Electron temperature and emission measure of HII regions in the central molecular zone (CMZ) from H40 alpha recombination line and continuum emissions by ALMA CMZ Exploration Survey - ACES -

The paper derives the overall distribution of HII regions in the Central Molecular Zone (CMZ) and their fundamental properties—electron temperature (Te) and emission measure (EM), and hence electron density—by analyzing the ACES (ALMA CMZ Exploration Survey) H40α (99.02 GHz) recombination line and 99.6 GHz continuum emission data with synthesized beam widths of 2″.45 (0.097 pc at 8.2 kpc) and 2″.14, respectively. The authors apply the “TeEM” method (Te–EM mapping), which creates Te and EM maps from input 2D maps of the continuum and integrated line intensity. The analysis covers the entire ACES field from l ∼ −0°.6 to +0°.8 and from b ∼ −0°.2 to +0°.1. The area analyzed is complete and includes previously known HII regions such as Sgr B2, Sgr B1, the Sickle, the Pistol, thermal filaments (Bridges), Sgr A HII regions, the Minispiral, and many other known HII regions. Sgr C is not included in the analysis due to the insufficient signal-to-noise ratio in the recombination line map.

The mean electron temperature over the CMZ is determined to be ⟨Te⟩CMZ = 5872 ± 78 (SE) ± 3682 (SD) K (SE: standard error of the mean, SD: pixel-to-pixel standard deviation). Some HII regions, such as Sgr B2 Main and the Minispiral, exhibit large scatter and an internal Te gradient of several thousand K per parsec. The EM distribution is more diverse, varying by orders of magnitude from ∼ 10 5 to ∼ 3 × 10 8 pc cm−6 within the CMZ, as well as within individual HII regions.

The paper introduces the TeEM algorithm, which uses 2D maps of continuum intensity, line peak intensity, and line integrated intensity to create continuum-to-line intensity ratio, velocity width, Te, and EM maps. The method skips the sophisticated and time-consuming analysis of fitting individual line profiles. The authors discuss limitations including effects of multiple velocity components, smaller contamination of synchrotron emission at higher frequencies (an advantage of mm-waves), contamination by dust continuum emission (masked where spectral index α > +2), side lobes, contamination by molecular lines, and edging effects of the line intensity map.

For individual regions, the paper reports: Sgr B2 Main has Te = 8926 ± 139 K with high EM ∼ 7 × 10 7 pc cm−6, and a steep Te gradient of dTe/dx ≃ 3000 K pc−1 with an extremely high peak EM ∼ 3 × 10 8 pc cm−6, corresponding to electron density as high as ne ∼ 4 × 10 4 cm−3. Sgr B1 has Te = 4100–7800 K with EM ∼ 1–6 × 10 5 pc cm−6. The Sickle has Te = 5800–7900 K with EM ∼ 2 × 10 5 pc cm−6, while the Pistol has Te = 4800 K with EM ∼ 4 × 10 5 pc cm−6 and ne ∼ 2000 cm−3. The thermal filaments (Bridges) have Te ∼ 4800–6000 K and EM ∼ 2 × 10 6 pc cm−6. Sgr A HII regions have Te ∼ 5300 K in the north-east clump, ∼ 6900 K in the companion, and ∼ 4250 K in the south-east clump, with the brightest north-east HII region having EM ∼ (1.46 ± 0.04) × 10 6 pc cm−6. The Minispiral shows the eastern arm as cool as Te ∼ 2542 ± 33 K, the western arm at 4712 ± 35 K, and the central vertical (NS) arm closest to Sgr A∗ with extraordinarily high Te = 11980 ± 324 K and EM ∼ 3 × 10 6 pc cm−6, increasing to ∼ 3 × 10 7 pc cm−6 toward the nucleus. Sgr A∗ appears as a sharp absorption hole in the H40α peak intensity map, with optical depth τ ∼ 4 × 10−3.

The paper also discusses the radial variation of Te as a function of projected distance from Sgr A∗, finding Te nearly constant at ∼ 6000 K within the CMZ, and combining with wider Galactic disk studies to show Te increases from the GC value up to ∼ 10 4 K in the outer regions at a rate of dTe/dR ≃ +300 K kpc−1, approximated by Te ∼ 5872 + 300(R/kpc) [K]. The distribution of HII regions is extremely asymmetric about Sgr A∗, with no significant HII region at negative longitudes except Sgr C (which is not detected in the analysis). The paper concludes that the global electron temperature is relatively constant at ⟨Te⟩CMZ = 5872 ± 78 (SE) ± 3682 (SD) K inside the CMZ, while individual HII regions exhibit higher-order variation across each region, with the highest Te ∼ 12000–18000 K recorded toward the Minispiral along the North-South arm closest to Sgr A∗, and the highest EM ∼ 2 × 10 8 pc cm−6 recorded towards Sgr B2 Main.

Improvements to AI Systems Based on This Paper:

1. Physics-constrained image-to-image translation for radio astronomy maps

• Train a neural network to directly convert 2D continuum + integrated line intensity maps into Te and EM maps, using the TeEM algorithm as a differentiable physics layer. The AI can then predict Te/EM for new fields (e.g., Sgr C or external galaxies) without per-pixel spectral fitting, reducing compute by orders of magnitude while preserving accuracy.

2. Uncertainty-aware segmentation and classification of HII regions

• Build a model that automatically detects and classifies HII regions (e.g., Sgr B2, Minispiral, Sickle) from low-SNR recombination line maps, with pixel-level uncertainty estimates (e.g., Monte Carlo dropout) to flag regions where Te/EM are unreliable (like Sgr C). This enables complete, unbiased surveys of star-forming regions in crowded galactic centers.

3. Gradient and morphology feature extraction for physical diagnostics

• Develop an AI that extracts local spatial gradients (e.g., dTe/dx, dEM/dx) from predicted maps, automatically identifying steep temperature gradients (like Sgr B2 Main’s 3000 K pc−1) or sharp absorption features (Sgr A* hole). This can be used to infer gas dynamics, shocks, or embedded heating sources without manual inspection.

4. Multi-scale fusion for contamination mitigation

• Implement a transformer-based model that fuses multi-frequency data (e.g., 99 GHz with lower-frequency synchrotron maps and higher-frequency dust maps) to automatically mask dust-contaminated pixels (where spectral index α > +2) and subtract side-lobe artifacts. This improves Te/EM accuracy in complex regions like the CMZ, where molecular line contamination is severe.

5. Galactic-scale extrapolation and radial trend prediction

• Train a regression model on the derived Te vs. projected radius relation (Te 5872 + 300 R/kpc) to predict electron temperatures for HII regions across the entire Milky Way disk, using only continuum and line data from future surveys (e.g., ALMA, ngVLA). The AI can flag outliers (e.g., Minispiral’s 12000 K) for follow-up.

6. Simulation-to-real domain adaptation for synthetic CMZ generation

• Use the derived Te/EM distributions (mean, variance, gradients) as priors to train a generative model (e.g., GAN or diffusion) that creates realistic synthetic CMZ maps. This AI can augment training data for other radio-astronomy models, test instrument designs, or simulate observations under different beam sizes (e.g., 2″ vs. 10″).

7. Real-time spectral line analysis pipeline

• Replace the time-consuming line-profile fitting with a lightweight neural network that takes 3D cubes (position-position-velocity) and outputs Te/EM maps in real time, using the TeEM method as a fast approximation. This enables on-the-fly analysis during ALMA observations, allowing dynamic target-of-opportunity follow-ups.

8. Anomaly detection for extreme physical conditions

• Train an autoencoder on the Te/EM maps to identify statistically rare features (e.g., the 11980 K NS arm near Sgr A* or the 3×108 pc cm−6 peak in Sgr B2 Main). The AI can automatically flag these as candidates for high-resolution follow-up or theoretical modeling of extreme star formation feedback.

What the improved AI system can do:

• Process full ALMA survey fields (e.g., 1° × 0.3°) into Te/EM maps in minutes instead of weeks.

• Provide reliable physical properties for thousands of HII regions, even in low-SNR or contaminated areas, with calibrated uncertainties.

• Automatically discover new HII regions and thermal filaments in the CMZ or other galactic centers, and predict their Te/EM without manual spectral fitting.

• Generate synthetic training data for next-generation telescopes (ngVLA, SKA) to optimize survey strategies.

• Enable real-time decision-making during observations, such as re-pointing to regions with anomalous Te or EM.

Star formation activity in the Central Molecular Zone (CMZ) directly manifests itself as radio continuum free-free emission (Bremsstrahlung) and radio recombination line emission from HII regions surrounding newly formed massive stars. We derive the overall distribution of the HII regions and their fundamental properties: electron temperature and emission measure (EM), and hence electron density in the form of two dimensional distribution maps over the CMZ by analyzing the ACES (ALMA CMZ Exploration Survey) 40 (99.02 GHz) recombination line and 99.6 GHz continuum emission data with synthesized beam widths of 2''.45 (0.097 pc at 8.2 kpc) and 2''.14, respectively. We apply the 'TeEM' method (-- EM mapping), which creates and EM maps from input 2D maps of the continuum and integrated line intensity. The analysis covers the entire ACES field from l about-0.6 to +0.8 and from b about-0.2 to +0.1. The area analyzed is complete and includes previously known HII regions such as Sgr B2, Sgr B1, the Sickle, the Pistol, thermal filaments (Bridges), Sgr A HII regions, the Minispiral, and many other known HII regions. Sgr C is not included in the analysis due to the insufficient signal-to-noise ratio in the recombination line map. The mean electron temperature over the CMZ is determined to be = 5872 plus or minus 78 (SE) plus or minus 3682 (SD) K (SE:standard error of the mean, SD: pixel-to-pixel standard deviation). Some HII regions, such as Sgr B2 Main and the Minispiral, exhibit large scatter and an internal gradient of several thousand K per parsec. The EM distribution is more diverse, varying by orders of magnitude from about 10 5 to about 3 times 10 8 within the CMZ, as well as within individual HII regions.