Bayesian analysis of proton-proton fusion in chiral effective field theory

This paper presents a Bayesian analysis of proton-proton fusion in chiral effective field theory, calculating the astrophysical $S$-factor at low energies with percent-level accuracy and providing the first theoretical uncertainty estimate due to the truncation of the chiral expansion of nuclear currents.

The Gist

Imagine the Sun as a giant, cosmic power plant. Its fuel isn't coal or gas; it's hydrogen atoms smashing into each other. The very first step in this process is two protons (hydrogen nuclei) fusing together to form a deuteron (a hydrogen isotope), releasing a positron and a neutrino. This is called proton-proton fusion.

This reaction is so rare and difficult that it happens incredibly slowly, which is actually a good thing for us—it means the Sun burns its fuel slowly and has been shining for billions of years. However, to understand how stars work, how old they are, and how much energy they produce, scientists need to know the exact "speed limit" of this reaction.

In physics, this speed limit is called the S-factor.

The Problem: A Missing Piece of the Puzzle

For decades, scientists have tried to calculate this S-factor. It's like trying to predict the exact speed of a car without a speedometer, using only a map and a guess. Previous calculations gave us a number, but they came with a big "maybe."

The main issue was uncertainty. Scientists knew their math was based on a theory called Chiral Effective Field Theory (χEFT). Think of this theory like a recipe for a cake.

• Leading Order (LO): The basic recipe (flour, eggs, sugar).
• Next-to-Leading Order (NLO): Adding a pinch of salt and vanilla.
• Next-to-Next-to-Leading Order (N2LO): Adding a specific type of baking powder.
• And so on...

The problem is that you can't bake an infinite cake. At some point, you have to stop adding ingredients and say, "This is good enough." But how do you know how much error you introduced by stopping? If you stop too early, your cake might be flat. If you stop too late, you've wasted time.

The Solution: A Bayesian "Crystal Ball"

This new paper, by a team of physicists, introduces a clever new way to measure that error. They used a statistical method called Bayesian Analysis.

Imagine you are trying to guess the weight of a mystery object.

1. Old Way: You weigh it once, then weigh it again with a slightly different scale, and if the numbers are close, you say, "It's probably around 5kg, give or take 1kg."
2. This Paper's Way (Bayesian): You build a "smart computer model" (a Gaussian Process) that learns the pattern of how the weight changes as you add more ingredients to your recipe. It doesn't just guess; it learns the shape of the uncertainty. It looks at the ingredients you did add and predicts what the missing ingredients would have done, giving you a very precise "confidence interval."

The authors applied this to the proton-proton fusion. They didn't just stop at one recipe; they tested six different "kitchens" (different nuclear models) to see if the result changed depending on which kitchen you cooked in.

The Big Discovery: Local vs. Non-Local Kitchens

One of the most interesting findings in the paper is about the "kitchens" themselves.

• Non-Local Kitchens (EMN models): These are like high-tech, futuristic kitchens where ingredients can interact across the room instantly.
• Local Kitchens (NV models): These are like traditional kitchens where ingredients only interact if they are touching.

The scientists found that the "futuristic" kitchens gave a slightly different result than the "traditional" ones.

• The traditional kitchens (which are similar to older, famous models) gave a result of 4.05.
• The futuristic kitchens gave a result of 4.09.
• The new, combined "Bayesian average" landed right in the middle at 4.068.

This explains why previous studies had slightly different numbers: they were mostly using only one type of kitchen. By mixing them all together and using the Bayesian "smart model" to weigh them, they got a more honest, robust answer.

Why Should You Care? (The Astrophysical Impact)

You might ask, "Does a difference between 4.05 and 4.09 really matter?"

The authors ran simulations to see what happens if this number changes:

1. Star Ages: If we change the fusion speed, does it change how old we think star clusters are? No. The change is so tiny (less than 1%) that it's lost in the noise of other uncertainties. The stars are still the same age.
2. Solar Neutrinos: These are ghostly particles streaming from the Sun. A change in the fusion rate changes the temperature of the Sun's core, which changes the neutrino output.
   • The paper found that even with a "worst-case" error margin, the neutrino flux changes by only about 2% to 5%.
   • While 5% sounds like a lot, it's actually quite small compared to other factors in solar models.

The Bottom Line

This paper is a triumph of precision.

• They calculated the most accurate value for the Sun's first fusion step: 4.068 (with a tiny error margin of 0.6%).
• They used a "smart" statistical method (Bayesian analysis) to prove that their error estimate is trustworthy, not just a guess.
• They showed that while different mathematical models give slightly different answers, the truth lies in the middle, and the uncertainty is now under control.

In simple terms: They finally put a very accurate speedometer on the Sun's engine. They found out that while the engine runs slightly differently depending on how you look at it, the overall speed is consistent, and we can now trust our calculations of how the Sun shines and how old the stars are.

Technical Summary

1. Problem Statement

The proton-proton ($pp$) fusion reaction ($p + p \to d + e^- + \bar{\nu}_e$) is the primary energy source for Sun-like stars. The reaction rate is governed by the astrophysical $S$-factor, $S(E)$. Because the cross-section at solar energies (around the Gamow peak, $E \approx 6$ keV) is too small to be measured directly in terrestrial laboratories, it must be calculated theoretically.

Previous theoretical efforts (summarized in the "Solar Fusion" reviews, specifically SFIII) have provided estimates for $S(0)$ (the $S$-factor at zero energy) and its derivatives. However, these estimates suffer from two main limitations:

1. Inconsistent Uncertainty Quantification: Previous error estimates relied on comparing different phenomenological models rather than a rigorous statistical framework for truncation errors.
2. Model Discrepancies: There is a persistent, unexplained discrepancy between results derived from local potentials (like AV18) and non-local chiral potentials. The AV18-based value for $S(0)$ is systematically lower than chiral effective field theory ($\chi$EFT) results.

This work aims to resolve these issues by calculating $S(E)$ using a wide variety of $\chi$EFT interactions (both local and non-local) and, for the first time, applying a Bayesian analysis to rigorously quantify theoretical uncertainties arising from the truncation of the chiral expansion and model dependence.

2. Methodology

Theoretical Framework

The authors perform calculations within Chiral Effective Field Theory ($\chi$EFT).

• Interactions: They utilize two distinct families of nuclear potentials:
  • EMN (Entem-Machleidt-Nosyk): Non-local, $\Delta$-less potentials derived in momentum space, available from Leading Order (LO) to Next-to-Next-to-Next-to-Leading Order (N3LO).
  • NV (Norfolk): Local, $\Delta$-full potentials derived in coordinate space at N3LO.
• Currents: Consistent weak axial currents are used, developed by the JLab-Pisa group, up to N3LO.
  • The low-energy constant (LEC) $z_0$ (entering the contact term of the axial current) is calibrated to reproduce the Gamow-Teller matrix element in Tritium ($^3$H) $\beta$-decay.
• Electromagnetic Corrections: To achieve percent-level accuracy, the calculation includes electromagnetic effects beyond the leading Coulomb interaction, specifically two-photon exchange and vacuum polarization.
• Partial Waves: Only the $^1S_0$ partial wave is included, as higher-order contributions are negligible at low energies.

Bayesian Uncertainty Analysis

The core innovation of this paper is the application of Bayesian statistics to estimate truncation errors:

1. Truncation Error Modeling: The expansion of the $S$-factor is treated as a series where missing higher-order terms are modeled using a Gaussian Process (GP). The GP is trained on a subset of calculated data points (LO, NLO, N2LO, N3LO) to predict the distribution of the missing contributions.
2. Separate Analysis: Truncation errors for the nuclear interaction and the nuclear current are analyzed separately and then combined.
3. Model Dependence: To address the uncertainty arising from the choice of potential model, the authors calculate a weighted average of results from all potential models (EMN and NV), assigning equal probability to the two families (local vs. non-local).
4. Validation: The GP emulator is validated using statistical diagnostics (Mahalanobis distance, Pivoted Cholesky, and credible intervals) to ensure the predicted error bands are statistically robust.

3. Key Contributions

• First Bayesian Error Estimate for $pp$ Fusion: This is the first study to apply Bayesian methods to quantify the theoretical uncertainty of the $pp$ fusion $S$-factor due to chiral expansion truncation.
• Unified Treatment of Local and Non-Local Potentials: The study systematically compares local (NV) and non-local (EMN) interactions within the same $\chi$EFT framework, addressing the historical discrepancy between AV18 (local) and $\chi$EFT results.
• Identification of Wave Function Origins: The authors identify that the systematic difference in $S(0)$ values between local and non-local potentials arises from differences in the short-range structure of the deuteron wave function (specifically the $d$-wave component), which is influenced by the regularization schemes used in the potentials.
• Refined Derivatives: The paper provides precise values for the first ($S'(0)$) and second ($S''(0)$) derivatives of the $S$-factor, essential for Taylor expansions used in astrophysical modeling.

4. Results

• $S(0)$ Value: The final recommended value for the astrophysical $S$-factor at zero energy is:
  $$S(0) = (4.068 \pm 0.025) \times 10^{-25} \text{ MeV b}$$
  This corresponds to a relative uncertainty of 0.6%.
• Comparison with Previous Work:
  • The central value (4.068) is slightly lower than the SFIII recommendation (4.090) but is in excellent agreement with the SFII value and previous non-local $\chi$EFT calculations.
  • The result is consistent with the SFIII value within $1\sigma$.
  • The lower central value is attributed to the inclusion of local potentials (NV), which yield lower $S(0)$ values than non-local potentials, bringing the average closer to the AV18 result.
• Derivatives:
  • $S'(0)/S(0) = 10.60 \pm 0.01 \text{ MeV}^{-1}$
  • $S''(0)/S(0) = 347.9 \pm 0.9 \text{ MeV}^{-2}$
  • These uncertainties are significantly smaller than those in previous reviews, reflecting the improved statistical treatment.
• Convergence: The analysis demonstrates good convergence of the chiral expansion. As the order increases (LO $\to$ N3LO), the uncertainty bands shrink, and results remain compatible with previous orders.

5. Significance and Astrophysical Implications

• Stellar Evolution: The authors investigated the impact of the new $S(0)$ value (varying by $\pm 3\sigma$) on stellar cluster age determinations. They found that even a $3\sigma$ variation in $S(0)$ changes the turn-off luminosity by only $\sim 2-2.5\permil$, which is negligible compared to other sources of uncertainty in stellar models.
• Solar Neutrinos: Within the Standard Solar Model, a variation in $S(0)$ is compensated by a change in the central temperature ($T_c$).
  • A $3\sigma$ variation in $S(0)$ leads to a change in $T_c$ of less than $2\permil$.
  • Consequently, helioseismic quantities and surface abundances remain unaffected.
  • The most sensitive neutrino fluxes ($^8$B, $^{15}$O, $^{17}$F) vary by at most $\sim 5\%$ at $3\sigma$. While non-negligible, this is smaller than the impact of varying other solar model inputs.
• Future Outlook: The authors conclude that a substantially more precise determination of $S(E)$ is not currently the limiting factor for solar neutrino physics or stellar age dating. Future progress requires comparable reductions in uncertainties of other theoretical inputs and percent-level accuracy in future neutrino flux measurements.

In summary, this paper provides a robust, statistically rigorous determination of the proton-proton fusion rate, resolving long-standing discrepancies between local and non-local potential models and establishing a new standard for theoretical uncertainty quantification in nuclear astrophysics.