Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Estimate Sonic Mach Number in the Interstellar Medium with Convolutional Neural Network

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read A convolutional neural network can estimate the sonic Mach number of interstellar gas directly from spectroscopic maps.

desk verdict First CNN for sonic Mach number is an incremental but sound proof-of-concept; the abstract's generalization claim outruns the retrain-per-condition experiments. read the letter →

arxiv 2411.11157 v2 pith:DQ4W34JT submitted 2024-11-17 astro-ph.GA

classification astro-ph.GA
keywords sonicMachnumberinterstellarturbulenceconvolutionalneuralnetworkmagnetohydrodynamicsimulationsspectroscopicobservationsmediumvelocitychannelmapsdeeplearning
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 tries to establish that a convolutional neural network (CNN) can read the sonic Mach number $M_s$ of turbulent interstellar gas directly from the morphology of spectroscopic maps (integrated intensity, velocity centroid, and thin velocity channels), without fitting probability density functions or assuming a lognormal column-density distribution. It argues that the physical link is the shock-driven small-scale filamentary structure that becomes increasingly prominent as $M_s$ rises, and that the network learns this link from synthetic observations built from MHD turbulence simulations. If correct, the method gives a fast, spatially resolved estimator of $M_s$ that works under different magnetic-field inclinations, in the presence of observational noise, and even when low spatial frequencies are missing from interferometric data. The authors report a median prediction uncertainty of about 0.5 to 1.5 depending on noise, with intensity maps the most accurate but channel maps uniquely able to provide the three-dimensional $M_s$ distribution needed for 3D magnetic-field estimation.

What carries the argument

The load-bearing mechanism is the morphological encoding of $M_s$ in projected maps, captured by a CNN. The paper identifies the physical correlation as the growth of small-scale density fluctuations and filamentary structures when $M_s$ increases, driven by shocks; intensity maps carry this most directly, while velocity centroids mix in intensity-weighted velocity information and thin velocity channels are dominated by velocity caustics, which makes them less sensitive to $M_s$ but more useful for 3D mapping. The CNN architecture, with convolution, pooling, batch normalization, and fully connected layers, learns the mapping from 32x32-cell map patches to a scalar $M_s$ prediction using mean-squared-error training on roughly 0.6 million subfields per iteration.

What would settle it

Compare CNN predictions, trained only on idealized isothermal solenoidally driven simulations, against independently measured sonic Mach numbers in real molecular clouds (for example, from column-density PDF widths or from Zeeman-inferred fields combined with velocity dispersions), and check whether the median absolute error stays within the claimed 0.5 to 1.5 range; a second test is to run the network on a simulation that includes radiative transfer and self-gravity and measure the error increase.

Watch

Extended reading notes

Core claim

The central claim is that the spatial pattern of gas emission encodes the sonic Mach number, and a CNN can decode it. Using six $512^{3}$ isothermal ideal-MHD simulations with similar Alfvénic Mach number but sonic Mach numbers from 2.17 to 11.02, the authors generate synthetic position-position-velocity cubes, cut them into 32x32-cell subfields, and train a CNN to predict the local $M_s^{\rm sub}$ defined by the line-of-sight velocity dispersion in each subfield. The network learns that higher $M_s$ produces more small-scale, shock-dominated filamentary structures in intensity and centroid maps, and it transfers this association across different magnetic field strengths, inclinations, noise levels, and missing low spatial frequencies. The paper reports median absolute errors around 0.5 in the noise-free perpendicular-field case, generally below 2 even at 100% noise, and below 1 when low spatial frequencies are removed, supporting the claim that the morphologies are robustly correlated with $M_s$.

Load-bearing premise

The model's predictive power depends on the simulated MHD maps being representative of real spectroscopic observations, since the network learns only morphologies that appear in the six idealized isothermal, solenoidally driven simulations used for training.

Editorial extensions

If this is right

  • Combined with CNN-estimated $M_A$, the predicted $M_s$ yields magnetic field strength through $B = c_s\sqrt{4\pi\rho}M_s M_A^{-1}$, enabling magnetic-field mapping in molecular clouds.
  • Using thin velocity channel maps rather than intensity maps makes it possible to predict the 3D distribution of $M_s$, which is needed for 3D Galactic magnetic field reconstruction when combined with the rotation curve.
  • The network remains accurate when low spatial frequencies are removed (median errors below 1 for filters out to wavenumber 30), so it can be applied to interferometric observations with missing short baselines.
  • Adding Gaussian noise up to a noise ratio of 100% keeps median errors below about 2, and intensity maps remain the most accurate input type across noise levels.

Reading between the lines

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

  • If the morphology-to-$M_s$ mapping generalizes, the same CNN architecture could be retrained to output $M_s$ and $M_A$ simultaneously from one set of maps, giving a direct pixel-by-pixel estimate of the compressibility parameter $\beta = 2(M_A/M_s)^2$.
  • A natural test of the paper's hidden assumption is to train on the idealized isothermal simulations and then evaluate on simulations that include radiative transfer, self-gravity, and outflow feedback; the Appendix's unseen-data result suggests the error increase would be measurable and would indicate when retraining is needed.
  • A practical consequence for surveys: the reported median errors of 0.5 to 1.5 imply the method can reliably separate subsonic, transonic, and supersonic regions, but may not resolve fine differences in $M_s$ for precision turbulence studies.
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

3 major / 6 minor

Summary. The paper trains a convolutional neural network on 32x32 subfields drawn from six isothermal, solenoidally driven MHD turbulence simulations, using integrated intensity, velocity centroid, and velocity channel maps as inputs to predict the sonic Mach number Ms. The target label is a local Ms defined through the line-of-sight velocity dispersion of each subfield. The authors report median absolute errors of roughly 0.5-2 across different Ms, magnetic field inclination angles, added Gaussian noise levels, and removal of low spatial frequencies, and conclude that the CNN can predict Ms under a variety of observational conditions. An appendix tests one additional simulation not used in training and reports that prediction errors increase for that unseen simulation.

Significance. If the generalization claim held, this would be a practically useful tool: it would allow spatially resolved Ms estimates from spectroscopic maps without region-by-region PDF fitting, and channel-map predictions could in principle support 3D magnetic field strength measurements. The paper is also a useful proof of concept for applying CNNs to synthetic PPV observations, and it explicitly compares three commonly used map types. However, the quantitative evidence in the main text comes from subfields of the same six simulations used for training, and each robustness experiment retrains a new model for the condition being tested. The single out-of-distribution test in Appendix B shows degraded performance. The central claim is therefore not yet supported at the level stated in the abstract; additional fixed-network or leave-one-out evaluations are needed.

major comments (3)
  1. [Abstract; Sections 3.3-3.5] The abstract claims that the CNN can predict Ms 'under various conditions, including different magnetic fields and levels of observational noise,' but the experiments in Sections 3.3-3.5 retrain the network separately for each condition: Section 3.3 states that the simulation boxes are rotated 'and retrain the CNN model accordingly,' Section 3.4 says the model 'is trained to make predictions' at each noise level, and Section 3.5 applies the k-space filter 'before CNN training.' Figures 5-7 therefore measure how well a freshly fitted model performs on data drawn from the same distribution as its training set, not whether a single trained model transfers across conditions. Real observations will have unknown combinations of inclination, noise, and spatial filtering, so the paper should include a fixed-network evaluation across a parameter sweep, or at minimum a leave-one-condition-out test, before claiming robustness under varied conditions.
  2. [Section 3.2 and Appendix B] The main demonstration of predictive accuracy, including Figure 3 and Figure 4, uses simulation min06 (Ms = 11.02), which Appendix B explicitly identifies as included in the training process. The headline error statistics are therefore in-sample or near-in-sample. The only genuinely held-out test, Appendix B's simulation min07 (Ms = 10.50, MA = 0.81), shows increased prediction error, but the manuscript reports this only as qualitative histograms without quoting the median and quartile errors for seen versus unseen data. Please quantify the degradation for each map type and present this as a central robustness metric rather than a deferred appendix result.
  3. [Section 2.4, Eq. (4)] The training label M_sub_s is defined using the line-of-sight velocity dispersion v_sub in each 32x32 subfield. The paper does not justify that this is an unbiased proxy for the true local sonic Mach number of the subfield. Along a given LOS, the observed velocity dispersion mixes turbulent velocity fluctuations with projection and density-weighting effects, and the relation between v_sub and the actual 3D velocity dispersion may vary with Ms and with magnetic field geometry. Since the label itself is constructed from the same PPV data that provide the input maps, the measured prediction errors partly reflect the correlation between the label and the local statistics of the input rather than an independent physical quantity. Please add a validation against the true 3D velocity dispersion or another independently defined Ms for the subfields, and report the resulting bias.
minor comments (6)
  1. [Section 3.4, first paragraph] The text says 'Fig. 5 presents the absolute error' for the noise analysis, but the figure shown for noise is Figure 6; Figure 5 is the inclination-angle figure. Please correct the cross-reference.
  2. [Abstract and Section 3.4] The abstract states that the median uncertainty ranges from 0.5 to 1.5, while Section 3.4 reports median errors rising to approximately 2-2.25 in the highest noise cases. These numbers should be reconciled or the abstract should be qualified by noise level.
  3. [Section 3.5] The removed spatial frequency ranges '0 - 10, 0 - 20, and 0 - 30' are not defined in physical units; please state the normalization of k (for example, k in units of 2π/L_box) and, for the interferometric motivation, explain how these cutoffs map to baseline coverage.
  4. [Section 2.3 and Figure 1] The architecture description lacks quantitative details: the number and size of convolution kernels, pooling sizes, number of fully connected units, activation functions, optimizer, learning rate, batch size, and the exact training/validation split are not stated. These details are needed to reproduce the results and to assess possible overfitting.
  5. [Section 2.5] The description of random rotation is ambiguous: it should state whether rotations are restricted to multiples of 90 degrees or are arbitrary angles, since arbitrary rotations require interpolation and effectively change the pixel statistics.
  6. [Table 1] There is a formatting error in the caption ('T able 1'), and the superscript/subscript notation for M_sub_s should be made consistent throughout the text and figures.

Circularity Check

3 steps flagged · score 6.0 of 10

Main-text 'predictions' are in-sample or per-condition refits; the only held-out test (Appendix B) shows degraded accuracy, so the robustness claim is partly an artifact of the training protocol.

  1. fitted input called prediction [Section 2.5; Section 3.2; Table 1; Appendix B]
    "the original non-rotated datasets can be used as validation and a method of creating prediction test scenarios."

    The 'prediction test scenarios' are drawn from the same six simulations (Table 1) that supplied the ~0.6 million training subfields; Section 3.2 evaluates on simulation Ms = 11.02 (min06 in Table 1) without a held-out split. The only explicitly unseen simulation, min07 in Appendix B, is reported to have increased error. The headline median uncertainties (0.5-1.5) are therefore in-sample fit residuals, not out-of-sample prediction errors, so calling them 'CNN-predicted Ms' overstates what was measured.

  2. fitted input called prediction [Section 3.3, Fig. 5]
    "we rotate the simulation boxes to achieve four different inclination angles (90, 60, 30, and 0 degrees), generate synthetic spectroscopic data, and retrain the CNN model accordingly."

    A separate CNN is fitted for each inclination angle, and the box-plot errors in Fig. 5 are computed on data from that same inclination. The experiment thus measures how well each condition-specific fit matches its own training condition, not how a single network trained on one geometry transfers to other magnetic-field orientations. The abstract's claim of prediction 'under various conditions, including different magnetic fields' is not established by this protocol.

1 more flagged steps
  1. fitted input called prediction [Section 3.4 and 3.5, Figs. 6-7]
    "This analysis is conducted for three types of input maps—intensity, centroid, and channel maps, all with an inclination angle of 90 degrees—on which the CNN model is trained to make predictions."

    For each noise ratio the model is retrained on that noise level, and in Section 3.5 the k-space filter is applied 'before CNN training,' so each reported error is a per-condition training residual. The claimed robustness to noise and missing low spatial frequencies is therefore a property of refitting a model to each condition, not of a single predictive model transferred across conditions.

full rationale

No circularity is found in the physical derivation: the CNN is a supervised regressor from synthetic spectroscopic maps to a locally defined M_sub^s, and no load-bearing argument reduces to a self-citation or to a definitional identity. The circularity is in the evaluation protocol. The main-text accuracy figures are in-sample (same six simulations used for training, with augmentation) or per-condition refits (Sections 3.3-3.5), so the errors reported are fit residuals rather than predictions. The paper itself supplies the decisive control: Appendix B tests a genuinely unseen simulation (min07) and reports that prediction error increases, which confirms that the earlier numbers were not out-of-sample. This is why the score is 6 rather than higher; the appendix prevents the claim from being entirely tautological. The simulation-realism concerns (radiative transfer, gravity, feedback deferred in Section 4.1) are generalization risks, not circularity, and are not scored here.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the assumption that idealized isothermal MHD simulations represent the observable morphology of the ISM, that the local LOS velocity dispersion is an adequate target label, and that the chosen CNN training setup learns a generalizable mapping. No new physical entities are introduced. The main free parameters are the network hyperparameters and preprocessing choices, which are not fully specified.

free parameters (5)
  • CNN weights (trainable parameters) = learned from ~0.6 million subfields per training iteration
    The prediction is the output of a supervised regressor fitted to simulation-derived labels, so the headline results depend on these fitted weights.
  • CNN architecture hyperparameters = not reported (e.g., optimizer, learning rate, kernel counts)
    The network design is shown in Fig. 1 but the exact hyperparameters needed for reproduction are not listed.
  • Input map normalization = normalized by maximum map value
    Chosen by hand; removes absolute intensity scale and makes morphology the primary signal, which is an assumption about what features carry Ms information.
  • Channel width Delta v = chosen such that Delta v < sqrt(delta v^2); exact value not stated
    Controls how much velocity caustic structure appears in channel maps, affecting the CNN input.
  • Training set size and iterations = ~0.6 million subfields per iteration; at least 20 iterations
    These training choices are empirical and not justified by a convergence criterion beyond MSE saturation.
assumptions (3)
  • domain assumption ZEUS-MP/3D isothermal, solenoidally driven ideal MHD simulations represent the density and velocity structure of the ISM.
    All training and testing data come from one simulation code and one forcing scheme; real ISM includes additional physics acknowledged in Section 4.1.
  • domain assumption The local line-of-sight velocity dispersion in a 32x32 subfield is an adequate proxy for the sonic Mach number of that subfield.
    The ground truth Ms^sub in Eq. 4 is defined this way; it assumes isotropy and that the LOS velocity dispersion reflects the 3D turbulent velocity.
  • domain assumption Morphological features in intensity, centroid, and channel maps are sufficiently determined by Ms that a CNN can learn a mapping that generalizes across magnetic field strengths, noise, and missing spatial frequencies.
    This is the central learning assumption; the paper tests limited variations of these conditions and retrains the CNN for each condition.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Estimate Sonic Mach Number in the Interstellar Medium with Convolutional Neural Network." pith.science (2026). https://pith.science/paper/DQ4W34JT

@misc{pith2026241111157,
  author       = {Pith},
  title        = {Pith review of: Estimate Sonic Mach Number in the Interstellar Medium with Convolutional Neural Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DQ4W34JT}},
  note         = {Machine review of arXiv:2411.11157}
}
abstract

Understanding the role of turbulence in shaping the interstellar medium (ISM) is crucial for studying star formation, molecular cloud evolution, and cosmic ray propagation. Central to this is the measurement of the sonic Mach number ($M_s$), which quantifies the ratio of turbulent velocity to the sound speed. In this work, we introduce a convolutional neural network (CNN)-based approach for estimating $M_s$ directly from spectroscopic observations. The approach leverages the physical correlation between increasing $M_s$ and the shock-induced small-scale fluctuations that alter the morphological features in intensity, velocity centroid, and velocity channel maps. These maps, derived from 3D magnetohydrodynamic (MHD) turbulence simulations, serve as inputs for the CNN training. By learning the relationship between these structural features and the underlying turbulence properties, CNN can predict $M_s$ under various conditions, including different magnetic fields and levels of observational noise. The median uncertainty of the CNN-predicted $M_s$ ranges from 0.5 to 1.5 depending on the noise level. While intensity maps offer lower uncertainty, channel maps have the advantage of predicting the 3D $M_s$ distribution, which is crucial in estimating 3D magnetic field strength. Our results demonstrate that machine-learning-based tools can effectively characterize complex turbulence properties in the ISM.

Figures

Figures reproduced from arXiv: 2411.11157 by the authors.

Figure 1
Figure 1. CNN model architecture utilized in this study. The input image is a 32 × 32 cells map cropped from either a channel, centroid, or intensity map. The CNN model then outputs a prediction of the sonic Mach numbers Ms. Simulation Ms MA Range of Msub s min01 2.17 0.51 1.36 - 2.84 min02 4.35 0.52 2.78 - 6.32 min03 6.53 0.52 3.85 - 8.97 min04 8.55 0.51 5.07 - 11.46 min05 10.86 0.69 6.18 - 13.68 min06 11.02 0.43 7.11 - 15.6… view at source ↗
Figure 2
Figure 2. Maps of the integrated intensity (top), velocity centroid (middle), and thin velocity channel (bottom). Three simulations with different Ms = 2.17, 4.35, and 6.53 are given here for comparison. From left to right Ms is increasing. ity fluctuations, though density fluctuations also con￾tribute. Compared to intensity and centroid maps, thin channel maps exhibit less sensitivity to changes in Ms. 3.2. Sonic Mach Number… view at source ↗
Figure 3
Figure 3. The right column presents a comparison of the CNN-predicted sonic Mach number Ms,CNN distribution from intensity (top), centroid (middle), and channel input maps (bottom) and the actual Msub s distribution in the left column. The supersonic simulation with Ms = 11.02 is used as an example here. highlighting the CNN model’s effectiveness, albeit with some scatter that reflects deviations from the actual val￾ues. The … view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: 2D histograms comparing CNN-predicted Ms,CNN and actual Ms of three simulations Ms = 2.17, 6.53, and 11.02. Ms,CNN is predicted from either intensity map (top), velocity channel map (middle), or centroid map (bottom). The dashed reference line refers to the one-to-one …
Figure 5
Figure 5. Figure 5: Box plots of the absolute difference σMs between CNN-predicted Ms and actual Msub s values for various inclination angles of mean magnetic field. Three simulations Ms = 2.17, 6.53, and 11.02 are used. Ms,CNN is predicted from either intensity map (top), velocity channe…
Figure 6
Figure 6. Figure 6: Box plots of the absolute difference σMs between CNN-predicted Ms and actual Msub s values for various levels of noise. Three simulations Ms = 2.17, 6.53, and 11.02 with an inclination angle of 90 degrees are used. Ms,CNN is predicted from either intensity map (top), v…
Figure 7
Figure 7. Figure 7: Box plots of the absolute difference σMs between CNN-predicted Ms and actual Msub s values for various levels of removed spatial frequencies. Kmin represents minimum wavenumber remaining in the filtered input map and Lbox is the size of simulation box. Each of the thre…
Figure 8
Figure 8. Figure 8: The right column presents a comparison of the CNN-predicted sonic Mach number Ms,CNN distribution from intensity (top), centroid (middle), and channel input maps (bottom) and the actual Msub s distribution in the left column. The supersonic simulation with Ms = 11.02 a…
Figure 9
Figure 9. Figure 9: Histograms comparing the error σMs in predictions from simulation min06 (Ms = 11.02, MA = 0.43) and from the unseen simulation min07 (Ms = 10.50, MA = 0.81). B. TEST WITH UNSEEN DATA [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Characterizing 3D Magnetic Fields and Turbulence in H I Clouds

    astro-ph.GA 2025-05 conditional novelty 5.0 of 10

    A neural network maps H I spectral cubes to 3D magnetic field orientation, strength, sonic and Alfven Mach numbers, and it is applied to two FAST-survey clouds in Monoceros.

Reference graph

Works this paper leans on

77 extracted references · 15 canonical work pages · cited by 1 Pith paper

  1. [1]

    2015, TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems

    Abadi, M., Agarwal, A., Barham, P., et al. 2015, TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems. https://www.tensorflow.org/

  2. [2]

    W., Rickett, B

    Armstrong, J. W., Rickett, B. J., & Spangler, S. R. 1995, ApJ, 443, 209, doi: 10.1086/175515

  3. [3]

    2011, MNRAS, 416, 1436, doi: 10.1111/j.1365-2966.2011.19141.x

    Ballesteros-Paredes, J., V´ azquez-Semadeni, E., Gazol, A., et al. 2011, MNRAS, 416, 1436, doi: 10.1111/j.1365-2966.2011.19141.x

  4. [4]

    R., & Federrath, C

    Beattie, J. R., & Federrath, C. 2019, Monthly Notices of the Royal Astronomical Society, 492, 668, doi: 10.1093/mnras/stz3377

  5. [5]

    R., Krumholz, M

    Beattie, J. R., Krumholz, M. R., Federrath, C., Sampson, M. L., & Crocker, R. M. 2022, Frontiers in Astronomy and Space Sciences, 9, doi: 10.3389/fspas.2022.900900

  6. [6]

    2005, ApJL, 624, L93, doi: 10.1086/430702

    Beresnyak, A., Lazarian, A., & Cho, J. 2005, ApJL, 624, L93, doi: 10.1086/430702

  7. [7]

    M., & Fletcher, A

    Berkhuijsen, E. M., & Fletcher, A. 2008, Monthly Notices of the Royal Astronomical Society: Letters, 390, L19, doi: 10.1111/j.1745-3933.2008.00526.x

  8. [8]

    2019, Astronomy & Astrophysics, 632, A17, doi: 10.1051/0004-6361/201935951

    Bracco, A., Ghosh, T., Boulanger, F., & Aumont, J. 2019, Astronomy & Astrophysics, 632, A17, doi: 10.1051/0004-6361/201935951

Show all 77 references
  1. [9]

    Brunt, C. M. 2010, A&A, 513, A67, doi: 10.1051/0004-6361/200913506

  2. [10]

    2018, ApJ, 863, 118, doi: 10.3847/1538-4357/aad002

    Burkhart, B. 2018, ApJ, 863, 118, doi: 10.3847/1538-4357/aad002

  3. [11]

    2012, The Astrophysical Journal Letters, 755, L19, doi: 10.1088/2041-8205/755/1/L19

    Burkhart, B., & Lazarian, A. 2012, The Astrophysical Journal Letters, 755, L19, doi: 10.1088/2041-8205/755/1/L19

  4. [12]

    2009, The Astrophysical Journal, 708, 1204, doi: 10.1088/0004-637X/708/2/1204

    Burkhart, B., Stanimirovi´ c, S., Lazarian, A., & Kowal, G. 2009, The Astrophysical Journal, 708, 1204, doi: 10.1088/0004-637X/708/2/1204

  5. [13]

    2010, ApJ, 710, 853

    Chepurnov, A., & Lazarian, A. 2010, ApJ, 710, 853

  6. [14]

    Crutcher, R. M. 2004, in The Magnetized Interstellar Medium, ed. B. Uyaniker, W. Reich, & R. Wielebinski, 123–132

  7. [15]

    Crutcher, R. M. 2012, ARA&A, 50, 29, doi: 10.1146/annurev-astro-081811-125514

  8. [16]

    Elmegreen, B. G. 1993, ApJL, 419, L29, doi: 10.1086/187129

  9. [17]

    2010, The Astrophysical Journal, 710, 125, doi: 10.1088/0004-637X/710/1/125

    Esquivel, A., & Lazarian, A. 2010, The Astrophysical Journal, 710, 125, doi: 10.1088/0004-637X/710/1/125

  10. [18]

    2015, MNRAS, 448, 3297, doi: 10.1093/mnras/stv180

    Federrath, C., & Banerjee, S. 2015, MNRAS, 448, 3297, doi: 10.1093/mnras/stv180

  11. [19]

    Federrath, C., & Klessen, R. S. 2012, ApJ, 761, 156, doi: 10.1088/0004-637X/761/2/156

  12. [20]

    S., & Schmidt, W

    Federrath, C., Klessen, R. S., & Schmidt, W. 2008, ApJL, 688, L79, doi: 10.1086/595280 16

  13. [21]

    C., Ruffolo, D., Chuychai, P., et al

    Ghilea, M. C., Ruffolo, D., Chuychai, P., et al. 2011, ApJ, 741, 16, doi: 10.1088/0004-637X/741/1/16

  14. [22]

    1995, ApJ, 438, 763, doi: 10.1086/175121

    Goldreich, P., & Sridhar, S. 1995, ApJ, 438, 763, doi: 10.1086/175121

  15. [23]

    2022, ApJ, 934, 7, doi: 10.3847/1538-4357/ac76bf

    Ha, T., Li, Y., Kounkel, M., et al. 2022, ApJ, 934, 7, doi: 10.3847/1538-4357/ac76bf

  16. [24]

    C., Norman, M

    Hayes, J. C., Norman, M. L., Fiedler, R. A., et al. 2006, The Astrophysical Journal Supplement Series, 165, 188, doi: 10.1086/504594

  17. [25]

    S., Benjamin, R

    Hill, A. S., Benjamin, R. A., Kowal, G., et al. 2008, The Astrophysical Journal, 686, 363, doi: 10.1086/590543

  18. [26]

    Hu, Y., Federrath, C., Xu, S., & Mathew, S. S. 2022, MNRAS, 513, 2100, doi: 10.1093/mnras/stac972

  19. [27]

    2023, Monthly Notices of the Royal Astronomical Society, 524, 2379–2394, doi: 10.1093/mnras/stad1996

    Hu, Y., & Lazarian, A. 2023, Monthly Notices of the Royal Astronomical Society, 524, 2379–2394, doi: 10.1093/mnras/stad1996

  20. [28]

    2024a, ApJ, 975, 66, doi: 10.3847/1538-4357/ad7950 —

    Hu, Y., & Lazarian, A. 2024a, ApJ, 975, 66, doi: 10.3847/1538-4357/ad7950 —. 2024b, arXiv e-prints, arXiv:2410.09294, doi: 10.48550/arXiv.2410.09294

  21. [29]

    Hu, Y., Lazarian, A., Alina, D., Pogosyan, D., & Ho, K. W. 2023, MNRAS, 524, 2994, doi: 10.1093/mnras/stad1924

  22. [30]

    2022, The Astrophysical Journal, 941, 92, doi: 10.3847/1538-4357/ac9df0

    Hu, Y., Lazarian, A., Beck, R., & Xu, S. 2022, The Astrophysical Journal, 941, 92, doi: 10.3847/1538-4357/ac9df0

  23. [31]

    2020, ApJ, 905, 129, doi: 10.3847/1538-4357/abc3c6

    Hu, Y., Lazarian, A., & Bialy, S. 2020, ApJ, 905, 129, doi: 10.3847/1538-4357/abc3c6

  24. [32]

    2021, The Astrophysical Journal, 912, 2, doi: 10.3847/1538-4357/abedb7

    Hu, Y., Lazarian, A., & Stanimirovi´ c, S. 2021, The Astrophysical Journal, 912, 2, doi: 10.3847/1538-4357/abedb7

  25. [33]

    2024, MNRAS, 527, 11240, doi: 10.1093/mnras/stad3766

    Hu, Y., Lazarian, A., Wu, Y., & Fu, C. 2024, MNRAS, 527, 11240, doi: 10.1093/mnras/stad3766

  26. [34]

    2022a, MNRAS, 512, 2111, doi: 10.1093/mnras/stac319

    Hu, Y., Lazarian, A., & Xu, S. 2022a, MNRAS, 512, 2111, doi: 10.1093/mnras/stac319

  27. [35]

    M., & Lazarian, A

    Hu, Y., Xu, S., Stone, J. M., & Lazarian, A. 2022b, ApJ, 941, 133, doi: 10.3847/1538-4357/ac9ebc

  28. [36]

    H., & Lazarian, A

    Hu, Y., Yuen, K. H., & Lazarian, A. 2019a, ApJ, 886, 17, doi: 10.3847/1538-4357/ab4b5e

  29. [37]

    H., Lazarian, V., et al

    Hu, Y., Yuen, K. H., Lazarian, V., et al. 2019b, Nature Astronomy, 3, 776, doi: 10.1038/s41550-019-0769-0

  30. [38]

    Jokipii, J. R. 1966, ApJ, 146, 480, doi: 10.1086/148912

  31. [39]

    2016, Monthly Notices of the Royal Astronomical Society, 461, 1227, doi: 10.1093/mnras/stw1296

    Kandel, D., Lazarian, A., & Pogosyan, D. 2016, Monthly Notices of the Royal Astronomical Society, 461, 1227, doi: 10.1093/mnras/stw1296

  32. [40]

    D., & Kamionkowski, M

    Kovetz, E. D., & Kamionkowski, M. 2015, Phys. Rev. D, 91, 081303, doi: 10.1103/PhysRevD.91.081303

  33. [41]

    2007, The Astrophysical Journal, 658, 423, doi: 10.1086/511515

    Kowal, G., Lazarian, A., & Beresnyak, A. 2007, The Astrophysical Journal, 658, 423, doi: 10.1086/511515

  34. [42]

    R., & McKee, C

    Krumholz, M. R., & McKee, C. F. 2005, ApJ, 630, 250, doi: 10.1086/431734

  35. [43]

    2007, in Proceedings of the 24th international conference on Machine learning, ICML ’07 & ILP ’07 (ACM), doi: 10.1145/1273496.1273556

    Bengio, Y. 2007, in Proceedings of the 24th international conference on Machine learning, ICML ’07 & ILP ’07 (ACM), doi: 10.1145/1273496.1273556

  36. [44]

    Larson, R. B. 1981, MNRAS, 194, 809, doi: 10.1093/mnras/194.4.809

  37. [45]

    2014, SSRv, 181, 1, doi: 10.1007/s11214-013-0031-5

    Lazarian, A. 2014, SSRv, 181, 1, doi: 10.1007/s11214-013-0031-5

  38. [46]

    2012, ApJ, 757, 154, doi: 10.1088/0004-637X/757/2/154

    Lazarian, A., Esquivel, A., & Crutcher, R. 2012, ApJ, 757, 154, doi: 10.1088/0004-637X/757/2/154

  39. [47]

    2000, The Astrophysical Journal, 537, 720, doi: 10.1086/309040

    Lazarian, A., & Pogosyan, D. 2000, The Astrophysical Journal, 537, 720, doi: 10.1086/309040

  40. [48]

    Lazarian, A., & Vishniac, E. T. 1999, ApJ, 517, 700, doi: 10.1086/307233

  41. [49]

    2023, The Astrophysical Journal, 956, 63, doi: 10.3847/1538-4357/acea5c

    Lazarian, A., & Xu, S. 2023, The Astrophysical Journal, 956, 63, doi: 10.3847/1538-4357/acea5c

  42. [50]

    Lazarian, A., & Yuen, K. H. 2018, The Astrophysical Journal, 853, 96, doi: 10.3847/1538-4357/aaa241

  43. [51]

    H., & Pogosyan, D

    Lazarian, A., Yuen, K. H., & Pogosyan, D. 2020, Obtaining magnetic field strength using differential measure approach and velocity channel maps. https://arxiv.org/abs/2002.07996 —. 2022, The Astrophysical Journal, 935, 77, doi: 10.3847/1538-4357/ac6877

  44. [52]

    H., & Pogosyan, D

    Lazarian, A., Yuen, K. H., & Pogosyan, D. 2024, ApJ, 974, 237, doi: 10.3847/1538-4357/ad6d62

  45. [53]

    1998, Proceedings of the IEEE, 86, 2278, doi: 10.1109/5.726791 Mac Low, M.-M., & Klessen, R

    Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. 1998, Proceedings of the IEEE, 86, 2278, doi: 10.1109/5.726791 Mac Low, M.-M., & Klessen, R. S. 2004, Reviews of Modern Physics, 76, 125, doi: 10.1103/RevModPhys.76.125

  46. [54]

    F., & Ostriker, E

    McKee, C. F., & Ostriker, E. C. 2007, Annual Review of Astronomy and Astrophysics, 45, 565, doi: https: //doi.org/10.1146/annurev.astro.45.051806.110602

  47. [55]

    1966, Monthly Notices of the Royal Astronomical Society, 133, 265, doi: 10.1093/mnras/133.2.265

    Mestel, L. 1966, Monthly Notices of the Royal Astronomical Society, 133, 265, doi: 10.1093/mnras/133.2.265

  48. [56]

    1956, MNRAS, 116, 503, doi: 10.1093/mnras/116.5.503

    Mestel, L., & Spitzer, L., J. 1956, MNRAS, 116, 503, doi: 10.1093/mnras/116.5.503

  49. [57]

    2023, A&A, 673, A134, doi: 10.1051/0004-6361/202345845

    Mignon-Risse, R., Gonz´ alez, M., & Commer¸ con, B. 2023, A&A, 673, A134, doi: 10.1051/0004-6361/202345845

  50. [58]

    C., Stone, J

    Ostriker, E. C., Stone, J. M., & Gammie, C. F. 2001, ApJ, 546, 980, doi: 10.1086/318290

  51. [59]

    Padoan, P., Nordlund, A., & Jones, B. J. T. 1997, MNRAS, 288, 145, doi: 10.1093/mnras/288.1.145

  52. [60]

    1998, Phys

    Passot, T., & V´ azquez-Semadeni, E. 1998, Phys. Rev. E, 58, 4501, doi: 10.1103/PhysRevE.58.4501

  53. [61]

    Peek, J. E. G., & Burkhart, B. 2019, ApJL, 882, L12, doi: 10.3847/2041-8213/ab3a9e Planck Collaboration:, Ade, P. A. R., Aghanim, N., et al. 2015, A&A, 576, A104, doi: 10.1051/0004-6361/201424082 17

  54. [62]

    2014, Journal of Physics: Conference Series, 496, 012025, doi: 10.1088/1742-6596/496/1/012025

    Pogosian, L. 2014, Journal of Physics: Conference Series, 496, 012025, doi: 10.1088/1742-6596/496/1/012025

  55. [63]

    2024, Monthly Notices of the Royal Astronomical Society, 528, 3897, doi: 10.1093/mnras/stae146

    Schmaltz, T., Hu, Y., & Lazarian, A. 2024, Monthly Notices of the Royal Astronomical Society, 528, 3897, doi: 10.1093/mnras/stae146

  56. [64]

    2020, IEEE Transactions on Circuits and Systems for Video Technology, 30, 2917–2931, doi: 10.1109/tcsvt.2019.2935128

    Takahashi, R., Matsubara, T., & Uehara, K. 2020, IEEE Transactions on Circuits and Systems for Video Technology, 30, 2917–2931, doi: 10.1109/tcsvt.2019.2935128

  57. [65]

    M., Burkhart, B., & Lazarian, A

    Tofflemire, B. M., Burkhart, B., & Lazarian, A. 2011, The Astrophysical Journal, 736, 60, doi: 10.1088/0004-637X/736/1/60 van Dyk, D. A., & Meng, X.-L. 2001, Journal of Computational and Graphical Statistics, 10, 1–50, doi: 10.1198/10618600152418584 Van Rossum, G., & Drake, F....

  58. [66]

    1994, ApJ, 423, 681, doi: 10.1086/173847 V´ azquez-Semadeni, E., Hu, Y., Xu, S., Guerrero-Gamboa, R., & Lazarian, A

    Vazquez-Semadeni, E. 1994, ApJ, 423, 681, doi: 10.1086/173847 V´ azquez-Semadeni, E., Hu, Y., Xu, S., Guerrero-Gamboa, R., & Lazarian, A. 2024, MNRAS, 530, 3431, doi: 10.1093/mnras/stae1085

  59. [67]

    2018, Frontiers in Astronomy and Space Sciences, 5, doi: 10.3389/fspas.2018.00039

    Wurster, J., & Li, Z.-Y. 2018, Frontiers in Astronomy and Space Sciences, 5, doi: 10.3389/fspas.2018.00039

  60. [68]

    2022, ApJ, 934, 136, doi: 10.3847/1538-4357/ac7c68

    Xu, S. 2022, ApJ, 934, 136, doi: 10.3847/1538-4357/ac7c68

  61. [69]

    2021, ApJ, 910, 88, doi: 10.3847/1538-4357/abe403

    Xu, S., & Hu, Y. 2021, ApJ, 910, 88, doi: 10.3847/1538-4357/abe403

  62. [70]

    2019, ApJ, 878, 157, doi: 10.3847/1538-4357/ab21be

    Xu, S., Ji, S., & Lazarian, A. 2019, ApJ, 878, 157, doi: 10.3847/1538-4357/ab21be

  63. [71]

    2013, ApJ, 779, 140, doi: 10.1088/0004-637X/779/2/140

    Xu, S., & Yan, H. 2013, ApJ, 779, 140, doi: 10.1088/0004-637X/779/2/140

  64. [72]

    2017, ApJ, 835, 2, doi: 10.3847/1538-4357/835/1/2

    Xu, S., & Zhang, B. 2017, ApJ, 835, 2, doi: 10.3847/1538-4357/835/1/2

  65. [73]

    2002, Phys

    Yan, H., & Lazarian, A. 2002, Phys. Rev. Lett., 89, 281102, doi: 10.1103/PhysRevLett.89.281102

  66. [74]

    2008, ApJ, 673, 942, doi: 10.1086/524771

    Yan, H., & Lazarian, A. 2008, ApJ, 673, 942, doi: 10.1086/524771

  67. [75]

    H., Ho, K

    Yuen, K. H., Ho, K. W., & Lazarian, A. 2021, ApJ, 910, 161, doi: 10.3847/1538-4357/abe4d4

  68. [76]

    H., & Lazarian, A

    Yuen, K. H., & Lazarian, A. 2020, ApJ, 898, 65, doi: 10.3847/1538-4357/ab9307

  69. [77]

    2024, arXiv e-prints, arXiv:2411.07080, doi: 10.48550/arXiv.2411.07080

    Zhang, J., Hu, Y., & Lazarian, A. 2024, arXiv e-prints, arXiv:2411.07080, doi: 10.48550/arXiv.2411.07080

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.