Testing general relativity with gravitational waves -- improving and extending Modified Dispersion Relation tests

This paper presents an improved methodology for testing general relativity using gravitational waves, featuring group velocity parametrization and enhanced sampling, which yields tighter constraints on modified dispersion relations and a reduced upper bound on the graviton mass when reanalyzing GWTC-3 events, while finding no evidence for violations of general relativity for negative momentum exponents.

The Gist

Imagine the universe as a giant, invisible ocean. For over a century, we've believed that ripples in this ocean—called gravitational waves—travel at the exact same speed as light, no matter how "tall" or "short" the ripple is. This is a core rule of Einstein's General Relativity.

But what if that rule isn't perfect? What if the ocean has a hidden texture that makes some ripples travel slightly faster or slower than others? This is what the scientists in this paper are investigating.

Here is a simple breakdown of their work, using some everyday analogies.

1. The Goal: Checking the Rules of the Universe

The LIGO, Virgo, and KAGRA detectors (let's call them the "Cosmic Ears") have been listening to the universe for years, catching the sounds of black holes smashing together. Every time they catch a sound, they check: Did this sound travel exactly as Einstein predicted?

One specific test they run is looking for a "Modified Dispersion Relation" (MDR).

• The Analogy: Imagine you are at a concert. If the music travels perfectly, the bass (low notes) and the violin (high notes) arrive at your ears at the exact same time.
• The Twist: If the air in the concert hall was "thick" or "sticky," the bass might arrive a split-second later than the violin. That delay would prove the air isn't behaving normally.
• The Science: The scientists are checking if the "air" of space-time is sticky. If it is, it means Einstein's theory might need a tiny tweak, or perhaps a new particle (like a "massive graviton") is slowing things down.

2. The Problem: The Old Tools Were a Bit Clunky

In previous years (specifically the "GWTC-3" catalog), the scientists ran this test, but they were using some old, slightly broken tools.

• The "Bad Map": They were using a map that assumed the ripples traveled like individual particles. But waves actually travel like a group (a "group velocity"). It's like trying to navigate a river by tracking a single drop of water instead of the whole current. This led to some confusion in the math.
• The "Bad Filter": When they analyzed the data, they used a filter that threw away too much information. Imagine trying to hear a whisper in a noisy room, but your headphones accidentally muted 90% of the sound. You'd get a result, but it would be shaky and full of "static" (statistical noise).
• The Result: Their previous answers were okay, but they were a bit blurry and sometimes had "ghost peaks" (fake signals that looked like real discoveries).

3. The Upgrade: Sharper Tools and New Angles

In this new paper, the team (led by researchers from Utrecht, Szeged, London, and Birmingham) completely rebuilt their analysis toolkit.

• Better Physics: They switched to the "group velocity" model. Now, they are tracking the whole wave packet, not just a single particle. It's like upgrading from a blurry black-and-white photo to a high-definition color video.
• Smarter Math: They fixed the "filter" issue. Instead of throwing away data, they now weigh it correctly. This means their results are much sharper, with fewer fake "ghost peaks."
• Looking in the Dark: Previously, they only looked for "high-energy" deviations (like a sledgehammer hitting a glass window). Now, they are also looking for "low-energy" deviations (like a gentle breeze). They tested new mathematical scenarios where the exponent is negative (like -1, -2, -3). This is like checking if the ocean is sticky not just for fast swimmers, but also for slow drifters.

4. The Results: Einstein Still Wins (For Now)

After re-analyzing 43 black hole collision events with their new, super-sharp tools, here is what they found:

• Cleaner Data: The "ghost peaks" are gone. The data is much more reliable.
• Tighter Limits: Because their tools are better, they can set stricter rules. They can now say with more confidence: "If there is a massive graviton, it must be lighter than X."
  • The Number: They tightened the limit on the "mass of the graviton" from $2.42 \times 10^{-11}$ peV to $2.21 \times 10^{-11}$ peV.
  • The Analogy: It's like saying, "We used to think the thief could weigh up to 200 pounds. Now, with better scales, we know the thief can't weigh more than 190 pounds."
• No New Physics Found: Despite looking harder and in new directions (including the negative exponents), they found no evidence that Einstein was wrong. The gravitational waves still travel exactly as General Relativity predicts.

5. Why This Matters

You might think, "If they didn't find anything new, why write a paper?"

It's like a detective saying, "I checked the crime scene with a brand-new, high-tech microscope, and I can confirm with 100% certainty that the suspect wasn't there."

• Ruling Out Theories: By proving that Einstein's theory holds up even under this new, stricter scrutiny, they rule out many alternative theories of gravity.
• Future Proofing: They have built a better engine for the next generation of detectors. When the next big catalog of black hole collisions comes out (GWTC-4), they will be ready to find the truth faster and more accurately than ever before.

In a nutshell: The scientists upgraded their math and physics tools to look for cracks in Einstein's theory of gravity. They looked harder and in new places than ever before, but the theory still stands strong. The universe, it seems, is still playing by Einstein's rules.

Technical Summary

1. Problem Statement

The LIGO-Virgo-KAGRA (LVK) collaboration routinely tests General Relativity (GR) by searching for deviations in the propagation of gravitational waves (GWs). A primary test involves constraining a Modified Dispersion Relation (MDR), which posits that GWs may not travel at the speed of light ($c$) or may have a frequency-dependent speed due to effects like a massive graviton or Lorentz invariance violation.

The previous standard analysis, performed on the GWTC-3 catalog, suffered from several technical limitations:

• Inefficient Sampling: The analysis sampled in a logarithmic wavelength parameter ($\log_{10}\lambda_{eff}$) with a uniform prior, which led to poor sampling efficiency when reweighting to the physical amplitude parameter ($A_\alpha$). This resulted in a low effective sample size ($n_{eff}$) and artificial multimodalities in posterior distributions.
• Theoretical Inconsistency: The previous derivation of the MDR phase correction relied on particle velocity. However, for general MDRs (where $\alpha \neq 0$), GWs should be treated as wave packets propagating at group velocity. The previous method failed to account for this distinction, particularly for the $\alpha=1$ case.
• Limited Scope: The analysis was restricted to non-negative exponents ($\alpha \geq 0$), missing potential low-energy modifications predicted by certain dark energy models.
• Waveform Limitations: Previous tests used waveform models containing only the dominant $(2,2)$ mode, neglecting higher-order modes (HMs) which are crucial for events with significant subdominant mode content.
• Calibration Errors: The paper notes that previous results contained uncorrected errors regarding calibration uncertainties and data windowing, which have been rectified in this work via reweighting.

2. Methodology

The authors implemented a comprehensive overhaul of the MDR analysis pipeline using the Bilby software package and the Dynesty nested sampler. Key methodological improvements include:

• Group Velocity Parametrization:

  • The phase correction $\delta\Psi(f)$ is now derived based on group velocity ($v_g = d\omega/dk$) rather than particle velocity.
  • This introduces a re-scaling factor of $(1-\alpha)$ to the amplitude parameter $A_\alpha$.
  • Special Case ($\alpha=1$): The group velocity approach reveals that for $\alpha=1$, the phase shift is frequency-independent (a constant offset). This creates a degeneracy where the posterior for $A_1$ becomes non-normalizable (split into infinite branches). Consequently, the authors exclude constraints on $A_1$ in this study, unlike previous analyses.

• Improved Sampling Strategy:

  • Instead of sampling in $\log_{10}\lambda_{eff}$, the analysis now samples directly in the effective amplitude parameter ($A_{eff}$) with a uniform prior.
  • This eliminates the need for aggressive reweighting, drastically reducing sample rejection and increasing the effective sample size ($n_{eff}$).
  • For the massive graviton case ($\alpha=0$), the authors demonstrate that sampling in $A_{eff}$ and transforming to graviton mass ($m_g$) yields more stable results than sampling directly in $m_g$ (which suffers from boundary effects at $m_g=0$).

• Extension to Negative Exponents:

  • The test is extended to negative values of $\alpha$ ($\alpha \in \{-1, -2, -3\}$).
  • Motivation: Negative $\alpha$ corresponds to low-frequency modifications, potentially arising from dark energy fields acting as a medium for GW propagation. The authors model the transition of GW speed back to $c$ at high frequencies (LVK band) as a power-law tail, which manifests as negative $\alpha$ in the MDR framework.

• Higher Order Modes (HMs):

  • The analysis utilizes the IMRPhenomXPHM waveform model, which includes higher-order modes, replacing the previous IMRPhenomXP (dominant mode only).
  • This is critical for events like GW190412 and GW190814, where HMs break degeneracies and improve parameter estimation.

• Reprocessing:

  • The entire GWTC-3 catalog (43 events) was re-analyzed.
  • Corrections were applied for known calibration uncertainties and data windowing issues via reweighting.

3. Key Contributions

1. Theoretical Correction: Shifted the MDR derivation from particle velocity to group velocity, correcting the phase evolution for general $\alpha$ and identifying the untestable nature of $\alpha=1$ under group velocity parametrization.
2. Algorithmic Efficiency: Developed a sampling scheme in $A_{eff}$ that avoids the "railing" and low effective sample size issues of the previous $\log_{10}\lambda_{eff}$ method.
3. New Parameter Space: First comprehensive test of MDRs with negative exponents ($\alpha < 0$), probing low-energy physics and dark energy models.
4. Robustness Verification: Conducted extensive injection campaigns (100 BBH signals) and pp-plots to verify that the new pipeline produces unbiased, well-behaved posteriors.
5. Code Migration: Migrated the MDR test code from the legacy C-based LALInference to the Python-based Bilby framework, facilitating future updates and integration with other GR tests.

4. Results

The re-analysis of the 43 GWTC-3 events yielded the following results:

• Posterior Quality:

  • Individual event posteriors show significantly higher effective sample sizes and fewer multimodalities compared to GWTC-3.
  • The "artificial" sharp cutoffs and secondary peaks seen in GWTC-3 (caused by low $n_{eff}$ and bounded KDEs) are largely eliminated.
  • Combined posteriors on amplitude parameters $A_\alpha$ are, on average, 19% narrower than the GWTC-3 results.

• Graviton Mass Constraint:

  • The 90% upper bound on the graviton mass ($m_g$) improved from $2.42 \times 10^{-11}$ peV (GWTC-3) to $2.21 \times 10^{-11}$ peV.
  • The authors note that the combined result is surprisingly close to the old result, suggesting that the errors in individual GWTC-3 posteriors (low $n_{eff}$) largely canceled out when combined.

• Negative $\alpha$ Tests:

  • For $\alpha \in \{-1, -2, -3\}$, no evidence of GR violation was found. The GR value ($A_\alpha = 0$) was recovered within the expected quantiles.
  • Injection tests confirmed that the pipeline can recover deviations even if the assumed $\alpha$ in the template mismatches the injected $\alpha$ (within the tested range), though precise determination of the specific $\alpha$ value remains difficult.

• Higher Order Modes:

  • Including HMs (IMRPhenomXPHM) significantly narrowed the posteriors for events with significant subdominant mode content (e.g., GW190412), particularly for $\alpha < 1$.
  • For $\alpha=0$, the inclusion of HMs narrowed the posterior enough that the GR value ($A_0=0$) fell outside the 90% credible interval for GW190412, highlighting the importance of HM modeling.

• Tension with GR:

  • For $\alpha=1.5$ and $\alpha=2.5$, the combined posteriors show a slight tension with GR (recovering GR outside the 90% CI). However, this is driven primarily by two specific events (GW200219 and GW200225). The authors attribute this to statistical fluctuations rather than a systematic failure, noting that adding more events in future catalogs (GWTC-4) should resolve this.

5. Significance

This paper represents a critical refinement of the LVK's standard GR tests. By correcting theoretical derivations (group vs. particle velocity) and fixing sampling inefficiencies, the authors have established a more robust and accurate baseline for testing gravity with gravitational waves.

• Reliability: The removal of artificial multimodalities and boundary artifacts ensures that future constraints on modified gravity are not biased by statistical sampling errors.
• New Physics Reach: The extension to negative $\alpha$ opens a new window for testing dark energy models and low-frequency dispersion, complementing cosmological observations.
• Future Readiness: The improved pipeline is ready for the upcoming GWTC-4 catalog. With the expected doubling of events, the authors anticipate significantly tighter constraints on the graviton mass and MDR amplitudes, potentially resolving current tensions or definitively ruling out specific deviations from General Relativity.