⚛️ general relativity

Bayesian and Machine-Learning Analyses of Nonminimal $f(Q)$ Gravity and $H_0$ Tension

This study employs Bayesian and machine-learning analyses on nonminimally coupled $f(Q)$ gravity within the metric-affine formalism to demonstrate that the framework offers a flexible approach for late-time cosmology that partially alleviates the $H_0$ tension while remaining consistent with diverse observational datasets.

Original authors: Simran Arora, Mridul Patel

Published 2026-07-30

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Simran Arora, Mridul Patel

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

Imagine the universe as a giant, expanding balloon. For decades, scientists have had a very reliable recipe for how that balloon inflates, called the "Standard Model" of cosmology. This recipe relies on two main ingredients: invisible stuff called "Dark Matter" that holds galaxies together, and a mysterious push called "Dark Energy" that makes the balloon expand faster and faster. This recipe works so well that it explains almost everything we see, from the afterglow of the Big Bang to the way stars move.

But recently, a tiny crack appeared in the recipe. When scientists measured how fast the universe is expanding right now using two different methods, they got two different answers. One method, looking at the baby universe (the Big Bang's leftovers), says the expansion rate is about 67.4 km/s/Mpc. The other method, looking at the "grown-up" universe nearby using exploding stars, says it's about 73.04 km/s/Mpc. It's like two chefs measuring the same cake and getting different sizes; the difference is small in absolute terms, but statistically, it's a huge deal. This disagreement is known as the "Hubble Tension," and it suggests our recipe might be missing a secret ingredient or that the rules of gravity we've been using aren't the whole story.

This paper dives into a new, slightly weird recipe for gravity to see if it can fix that cake measurement. Instead of the usual rules of gravity (General Relativity), the authors explore a theory called f(Q) gravity. In this theory, gravity isn't just about the bending of space (curvature) or the twisting of space (torsion), but about something called nonmetricity. Think of nonmetricity as a "stretchiness" or a change in the ruler we use to measure distances. In this new framework, the authors propose that this stretchiness doesn't just sit there; it talks directly to the matter in the universe (like stars and gas), creating a "nonminimal coupling." It's as if the ruler itself is whispering to the cake, changing how it expands. The authors built a specific mathematical model for this, ran it through a massive computer simulation using real data from telescopes, and even used artificial intelligence to double-check their work.

The Investigation: A New Gravity Recipe

The authors started by constructing a specific version of this "stretchy" gravity theory. They imagined a universe where the nonmetricity scalar (let's call it Q, the measure of that stretchiness) is linked directly to the matter in the universe. In standard physics, matter and geometry (space) usually follow their own separate rules. Here, they are tied together like dance partners. The authors chose a simple mathematical shape for this relationship: a mix of a straight line and a curve (a power-law model). They then asked: "If we use this new gravity recipe, can we get the two different measurements of the universe's expansion rate to agree better?"

To find out, they didn't just guess; they brought in the heavy artillery of data. They combined information from four different cosmic sources:

Cosmic Chronometers: These are ancient, passive galaxies that act like cosmic stopwatches, telling us how fast the universe was expanding at different times in the past.
Type Ia Supernovae: These are exploding stars that act as "standard candles." Because we know how bright they should be, we can tell how far away they are and how fast they are moving away from us. The team used three different catalogs of these stars (Pantheon+, DESY5, and Union3) to be extra sure.
DESI BAO: This uses the "Baryon Acoustic Oscillations," which are frozen sound waves from the early universe. They act like a cosmic ruler, measuring the distance between galaxies.
CMB (Cosmic Microwave Background): This is the "baby picture" of the universe, the leftover heat from the Big Bang, which gives us a baseline for how the universe started.

The Results: A Partial Fix, Not a Magic Wand

After crunching the numbers using a powerful statistical method called MCMC (which is like trying millions of different ingredient combinations to find the perfect recipe), the authors found some interesting results.

First, the new gravity model does work. It fits the data almost as well as the standard recipe. The universe in their model expands in a way that looks very similar to what we observe. The "stretchiness" of space (the nonmetricity) seems to provide just enough extra push to change the expansion history slightly.

Second, and most importantly, this new model partially alleviates the Hubble Tension. When they calculated the expansion rate ( $H_0$ ) using their new model, they got a value around 68.4 to 68.8 km/s/Mpc.

This is higher than the "baby universe" measurement (67.4), which is good.
But it is still lower than the "grown-up universe" measurement (73.04).
The gap between the model and the "grown-up" measurement is reduced from about 4.4 standard deviations (a huge disagreement) to about 3 standard deviations.

Think of it like this: The standard recipe said the cake was size 67. The local measurement said it was size 73. The new recipe suggests the cake is actually size 68.5. It's not a perfect match for the local measurement, but it's a much better compromise than before. The tension is still there, but it's less severe. The authors note that while this is a promising step, it doesn't completely solve the mystery. The model is consistent with the data, but the standard model (without the extra stretchiness) is still statistically very competitive.

The AI Check: Teaching Machines to See the Pattern

To make sure their math wasn't just a fluke, the authors brought in a third party: Machine Learning. They treated the expansion history predicted by their new gravity model as a "teacher" and asked three different types of AI algorithms to learn from it and predict the future expansion.

The three "students" were:

Linear Regression: A simple student that looks for straight-line patterns.
Random Forest: A student that uses many decision trees to find complex patterns.
Support Vector Regression (SVR): A sophisticated student that can handle very complex, curved relationships.

The results were clear: The SVR with an RBF kernel (a specific type of mathematical trick) was the star student. It predicted the expansion history with near-perfect accuracy (99.93% accuracy). It was far better than the simple linear student and even better than the Random Forest. This confirmed that the expansion history generated by their new gravity model is smooth, consistent, and follows a pattern that advanced AI can easily recognize. It's like the AI looked at the new gravity recipe and said, "Yes, this makes sense; the pattern is solid."

The Verdict

So, what's the final takeaway? The paper suggests that a universe where space has a "stretchy" quality that talks directly to matter is a viable candidate for explaining our cosmos. It offers a flexible framework that can mimic the standard model but also tweak the expansion rate to help ease the tension between different measurements.

However, the authors are careful not to declare victory. They point out that while the tension is reduced, it isn't gone. The new model is a "promising and flexible framework," but it requires more testing. The Bayesian analysis (a way of weighing how likely a theory is compared to another) suggests that while the new model fits the data, the extra complexity isn't strongly demanded by the current observations yet. It's a strong contender, but the standard model is still holding its ground.

In the end, this paper is a reminder that the universe is full of surprises. By tweaking the rules of gravity just a little bit—introducing a conversation between the geometry of space and the matter within it—we might be getting closer to understanding why the universe is expanding at the rate it is. The Hubble Tension remains a puzzle, but this new piece of the jigsaw fits in a way that makes the picture a little clearer.

Technical Summary: Bayesian and Machine-Learning Analyses of Nonminimal f(Q) Gravity and H0 Tension

Problem Statement
The paper addresses the persistent "Hubble tension," a significant discrepancy (currently $\sim 4.4\sigma$ to $5.3\sigma$ ) between early-universe determinations of the Hubble constant ( $H_0 \approx 67.4$ km s $^{-1}$ Mpc $^{-1}$ from Planck/ΛCDM) and late-universe direct measurements ( $H_0 \approx 73.04$ km s $^{-1}$ Mpc $^{-1}$ from SH0ES). While the standard $\Lambda$ CDM model successfully describes cosmic evolution, it struggles to reconcile these values without invoking new physics. The authors investigate whether nonminimally coupled $f(Q)$ gravity, a modification of General Relativity (GR) based on nonmetricity rather than curvature or torsion, can provide a consistent geometric framework that alleviates this tension while remaining compatible with other cosmological probes.

Methodology
The study employs a metric–affine formalism within the symmetric teleparallel framework, where the nonmetricity scalar $Q$ couples directly to the matter Lagrangian ( $L_m$ ).

Theoretical Framework: The authors construct an action $S = \int d^4x \sqrt{-g} [\frac{1}{2}f_1(Q) + f_2(Q)L_m]$ . They derive the field equations and continuity equations, noting that the nonminimal coupling leads to the non-conservation of the energy-momentum tensor, effectively introducing an exchange of energy and momentum between geometry and matter.
Cosmological Model: A specific power-law model is adopted: $f_1(Q) = -Q + \alpha Q^2$ and $f_2(Q) = 1 + \beta Q$ . The background cosmology is derived for a flat FLRW universe, yielding modified Friedmann equations. The model introduces coupling parameters $\alpha$ and $\beta$ , with $\beta$ constrained by current matter density and $H_0$ .
Statistical Analysis: A comprehensive Markov Chain Monte Carlo (MCMC) analysis is performed using the emcee sampler. The analysis utilizes a combination of observational datasets:
- Cosmic Chronometers (CC) for $H(z)$ .
- DESI BAO DR2 (Baryon Acoustic Oscillations).
- Type Ia Supernovae (SNeIa) from Pantheon+ (PP), DESY5 (DESY), and Union3.
- Compressed CMB data from Planck (distance priors $R$ , $l_A$ , and $\Omega_b h^2$ ).
- Model selection criteria (AIC, BIC, and Bayesian evidence) are used to compare the $f(Q)$ model against standard $\Lambda$ CDM.
Machine Learning (ML) Validation: To complement the statistical analysis, the authors apply supervised learning algorithms—Linear Regression (LR), Support Vector Regression (SVR) with RBF and polynomial kernels, and Random Forest (RF)—to the theoretical $H(z)$ data generated by the best-fit parameters. These models are trained on 80% of the data and tested on 20% to evaluate predictive performance and robustness.

Key Contributions and Results

Partial Alleviation of H0 Tension: The MCMC analysis yields a best-fit $H_0$ in the range of $68.4 $–$ 68.8$ km s $^{-1}$ Mpc $^{-1}$ across different dataset combinations. This value sits between the CMB and local distance-ladder measurements, representing a partial alleviation of the tension. The model reduces the discrepancy with SH0ES to approximately $3\sigma$ , while remaining consistent with other late-time probes (TRGB, GWTC-3, etc.) within $1\text{--}2\sigma$ .
Statistical Consistency: The nonminimal $f(Q)$ model provides a fit of comparable statistical quality to $\Lambda$ CDM. While the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) show mixed results depending on the dataset combination, the Bayesian evidence ( $\Delta \ln Z$ ) suggests that the current data does not strongly favor the additional complexity of the $f(Q)$ model over $\Lambda$ CDM (values of $|\Delta \ln Z| < 1$ indicate inconclusive evidence).
Cosmographic Evolution: The reconstructed expansion history shows a persistent accelerating phase ( $q \approx -0.8$ at low redshift) transitioning toward a de Sitter phase ( $q \to -1$ ) in the asymptotic future. The effective equation of state ( $w_{eff}$ ) remains close to $-0.9$, mimicking a dark-energy-dominated universe.
Machine Learning Performance: The ML analysis reveals a clear hierarchy in predictive accuracy. The SVR model with a Radial Basis Function (RBF) kernel outperforms all other methods, achieving an $R^2 \approx 0.9993$ and minimal Mean Squared Error (MSE) and Mean Absolute Error (MAE). Linear Regression also performs well ( $R^2 \approx 0.9976$ ), while SVR with a polynomial kernel and Random Forest show lower accuracy. This confirms the robustness of the reconstructed $H(z)$ dynamics derived from the $f(Q)$ model.

Significance and Claims
The paper claims that nonminimally coupled $f(Q)$ gravity offers a "promising and flexible framework" for late-time cosmology. Its primary significance lies in demonstrating that geometric modifications involving nonmetricity can naturally generate late-time acceleration and partially resolve the $H_0$ tension without violating constraints from CMB, BAO, or SNeIa.

The authors modestly conclude that while the model does not entirely resolve the Hubble tension, it successfully bridges the gap between early- and late-universe determinations. The study emphasizes that the nonminimal matter–geometry coupling provides a consistent realization of second-order field equations, avoiding higher-derivative instabilities often found in other modified gravity theories. The integration of machine learning serves as a complementary validation tool, confirming that the theoretical expansion history derived from the model is robust and predictable. The authors note that future work is required to investigate the model's impact on matter density perturbations and structure growth, which are currently outside the scope of this background-level analysis.

The Investigation: A New Gravity Recipe

The Results: A Partial Fix, Not a Magic Wand

The AI Check: Teaching Machines to See the Pattern

The Verdict

More like this