MOLLId: software for automatic identification of spectral molecular lines in the sub-millimeter and millimeter bands and its application to the spectra of protostars from the region RCW 120

Here is an explanation of the paper, translated into everyday language with some creative analogies.

The Big Picture: A Digital Detective for the Stars

Imagine you are standing in a massive, crowded library at night. The lights are off, but thousands of people are whispering simultaneously. Each whisper is a different language, and they are all overlapping. Your job is to figure out exactly who is saying what, and what they are saying, just by listening to the noise.

This is essentially what astronomers face when they look at protostars (baby stars) in the RCW 120 region of space. These baby stars are surrounded by thick clouds of gas and dust. As they heat up, they release a "forest" of radio waves—millions of tiny signals, each representing a specific molecule (like water, alcohol, or complex organic compounds).

The problem? It's a chaotic mess. The signals overlap, and there are so many of them that a human trying to identify them one by one would need a lifetime and a lot of coffee.

The Solution: Enter MOLLId

The authors of this paper built a new piece of software called MOLLId (MOLecular Line Identification). Think of MOLLId as a super-powered, hyper-organized librarian who can read the entire library in seconds.

Here is how it works, step-by-step:

The "Noise" Filter: First, the software looks at the chaotic radio signal. It ignores the static (the background noise) and focuses only on the "whispers" that are loud enough to be heard.
The "Shape" Matcher: In space, these molecular whispers usually have a specific shape (like a bell curve). MOLLId tries to fit a perfect bell curve over every whisper it finds. If the shape matches, it knows, "Okay, this is a real signal, not just random static."
The "ID Card" Check: Once it isolates a signal, MOLLId checks its "ID card" (its frequency). It compares this ID against a massive digital database of known molecules (like the JPL and CDMS catalogs). It asks: "Does this frequency match Methanol? Or maybe Acetone?"
The "Best Guess" Logic: Sometimes, two different molecules might have frequencies that are very close. MOLLId uses a smart, multi-level strategy. It starts by looking for the most obvious matches (low energy, high probability). If it can't find a match, it widens its search net to look for more exotic, high-energy molecules. It's like a detective who first checks the most likely suspects before looking for the rare ones.

The Experiment: Testing on Two Baby Stars

The team tested MOLLId on two specific baby stars in the RCW 120 region, which they named S1 and S2. These stars are located near the edge of a "Photo-Dissociation Region" (PDR)—a fancy way of saying they are on the border of a zone where intense starlight is breaking apart molecules.

Star S1 (The Quiet One): This star is a bit younger and cooler. MOLLId listened to its signals and found 100 distinct lines belonging to 41 different molecules.
Star S2 (The Party Animal): This star is much more active and hotter. It's in a "Hot Core" stage, meaning the dust around it is evaporating, releasing a huge cocktail of complex chemicals into the gas. MOLLId went wild here, identifying 407 distinct lines belonging to 79 different molecules.

The Speed: Doing this manually would take a human weeks. MOLLId did the job for S1 in 6 minutes and S2 in 8 minutes on a standard computer.

What Did They Find?

The results were like finding a treasure chest of chemistry:

The "Alcohol" King: The most common molecule found was Methanol (CH₃OH). It's essentially wood alcohol, but in space, it's a building block for life. In the active star S2, Methanol was everywhere.
Complex Ingredients: They found "complex organic molecules" (COMs). These are molecules with carbon chains, like Methyl Formate (which smells like rum) and Dimethyl Ether. Finding these is crucial because they are the ingredients needed to eventually make amino acids—the building blocks of life.
The "Hot" vs. "Cold" Split: By analyzing the signals, the team realized that Star S2 isn't just one uniform cloud. It has two layers:
- A cold outer layer (like the winter coat of the star).
- A hot inner core (the star's fiery heart).
  The software detected that high-energy molecules were coming from the hot core, while low-energy ones were from the cooler outer shell. This confirmed that the star is hot enough to be evaporating dust grains, releasing trapped chemicals into the gas.

Why Does This Matter?

Think of the universe as a giant kitchen. For a long time, we could only taste the soup (the gas) and guess what was in it.

Before MOLLId: We were trying to taste the soup with a blindfold on, guessing, "Maybe that's salt? Maybe that's pepper?" It was slow and often wrong.
With MOLLId: We now have a digital food analyzer. We can instantly tell, "Ah, that's definitely salt, and there's a pinch of saffron, and a hint of vanilla."

This paper proves that we can now automatically and quickly decode the chemical recipes of baby stars. By understanding what chemicals are present and how hot the environment is, we get a better picture of how stars form and, more importantly, how the chemical ingredients for life are created in the cradle of the universe.

The Bottom Line

The authors built a smart robot librarian (MOLLId) that can read the chaotic whispers of baby stars faster and more accurately than any human. They used it to prove that one of the stars they studied is a "hot core" factory, churning out complex organic molecules that could one day become part of a living planet.

Here is a detailed technical summary of the paper "MOLLId: software for automatic identification of spectral molecular lines in the sub-millimeter and millimeter bands and its application to the spectra of protostars from the region RCW 120."

1. Problem Statement

Observations in the sub-millimeter and millimeter ranges of star-forming regions yield high-resolution spectra containing a dense "forest" of molecular emission lines. Analyzing these spectra is labor-intensive and prone to error when performed manually.

Challenges: Manual identification is time-consuming, often fails to account for excitation conditions, and struggles with overlapping lines or high noise levels. Existing automated tools (e.g., XCLASS, MADCUBA, GAUSSPY) often require manual selection of molecules, assume Local Thermodynamic Equilibrium (LTE) from the start, or rely on specific decomposition algorithms that may not handle complex line forests efficiently without prior constraints.
Goal: To develop an automated software package capable of approximating spectral line profiles and identifying molecules with high efficiency and accuracy, minimizing human intervention while handling the complexity of crowded spectra.

2. Methodology

The authors developed MOLLId (MOLecular Line Identification), a Python-based software package designed for two sequential stages:

A. Spectral Line Profile Approximation

Baseline Subtraction: Uses the arpls function (from the pybaselines package) based on asymmetric weighted least squares to remove continuum baselines.
Gaussian Fitting: The algorithm iteratively identifies the most intense line in a spectrum (above a $3\sigma$ noise threshold) and fits it with a single Gaussian profile.
- Parameters: Central frequency ( $x_0$ ), half-width at half-maximum ( $\sigma'$ ), and line intensity.
- Optimization: Uses the scipy.optimize.least_squares function with the trf method to minimize residuals.
- Iterative Subtraction: Once a line is fitted, it is subtracted from the spectrum. The process repeats on the residual spectrum until no lines above the $3\sigma$ level remain.
- Outlier Filtering: Random noise spikes are filtered out using the coefficient of determination ( $R^2$ ); fits with $R^2 < 0.1$ are discarded.
- Adaptive Width: The width of the first detected line is used to refine initial guesses for subsequent lines.

B. Molecular Identification

Frequency Matching: The central frequencies derived from the Gaussian fits are compared against spectroscopic databases (primarily JPL and CDMS).
Multi-Level Selection Criteria: To handle line crowding and ambiguity, the software employs a four-level search strategy based on:
1. Frequency deviation ( $\Delta\nu$ ) from the database.
2. Upper-level energy ( $E_u$ ).
3. Logarithm of the Einstein spontaneous decay coefficient ( $\log A_{ij}$ ).
- Level 1: Strictest criteria ( $\Delta\nu \le 0.0005$ GHz, $E_u \le 200$ K).
- Levels 2–4: Gradually relaxed criteria if no match is found in the previous level.
Candidate Selection: If multiple molecules match a frequency, the algorithm selects the candidate with the smallest frequency deviation, lowest $E_u$ , and highest $\log A_{ij}$ .
Physical Parameter Estimation (LTE): For identified molecules with multiple transitions, the software calculates the excitation temperature ( $T_{ex}$ ) and column density ( $N$ ) using the rotational diagram method under the LTE approximation.

3. Key Contributions

Development of MOLLId: A fully automated pipeline for line approximation and identification that does not require pre-selection of target molecules.
Efficiency: The software significantly reduces processing time compared to manual methods.
Application to RCW 120: The tool was successfully applied to high-sensitivity observational data of two Young Stellar Objects (YSOs), RCW 120 YSO S1 and RCW 120 YSO S2, located near the photodissociation region (PDR) of RCW 120.

4. Results

The software was tested on data obtained with the APEX telescope (nFLASH230 receiver) across six frequency bands (202–261 GHz).

Performance Metrics

Processing Time: ~6 minutes for S1 and ~8 minutes for S2 per spectral range on an Intel Core i7-12700K CPU.
Identification Rate:
- RCW 120 YSO S2: Identified 407 lines belonging to 79 molecules. Only 3 lines (0.7%) remained unidentified.
- RCW 120 YSO S1: Identified 100 lines belonging to 41 molecules. Only 1 line (1%) remained unidentified.
Dominant Species: Methanol ( $CH_3OH$ ) was the most abundant, accounting for ~14% of lines in S2. Other major species included methyl formate ( $CH_3OCHO$ ), dimethyl ether ( $CH_3OCH_3$ ), and methyl cyanide ( $CH_3CN$ ).
Chemical Composition:
- Carbon-containing species dominated (84.5% in S2).
- Sulfur-containing species were notably abundant (23.8%), contrasting with some previous studies where nitrogen species were more prevalent.

Physical Analysis (RCW 120 YSO S2)

Two-Component Structure: Analysis of $CH_3OH$ $C H_{3} O H$ , $CH_3CN$ $C H_{3} C N$ , and $CH_3CCH$ $C H_{3} C C H$ revealed a distinct separation between low-energy and high-energy transitions.
- Low-Energy Component: Cold gas ( $T_{ex} \approx 27-70$ K).
- High-Energy Component: Hot gas ( $T_{ex} \approx 232-238$ K).
Hot Core Confirmation: The detection of high-energy lines ( $E_u > 100$ K) and the derived hot temperatures confirm that RCW 120 YSO S2 is in the "hot core" stage, where dust grain evaporation releases complex organic molecules into the gas phase.
Evolutionary Difference: RCW 120 YSO S1 showed no high-energy lines, suggesting it is at an earlier evolutionary stage (cold core) compared to S2.

5. Significance

Automation in Astrochemistry: MOLLId demonstrates that automated, multi-level frequency matching can effectively replace laborious manual searches in crowded spectra, achieving near-perfect identification rates (<1% unassigned lines).
Insight into Star Formation: The study provides robust evidence for the evolutionary divergence between S1 and S2 within the same region, highlighting the transition from cold cores to hot cores.
Chemical Complexity: The high abundance of sulfur-bearing species and complex organic molecules (COMs) in S2 offers new constraints on chemical models of hot cores and PDR boundaries.
Future Improvements: The authors note that future iterations will integrate full LTE spectral synthesis directly into the identification loop to better resolve overlapping lines and refine physical parameter estimates.

In conclusion, MOLLId represents a significant advancement in the automated analysis of millimeter/sub-millimeter spectra, enabling rapid and accurate chemical characterization of star-forming regions.