Gravitational Anomaly Measurement in Wide Binaries is Sensitive to Orbital Modeling

This paper demonstrates that the previously reported gravitational anomaly in wide binaries, which supported Modified Newtonian Dynamics, is likely an artifact of orbital modeling choices, as a hierarchical Bayesian analysis accounting for three-dimensional orbital elements yields results consistent with standard Newtonian gravity.

The Gist

Here is an explanation of the paper, translated into everyday language with some creative analogies.

The Big Question: Is Gravity Broken?

Imagine you are a detective trying to solve a mystery about how the universe works. For a long time, we've believed in Newton's Law of Gravity, which says that the heavier two objects are, and the closer they are, the stronger they pull on each other.

But recently, some astronomers noticed something weird. In the vast emptiness between stars (where gravity is very weak), stars seem to be moving faster than Newton's laws predict. It's like two dancers holding hands, spinning so fast that they should fly apart, but they don't.

Some scientists, led by Chae et al. (2026), looked at 36 pairs of these "wide binary" stars. They concluded: "Newton is wrong! Gravity is actually about 60% stronger than we thought in these quiet zones." They called this a "gravitational anomaly" and said it proves a theory called MOND (Modified Newtonian Dynamics), which suggests gravity changes its rules when things get very slow or far apart.

The New Investigation: A Different Way to Measure

The authors of this new paper (Saad & Ting) decided to take a second look at the exact same 36 star pairs. They used the same data but a different mathematical "magnifying glass" (a statistical model).

Think of it like trying to guess the speed of a car based on a blurry photo.

• Chae's approach was like saying: "I see the car is 100 meters away in the photo. I'll assume that's exactly how far it is, and calculate the speed from there."
• Saad & Ting's approach was like saying: "I see the car is 100 meters away in the photo, but I also know cars have different sizes and speeds. Let me calculate the actual distance the car traveled, which might be different from what the photo shows, and then calculate the speed."

The "Aha!" Moment: It's All About the Ruler

When Saad & Ting ran their new model, they found something surprising.

1. The "Strict Ruler" Method (Like Chae): When they forced the model to treat the distance seen in the photo as the only truth (ignoring the fact that orbits are 3D and we are seeing them from an angle), they got the same result as Chae: Gravity is boosted (γ ≈ 1.6).
2. The "Flexible Ruler" Method (Their Baseline): When they allowed the model to realize that the "distance" in the photo is just a 2D shadow of a 3D orbit, and let the math figure out the real distance (the semi-major axis) as a separate variable, the result changed completely. They found: Gravity is normal (γ ≈ 1.12).

The Result: Their result is consistent with Newton's original laws. There is no need for a "gravity boost."

The Metaphor: The Shadow Puppet

Imagine you are looking at a shadow puppet on a wall.

• The shadow looks small.
• If you assume the puppet is exactly the size of the shadow, you might think the puppet is tiny and moving incredibly fast to cast that shadow.
• But if you realize the puppet is actually a large hand held far away from the light, the "fast movement" disappears. The hand is just moving at a normal speed; it's just the perspective that made it look weird.

The authors argue that the previous study (Chae et al.) was looking at the shadow (the projected separation) and assuming it was the whole story. This made the stars look like they were moving too fast, creating a fake "gravity boost."

When Saad & Ting accounted for the 3D reality (the actual size of the hand), the "too fast" movement vanished, and the stars were behaving exactly as Newton predicted.

Why This Matters

This paper doesn't just say "Chae was wrong." It says, "The answer depends entirely on how you measure the distance."

• If you measure distance in a specific, rigid way, you find a mystery (Modified Gravity).
• If you measure distance in a more flexible, realistic way, the mystery disappears (Standard Gravity).

The Bottom Line

The evidence for "broken gravity" in these 36 star systems is not solid. It turns out that the "anomaly" was likely an illusion caused by how the math handled the 3D shape of the stars' orbits.

Before we rewrite the laws of physics, we need to make sure we aren't just misreading the shadows. For now, Newton is still the king of gravity, at least for these star pairs.

Technical Summary

Here is a detailed technical summary of the paper "Gravitational Anomaly Measurement in Wide Binaries is Sensitive to Orbital Modeling" by Saad & Ting (2026).

1. Problem Statement

Recent studies, most notably Chae et al. (2026), have reported a gravitational anomaly in a sample of 36 wide-binary star systems. By analyzing high-precision 3D velocities, Chae et al. inferred a gravity boost factor of $\gamma \equiv G_{\text{eff}}/G_N \approx 1.60^{+0.17}_{-0.14}$, suggesting a deviation from Newtonian gravity consistent with Modified Newtonian Dynamics (MOND) predictions in the low-acceleration regime.

However, other studies (e.g., Banik et al. 2024; Saglia et al. 2025) using similar datasets or methodologies have found results consistent with Newtonian gravity ($\gamma = 1$). This conflict raises a critical question: Is the reported anomaly a genuine physical effect, or is it an artifact of the statistical methodology and orbital modeling assumptions used?

2. Methodology

The authors reanalyze the exact same dataset of 36 wide binaries from Chae et al. (2026) using a hierarchical Bayesian framework adapted from their previous work (S&T25).

• Global Parameter Inference: Unlike Chae et al. (2026), who inferred a parameter $\Gamma$ for each system individually and then combined them, Saad & Ting infer a single global gravity boost factor ($\gamma$) shared across all 36 systems.

• Orbital Modeling: The model treats each binary as a two-body Keplerian system with six orbital elements ($a, e, i, \Omega, \omega, M$) and stellar masses. The orbital velocity is modified by $\gamma$ via $v = \sqrt{\gamma G_N M_{\text{tot}}/r}$.

• Two Model Variants: To isolate the source of the discrepancy, the authors compare two specific approaches to determining the three-dimensional (3D) orbital separation ($r_{\text{true}}$):

  1. Baseline Model (Independent Semi-Major Axis): The semi-major axis ($a$) is treated as a free parameter with a prior distribution dependent on eccentricity. The 3D separation $r_{\text{true}}$ is calculated via Kepler's equation, and the observed projected separation ($r_\perp$) is treated as an observable with Gaussian uncertainty. This allows the model to infer an orbital scale independent of the instantaneous projected separation.
  2. Geometric De-projection Model (Chae-style): The semi-major axis is not an independent parameter. Instead, the 3D separation is derived directly from the observed projected separation ($r_\perp$) using a geometric de-projection formula: $r_{\text{true}} = r_\perp / \sqrt{\cos^2 \phi + \cos^2 i \sin^2 \phi}$. This tightly couples the orbital scale to the observed projection.

• Inference Engine: The analysis uses Hamiltonian Monte Carlo (HMC) with the No-U-Turn Sampler (NUTS) implemented in PyMC to explore the high-dimensional posterior distributions.

3. Key Contributions

• Isolation of Modeling Sensitivity: The study demonstrates that the inferred value of $\gamma$ is highly sensitive to how the 3D orbital separation is modeled, specifically the treatment of the semi-major axis.
• Reproduction of Conflicting Results: By simply switching the orbital parameterization while keeping the data and likelihood function otherwise identical, the authors successfully reproduce both the Newtonian-consistent result and the MOND-consistent anomaly.
• Statistical Rigor: The paper provides a rigorous comparison of hierarchical global inference versus per-system consolidation, showing that the choice of orbital parameterization is the dominant factor in the discrepancy, rather than the inference algorithm (NUTS vs. emcee) or the treatment of projected separation uncertainties.

4. Results

The results of the two model variants are starkly different:

Model Variant	Inferred $\gamma$	68% Credible Interval	Consistency with Newtonian ($\gamma=1$)	Consistency with Chae et al. (2026)
Baseline (Independent $a$)	$1.12^{+0.27}_{-0.22}$|$[0.90, 1.38]$| **Consistent** ($\sim 0.4\sigma$)	Inconsistent
Geometric De-projection	$1.56^{+0.21}_{-0.18}$|$[1.38, 1.77]$| Inconsistent | **Consistent** (Matches$\approx 1.60$)

• Baseline Model: When the semi-major axis is allowed to vary independently, the data favors $\gamma \approx 1.12$, which is fully consistent with Newtonian gravity. The probability that $\gamma > 1$ is only 0.70.
• Geometric Model: When the 3D separation is forced to be a geometric de-projection of the observed projected separation (mimicking Chae et al.'s approach), the inferred $\gamma$ shifts to $1.56$, with a probability of$\gamma > 1$ effectively equal to 1.0. This matches the anomaly reported by Chae et al.
• Sensitivity: The shift in $\gamma$ ($\Delta \gamma \approx 0.44$) is driven entirely by the treatment of the orbital scale. In the geometric model, any velocity excess that cannot be explained by the fixed geometric relationship is absorbed by the gravity boost factor $\gamma$. In the baseline model, the independent semi-major axis parameter can account for these velocity variations without invoking non-Newtonian gravity.

5. Significance and Conclusion

The paper concludes that the evidence for a gravitational anomaly in the Chae et al. (2026) sample is not robust to reasonable changes in orbital modeling.

• Methodological Artifact: The reported detection of $\gamma \approx 1.6$ appears to be a consequence of the specific geometric de-projection method used, which lacks an independent semi-major axis parameter. This parameterization may force the model to attribute kinematic discrepancies (which could be due to orbital geometry or measurement noise) to a modification of gravity.
• Implications for MOND Tests: Future tests of gravity in wide binaries must carefully treat the relationship between projected separation, true separation, and the semi-major axis. The inclusion of an independent semi-major axis parameter is crucial to avoid spurious detections of gravity boosts.
• Broader Context: The findings align with other studies (Banik et al. 2024; Saglia et al. 2025) that favor Newtonian gravity, suggesting that the "anomaly" reported by Chae et al. may be a modeling artifact rather than new physics.

In summary, Saad & Ting (2026) provide a critical methodological check, showing that the claimed detection of Modified Newtonian Dynamics in wide binaries is highly dependent on the statistical treatment of orbital elements, specifically the semi-major axis.