulations, degrade the inputs with Lorentz-TEM-like noise and blur, and check how much the parameter estimates drift; this would directly measure the domain gap that currently separates simulated and experimental images.
  • If the transfer assumption holds, the approach could be extended to spatially resolved parameter maps, estimating $A_{ex}$ or $DMI$ at different positions in a heterogeneous image, since the sliding window already produces local patches.
  • The reported relative errors, roughly 2 percent for $A_{ex}$ and 10 percent for $DMI$ and $M_{sat}$, suggest a natural comparison: feed the same experimental image through networks trained at different cell sizes or resolutions to see which microstructural scale carries the parameter information.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes a protocol for estimating micromagnetic Hamiltonian parameters (Aex, DMI, and Msat) from a single experimental spin-configuration image. A convolutional neural network is trained on a small set of micromagnetic simulation images generated under the same nominal temperature and magnetic field as the experiment, with a sliding-window augmentation to enlarge the effective training set. The trained CNN is then applied to experimental Lorentz TEM images, and the estimated parameters are used to reproduce the spin configuration and to predict the hysteresis loop and the sample volume. The in-simulation tests in Figs. 3 and 4 show near-diagonal estimation for seen and unseen parameter grids and for varied image sizes, with reported relative errors around 2% for Aex and around 10% for DMI and Msat.

Significance. The in-simulation validation is a genuine strength: the tests in Figs. 3 and 4 cover parameter interpolation, extrapolation to unseen parameter sets, and image-size generalization, and the reported errors are modest. The idea of exploiting the spatially homogeneous distribution of parameter information in spin images to augment a small labeled set via sliding windows is sensible and potentially useful for other image-to-parameter regression problems. However, the experimental transfer claim, which is the headline contribution, is not established by the evidence as presented: the external-condition mismatch for the FeGe demonstration and the circular volume estimate prevent the experimental results from validating the method. If the conditions are corrected, the volume claim is reframed, and a sensitivity analysis is supplied, the approach could be valuable for the community.

major comments (4)
  1. [Section III, Figure 5, Appendix C] The paper states in Section III and Figure 5(b) that the FeGe skyrmion lattice was observed and simulated at 265 K and 0.18 T, but Appendix C says 'FeGe spin configuration is observed at 265 K under 50 mT.' These two statements are mutually inconsistent, with a 130 mT discrepancy. Because the equilibrium spin configuration depends strongly on the applied field, this discrepancy directly breaks the 'same conditions' premise that the method relies on for simulation-to-experiment transfer. Please correct the stated field, and provide a sensitivity analysis showing how the estimated parameters depend on a field mismatch, or retrain the CNN at the actual experimental field.
  2. [Section III, Figure 5(b), Appendix C] The predicted hysteresis loop is labeled as being at 265 K, but the experimental hysteresis loop from Ref. 32 that it is compared against was measured at 250 K, as stated in Appendix C. This 15 K temperature mismatch is not discussed, and no sensitivity analysis is provided. Since magnetization and magnetic interactions are temperature dependent, the agreement in Fig. 5(b) does not by itself validate the estimated parameters or the predictive claim.
  3. [Section III, volume estimate] The manuscript says 'we vary the sample volume in our simulation to fit the experimental value of the magnetic moment' and then reports the fitted volume as an estimate. This is a circular procedure, not a prediction, and it cannot be used as evidence that the estimated parameters are correct. The abstract's claim that the approach 'predict[s] ... the volume of the experimental sample' overstates what is done. The volume fitting should be presented as a calibration step, and the abstract and conclusion should be revised accordingly.
  4. [Section III, Figure 5(a)] The experimental transfer for FeGe0.5Si0.5 is validated only by visual similarity between the simulated reproduction (Fig. 5(a2)) and the input image (Fig. 5(a1)). No quantitative image-difference metric is reported, and no independent parameter measurement is available for this specimen. Given the concern that the CNN may rely on simulation-specific features, visual similarity alone is insufficient to establish that the estimated parameters are physically correct.
minor comments (4)
  1. [Abstract and Section III] The abstract states that the method 'predict[s] ... the volume of the experimental sample,' but the volume is obtained by fitting the simulated magnetic moment to the measured value, not predicted from the image alone. Please rephrase to avoid overclaiming.
  2. [Figure 2 caption] The sentence 'The output layer is set as a estimator actived by sigmoid' contains a typo; it should read 'an estimator activated by sigmoid'.
  3. [Reference list] Reference 33 has garbled author formatting ('X. Z. Yu, . Onose, Y ., . Kanazawa, N., ...') and needs to be corrected.
  4. [Appendix B, Table I] The table lists the original image input as '512*512*3 image of PNG format,' but it is unclear how experimental images with different resolutions (2.34 nm/pixel and 0.54 nm/pixel in Appendix C) are resampled to the simulation pixel size. Please specify the preprocessing steps for experimental images.

Circularity Check

1 steps flagged · score 6.0 of 10

The central parameter estimation is not circular, but one claimed prediction (the experimental sample volume) is fitted to the measured magnetic moment and therefore reduces by construction to its target.

  1. fitted input called prediction [Section III, FeGe paragraph discussing Figure 5(b) and the inset table, after the hysteresis-loop comparison.]
    "Since we are not able to get access to the volume of the experimental sample, we vary the sample volume in our simulation to fit the experimental value of the magnetic moment, which is an extensive property. So that we can estimate the actual volume of the experimental sample around 1mm × 1mm × 3nm, which is reasonable for a SQUID measurement."

    The paper presents the sample-volume value as an outcome of the protocol, but the preceding sentence defines the procedure as varying the volume until the simulated magnetic moment matches the measured value. For a fixed spin configuration and fixed saturation magnetization, the magnetic moment is volume times the average magnetization, so the volume obtained is uniquely the one that reproduces the target moment; it is a one-parameter fit, not a prediction. Nothing independent is tested, and the quoted 'estimate' is therefore equivalent to its input by construction. This does not infect the Aex/DMI/Msat estimates, which are supported by independent microwave-spectroscopy values and an external hysteresis loop.

full rationale

The CNN parameter-estimation chain is largely self-contained: training and testing both use micromagnetically simulated spin configurations, test B uses parameter sets absent from training, and the experimental outputs are compared against external microwave-spectroscopy values (Ref. 31) and a published hysteresis loop (Ref. 32), so the central Aex/DMI/Msat estimation is not circular. The one clear reduction-by-construction is the volumetric side-result: the volume is varied to fit the experimental magnetic moment and then reported as an estimated prediction, which is a fitted input relabeled as a prediction. Separately, the simulation-to-experiment transfer is weakened by an internal contradiction in the stated 'same conditions' premise: Section III and Fig. 5(c) say the FeGe image and the training simulations were at 265 K and 0.18 T, while Appendix C says the FeGe image was observed at 265 K under 50 mT, and the experimental hysteresis used for validation was measured at 250 K rather than 265 K. These mismatches are not circular reductions, but they are missing-support issues that should lower confidence in the transfer claim. Overall circularity score is 6 because one claimed prediction reduces by construction while the central parameter estimation retains independent content.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the fidelity of the micromagnetic model, the invertibility of the spin configuration mapping, the direct transferability of simulated images to experimental TEM images, and the validity of comparing simulations and experiments at slightly different temperatures. The only fitted physical quantity is the sample volume, tuned to match the measured magnetic moment.

free parameters (1)
  • sample volume = approximately 1 mm x 1 mm x 3 nm
    Vary volume in simulation to fit the experimental magnetic moment, then presented as an estimated prediction.
assumptions (4)
  • domain assumption Micromagnetic model (MuMax3) accurately represents the spin configurations of FeGe and FeGe0.5Si0.5 at the relevant temperatures and fields.
    The training data and final hysteresis predictions are generated by MuMax3; the paper assumes model fidelity to real materials.
  • domain assumption A spin configuration image, or any sufficiently large patch, contains enough information to uniquely determine Aex, DMI, and Msat.
    The CNN regression assumes the mapping from spin configuration to parameters is learned from local patches; the paper states information is distributed evenly.
  • domain assumption Experimental Lorentz TEM images are in the same data distribution as simulated spin orientation maps when external conditions match.
    The CNN is applied directly to the experimental image without domain adaptation; this is the weakest assumption.
  • ad hoc to paper The micromagnetic simulation at 265 K can be compared to the experimental hysteresis loop measured at 250 K.
    The paper predicts the hysteresis using estimated parameters at 265 K but validates against experimental data measured at 250 K, without correcting for the temperature difference.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine learning magnetic parameters from spin configurations." pith.science (2026). https://pith.science/paper/VDZLUTEN

@misc{pith2026190805829,
  author       = {Pith},
  title        = {Pith review of: Machine learning magnetic parameters from spin configurations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VDZLUTEN}},
  note         = {Machine review of arXiv:1908.05829}
}
read the original abstract

Hamiltonian parameter estimation is crucial in condensed matter physics, but time and cost consuming in terms of resources used. With advances in observation techniques, high-resolution images with more detailed information are obtained, which can serve as an input to machine learning (ML) algorithms to extract Hamiltonian parameters. However, the number of labeled images is rather limited. Here, we provide a protocol for Hamiltonian parameter estimation based on a machine learning architecture, which is trained on a small amount of simulated images and applied to experimental spin configuration images. Sliding windows on the input images enlarges the number of training images; therefore we can train well a neural network on a small dataset of simulated images which are generated adaptively using the same external conditions such as temperature and magnetic field as the experiment. The neural network is applied to the experimental image and estimates magnetic parameters efficiently. We demonstrate the success of the estimation by reproducing the same configuration from simulation and predict a hysteresis loop accurately. Our approach paves a way to a stable and general parameter estimation.

Figures

Figures reproduced from arXiv: 1908.05829 by the authors.

Figure 1
Figure 1. FIG. 1. Flow chart of our approach. (a) For a particular experimental observation, a training dataset is formed using simulated images of [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. The implementation of a convolutional neural network ( [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. ( [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: FIG. 5. ( [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: In the experimental image estimation part, the environment parameters and geometry parameters have been set the same as the observation conditions [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Illustration of our data used. Blue blocks represent train [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 33 canonical work pages

  1. [1]

    Cartwright ,\ @noop title How the laws of physics lie \ ( year 1983 ) NoStop

    author author N. Cartwright ,\ @noop title How the laws of physics lie \ ( year 1983 ) NoStop

    work page 1983

  2. [2]

    author author J. A. \ Burgess , author A. E. \ Fraser , author F. F. \ Sani , author D. Vick , author B. D. \ Hauer , author J. P. \ Davis , \ and\ author M. R. \ Freeman ,\ @noop journal journal Science \ volume 339 ,\ pages 1051 ( year 2013 ) NoStop

    work page 2013

  3. [3]

    Leliaert , author M

    author author J. Leliaert , author M. Dvornik , author J. Mulkers , author J. De Clercq , author M. Milo s evi \'c , \ and\ author B. Van Waeyenberge ,\ @noop journal journal Journal of Physics D: Applied Physics \ volume 51 ,\ pages 123002 ( year 2018 ) NoStop

    work page 2018

  4. [4]

    Thiaville , author J

    author author A. Thiaville , author J. M. \ Garcı́A , \ and\ author J. Miltat ,\ @noop journal journal Journal of Magnetism & Magnetic Materials \ volume 242 ,\ pages 1061 ( year 2002 ) NoStop

    work page 2002

  5. [5]

    Sampaio , author V

    author author J. Sampaio , author V. Cros , author S. Rohart , author A. Thiaville , \ and\ author A. Fert ,\ @noop journal journal Nature Nanotechnology \ volume 8 ,\ pages 839 ( year 2013 ) NoStop

    work page 2013

  6. [6]

    Zhou , author E

    author author Y. Zhou , author E. Iacocca , author A. A. \ Awad , author R. K. \ Dumas , author F. C. \ Zhang , author H. B. \ Braun , \ and\ author J. Åkerman ,\ @noop journal journal Nature Communications \ volume 6 ,\ pages 8193 ( year 2015 ) NoStop

    work page 2015

  7. [7]

    Farhan , author P

    author author A. Farhan , author P. M. \ Derlet , author A. Kleibert , author A. Balan , author R. V. \ Chopdekar , author M. Wyss , author L. Anghinolfi , author F. Nolting , \ and\ author L. J. \ Heyderman ,\ @noop journal journal Nature Physics \ volume 9 ,\ pages 375 ( year 2013 ) NoStop

    work page 2013

  8. [8]

    author author A. P. \ Ramirez , author A. Hayashi , author R. J. \ Cava , author R. Siddharthan , \ and\ author B. S. \ Shastry ,\ @noop journal journal Nature \ volume 399 ,\ pages 333 ( year 1999 ) NoStop

    work page 1999

Show all 34 references
  1. [9]

    author author R. F. \ Wang , author C. Nisoli , author R. S. \ Freitas , author J. Li , author W. Mcconville , author B. J. \ Cooley , author M. S. \ Lund , author N. Samarth , author C. Leighton , \ and\ author V. H. \ Crespi ,\ @noop journal journal Nature \ volume 439 ,\ pa...

    2006

  2. [10]

    Ladak , author D

    author author S. Ladak , author D. E. \ Read , author G. K. \ Perkins , author L. F. \ Cohen , \ and\ author W. R. \ Branford ,\ @noop journal journal Nature Physics \ volume 6 ,\ pages 359 ( year 2010 ) NoStop

    2010

  3. [11]

    Tchernyshyov ,\ @noop journal journal Nature Physics \ volume 6 ,\ pages 323 ( year 2010 ) NoStop

    author author O. Tchernyshyov ,\ @noop journal journal Nature Physics \ volume 6 ,\ pages 323 ( year 2010 ) NoStop

    2010

  4. [12]

    Zhang \ and\ author M

    author author J. Zhang \ and\ author M. Sarovar ,\ @noop journal journal Physical review letters \ volume 113 ,\ pages 080401 ( year 2014 ) NoStop

    2014

  5. [13]

    Eyrich , author W

    author author C. Eyrich , author W. Huttema , author M. Arora , author E. Montoya , author F. Rashidi , author C. Burrowes , author B. Kardasz , author E. Girt , author B. Heinrich , author O. Mryasov , et al. ,\ @noop journal journal Journal of Applied Physics \ volume 111 ,\...

    2012

  6. [14]

    Buford , author P

    author author B. Buford , author P. Dhagat , \ and\ author A. Jander ,\ @noop journal journal IEEE Magnetics Letters \ volume PP ,\ pages 1 ( year 2016 ) NoStop

    2016

  7. [15]

    Burgarth \ and\ author A

    author author D. Burgarth \ and\ author A. Ajoy ,\ 10.1103/PhysRevLett.119.030402 journal journal Phys. Rev. Lett. \ volume 119 ,\ pages 030402 ( year 2017 ) NoStop

    2017 doi

  8. [16]

    author author L. A. \ Clark \ and\ author D. Pregibon ,\ in\ @noop booktitle Statistical models in S \ ( publisher Routledge ,\ year 2017 )\ pp.\ pages 377--419 NoStop

    2017

  9. [17]

    Xue , author P

    author author D. Xue , author P. V. \ Balachandran , author J. Hogden , author J. Theiler , author D. Xue , \ and\ author T. Lookman ,\ @noop journal journal Nature communications \ volume 7 ,\ pages 11241 ( year 2016 ) NoStop

    2016

  10. [18]

    Carrasquilla \ and\ author R

    author author J. Carrasquilla \ and\ author R. G. \ Melko ,\ @noop journal journal Nature Physics \ ( year 2017 ) NoStop

    2017

  11. [19]

    van Nieuwenburg , author Y

    author author E. van Nieuwenburg , author Y. H. \ Liu , \ and\ author S. Huber ,\ @noop journal journal Nature Physics \ ( year 2016 ) NoStop

    2016

  12. [20]

    a ming , author M. Tarnowski , author L. Asteria , author N. Fl \

    author author B. S. \ Rem , author N. K \"a ming , author M. Tarnowski , author L. Asteria , author N. Fl \"a schner , author C. Becker , author K. Sengstock , \ and\ author C. Weitenberg ,\ @noop journal journal Nature Physics \ volume 15 ,\ pages 917 ( year 2019 ) NoStop

    2019

  13. [21]

    Zhang , author A

    author author Y. Zhang , author A. Mesaros , author K. Fujita , author S. Edkins , author M. Hamidian , author K. Ch’ng , author H. Eisaki , author S. Uchida , author J. S. \ Davis , author E. Khatami , et al. ,\ @noop journal journal Nature \ volume 570 ,\ pages 484 ( year 20...

    2019

  14. [22]

    Carleo \ and\ author M

    author author G. Carleo \ and\ author M. Troyer ,\ @noop journal journal Science \ volume 355 ,\ pages 602 ( year 2016 ) NoStop

    2016

  15. [23]

    Bohrdt , author C

    author author A. Bohrdt , author C. S. \ Chiu , author G. Ji , author M. Xu , author D. Greif , author M. Greiner , author E. Demler , author F. Grusdt , \ and\ author M. Knap ,\ @noop journal journal Nature Physics \ volume 15 ,\ pages 921 ( year 2019 ) NoStop

    2019

  16. [24]

    Greplova , author C

    author author E. Greplova , author C. K. \ Andersen , \ and\ author K. M lmer ,\ @noop journal journal arXiv preprint arXiv:1711.05238 \ ( year 2017 ) NoStop

    2017 arXiv

  17. [25]

    Valenti , author E

    author author A. Valenti , author E. van Nieuwenburg , author S. Huber , \ and\ author E. Greplova ,\ @noop journal journal arXiv preprint arXiv:1907.02540 \ ( year 2019 ) NoStop

    1907 arXiv

  18. [26]

    Ch'ng , author J

    author author K. Ch'ng , author J. Carrasquilla , author R. G. \ Melko , \ and\ author E. Khatami ,\ 10.1103/PhysRevX.7.031038 journal journal Phys. Rev. X \ volume 7 ,\ pages 031038 ( year 2017 ) NoStop

    2017 doi

  19. [27]

    Chebotar , author A

    author author Y. Chebotar , author A. Handa , author V. Makoviychuk , author M. Macklin , author J. Issac , author N. Ratliff , \ and\ author D. Fox ,\ in\ @noop booktitle 2019 International Conference on Robotics and Automation (ICRA) \ ( organization IEEE ,\ year 2019 )\ pp....

    2019

  20. [28]

    Mulkers , author M

    author author J. Mulkers , author M. V. \ Milošević , \ and\ author B. V. \ Waeyenberge ,\ @noop journal journal Phys.rev.b \ volume 93 ,\ pages 214405 ( year 2016 ) NoStop

    2016

  21. [29]

    Matsumoto , author Y

    author author T. Matsumoto , author Y. G. \ So , author Y. Kohno , author H. Sawada , author Y. Ikuhara , \ and\ author N. Shibata ,\ @noop journal journal Science Advances \ volume 2 ,\ pages e1501280 ( year 2016 ) NoStop

    2016

  22. [30]

    author author B. D. \ Esser ,\ title High Resolution Characterization of Magnetic Materials for Spintronic Applications ,\ @noop Ph.D. thesis ,\ school The Ohio State University ( year 2018 ) NoStop

    2018

  23. [31]

    Takagi , author D

    author author R. Takagi , author D. Morikawa , author K. Karube , author N. Kanazawa , author K. Shibata , author G. Tatara , author Y. Tokunaga , author T. Arima , author Y. Taguchi , author Y. Tokura , \ and\ author S. Seki ,\ 10.1103/PhysRevB.95.220406 journal journal Phys....

    doi

  24. [32]

    author author S. X. \ Huang \ and\ author C. L. \ Chien ,\ @noop journal journal Physical Review Letters \ volume 108 ,\ pages 267201 ( year 2012 ) NoStop

    2012

  25. [33]

    author author X. Z. \ Yu , author . Onose , Y. , author . Kanazawa , N. , author J. H. \ Park , author J. H. \ Han , author . Matsui , Y. , author . Nagaosa , N. , \ and\ author . Tokura , Y. ,\ @noop journal journal Nature \ volume 465 ,\ pages 901 ( year 2010 ) NoStop

    2010

  26. [34]

    Vansteenkiste , author J

    author author A. Vansteenkiste , author J. Leliaert , author M. Dvornik , author F. Garcia-Sanchez , \ and\ author B. V. \ Waeyenberge ,\ @noop journal journal Aip Advances \ volume 4 ,\ pages 323001 ( year 2014 ) NoStop

    2014

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.