🔭 astrophysics

The IRIS $^{2+}$ inversion tool: recovering the radiative losses and the thermodynamics in the lower solar atmosphere

This paper introduces IRIS $^{2+}$ , a fast and reliable k-nearest neighbor inversion tool that utilizes a large database of synthetic profiles to efficiently recover the thermodynamic properties and integrated radiative losses of the lower solar atmosphere from IRIS observations, yielding results comparable to the state-of-the-art STiC code.

Original authors: Alberto Sainz Dalda, Jaime de la Cruz Rodríguez, Viggo Hansteen, Bart De Pontieu, Milan Gošić

Published 2026-01-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Alberto Sainz Dalda, Jaime de la Cruz Rodríguez, Viggo Hansteen, Bart De Pontieu, Milan Gošić

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: A Solar Weather Forecast Tool

Imagine the Sun's lower atmosphere (the part closest to us) as a complex, chaotic kitchen. Inside, there are swirling gases, intense heat, and magnetic fields. Scientists want to know exactly what is happening in this kitchen: How hot is it? How fast is the gas moving? How much energy is being lost as light?

For years, figuring this out has been like trying to guess the ingredients of a soup just by looking at the steam, but doing the math so slowly that you can only taste one spoonful a day. This new paper introduces IRIS2+, a new tool that acts like a "super-fast recipe decoder." It can look at the light coming from the Sun and instantly tell us the temperature, speed, and energy loss of the solar atmosphere, doing in minutes what used to take days.

The Problem: The "Slow Cooker" vs. The "Instant Pot"

To understand the Sun, scientists use a method called "inversion." Think of it like this:

The Observation: You see a spectral line (a specific color of light) coming from the Sun.
The Goal: You want to know the physical conditions (temperature, speed, etc.) that created that specific color.

The "gold standard" tool for this is a program called STiC. STiC is like a master chef who tastes the soup, calculates the chemistry, simulates the cooking process, tastes it again, and adjusts the recipe until it matches perfectly. It is incredibly accurate, but it is also very slow. If you have a whole pot of soup (a large dataset of the Sun), STiC might take days to figure out the recipe for every single drop.

The Solution: The "Recipe Database" (IRIS2+)

The authors, led by Alberto Sainz Dalda, created a new tool called IRIS2+. Instead of calculating the recipe from scratch every time, IRIS2+ uses a massive library of pre-cooked recipes.

Here is how it works:

The Library: The team used the slow, careful chef (STiC) to create a database of 135,472 different "representative profiles." Think of these as 135,000 different "standard soup recipes" that cover almost every possible weather condition on the Sun, from quiet zones to stormy active regions.
The Match: When a scientist observes the Sun, IRIS2+ doesn't do the hard math. Instead, it looks at the light and says, "Hey, this looks exactly like Recipe #42,091 in our library!"
The Result: It instantly assigns the physical conditions (temperature, speed, etc.) associated with that recipe to the observation.

Because it's just looking up a match in a library rather than cooking from scratch, IRIS2+ is thousands of times faster than STiC. It can process a whole solar map in minutes.

What Does It Tell Us?

IRIS2+ doesn't just guess the temperature; it gives a full "weather report" for the lower solar atmosphere:

Temperature: How hot is the gas at different heights?
Velocity: How fast is the gas moving toward or away from us?
Turbulence: How chaotic is the movement?
Radiative Loss: How much energy is the Sun losing as light? This is crucial because it tells us the minimum amount of energy the Sun needs to generate to keep that layer of the atmosphere hot.

Did It Work? (The Taste Test)

The authors wanted to know if this "fast library method" was as good as the "slow chef method." They took 10 different solar observations and ran them through both STiC and IRIS2+.

The Verdict: The results were very comparable.
- For Radiative Loss (energy loss) and Temperature, the two tools agreed very well. The fast tool found the right "recipe" almost as often as the slow chef.
- For Speed and Turbulence, there were some differences. IRIS2+ sometimes gave slightly lower values than STiC, especially in the lower layers of the atmosphere. The authors suggest this is because the library might need a few more "recipes" to cover every tiny variation in speed, but overall, the results are reliable.

The "Magic" of the Library

One of the coolest features of IRIS2+ is its flexibility.

The "Default" Setting: It comes pre-loaded with a smart way to weigh different parts of the light spectrum, so it works well out of the box.
The "Custom" Setting: If a scientist only has data for certain colors (lines), IRIS2+ can adjust its library search to ignore the missing colors and still find the best match.

The Bottom Line

The paper concludes that IRIS2+ is a game-changer. It allows scientists to analyze huge amounts of solar data quickly without sacrificing too much accuracy.

Analogy: If STiC is a master chef cooking a single dish perfectly but slowly, IRIS2+ is a high-tech food delivery service that has a database of every dish imaginable. It can deliver the correct meal to your door in seconds. While it might not be quite as perfect as the chef's custom cooking in every single scenario, it is good enough for almost everything, and it's fast enough to feed the whole neighborhood.

The authors plan to make the library even bigger (up to 500,000 recipes) to make the matches even better, but even in its current form, it is a powerful new tool for understanding our Sun.

Technical Summary: The IRIS2+ Inversion Tool

Problem Statement
Inferring the physical state of the chromospheric plasma from observations remains a significant challenge in solar physics. Due to low collisional rates in the chromosphere, a non-local thermodynamical equilibrium (non-LTE) description is required to explain emerging intensities, preventing a straightforward translation from observed intensities to physical parameters. While modern non-LTE inversion codes (e.g., STiC) have improved fidelity by incorporating multiple spectral lines to increase depth resolution, they are computationally intensive. Furthermore, calculating integrated radiative losses (IRL)—crucial for understanding the energy balance and heating mechanisms in the chromosphere—requires complex post-processing of the inverted atmospheric models. There is a need for a fast, reliable tool capable of recovering thermodynamic parameters and radiative losses for large datasets observed by the Interface Region Imaging Spectrograph (IRIS) without the prohibitive computational cost of full iterative inversions for every pixel.

Methodology
The authors introduce IRIS2+, an improved inversion tool based on a nearest-neighbor (k-nearest neighbor or k-nn) algorithm. The core of this method is the IRIS2+ database, an extension of the previous IRIS2 database, which now contains 135,472 synthetic Representative Profiles (RPs).

Database Construction: The database was built by selecting 550 IRIS datasets covering various solar features (umbra, pore-like, penumbra, quiet sun, plage). These were pre-sorted into five classes. For each dataset, 320 multi-line RPs were calculated using unsupervised k-means clustering. Each RP consists of 6 chromospheric lines (C II 1334 & 1335 Å, Mg II h&k, Mg II UV triplet) and 6 photospheric lines (Ti II, Fe I, C I, Ni II). These RPs were inverted using the state-of-the-art STiC code to generate Representative Model Atmospheres (RMAs) containing temperature ( $T$ ), line-of-sight velocity ( $v_{los}$ ), turbulent velocity ( $v_{turb}$ ), and electron density ( $n_e$ ) across 39 optical depths.
Radiative Loss Calculation: The database includes Representative Integrated Radiative Loss (RIRL) values. These were calculated using STiC (utilizing the RH code) by integrating the divergence of the radiative flux over the chromosphere (defined as $4 \le T \le 18$ kK). To manage computational complexity during the database creation, radiative losses were computed by treating species (H, Mg II, Ca II) independently (one in non-LTE, others in LTE) rather than simultaneously, a method validated to overestimate total losses by only $\approx 15\%$ compared to full non-LTE calculations.
Inversion Process: For a new observation, IRIS2+ transforms the observed spectral sampling and weights to match the database. It then employs a k-nn algorithm (with $k=1$ ) to find the closest synthetic RP in the database to the observed profile. The corresponding RMA and RIRL are assigned to the observed location. The tool offers flexible weighting options ("mean," "default," "array," "manual") to handle different line combinations and observational setups.

Key Contributions

Expanded Database: The IRIS2+ database is more than three times larger than its predecessor (135,472 vs. 40,320 RPs), covering a broader range of solar conditions from the photosphere to the transition region.
Integrated Radiative Losses: Unlike previous tools, IRIS2+ directly provides the integrated radiative losses for the chromosphere alongside thermodynamic parameters, derived from the RMA database.
Computational Efficiency: The tool is implemented in Python and utilizes standard machine learning libraries (Scikit-learn). It reduces execution time by orders of magnitude ( $O(10^5-10^6)$ ) compared to STiC, allowing typical raster scans to be inverted in minutes.
Interactive Inspection Tool: A companion tool allows users to visualize observed vs. synthetic profiles, thermodynamic variables along optical depth, IRL maps, and fit quality ( $\chi^2$ ).

Results
The authors compared IRIS2+ results against STiC inversions for 10 IRIS datasets:

Radiative Losses (IRL): IRIS2+ and STiC results show a strong linear correlation ( $r_p \approx 0.9$ ) after filtering outliers. Spatial maps are visually equivalent, with discrepancies largely attributed to outliers or episodic events requiring more sophisticated inversion schemes. The tool performs best when C II and Mg II lines are inverted simultaneously.
Thermodynamics:
- Temperature ( $T$ ): Correlations are moderate to strong ( $r_p \gtrsim 0.5$ ) between $\log(\tau) = -7$ and $-1$. IRIS2+ tends to produce slightly cooler values at high chromospheric heights ( $\log(\tau) = -6$ ) but slightly hotter values in the mid-chromosphere and photosphere ( $\log(\tau) = -4, -1$ ), likely due to bimodal distributions in the database fitting.
- Velocities ( $v_{los}, v_{turb}$ ): Strong correlations exist in the high and mid-chromosphere ( $r_p \approx 0.7$ ). However, in the low chromosphere and photosphere, correlations weaken, and IRIS2+ generally yields lower values than STiC. The authors attribute this to the artificial limit of $15$ km s $^{-1}$ imposed on $v_{turb}$ in the database and the limited sensitivity of the selected lines to thermodynamic changes in the photosphere.
Weighting Sensitivity: The study confirms that using all available lines (chromospheric and photospheric) yields the best agreement with STiC. Relying solely on photospheric lines or only C II lines results in poor correlations for chromospheric parameters.

Significance and Claims
The paper claims that IRIS2+ is a fast and reliable inversion tool capable of providing approximate values of the thermodynamic state and radiative losses for a wide variety of solar scenes observed by IRIS. The authors emphasize that while IRIS2+ is not a replacement for the rigorous, iterative STiC code in all scenarios, it provides results that are "comparable in most cases" to STiC when all lines are inverted simultaneously.

The significance lies in the ability to process large volumes of IRIS data efficiently, enabling statistical studies of chromospheric heating and radiative losses that were previously computationally prohibitive. The authors modestly note that the tool's accuracy is bounded by the sensitivity of the observed lines and the diversity of the database; they suggest that expanding the database to 300,000–500,000 elements could further improve coverage and reduce degeneracy. The tool is presented as a practical solution for the solar physics community to access thermodynamic and radiative loss data without the heavy computational overhead of full non-LTE inversions.