Testing the Constancy of Type Ia Supernova Luminosities with Gaussian Process

Using a model-independent Gaussian Process reconstruction of cosmic expansion history from cosmic chronometer data, this study tests the constancy of Type Ia supernova luminosities and finds them generally consistent with standard candles within 1σ, while identifying localized, non-statistical deviations in both Pantheon+ and DES 5YR datasets that suggest a possible non-monotonic luminosity evolution driven by varying physical mechanisms across different redshifts.

The Gist

The Big Picture: Are Our Cosmic Rulers Broken?

Imagine you are trying to measure the distance to a series of lighthouses across a vast ocean. You assume every lighthouse has the exact same brightness. If you see one that looks dimmer, you assume it's just further away. If you see one that looks brighter, you assume it's closer. This is how astronomers use Type Ia Supernovae (exploding stars) to measure the universe. They are our "standard candles."

For decades, we've assumed these cosmic lighthouses never change their brightness, no matter how far back in time (or how far away) we look. But what if they do change? What if a lighthouse built 5 billion years ago glows slightly differently than one built 1 billion years ago? If we don't account for that, our map of the universe is wrong.

This paper asks a simple but crucial question: Do these cosmic lighthouses stay the same brightness over time, or do they drift?

The Old Way vs. The New Way

The Old Way (The "Guessing Game"):
Previously, scientists tried to test this by assuming a specific shape for how the universe expands (like a straight line or a curve). They would compare the stars to that shape. The problem? If their guess about the universe's shape was slightly wrong, they might think the stars were changing brightness when they actually weren't. It's like trying to measure a runner's speed while standing on a moving treadmill; you might blame the runner for the treadmill's movement.

The New Way (The "Independent Witness"):
This paper uses a clever trick to avoid the treadmill. Instead of guessing the universe's shape, the authors use a completely different set of data called Cosmic Chronometers.

• The Analogy: Imagine you want to know how fast a car is driving. Instead of looking at the road markings (which might be painted wrong), you look at the driver's watch.
• The Method: They used the ages of old, passive galaxies (the "watches") to measure the expansion rate of the universe directly, without assuming any specific theory about dark energy or the universe's shape.

The Tool: Gaussian Processes (The "Smart Rubber Band")

To turn these "watch" measurements into a distance map, the authors used a mathematical tool called Gaussian Processes (GP).

• The Analogy: Imagine you have a few scattered points on a piece of paper and you want to draw a smooth line connecting them. A normal ruler forces a straight line. A "Smart Rubber Band" (the GP) stretches between the points, finding the smoothest, most natural curve that fits the data without forcing it into a rigid shape.
• Why it matters: This allows the data to speak for itself. It doesn't force the universe to fit a pre-made mold. It creates a "baseline" distance map that is purely based on observation, not theory.

The Experiment: Checking the Residuals

Once they had their "Smart Rubber Band" baseline (the expected brightness), they compared it to the actual supernova data from two massive surveys: Pantheon+ (1,701 stars) and DES (435 stars).

They looked at the difference (the "residuals") between what they expected and what they saw.

• If the stars are perfect: The difference should be a flat line at zero.
• If the stars change: The line would wiggle up or down.

They also looked at the slope (how fast the line is changing). This is like checking if the stars are getting steadily brighter or dimmer, or if they suddenly jump in brightness at a specific time.

The Findings: A "Wobble" in the Data

The results were fascinating:

1. Overall Success: For the most part, the stars are reliable. They behave like standard candles within a very small margin of error. The "rubber band" fits the data well.
2. The Glitch: However, they found small, localized "wobbles" that couldn't be ignored.
   • In the Pantheon+ data, there was a weird dip around a specific time in the universe's history (redshift $z \sim 1$).
   • In the DES data, there was a similar weird dip at a slightly different time ($z \sim 0.3–0.5$).

Why is this important?
Because two completely different surveys (one a mix of many telescopes, the other one specific telescope) saw similar "wobbles," it's unlikely to be a mistake or a fluke. It suggests something real is happening.

What Does It Mean? (The Story of the Stars)

The authors suggest these wobbles tell a story about how the "lighthouses" are built:

• Early Universe (Low Redshift): The stars seem to get slightly dimmer as we look back in time. This might be because the "ingredients" (metal content) in the stars were different back then, making them less efficient at exploding brightly.
• Middle Universe: Suddenly, the trend flips, and stars seem to get brighter. This could mean the "recipe" for these explosions changed. Maybe the types of stars that explode shifted from old, slow-burning ones to younger, more energetic ones.

It's like if you noticed that cars made in the 1970s got slightly less fuel-efficient, but then cars made in the 1990s suddenly got more efficient for a while, before settling down. It implies the "engine" of the universe (the stars) isn't running on a single, unchanging setting.

The Conclusion

This paper doesn't say our map of the universe is broken. It says, "Hey, our map is great, but there are tiny, interesting bumps we need to understand."

By using a method that doesn't rely on guessing the shape of the universe, the authors found subtle evidence that Type Ia supernovae might evolve slightly over time. This is a wake-up call for future astronomers: to get the most precise measurements of the universe's expansion, we need to understand the "personality" of these exploding stars better.

In short: The cosmic lighthouses are mostly reliable, but they might be changing their bulbs slightly over the eons, and we finally have a tool sensitive enough to see it.

Technical Summary

1. Problem Statement

Type Ia supernovae (SNe Ia) serve as the primary "standard candles" for measuring cosmic expansion and constraining dark energy parameters (such as $H_0$ and $w$). This utility relies on the fundamental assumption that their absolute magnitude, $M_B$, is constant across cosmic time (redshift $z$) after standardization.

However, this assumption is not guaranteed. Astrophysical factors—including progenitor demographics (single vs. double degenerate channels), host galaxy properties (mass, star formation rate), metallicity evolution, and intergalactic dust—could induce a redshift-dependent drift in luminosity. Even a small drift ($\sim 0.01\text{--}0.05$ mag) can bias cosmological inferences. Previous studies have largely relied on parametric models (linear, quadratic, or power-law) to test for evolution, which may fail to detect subtle, non-monotonic, or scale-dependent variations. The paper aims to test the constancy of SNe Ia luminosities using a model-independent, non-parametric approach to avoid the biases inherent in assuming a specific cosmological model or functional form for evolution.

2. Methodology

The core innovation of this work is a two-tiered diagnostic framework that decouples the test of luminosity evolution from assumptions about the cosmological model.

A. Model-Independent Baseline via Gaussian Processes (GP)

Instead of assuming a fiducial cosmology (e.g., $\Lambda$CDM) to predict theoretical distances, the authors reconstruct the expansion history directly from Cosmic Chronometer data ($H(z)$ measurements).

1. GP Reconstruction: A Gaussian Process regression is applied to 32 $H(z)$ data points (redshift range $0.07 \le z \le 1.965$) using a squared exponential kernel. This provides a smooth, non-parametric reconstruction of the Hubble parameter $H(z)$.
2. Derivative and Integration: The GP framework allows for the analytic calculation of derivatives ($dH/dz$) and numerical integration to obtain the comoving distance $r(z) = \int_0^z \frac{dz'}{H(z')}$.
3. Uncertainty Propagation: To rigorously handle uncertainties, the authors generate 2,000 Monte Carlo realizations of the $H(z)$ posterior. These are evaluated on a Chebyshev grid to minimize interpolation errors and Runge instabilities.
4. Baseline Distance Modulus: The luminosity distance $d_L(z)$ and the corresponding distance modulus $\mu_{GP}(z)$ are derived from these realizations, creating a cosmology-independent baseline.

B. Diagnostic: Residuals and Derivatives

The authors define the deviation between observed supernova data and the GP baseline:
$$\Delta M_B(z) \equiv \mu_{obs}(z) - \mu_{GP}(z)$$
where $\mu_{obs}(z)$ is the standardized distance modulus from supernova surveys.

• $\Delta M_B(z)$: A non-zero, redshift-dependent trend indicates a drift in absolute magnitude.
• $d(\Delta M_B)/dz$: The derivative of the residuals provides a sharper diagnostic. It reveals the rate of change in luminosity evolution, distinguishing between monotonic drifts (gradual aging) and localized features (transitions in progenitor channels or environmental effects).

C. Datasets

The analysis utilizes two independent supernova compilations to cross-validate findings:

1. Pantheon+ (Primary): 1,701 spectroscopically confirmed SNe Ia ($0.001 \le z \le 2.261$).
2. DES 5YR (Cross-check): 435 spectroscopically confirmed SNe Ia from the Dark Energy Survey ($0.02 \le z \le 0.72$). This dataset offers a homogeneous, single-survey calibration to check for survey-specific systematics.

3. Key Contributions

1. Non-Parametric Framework: The paper introduces a fully model-independent test for SN Ia luminosity evolution by anchoring the distance scale to cosmic chronometers via Gaussian Processes, eliminating degeneracy with cosmological parameters.
2. Derivative-Based Diagnostic: By analyzing the derivative $d(\Delta M_B)/dz$, the study enhances sensitivity to subtle evolutionary signatures that might be smoothed out in cumulative residual plots.
3. Robust Uncertainty Handling: The use of Monte Carlo realizations on a Chebyshev grid ensures high numerical stability and accurate propagation of correlated systematics from $H(z)$ measurements to the derived distance modulus.
4. Cross-Survey Validation: The simultaneous application to Pantheon+ and DES 5YR allows for the differentiation between genuine astrophysical evolution and survey-specific calibration artifacts.

4. Results

• Overall Consistency: Both datasets are consistent with the standard candle hypothesis within $1\sigma$ confidence intervals. The mean residuals $\Delta M_B(z)$ are generally close to zero.
• Localized Departures: Despite overall consistency, both datasets exhibit statistically significant, localized deviations:
  • Pantheon+: A departure is observed near $z \sim 1$.
  • DES 5YR: A departure is observed in the range $z \sim 0.3\text{--}0.5$.
• Derivative Analysis: The derivative $d(\Delta M_B)/dz$ reveals a non-monotonic pattern:
  • Low Redshift ($z < 0.5$): Negative slope ($d(\Delta M_B)/dz < 0$), implying $dM_B/dz > 0$. This suggests supernovae become fainter at higher redshifts within this range, consistent with older progenitors or lower metallicity producing less $^{56}\text{Ni}$.
  • Intermediate/High Redshift: The slope turns positive ($d(\Delta M_B)/dz > 0$), implying $dM_B/dz < 0$. This suggests supernovae become brighter at higher redshifts, potentially indicating a transition to younger, more massive progenitors in star-forming environments.
• Robustness: The fact that these specific features appear in two independent surveys with different selection functions and calibration pipelines suggests they are not purely statistical fluctuations or calibration errors.

5. Significance and Implications

• Astrophysical Insight: The results point toward a non-monotonic luminosity evolution, where different physical drivers dominate at different cosmic epochs. The transition from a "fading" trend to a "brightening" trend may reflect a shift in the dominant progenitor population (e.g., from old, passive galaxies to young, star-forming galaxies).
• Cosmological Impact: Even a modest drift of $\sim 0.05$ mag can introduce percent-level biases in the determination of the Hubble constant ($H_0$) and the dark energy equation of state ($w$). Ignoring these subtle, redshift-dependent effects could lead to incorrect conclusions about the nature of dark energy.
• Methodological Advance: The study demonstrates that Gaussian Processes are a powerful tool for precision cosmology, capable of revealing fine structures in data that parametric models might miss. It provides a robust, independent check on the standard candle assumption, essential for the upcoming Stage IV surveys (e.g., LSST, Euclid, Roman) where systematic errors must be controlled below statistical precision.

In conclusion, while Type Ia supernovae remain highly reliable standardizable candles, this work provides evidence for subtle, non-monotonic luminosity evolution that warrants deeper astrophysical investigation and careful modeling in future cosmological analyses.