Exploring the origins of high-velocity features in SNe Ia with the spectral synthesis code TARDIS

Using the TARDIS radiative-transfer code combined with neural network emulation and MCMC inference, this study demonstrates that single high-velocity density enhancements cannot simultaneously reproduce the observed silicon and calcium high-velocity features in Type Ia supernovae, suggesting that current delayed- and double-detonation explosion models are incomplete.

The Gist

Imagine a Type Ia supernova as a cosmic firework. When a white dwarf star explodes, it doesn't just burst into a uniform cloud of ash; it shoots out layers of material at different speeds, like a multi-layered onion being peeled apart by an explosion.

For decades, astronomers have noticed something strange in the "smoke" of these explosions. While looking at the light from these stars, they see a main absorption line (a dark mark in the spectrum) moving at a certain speed. But often, there's a second, fainter mark even further to the blue side, indicating a chunk of material moving much faster than the rest. These are called High-Velocity Features (HVFs).

Think of it like this: If the main explosion is a car driving down the highway at 60 mph, the HVF is a sports car zooming past it at 150 mph, detached from the main pack.

The Mystery

The big question is: Where does this fast sports car come from?
Is it a clump of extra fuel? A piece of the star that was thrown out harder than the rest? Or is it a different layer of the explosion entirely?

To solve this, the authors of this paper acted like cosmic detectives. They picked six famous supernovae that showed these fast "sports cars" clearly and tried to rebuild the explosion in a computer to see what ingredients were needed to create them.

The Investigation: Building a Virtual Explosion

The team used a powerful computer code called TARDIS (which stands for Time-Dependent Radiative Transfer). Think of TARDIS as a super-advanced flight simulator for stars.

1. The Base Model: First, they built a standard model of the explosion to match the "normal" part of the light (the 60 mph car). This was their "photosphere" model.
2. The Problem: The standard model couldn't produce the fast "sports car" (the HVF). The fast material just wasn't there in the simulation.
3. The Hypothesis: They guessed that maybe there was a density bump in the outer layers. Imagine the explosion isn't a smooth cloud, but has a thick, dense knot of material floating way out in front.
4. The Experiment: They added "Gaussian bumps" (mathematical hills of extra density) to their simulation at high speeds. They ran thousands of simulations, tweaking the size, shape, and location of these bumps, trying to see which one made the "sports car" appear in the light spectrum.

The AI Shortcut

Running these simulations one by one would take years. So, the team trained Neural Networks (a type of Artificial Intelligence).

• Think of the AI as a super-fast translator. It learned the relationship between "adding a density bump" and "what the light spectrum looks like."
• Once trained, the AI could predict the result of a simulation in a fraction of a second, allowing the team to run millions of scenarios to find the perfect match.

The Findings: What They Discovered

1. The "Sports Car" Needs a Boost
To get the fast material to show up in the light, they had to add a significant amount of extra mass (density) far out in the explosion. You can't just have a little bit of extra stuff; you need a massive, dense knot of material flying ahead of the main explosion.

2. The Silicon vs. Calcium Puzzle
Here is where it gets tricky.

• They successfully modeled the fast Silicon (Si) features.
• However, when they tried to use that same dense knot to explain the fast Calcium (Ca) features, it failed.
• The Analogy: It's like trying to explain why a fast runner is wearing red shoes and a fast cyclist is wearing blue shoes by saying "they are both wearing the same pair of shoes." It doesn't work.
• The Conclusion: The fast Silicon and the fast Calcium likely come from two different layers of the explosion. There might be a "Silicon knot" and a separate, even faster "Calcium knot" further out.

3. The Explosion Theories Don't Fit
Astronomers have two main theories for how these stars explode:

• Delayed Detonation: A slow burn that turns into a fast explosion.
• Double Detonation: A surface explosion that triggers a core explosion.

The team tested both theories. They found that neither of these standard explosion models naturally creates the dense, fast-moving knots required to explain the observations.

• It's like trying to bake a cake using a recipe that calls for flour and sugar, but the cake you end up with tastes like it needs a secret ingredient that isn't in the recipe book.
• To make the "Double Detonation" model work, they would have to increase the explosion's energy by 300%, which is physically unrealistic.

The Big Takeaway

This paper tells us that our current understanding of how Type Ia supernovae explode is incomplete.

The "fast sports cars" (HVFs) are real, and they are common. But the standard explosion recipes we have in our textbooks cannot produce them. There is something missing from the physics of these explosions—perhaps a hidden asymmetry, a weird interaction with the star's companion, or a completely new mechanism we haven't discovered yet.

The authors conclude that we need to look deeper into the "outer edges" of these explosions and perhaps update our physics models to include more complex effects (like how atoms behave in extreme heat) to finally solve the mystery of the high-velocity features.

Technical Summary

Here is a detailed technical summary of the paper "Exploring the origins of high-velocity features in SNe Ia with the spectral synthesis code tardis."

1. Problem Statement

Type Ia supernovae (SNe Ia) frequently exhibit High-Velocity Features (HVFs): secondary, blue-shifted absorption components in spectral lines (most notably Si II $\lambda$6355 and Ca II) that appear detached from the main photospheric velocity (PV) components. These features typically appear several thousand km s$^{-1}$ to the blue of the photosphere.

• The Mystery: While HVFs are ubiquitous (observed in $\sim$78% of early-time spectra), their physical origin remains debated. Proposed mechanisms include abundance enhancements, density enhancements, ionization state changes, or interaction with circumstellar material (CSM).
• The Gap: Previous studies have often focused on Ca II HVFs or used simplified 1D models. There is a lack of rigorous, quantitative modeling that simultaneously reproduces the evolution of Si II HVFs across multiple epochs to constrain the physical parameters (density, abundance, geometry) of the outer ejecta. Furthermore, it is unclear if standard explosion mechanisms (Delayed-Detonation or Double-Detonation) can naturally produce the required conditions.

2. Methodology

The authors utilized the radiative transfer code TARDIS combined with Machine Learning (Neural Networks) and Markov-Chain Monte Carlo (MCMC) inference to model six well-observed SNe Ia (SN 1994D, SN 2009ig, SN 2012fr, SN 2018cnw, SN 2021fxy, SN 2021aefx).

A. Sample and Data

• Sample: Six SNe Ia with high-cadence spectroscopy in the first two weeks post-explosion, showing distinct Si II $\lambda$6355 HVFs.
• Data: 27 spectra total, flux-calibrated and corrected for Milky Way extinction.

B. Modeling Workflow

1. Base Photospheric (PV) Models:

   • Custom 1D TARDIS models were created for each SN to reproduce the photospheric Si II features and overall spectral shape.
   • Abundance Tomography: Used to derive custom abundance profiles (O, Mg, Si, S, Ca, Ti, Cr, Fe, Ni) for the outer ejecta shells.
   • Density Profile: A power-law density profile ($\rho \propto v^m$) was adopted, tuned to match the N100 delayed-detonation model in the high-velocity regime ($v > 10,000$ km s$^{-1}$).
   • Carbon Exclusion: Carbon was initially excluded from PV models to prevent spurious C II $\lambda$6580 absorption that could contaminate the Si II fit.

2. Density Enhancement Grid:

   • To explain HVFs, a Gaussian density enhancement was added to the base density profile:
     $$\rho(v) = \rho_0(v) + r \rho_0(b) \exp\left(-\frac{(v-b)^2}{2c^2}\right)$$
     Where $r$ is amplitude, $b$ is centroid velocity, and $c$ is width.
   • A 4D grid of simulations was run for each SN, varying:
     • Amplitude ($r$): 0.5 to 16.
     • Width ($c$): 200–2000 km s$^{-1}$.
     • Centroid ($b$): 5 values based on observed HV velocities.
     • Outer Silicon Abundance ($X_{Si}$): 1% to 6%.
   • Total: $\sim$1,800 TARDIS simulations per SN.

3. Neural Network (NN) Emulators:

   • To make the parameter space exploration computationally feasible, Neural Networks were trained to emulate the TARDIS output.
   • Architecture: 2 hidden layers (200 neurons each, softplus activation).
   • Input: 4 parameters ($X_{Si}, r, b, c$).
   • Output: 75 wavelength points around the Si II $\lambda$6355 feature.
   • Performance: The NNs achieved high accuracy ($\text{MSE} \sim 10^{-4}$), emulating TARDIS $\sim$10,000 times faster.

4. MCMC Inference:

   • The trained NNs were used within an MCMC framework (using emcee) to infer the posterior distributions of the density enhancement parameters that best fit the observed spectral evolution for each SN.

3. Key Contributions

• First Systematic MCMC Analysis of HVFs: This work applies a rigorous statistical framework (MCMC + NN emulators) to constrain the physical properties of HVFs in SNe Ia, moving beyond visual inspection or single-epoch fitting.
• Decoupling Si and Ca HVFs: The study explicitly tests whether a single density enhancement can explain both Si II and Ca II HVFs, finding it impossible.
• Testing Explosion Mechanisms: The derived density enhancements were compared against angle-averaged and line-of-sight profiles from Delayed-Detonation (Chandrasekhar mass) and Double-Detonation (sub-Chandrasekhar) hydrodynamical models.

4. Key Results

A. Density Enhancement Requirements

• Success: The models successfully reproduced the Si II $\lambda$6355 HVF evolution for all six SNe.
• Parameter Correlations:
  • Large Velocity Separation (e.g., SN 2012fr): Requires strong, wide density enhancements located far from the photosphere ($\sim$23,000 km s$^{-1}$) with low silicon abundance.
  • Small Velocity Separation (e.g., SN 1994D): Requires weak density enhancements closer to the photosphere with higher silicon abundance.
• Mass: The density enhancements imply additional mass in the outer ejecta ranging from $\sim 0.001$ to $0.09 M_\odot$.

B. The Si vs. Ca HVF Discrepancy

• Single Enhancement Failure: When the best-fitting Si II density enhancement was applied to Ca II, the resulting Ca II HVFs formed at velocities significantly lower than observed.
• Three-Line-Forming Regions Hypothesis: The authors propose a structure with three distinct regions:
  1. Photosphere: Standard PV components.
  2. Si-Dominated Enhancement: Produces Si II HVFs.
  3. Ca-Dominated Enhancement: A second, higher-velocity density enhancement required to produce Ca II HVFs.
• Abundance Ratio: This implies a drop in Si abundance and a rise in the Ca/Si ratio at the highest velocities. This trend matches Double-Detonation models (which show double-peaked abundance profiles) but contradicts Delayed-Detonation models (where Ca peaks at lower velocities than Si).

C. Comparison with Hydrodynamical Models

• Delayed-Detonation: Fails to produce the necessary density bumps at the required high velocities ($>19,000$ km s$^{-1}$).
• Double-Detonation: While these models show density enhancements, they occur at much lower velocities ($\sim 10,000\text{--}15,000$ km s$^{-1}$).
  • Scaling Test: To match the observed HVF velocities, Double-Detonation models would require a 300% increase in kinetic energy, which is physically unrealistic.
• Conclusion: Neither standard Delayed-Detonation nor Double-Detonation mechanisms can currently explain the origin of the observed HVFs.

D. Physics Limitations

• The study notes that the Nebular Ionization and Dilute-LTE approximations used in TARDIS may introduce uncertainties. Full Non-LTE (NLTE) modeling is required to confirm if ionization effects alone could mimic density enhancements, though the authors argue the density enhancement conclusion remains robust.

5. Significance

• Constraint on Progenitors: The inability of current explosion models to reproduce the required high-velocity density structures suggests that our understanding of the outer ejecta in SNe Ia is incomplete.
• Asymmetry and Geometry: The diversity of HVF parameters (strength, location) across the sample suggests significant asphericity or line-of-sight variations in the ejecta, potentially linked to the explosion mechanism or progenitor system.
• Future Directions: The paper highlights the need for:
  1. Full NLTE modeling to decouple ionization effects from density effects.
  2. Simultaneous modeling of Si II and Ca II HVFs to constrain the Ca/Si abundance ratio in the outermost layers.
  3. Advanced Hydrodynamical Simulations that can produce high-velocity density enhancements without requiring unphysical kinetic energies.

In summary, this paper demonstrates that while density enhancements can successfully reproduce observed Si II HVFs, the specific requirements (high velocity, specific mass) challenge existing explosion paradigms, indicating that a "missing piece" in our theoretical models of SNe Ia explosions is likely required.