pith. sign in

arxiv: 2607.01338 · v1 · pith:DYGPCMNInew · submitted 2026-07-01 · 🌌 astro-ph.GA

IRIS: Deciphering Spectral-Line Imagery of the Galactic Center by Machine-Learning on Simulations

B. L. DuBois , Cara Battersby , Jonah C. Baade , Dani R. Lipman , H Perry Hatchfield , Jack Sullivan , Russell Bentley , Stefan Reissl
show 6 more authors
Ralf S. Klessen Victor F. Ksoll Mattia C. Sormani Zi-Xuan Feng Adam Ginsburg Robin Tress
This is my paper

Pith reviewed 2026-07-03 19:35 UTC · model grok-4.3

classification 🌌 astro-ph.GA
keywords machine learninggalactic centerspectral linessimulations3D structureconvolutional neural networksynthetic observationsCentral Molecular Zone
0
0 comments X

The pith

A neural network trained on galaxy simulations can convert edge-on spectral-line observations of the galactic center into top-down maps.

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

The Central Molecular Zone of the Milky Way appears edge-on from our vantage point, so spectral-line data alone cannot directly reveal its three-dimensional layout. IRIS trains a convolutional neural network on large numbers of synthetic edge-on and top-down image pairs generated from AREPO simulations. The network learns to perform supervised reversion, turning new edge-on inputs into estimated top-down projections. When tested on synthetic data the approach works, and when applied to real SEDIGISM 13CO observations it yields candidate top-down views of the CMZ. The authors present the work as a proof-of-concept that this simulation-trained reversion can help recover the zone's spatial structure.

Core claim

IRIS trains a bespoke deep convolutional neural network exclusively on 100,000 synthetic spectral-line observations produced from AREPO galaxy simulations via the new IRIS-SO code. The trained network performs supervised reversion that maps edge-on inputs to top-down outputs. Applied to SEDIGISM 13CO(2-1) data, the network generates new top-down images of the CMZ, although outputs vary across independent training runs. The authors conclude that these results demonstrate a viable route to recovering the three-dimensional structure of the CMZ through supervised reversion.

What carries the argument

The IRIS convolutional neural network that learns to map edge-on spectral-line images to top-down projections by supervised training on matched pairs of synthetic observations.

If this is right

  • The method produces candidate top-down maps of the CMZ directly from existing SEDIGISM spectral-line cubes.
  • IRIS-SO generates training data up to 10,000 times faster than earlier synthetic-observation codes, enabling larger training sets.
  • Increasing the size and diversity of the simulation training set is expected to reduce variation across model runs.
  • The same supervised-reversion architecture can be retrained on other spectral-line surveys once matching synthetic data exist.

Where Pith is reading between the lines

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

  • If the reversion works, similar networks could be trained to recover additional quantities such as line-of-sight velocities or density gradients rather than only projected morphology.
  • Discrepancies between simulation physics and real interstellar-medium conditions could systematically bias the output maps, so cross-checks against independent tracers would be needed.
  • The open-source release of both the network and IRIS-SO allows other groups to test the approach on different simulation suites or observational datasets.

Load-bearing premise

Synthetic observations generated from AREPO simulations contain enough of the relevant physics, resolution, and noise properties that a model trained only on them will still produce useful reversions when given real telescope data.

What would settle it

If repeated independent training runs applied to the same SEDIGISM data produce top-down maps whose bright features differ by more than the known uncertainty in CMZ morphology, the claimed generalization would fail.

Figures

Figures reproduced from arXiv: 2607.01338 by Adam Ginsburg, B. L. DuBois, Cara Battersby, Dani R. Lipman, H Perry Hatchfield, Jack Sullivan, Jonah C. Baade, Mattia C. Sormani, Ralf S. Klessen, Robin Tress, Russell Bentley, Stefan Reissl, Victor F. Ksoll, Zi-Xuan Feng.

Figure 1
Figure 1. Figure 1: CMZ Overview: A visualization of the CMZ structure. Depicted in the left panel are scaled versions of the x2 orbital fittings determined in the 3D CMZ paper series by D. L. Walker et al. (2025) and D. R. Lipman et al. (2026). In the right panel are ℓ, b and ℓ, v mean projections of the continuum-subtracted Raleigh-Jeans brightness temperature (see subsection 4.10 and subsection 4.9) recorded in the SEDIGIS… view at source ↗
Figure 2
Figure 2. Figure 2: Simulation Overview: A visualization of the AREPO simulations due to Lipman et al. (in prep), as described in subsection 2.1, that we use in generating our training dataset (see [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Synthetic-Observation Radial-Resolution Comparison: [PITH_FULL_IMAGE:figures/full_fig_p032_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: RADMC3D-IRIS Side-by-Side: A comparison of the RADMC-3D (C. P. Dullemond et al. 2012) and IRIS synthetic-observation codes, in the 13CO (2–1) spectral line. We define a set of test conditions, as described in subsection 5.3, such that RADMC-3D and IRIS-SO compute like observations of the same physical tensor (see subsection 2.2 for physical-tensor definitions). The observed simulation is due to Lipman et a… view at source ↗
Figure 5
Figure 5. Figure 5: POLARIS-IRIS Side-by-Side: A comparison of the POLARIS and IRIS synthetic-observation codes, in the 13CO (2–1) spectral line. We define a set of test conditions, as described in subsection 5.4, such that POLARIS and IRIS-SO compute like observations. Since POLARIS directly observes the AREPO Voronoi mesh as opposed to the physical tensor observed by IRIS, this comparison provides insight into the error int… view at source ↗
Figure 6
Figure 6. Figure 6: Synthetic-Observation Optical-Depth Analysis (Line Only): [PITH_FULL_IMAGE:figures/full_fig_p038_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Synthetic Observation With and Without Dust (Standard Opacity): [PITH_FULL_IMAGE:figures/full_fig_p038_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Synthetic Observation With and Without Dust (Opacity Threshold for Visual Impact): [PITH_FULL_IMAGE:figures/full_fig_p039_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Synthetic Observation of the Continuum Temperature: [PITH_FULL_IMAGE:figures/full_fig_p041_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Deviation Below Expectation of the Background Temperature in the OT Level Balance: [PITH_FULL_IMAGE:figures/full_fig_p041_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Deviation Above Expectation of the Background Temperature in the OT Level Balance: [PITH_FULL_IMAGE:figures/full_fig_p042_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Synthetic Observation via the OT vs. LVG Level Balance: [PITH_FULL_IMAGE:figures/full_fig_p042_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Synthetic Observation via the OT vs. LTE Level Balance: [PITH_FULL_IMAGE:figures/full_fig_p043_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Synthetic Observation in Formal vs. Smooth Integration: [PITH_FULL_IMAGE:figures/full_fig_p044_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Density-Tracing vs Synthetic Observation Comparison: [PITH_FULL_IMAGE:figures/full_fig_p045_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: IRIS-SO Speed Testing: We summarize our speed comparisons between IRIS-SO and RADMC-3D, which we describe in subsection 5.11. Each RADMC-3D observation is computed on a single AMD EPYC 7713 CPU, whereas each IRIS observation is computed on one AMD EPYC 7713 CPU plus one 40GB NVIDIA A100 GPU. We ran a total of 81 separate speed tests, running two trials of each test, exploring the effect of varying resolut… view at source ↗
Figure 17
Figure 17. Figure 17: Reverter Architecture: A schematic diagram illustrating the architecture of our custom reversion model, or reverter, as described in subsection 6.6. The reverter is a deep convolutional neural network (CNN) with pixelwise self-attention. The architecture follows a classic encoder-decoder heuristic, mapping from the observed image space to the top-down image space via a fully featural latent space out of w… view at source ↗
Figure 18
Figure 18. Figure 18: Training Loss Convergence [PITH_FULL_IMAGE:figures/full_fig_p054_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Synthetic Reversions: A visualization of successful reversion of synthetic data, as described in subsection 7.2. Each of the six panels shows reversion of a physical tensor (subsection 2.2) generated from our training simulation due to Lipman et al. (in prep) but not included in our training dataset. We applied random distance perturbations with scaling constants sampled from the interval [0.4, 0.6) durin… view at source ↗
Figure 20
Figure 20. Figure 20: Failure Modes: A visualization of some of the failure modes of our trained reversion model, as described in subsection 7.3. In the left panel, we perform a reversion on an AREPO simulation with very different physics than the simulation due to Lipman et al. (in prep) that we use for our training dataset (subsection 2.1). The simulation we choose is a modified recreation of H. P. Hatchfield et al. (2021). … view at source ↗
Figure 21
Figure 21. Figure 21: SEDIGISM Reversions: A visualization of some reversions of the SEDIGISM 13CO (2–1) spectral-line survey of the CMZ (F. Schuller et al. 2021). As described in subsection 7.4, we trained four separate, randomly initialized instances of our reversion model under identical conditions. We then applied all four trained models to the SEDIGISM data separately, generating four separate predictions of the top-down … view at source ↗
read the original abstract

In understanding the 3D structure of the Milky Way's Central Molecular Zone (CMZ), we are limited by our edge-on perspective. Towards addressing this problem, we introduce Imagery Reversion Informed by Simulation (IRIS). IRIS is a novel machine-learning code base featuring a deep convolutional neural network (CNN), which we have designed to translate edge-on observations of our Milky Way Galaxy into top-down images by training on data generated from AREPO galaxy simulations and synthetic observations of those simulations. We develop a large custom dataset on which we train our bespoke model, and then test the trained model on synthetic data to probe the potential of this machine-learning method, which we call supervised reversion. We then apply our trained model to real observations from the SEDIGISM 13CO(2-1) survey, yielding new top-down views of our CMZ. Though our SEDIGISM reversions are not fully consistent across model training runs, we posit that this lack of convergence can be alleviated by expansion of the training dataset. We argue that these results represent a strong proof-of-concept for the use of supervised reversion to decipher our CMZ's 3D structure. Crucial in generating our training dataset's 100k synthetic observations, we introduce IRIS Synthetic Observation (IRIS-SO), a new GPU-accelerated and fully differentiable code implemented in PyTorch for the non-LTE synthetic observation of spectral lines and dust. We find that IRIS-SO provides up to 10,000x speedups in comparison to the synthetic-observation code RADMC-3D. We release all the IRIS code open-source at https://github.com/bldubois/IRIS.

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

2 major / 0 minor

Summary. The paper introduces IRIS, a CNN-based supervised reversion technique trained on synthetic spectral-line observations generated from AREPO simulations via a new GPU-accelerated, differentiable code IRIS-SO. After testing on held-out synthetic data, the model is applied to real SEDIGISM 13CO(2-1) observations to produce top-down views of the CMZ; the authors note run-to-run inconsistencies in these reversions but argue the work constitutes a strong proof-of-concept for recovering 3D structure, and they release the full codebase openly.

Significance. If the supervised reversion approach can be made robust, it would offer a new route to inferring three-dimensional Galactic Center structure from edge-on data. The reported 10,000x speedup of IRIS-SO over RADMC-3D and the open-source release of IRIS constitute concrete, reusable contributions to the community.

major comments (2)
  1. [Abstract] Abstract: the assertion of a 'strong proof-of-concept' for deciphering CMZ 3D structure is load-bearing on the reliability of the SEDIGISM reversions, yet the text states that 'our SEDIGISM reversions are not fully consistent across model training runs' without providing quantitative measures of that variability (e.g., pixel-wise standard deviation or structural similarity indices across runs). This directly weakens the central claim.
  2. [Abstract] Abstract: no quantitative validation metrics, error bars, or cross-validation results are reported for the application to real SEDIGISM data, only the qualitative statement of inconsistency. Without such metrics it is impossible to assess whether the learned mapping generalizes beyond the specific synthetic training realizations to the target domain.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important aspects of how the SEDIGISM results are presented. We address each major comment below and will revise the manuscript to incorporate quantitative measures of variability where feasible.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion of a 'strong proof-of-concept' for deciphering CMZ 3D structure is load-bearing on the reliability of the SEDIGISM reversions, yet the text states that 'our SEDIGISM reversions are not fully consistent across model training runs' without providing quantitative measures of that variability (e.g., pixel-wise standard deviation or structural similarity indices across runs). This directly weakens the central claim.

    Authors: We agree that quantitative measures would strengthen the presentation of the SEDIGISM results and the associated claim. In the revised manuscript we will report pixel-wise standard deviations computed across multiple independent training runs on the SEDIGISM data, together with structural similarity index (SSIM) values between the resulting top-down maps. These additions will provide an objective assessment of run-to-run consistency and better support the proof-of-concept statement. revision: yes

  2. Referee: [Abstract] Abstract: no quantitative validation metrics, error bars, or cross-validation results are reported for the application to real SEDIGISM data, only the qualitative statement of inconsistency. Without such metrics it is impossible to assess whether the learned mapping generalizes beyond the specific synthetic training realizations to the target domain.

    Authors: We acknowledge that the manuscript currently provides only a qualitative statement of inconsistency for the real-data application. Because no ground truth exists for the three-dimensional CMZ structure, conventional validation metrics or cross-validation against observations are not possible. The run-to-run variability will now be quantified (as described in the response to the first comment) and presented as a practical uncertainty estimate. We will also expand the discussion to explicitly address the domain-shift limitations between synthetic training data and real SEDIGISM observations, while retaining the proof-of-concept framing for the overall supervised-reversion approach. revision: partial

Circularity Check

0 steps flagged

No significant circularity; empirical ML pipeline is self-contained

full rationale

The paper's derivation chain consists of generating synthetic observations from AREPO simulations via the new IRIS-SO code, training a CNN on that dataset, validating on held-out synthetics, and applying the model to real SEDIGISM data. None of the load-bearing steps reduce by construction to the inputs: the CNN outputs on real data are not fitted parameters renamed as predictions, no self-citation chain justifies a uniqueness theorem or ansatz, and the method does not define its target via its own results. The reported run-to-run inconsistency on real data is presented as an empirical observation rather than concealed, and the proof-of-concept claim rests on external performance metrics rather than tautological equivalence. This is the standard case of an independent empirical pipeline.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The approach depends on the domain assumption that AREPO simulations plus IRIS-SO synthetic observations sufficiently represent real CMZ conditions for the CNN to learn a transferable mapping; the CNN itself contains many fitted weights but these are standard for the method.

axioms (1)
  • domain assumption AREPO galaxy simulations plus non-LTE synthetic observations accurately capture the relevant physics and observational effects of the real CMZ.
    Invoked when training the model on synthetic data and applying it to real SEDIGISM observations.
invented entities (1)
  • IRIS-SO independent evidence
    purpose: GPU-accelerated differentiable code for non-LTE synthetic observation of spectral lines and dust.
    New tool introduced to generate the 100k synthetic observations; independent evidence is the claimed 10,000x speedup over RADMC-3D.

pith-pipeline@v0.9.1-grok · 5905 in / 1391 out tokens · 23604 ms · 2026-07-03T19:35:58.059007+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

119 extracted references · 90 canonical work pages · 32 internal anchors

  1. [1]

    3D CMZ. V. A New Orbital Model of Our Galaxy's Center, Informed by Data Across the Electromagnetic Spectrum. , keywords =. doi:10.3847/1538-4357/ae561f , adsurl =

    work page doi:10.3847/1538-4357/ae561f

  2. [2]

    , keywords =

    Cosmological smoothed particle hydrodynamics simulations: a hybrid multiphase model for star formation. , keywords =. doi:10.1046/j.1365-8711.2003.06206.x , archivePrefix =. astro-ph/0206393 , primaryClass =

    work page doi:10.1046/j.1365-8711.2003.06206.x 2003

  3. [3]

    Revista Mexicana de Astronomia y Astrofisica Conference Series , year = 2008, series =

    Massive Star Evolution with Mass Loss and Rotation. Revista Mexicana de Astronomia y Astrofisica Conference Series , year = 2008, series =

    2008

  4. [4]

    , keywords =

    The SEDIGISM survey: First Data Release and overview of the Galactic structure. , keywords =. doi:10.1093/mnras/staa2369 , archivePrefix =. 2012.01527 , primaryClass =

    work page doi:10.1093/mnras/staa2369 2012

  5. [5]

    , year = 1957, month = nov, volume =

    A Mechanism of Non-Thermal Radio-Noise Origin. , year = 1957, month = nov, volume =. doi:10.1086/146421 , adsurl =

    work page doi:10.1086/146421 1957

  6. [6]

    An atomic and molecular database for analysis of submillimetre line observations

    An atomic and molecular database for analysis of submillimetre line observations. , keywords =. doi:10.1051/0004-6361:20041729 , archivePrefix =. astro-ph/0411110 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361:20041729

  7. [7]

    Quantum Calculation of Inelastic CO Collisions with H. II. Pure Rotational Quenching of High Rotational Levels. , keywords =. doi:10.1088/0004-637X/811/1/27 , archivePrefix =. 1511.04981 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/811/1/27

  8. [8]

    , keywords =

    BASECOL2023 scientific content. , keywords =. doi:10.1051/0004-6361/202348233 , adsurl =

    work page doi:10.1051/0004-6361/202348233

  9. [9]

    , keywords =

    Dust opacities for protostellar cores. , keywords =

  10. [10]

    , keywords =

    The complex multiscale structure in simulated and observed emission maps of the proto-cluster cloud G0.253+0.016 ('the Brick'). , keywords =. doi:10.1093/mnras/stad229 , archivePrefix =. 2104.09558 , primaryClass =

    work page doi:10.1093/mnras/stad229

  11. [11]

    arXiv e-prints , keywords =

    ALMA Central Molecular Zone Exploration Survey (ACES) I: Overview. arXiv e-prints , keywords =. doi:10.48550/arXiv.2602.20340 , archivePrefix =. 2602.20340 , primaryClass =

    work page doi:10.48550/arxiv.2602.20340

  12. [12]

    , keywords =

    The existence and shapes of dust lanes in galactic bars. , keywords =. doi:10.1093/mnras/259.2.345 , adsurl =

    work page doi:10.1093/mnras/259.2.345

  13. [13]

    GALACTICNUCLEUS: A high-angular-resolution JHK _ s imaging survey of the Galactic centre. IV. Extinction maps and de-reddened photometry. , keywords =. doi:10.1051/0004-6361/202140996 , archivePrefix =. 2107.00021 , primaryClass =

    work page doi:10.1051/0004-6361/202140996

  14. [14]

    SOFIA/HAWC+ Far-Infrared Polarimetric Large Area CMZ Exploration Survey. I. General Results from the Pilot Program. , keywords =. doi:10.3847/1538-4357/ad12b9 , archivePrefix =. 2306.01681 , primaryClass =

    work page doi:10.3847/1538-4357/ad12b9

  15. [15]

    ATLASGAL - The APEX Telescope Large Area Survey of the Galaxy at 870 microns

    ATLASGAL - The APEX telescope large area survey of the galaxy at 870 m. , keywords =. doi:10.1051/0004-6361/200811568 , archivePrefix =. 0903.1369 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/200811568

  16. [16]

    The HI/OH/Recombination line survey of the inner Milky Way (THOR)

    The HI/OH/Recombination line survey of the inner Milky Way (THOR). Survey overview and data release 1. , keywords =. doi:10.1051/0004-6361/201629143 , archivePrefix =. 1609.03329 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201629143

  17. [17]

    The VLA Galactic Plane Survey

    The VLA Galactic Plane Survey. , keywords =. doi:10.1086/505940 , archivePrefix =. astro-ph/0605422 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/505940

  18. [18]

    , keywords =

    Star Formation in Nuclear Rings with the TIGRESS Framework. , keywords =. doi:10.3847/1538-4357/abfa93 , archivePrefix =. 2104.10349 , primaryClass =

    work page doi:10.3847/1538-4357/abfa93

  19. [19]

    arXiv e-prints , keywords =

    The effects of turbulence on galactic nuclear gas rings. arXiv e-prints , keywords =. doi:10.48550/arXiv.2010.04170 , archivePrefix =. 2010.04170 , primaryClass =

    work page doi:10.48550/arxiv.2010.04170 2010

  20. [20]

    , keywords =

    Understanding the kinematics of Galactic Centre gas. , keywords =. doi:10.1093/mnras/252.2.210 , adsurl =

    work page doi:10.1093/mnras/252.2.210

  21. [21]

    Gas flow in barred potentials

    Gas flow in barred potentials. , keywords =. doi:10.1093/mnras/stv441 , archivePrefix =. 1502.02740 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stv441

  22. [22]

    Recognizing the fingerprints of the Galactic bar: a quantitative approach to comparing model (l,v) distributions to observation

    Recognizing the fingerprints of the Galactic bar: a quantitative approach to comparing model (l, v) distributions to observations. , keywords =. doi:10.1093/mnras/stu2316 , archivePrefix =. 1410.8073 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stu2316

  23. [23]

    , keywords =

    Orbits in barred galaxies. , keywords =. doi:10.1007/BF00873080 , adsurl =

    work page doi:10.1007/bf00873080

  24. [24]

    Rhea-RT: Dynamical impact of Central Molecular Zone conditions on the properties of the interstellar medium and stellar feedback coupling

    Rhea-RT: Dynamical impact of Central Molecular Zone conditions on ISM properties and stellar feedback coupling. arXiv e-prints , keywords =. doi:10.48550/arXiv.2512.09981 , archivePrefix =. 2512.09981 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2512.09981

  25. [25]

    Dense gas in the Galactic central molecular zone is warm and heated by turbulence

    Dense gas in the Galactic central molecular zone is warm and heated by turbulence. , keywords =. doi:10.1051/0004-6361/201526100 , archivePrefix =. 1509.01583 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201526100

  26. [26]

    The Dense Gas Fraction in Galactic Center Clouds

    The Dense Gas Fraction in Galactic Center Clouds. , keywords =. doi:10.3847/1538-4357/aae581 , archivePrefix =. 1810.00266 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-4357/aae581

  27. [27]

    Molecular gas kinematics within the central 250 pc of the Milky Way

    Molecular gas kinematics within the central 250 pc of the Milky Way. , keywords =. doi:10.1093/mnras/stw121 , archivePrefix =. 1601.03732 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stw121

  28. [28]

    The Bolocam Galactic Plane Survey. IX. Data Release 2 and Outer Galaxy Extension. , keywords =. doi:10.1088/0067-0049/208/2/14 , archivePrefix =. 1305.6622 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0067-0049/208/2/14

  29. [29]

    A 100-parsec elliptical and twisted ring of cold and dense molecular clouds revealed by Herschel around the Galactic Center

    A 100 pc Elliptical and Twisted Ring of Cold and Dense Molecular Clouds Revealed by Herschel Around the Galactic Center. , keywords =. doi:10.1088/2041-8205/735/2/L33 , archivePrefix =. 1105.5486 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/2041-8205/735/2/l33 2041

  30. [30]

    GLIMPSE. I. An SIRTF Legacy Project to Map the Inner Galaxy. , eprint =. doi:10.1086/376696 , adsurl =

    work page doi:10.1086/376696

  31. [31]

    3D CMZ. III. Constraining the 3D Structure of the Central Molecular Zone via Molecular Line Emission and Absorption. , keywords =. doi:10.3847/1538-4357/adb5ef , archivePrefix =. 2410.17320 , primaryClass =

    work page doi:10.3847/1538-4357/adb5ef

  32. [32]

    Protostars and Planets VII , year = 2023, editor =

    Star Formation in the Central Molecular Zone of the Milky Way. Protostars and Planets VII , year = 2023, editor =. doi:10.48550/arXiv.2203.11223 , archivePrefix =. 2203.11223 , primaryClass =

    work page doi:10.48550/arxiv.2203.11223 2023

  33. [33]

    Galactic-Center Molecular Arms, Ring, and Expanding Shell. I. Kinematical Structures in Longitude--Velocity Diagrams. , keywords =

  34. [34]

    , keywords =

    Three-dimensional structure of the central molecular zone. , keywords =. doi:10.1093/mnras/stac2243 , archivePrefix =. 2208.02451 , primaryClass =

    work page doi:10.1093/mnras/stac2243

  35. [35]

    , keywords =

    The Galactic Center arms inferred from the ALMA CMZ Exploration Survey (ACES). , keywords =. doi:10.1093/pasj/psaf034 , archivePrefix =. 2504.03331 , primaryClass =

    work page doi:10.1093/pasj/psaf034

  36. [36]

    Orbital structure and evolutionary timeline

    The dynamical evolution of molecular clouds near the Galactic Centre - I. Orbital structure and evolutionary timeline. , archivePrefix = "arXiv", eprint =. doi:10.1093/mnras/stu2526 , adsurl =

    work page doi:10.1093/mnras/stu2526

  37. [37]

    Gas dynamics

    Simulations of the Milky Way's central molecular zone - I. Gas dynamics. , keywords =. doi:10.1093/mnras/staa3120 , archivePrefix =. 2004.06724 , primaryClass =

    work page doi:10.1093/mnras/staa3120 2004

  38. [38]

    3D CMZ. I. Central Molecular Zone Overview. , keywords =. doi:10.3847/1538-4357/adb5f0 , archivePrefix =. 2410.17334 , primaryClass =

    work page doi:10.3847/1538-4357/adb5f0

  39. [39]

    3D CMZ. II. Hierarchical Structure Analysis of the Central Molecular Zone. , keywords =. doi:10.3847/1538-4357/adb844 , archivePrefix =. 2410.17332 , primaryClass =

    work page doi:10.3847/1538-4357/adb844

  40. [40]

    3D CMZ. IV. Distinguishing Near versus Far Distances in the Galactic Center Using Spitzer and Herschel. , keywords =. doi:10.3847/1538-4357/adb5ee , archivePrefix =. 2410.17321 , primaryClass =

    work page doi:10.3847/1538-4357/adb5ee

  41. [41]

    Inflow, star formation, and winds

    The life cycle of the Central Molecular Zone - I. Inflow, star formation, and winds. , keywords =. doi:10.1093/mnras/stz2880 , archivePrefix =. 1905.01309 , primaryClass =

    work page doi:10.1093/mnras/stz2880 1905

  42. [42]

    Star formation

    Simulations of the Milky Way's Central Molecular Zone - II. Star formation. , keywords =. doi:10.1093/mnras/staa1999 , archivePrefix =. 2004.06731 , primaryClass =

    work page doi:10.1093/mnras/staa1999 2004

  43. [43]

    , keywords =

    Dynamically Driven Inflow onto the Galactic Center and its Effect upon Molecular Clouds. , keywords =. doi:10.3847/1538-4357/ac1e89 , archivePrefix =. 2106.08461 , primaryClass =

    work page doi:10.3847/1538-4357/ac1e89

  44. [44]

    , keywords =

    Magnetic field morphology and evolution in the Central Molecular Zone and its effect on gas dynamics. , keywords =. doi:10.1051/0004-6361/202450035 , archivePrefix =. 2403.13048 , primaryClass =

    work page doi:10.1051/0004-6361/202450035

  45. [45]

    , keywords =

    E pur si muove: Galilean-invariant cosmological hydrodynamical simulations on a moving mesh. , keywords =. doi:10.1111/j.1365-2966.2009.15715.x , archivePrefix =. 0901.4107 , primaryClass =

    work page doi:10.1111/j.1365-2966.2009.15715.x 2009

  46. [46]

    M., Steeghs, D

    Magnetohydrodynamics on an unstructured moving grid. , keywords =. doi:10.1111/j.1365-2966.2011.19591.x , archivePrefix =. 1108.1792 , primaryClass =

    work page doi:10.1111/j.1365-2966.2011.19591.x 2011

  47. [47]

    Simulations of magnetic fields in isolated disk galaxies

    Simulations of magnetic fields in isolated disc galaxies. , keywords =. doi:10.1093/mnras/stt428 , archivePrefix =. 1212.1452 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stt428

  48. [48]

    Improving the convergence properties of the moving-mesh code AREPO

    Improving the convergence properties of the moving-mesh code AREPO. , keywords =. doi:10.1093/mnras/stv2380 , archivePrefix =. 1503.00562 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stv2380

  49. [49]

    , keywords =

    The AREPO Public Code Release. , keywords =. doi:10.3847/1538-4365/ab908c , archivePrefix =. 1909.04667 , primaryClass =

    work page doi:10.3847/1538-4365/ab908c 1909

  50. [50]

    , keywords =

    Testing kinematic distances under a realistic Galactic potential: Investigating systematic errors in the kinematic distance method arising from a non-axisymmetric potential. , keywords =. doi:10.1051/0004-6361/202450000 , archivePrefix =. 2403.18000 , primaryClass =

    work page doi:10.1051/0004-6361/202450000

  51. [51]

    arXiv e-prints , keywords =

    Introducing the Rhea simulations of Milky-Way-like galaxies I: Effect of gravitational potential on morphology and star formation. arXiv e-prints , keywords =. doi:10.48550/arXiv.2502.02646 , archivePrefix =. 2502.02646 , primaryClass =

    work page doi:10.48550/arxiv.2502.02646

  52. [52]

    M., Steeghs, D

    Approximations for modelling CO chemistry in giant molecular clouds: a comparison of approaches. , keywords =. doi:10.1111/j.1365-2966.2011.20260.x , archivePrefix =. 1102.0670 , primaryClass =

    work page doi:10.1111/j.1365-2966.2011.20260.x 2011

  53. [53]

    Simulating the Formation of Molecular Clouds. I. Slow Formation by Gravitational Collapse from Static Initial Conditions. , keywords =. doi:10.1086/512238 , archivePrefix =. astro-ph/0605120 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/512238

  54. [54]

    Simulating the Formation of Molecular Clouds. II. Rapid Formation from Turbulent Initial Conditions. , keywords =. doi:10.1086/512227 , archivePrefix =. astro-ph/0605121 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/512227

  55. [55]

    , keywords =

    The Dynamics of Low-Mass Molecular Clouds in External Radiation Fields. , keywords =. doi:10.1086/304167 , adsurl =

    work page doi:10.1086/304167

  56. [56]

    Non-Equilibrium Chemistry and Destruction of CO by X-ray Flares

    Non-equilibrium chemistry and destruction of CO by X-ray flares. , keywords =. doi:10.1093/mnras/stz902 , archivePrefix =. 1803.10367 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stz902

  57. [57]

    Mass inflow rate into the Central Molecular Zone: observational determination and evidence of episodic accretion

    Mass inflow rate into the Central Molecular Zone: observational determination and evidence of episodic accretion. , keywords =. doi:10.1093/mnras/stz046 , archivePrefix =. 1901.00867 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stz046 1901

  58. [58]

    , keywords =

    Nuclear rings are the inner edge of a gap around the Lindblad Resonance. , keywords =. doi:10.1093/mnras/stae082 , archivePrefix =. 2309.14093 , primaryClass =

    work page doi:10.1093/mnras/stae082

  59. [59]

    Radiative Processes in Astrophysics

  60. [60]

    , keywords =

    The size distribution of interstellar grains. , keywords =. doi:10.1086/155591 , adsurl =

    work page doi:10.1086/155591

  61. [61]

    Disentangling the independently controllable factors of variation by interacting with the world

    Interstellar Dust Grains. , keywords =. doi:10.1146/annurev.astro.41.011802.094840 , archivePrefix =. astro-ph/0304489 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1146/annurev.astro.41.011802.094840

  62. [62]

    Assessment of Models of Galactic Thermal Dust Emission Using COBE/FIRAS and COBE/DIRBE Observations

    Assessment of Models of Galactic Thermal Dust Emission Using COBE/FIRAS and COBE/DIRBE Observations. , keywords =. doi:10.3847/0004-637X/828/1/16 , archivePrefix =. 1606.08783 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/0004-637x/828/1/16

  63. [63]

    , keywords =

    CMZoom: Survey Overview and First Data Release. , keywords =. doi:10.3847/1538-4365/aba18e , archivePrefix =. 2007.05023 , primaryClass =

    work page doi:10.3847/1538-4365/aba18e 2007

  64. [64]

    A computer program for fast non-LTE analysis of interstellar line spectra

    A computer program for fast non-LTE analysis of interstellar line spectra. With diagnostic plots to interpret observed line intensity ratios. , keywords =. doi:10.1051/0004-6361:20066820 , archivePrefix =. 0704.0155 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361:20066820

  65. [65]

    doi:10.4159/harvard.9780674864658 , adsurl =

    Moving Envelopes of Stars. doi:10.4159/harvard.9780674864658 , adsurl =

    work page doi:10.4159/harvard.9780674864658

  66. [66]

    Astronomy and Computing , keywords =

    Magritte, a modern software library for spectral line radiative transfer. Astronomy and Computing , keywords =. doi:10.1016/j.ascom.2024.100889 , archivePrefix =. 2411.03040 , primaryClass =

    work page doi:10.1016/j.ascom.2024.100889 2024

  67. [67]

    arXiv e-prints , keywords =

    Revealing Fine Structure in Protoplanetary Disks with Physics Constrained Neural Fields. arXiv e-prints , keywords =. doi:10.48550/arXiv.2509.03623 , archivePrefix =. 2509.03623 , primaryClass =

    work page doi:10.48550/arxiv.2509.03623

  68. [68]

    arXiv e-prints , keywords =

    THOR: a GPU-accelerated and MPI-parallel radiative transfer code. arXiv e-prints , keywords =. doi:10.48550/arXiv.2507.11603 , archivePrefix =. 2507.11603 , primaryClass =

    work page doi:10.48550/arxiv.2507.11603

  69. [69]

    , year = 1994, month = jan, volume =

    Abundances in the Interstellar Medium. , year = 1994, month = jan, volume =. doi:10.1146/annurev.aa.32.090194.001203 , adsurl =

    work page doi:10.1146/annurev.aa.32.090194.001203 1994

  70. [70]

    Ultraviolet Detection of Interstellar 12C17O and the CO Isotopomeric Ratios toward X Persei

    Ultraviolet Detection of Interstellar ^ 12 C ^ 17 O and the CO Isotopomeric Ratios toward X Persei. , keywords =. doi:10.1086/342501 , archivePrefix =. astro-ph/0206449 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1086/342501

  71. [71]

    , keywords =

    12C/ 13C Isotope Ratio across the Galaxy from Observations of 13C 18O in Molecular Clouds. , keywords =. doi:10.1086/168935 , adsurl =

    work page doi:10.1086/168935

  72. [72]

    Tracing gas accretion in the Galactic center using isotopic ratios

    Tracing gas accretion in the Galactic center using isotopic ratios. , keywords =. doi:10.1051/0004-6361/201015008 , archivePrefix =. 1008.1873 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201015008

  73. [73]

    doi:10.1007/978-3-642-39950-3 , adsurl =

    Tools of Radio Astronomy. doi:10.1007/978-3-642-39950-3 , adsurl =

    work page doi:10.1007/978-3-642-39950-3

  74. [74]

    1980 , publisher =

    Introduction to numerical analysis , author =. 1980 , publisher =

    1980

  75. [75]

    1996 , doi =

    Hairer, Ernst and Wanner, Gerhard , title =. 1996 , doi =

    1996

  76. [76]

    , year = 1974, month = jun, volume =

    Aperture Synthesis with a Non-Regular Distribution of Interferometer Baselines. , year = 1974, month = jun, volume =

    1974

  77. [77]

    Radiative transfer with POLARIS. I. Analysis of magnetic fields through synthetic dust continuum polarization measurements. , keywords =. doi:10.1051/0004-6361/201424930 , archivePrefix =. 1604.05305 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201424930

  78. [78]

    Radiative Transfer with POLARIS. II. Modeling of Synthetic Galactic Synchrotron Observations. , keywords =. doi:10.3847/1538-4357/ab3664 , archivePrefix =. 1907.12617 , primaryClass =

    work page doi:10.3847/1538-4357/ab3664 1907

  79. [79]

    Magnetic fields in molecular clouds: Limitations of the analysis of Zeeman observations

    Magnetic fields in molecular clouds: Limitations of the analysis of Zeeman observations. , keywords =. doi:10.1051/0004-6361/201629001 , archivePrefix =. 1703.02745 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201629001

  80. [80]

    Journal of Geophysical Research (Atmospheres) , keywords =

    Optical constants of ice from the ultraviolet to the microwave: A revised compilation. Journal of Geophysical Research (Atmospheres) , keywords =. doi:10.1029/2007JD009744 , adsurl =

    work page doi:10.1029/2007jd009744

Showing first 80 references.