The NANOGrav 15-year data set: A Limit on Continuous Spin Gravitons

arXiv:2611.14613 · hep-ph, astro-ph.CO, gr-qc · Submitted 2026-11-10 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Phenomenology. The day in particle phenomenology: models, signatures and constraints.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "The NANOGrav 15-year data set".

Mira: Here's the gist of the paper. The NANOGrav 15-year data set constrains the graviton spin scale to ρg < 1.23 × 10−23 eV at the 95% credible level Continuous spin particles (CSPs) are massless particles characterized by a dimensionful spin scale, ρ, which, when non-zero, changes their helicity under Lorentz boosts Since CSPs reduce to ordinary particles when ρ → 0, the photon or graviton could be CSPs and only appear to behave as ordinary particles due to a small value of ρ Therefore, it is important to determine what value of ρ is experimentally allowed Recent work has shown that the physical effects of CSPs scale as powers of ρ/ω when ρ≪ ω, where ω is a frequency scale, making low-frequency probes particularly beneficial for highlighting these phenomena In this paper we use pulsar timing arrays to set constraints on the spin scale of the continuous spin graviton Modeling the Timing Residuals PTAs are exceptionally sensitive to changes in the arrival time of light emitted by pulsars Once ephemeris effects are accounted for, the arrival time of a light pulse that was emitted at time tem can be modeled as, tarr = tem + d + δt The purpose of this section is to derive δt at linear order in the GWB amplitude, starting with monochromatic GWs in Sec.

Kai: II A, and then focusing on a SGWB in Sec.

Mira: II B For a monochromatic GW with amplitude hA in the Ath polarization, A ∈ S+, and four-momentum kµ = ω (1, omegaˆ), the ρg-dependent photon Lagrangian to linear order in hA is given by Eq.

Kai: (4) The final CSG-induced timing residual is given by Eq.

Title and authors: Kai: So the title itself says exactly what we're looking at here: "The NANOGrav fifteen-year data set: A Limit on Continuous Spin Gravitons <ref:2611.14613#pg1,The NANOGrav 15-year data set: A Limit on Continuous Spin Gravitons>." It’s very direct about what they did and where they got their data.

Mira: And the authors are Ambasht, Bitcon, Trickle, and the NANOGrav Collaboration. It's a big collaboration pulling together all that pulsar timing array data we talked about earlier.

Lev: The implication is that this isn't just theoretical math; they took actual observational data from those arrays and put it to work constraining this new particle property.

Kai: They are essentially testing if the standard description of gravity holds up when we allow for these continuous spin particles with some spin scale, rho g.

Mira: And what they're doing is using the interferometric phenomenology derived in a previous paper, Ref. forty-two, to see how this spin scale affects the timing residuals in their data set, which is what they call the NANOGrav fifteen-year data set forty-three <ref:2611.14613#pg2,the NANOGrav 15-year data set 43>.

The paper's summary: Kai: So the summary boils down to using pulsar timing arrays to constrain the spin scale of a continuous spin graviton. They are looking at how this spin scale changes the timing residuals from gravitational waves.

Mira: They set up a model where they assume the photon’s spin scale is zero, and then they derive how this affects the timing delay when you have a monochromatic gravitational wave, or GW.

Lev: The key thing here is that they found that these effects appear at a level of one over rho g over omega squared when rho g is small compared to omega.

Kai: So PTAs are very sensitive because they probe those lowest frequency gravitational waves, which makes them ideal for sensing these tiny spin scales.

Mira: The preliminary expectation is that PTAs are sensitive to a spin scale around two pi over T, where T is the observation time of the PTA, normalized to fifteen years since that's when the NANOGrav collaboration started their observations <ref:2611.14613#pg2>.

Lev: That gives us a rough idea of the expected sensitivity based on observation time alone before you even look at the actual data constraints.

The paper's improvements: Kai: The authors introduce some modifications to how they model the timing residual pattern, specifically by modifying the overlap reduction function, which is called Gamma or Gamma with a rho g dependence.

Mira: They show that when rho g is not zero, this overlap reduction function changes in a specific way that depends on both the pulsar separation angle and the value of rho g over omega.

Lev: Because this modification enters as a term proportional to rho g over omega in their equations, they get two main effects: first, when rho g over omega is much larger than one, the correlation structure gets damped at all angles.

Kai: Second, for any fixed value of rho g, there's a frequency cutoff in that modified overlap reduction function at a frequency around rho g over two pi Hertz.

Mira: So they use this modified function to analyze the data and then run a marginalized likelihood analysis to find the allowed range for rho g.

Lev: The authors use MCMC sampling techniques, like PTArcade, to sample the posterior distribution of this spin scale based on their priors defined in Table I.

Conclusion: Kai: So to wrap up, the NANOGrav fifteen-year data set puts an upper limit on the continuous spin graviton spin scale at a nine five percent credible level of less than or equal to one point two three times ten to the negative twenty three electron volts <ref:2611.14613#pg1,the NANOGrav 15-year data set>.

Mira: That limit is independent of any prior range they choose, which means it’s a robust constraint derived directly from the data analysis.

Lev: It also notes that there is a peak in the posterior distribution around rho g equal to one point negative twenty three electron volts, but they say this isn't statistically significant compared to their prior limits.

Kai: So while they found a region where deviations from zero spin are slightly favored by the data, it’s not strong enough evidence on its own to claim anything new yet.

Mira: The implication is that for now, the graviton behaves very much like a standard massless particle in this context, but this sets an important benchmark for future experiments.

Lev: They suggest that future work could look at how a continuous spin graviton might interact with gravitational waves produced by things outside the Standard Model.

Kai: If there are other physics out there, like cosmologically produced gravitational waves from dark matter, some models predict enhanced low-frequency GW production that would be too large for the current PTA data.

Mira: But a non-zero rho g could actually suppress that signal, so it might be able to reconcile those tensions if we find a suitable spin scale.

Lev: They also mention that applying this exact framework to future, longer PTA datasets could potentially improve this upper limit on rho g even further.

Department of Physics, Grainger College of Engineering, University of Illinois Urbana-Champaign

hep-ph, astro-ph.CO, gr-qc

Submitted: 2026-11-10

Updated: 2026-11-10

Comments: 15 pages, 3 figures

Code: https://github.com/nanograv/enterprise_extensions

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 100/100

The gist: The NANOGrav 15-year data set constrains the graviton spin scale to ρg < 1.23 × 10−23 eV at the 95% credible level Introduction and Motivation Continuous spin particles (CSPs) are massless…

Key concepts

Continuous Spin Particles (CSPs)
These are massless particles that have a dimensionful spin scale, $ ho$. If $ ho$ is non-zero, these particles change their helicity when subjected to Lorentz boosts. They reduce to ordinary particles only when the scale $ ho$ approaches zero.
Pulsar Timing Arrays (PTAs)
PTAs are used to measure minute changes in the arrival time of light pulses from pulsars. These timing residuals are sensitive probes for gravitational waves, allowing scientists to search for signals from sources like continuous spin gravitons.
Overlap Reduction Function (ORF)
The ORF describes the correlation pattern of timing residuals between pairs of pulsars. When a continuous spin graviton exists ($ ho_g eq 0$), this function is modified, leading to two main effects: damping of correlations at all angles and a frequency cutoff in the signal.
Spin Scale ($ ho_g$)
This parameter represents the fundamental scale associated with the spin of a continuous spin graviton. The paper investigates how this non-zero scale affects gravitational wave signals, specifically how it suppresses PTA signals below a certain frequency.

Terminology

Summary

The NANOGrav 15-year data set constrains the graviton spin scale to ρg < 1. p. 023 × 10−23 eV at the 95% credible level

Introduction and Motivation

Continuous spin particles (CSPs) are massless particles characterized by a dimensionful spin scale, ρ, which, when non-zero, changes their helicity under Lorentz boosts Since CSPs reduce to ordinary particles when ρ → 0, the photon or graviton could be CSPs and only appear to behave as ordinary particles due to a small value of ρ Therefore, it is important to determine what value of ρ is experimentally allowed Recent work has shown that the physical effects of CSPs scale as powers of ρ/ω when ρ≪ ω, where ω is a frequency scale, making low-frequency probes particularly beneficial for highlighting these phenomena In this paper we use pulsar timing arrays to set constraints on the spin scale of the continuous spin graviton

Modeling the Timing Residuals

PTAs are exceptionally sensitive to changes in the arrival time of light emitted by pulsars Once ephemeris effects are accounted for, the arrival time of a light pulse that was emitted at time tem can be modeled as, tarr = tem + d + δt The purpose of this section is to derive δt at linear order in the GWB amplitude, starting with monochromatic GWs in Sec. II A, and then focusing on a SGWB in Sec. p. 1 II B For a monochromatic GW with amplitude hA in the Ath polarization, A ∈ S+, and four-momentum kµ = ω (1, omegaˆ), the ρg-dependent photon Lagrangian to linear order in hA is given by Eq. p. 1 (4) The final CSG-induced timing residual is given by Eq. p. 2 (7)

Impact of the Modified Overlap Reduction Function

The ORF, Γ(ξab), determines the timing residual correlation pattern in a PTA as a function of the separation angle, ξab, between pulsars a and b When ρg ̸= 0, the ORF is modified due to the timing residual changes in Eq. (9) The resulting modified ORF for a ≠ b is given by Eq. p. 3 (12) Since the effects of ρg enter as ρg/ω in Eq. p. 3 (12), there are two primary effects associated with a non-zero CSG spin-scale First, when ρg/ω ≫ 1 the correlation structure is damped at all angles Second, for fixed ρg there is a frequency cutoff in the modified ORF at f ∼ ρg/2π

Data Analysis and Constraints

The marginalized likelihood can be written as [47, 51, 52] With the priors and marginalized likelihood, posteriors can be computed The upper limit on the CSG spin scale at the 95% credible level from the NANOGrav 15-year data set is, ρg < 1.23 × 10−23 eV This limit is independent of the upper bound on the prior range

Conclusion and Future Work

Continuous spin particles are a theoretically selfconsistent generalization of massless particles [35] In this paper we used the NANOGrav 15-year data set [43] to place an upper limit at the 95% credible level of ρg < 1.23 × 10−23 eV Future work could study the implications a CSG would have on non-SMBHB produced GWBs, for example those produced cosmologically by physics beyond the Standard Model [10–14] Some of these models predict enhanced low-frequency GW production, yielding timing residuals too large to be consistent with PTA datasets However, a CSG might be able to reconcile this tension, since a non-zero ρg suppresses the PTA signal The upper limit on ρg could be further improved by applying the framework detailed here to future PTA datasets

Appendix A: Numerical Evaluation of the Modified Overlap Reduction Function

The modified ORF in Eq. (12) is calculated by numerically performing the integral over omegaˆ in the computational frame, i. p. 7e., with dˆ a, dˆ b given by Eq. p. 3 (13) Because the modified ORF depends solely on the angle between pulsars and ρg/ω, we avoid evaluating Eq. p. 7 (12) for every pair by precomputing Γ(ξab) on a grid and interpolating as needed The grid in angular separation is 500 linearly spaced points between 10−3 ≤ ξab ≤ π The grid in ρg/ω is built from three different grids First, a linearly spaced grid of 10 cubed points between 0 ≤ ρg/ω ≤ 10−1 is used to finely cover the range of small ρg/ω Second, a logarithmically spaced grid of 7 × 10 cubed points between 10−1 ≤ ρg/ω ≤ 10 cubed is used to capture the region with maximal Bessel function oscillations Third, we use a logarithmically spaced grid of 2 × 10 cubed points between 10 cubed ≤ ρg/ω ≤ 10 4 For ρg/ω > 10 4, the ORF is clipped to the value obtained from ρg/ω = 10 4 The absolute and relative errors of this look-up grid for the modified ORF are O(10−6), leading to negligible interpolation error

Appendix B: NANOGrav Collaboration Author List

• Gabriella Agazie Center for Gravitation, Cosmology and Astrophysics, Department of Physics and Astronomy, University of Wisconsin-Milwaukee, P.O. Box 413, Milwaukee, WI 53201, USA • Nikita Agarwal Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506, USA • Anjana Ashok Department of Physics, Oregon State University, Corvallis, OR 97331 • Jeremy G. p. 9 Baier Department of Physics, Oregon State University, Corvallis, OR 97331 • Paul T. p. 9 Baker Department of Physics and Astronomy, Widener University, One University Place, Chester, PA 19013 • Bence B´ecsy Institute for Gravitational Wave Astronomy and School of Physics and Astronomy, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK • Laura Blecha Physics Department, University of Florida, Gainesville, FL 32611 • Adam Brazier Cornell Center for Astrophysics and Planetary Science and Department of Astronomy, Cornell University, Ithaca, NY 14853 • Paul R. p. 9 Brook Institute for Gravitational Wave Astronomy and School of Physics and Astronomy, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK • Sarah Burke-Spolaor Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506 • Rand Burnette Department of Physics, Oregon State University, Corvallis, OR 97331 • Robin Case Department of Physics, Oregon State University, Corvallis, OR 97331

Appendix B: NANOGrav Collaboration Author List

• J. Andrew Casey-Clyde Department of Physics, University of Connecticut, 196 Auditorium Road, U-3046, Storrs, CT 06269-3046 • Maria Charisi Department of Physics and Astronomy, Washington State University, Pullman, WA 99163 • Shami Chatterjee Cornell Center for Astrophysics and Planetary Science and Department of Astronomy, Cornell University, Ithaca, NY 14853 • Tyler Cohen Department of Physics, New Mexico Institute of Mining and Technology, 801 Leroy Place, Socorro, NM 87801 • James M. p. 9 Cordes Cornell Center for Astrophysics and Planetary Science and Department of Astronomy, Cornell University, Ithaca, NY 14853 • Neil J. p. 9 Cornish Department of Physics, Montana State University, Bozeman, MT 59717 • Fronefield Crawford Department of Physics and Astronomy, Franklin & Marshall College, P. p. 9O. Box 3003, Lancaster, PA 17604 • H. p. 9 Thankful Cromartie Department of Physics and Astronomy, Vanderbilt University, 2301 Vanderbilt Place, Nashville, TN 37235 • Megan E. p. 9 DeCesar Department of Physics and Astronomy, George Mason University, Fairfax, VA 22030 • Paul B. p. 9 Demorest National Radio Astronomy Observatory, 1003 Lopezville Rd. p. 7, Socorro, NM 87801 • Heling Deng Columbia Astrophysics Laboratory, Columbia University, 538 West 120th Street, New York, NY 10027 • Lankeswar Dey Institute of Astrophysics, FORTH, GR-71110, Heraklion, Greece • Timothy Dolch Department of Physics and Astronomy, University of New Mexico, Albuquerque, NM 87131 • Department of Physics Hillsdale College, 33 E. p. 9 College Street, Hillsdale, MI 49242 • Eureka Scientific, 2452 Delmer Street, Suite 100, Oakland, CA 94602-3017 • SETI Institute, 339 N Bernardo Ave Suite 200, Mountain View, CA 94043 • Graham M. Doskoch Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506 • William Fiore Department of Physics and Astronomy, University of British Columbia, 6224 Agricultural Road, Vancouver, BC V6T 1Z1 • Emmanuel Fonseca Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506

Appendix B: NANOGrav Collaboration Author List

• Gabriel E. Freedman NASA Goddard Space Flight Center, Greenbelt, MD 20771 • Emiko C. p. 2 Gardiner Department of Astronomy, University of California, Berkeley, 501 Campbell Hall No. p. 93411, Berkeley, CA 94720 • Kyle A. Gersbach Department of Physics and Astronomy, Vanderbilt University, 2301 Vanderbilt Place, Nashville, TN 37235 • Joseph Glaser Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506 • Deborah C. p. 9 Good Department of Physics and Astronomy, University of Montana, 32 Campus Drive, Missoula, MT 59812 • Kayhan G¨ultekin Department of Astronomy and Astrophysics, University of Michigan, Ann Arbor, MI 48109 • Aiden Gundersen Department of Physics, Montana State University, Bozeman, MT 59717 • C. p. 9 J. Harris Department of Astronomy and Astrophysics, University of Michigan, Ann Arbor, MI 48109 • Jeffrey S. p. 9 Hazboun Department of Physics, Oregon State University, Corvallis, OR 97331 • Ross J. p. 9 Jennings Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506 • Aaron D. p. 9 Johnson Center for Gravitation, Cosmology and Astrophysics, Department of Physics and Astronomy, University of Wisconsin-Milwaukee, P. p. 9O. Box 413, Milwaukee, WI 53201 • David L. p. 9 Kaplan Center for Gravitation, Cosmology and Astrophysics, Department of Physics and Astronomy, University of Wisconsin-Milwaukee, P. p. 9O. Box 413, Milwaukee, WI 53201 • Anala K. p. 9 Sreekumar Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506 • Luke Zoltan Kelley Astrophysics Working Group NANOGrav Collaboration, Berkeley, CA, USA • Joey S. p. 9 Key University of Washington Bothell, 1815 Campus Way NE, Bothell, WA 98011 • Nima Laal Department of Physics and Astronomy, Vanderbilt University, 2301 Vanderbilt Place, Nashville, TN 37235

Appendix B: NANOGrav Collaboration Author List

• Michael T. Lam SETI Institute, 339 N Bernardo Ave Suite 200, Mountain View, CA 94043 • William G. p. 9 Lamb Department of Physics and Astronomy, Vanderbilt University, 2301 Vanderbilt Place, Nashville, TN 37235 • Bjorn Larsen Department of Physics, Yale University, New Haven, CT 06511 • T. p. 9 Joseph W. Lazio Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109 • Natalia Lewandowska Department of Physics and Astronomy, State University of New York at Oswego, Oswego, NY 13126 • Tingting Liu Department of Physics and Astronomy, Georgia State University, 25 Park Place, Suite 605, Atlanta, GA 30303 • Ryan S. p. 9 Lynch Green Bank Observatory, P. p. 6O. Box 2, Green Bank, WV 24944 • Chung-Pei Ma Department of Astronomy, University of California, Berkeley, 501 Campbell Hall No. p. 93411, Berkeley, CA 94720 • David R. Madison Department of Physics, Occidental College, 1600 Campus Road, Los Angeles, CA 90041 • Ashley Martsen Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506 • Cayenne Matt Department of Astronomy and Astrophysics, University of Michigan, Ann Arbor, MI 48109 • James W. p. 9 McKee Department of Physics and Astronomy, Union College, Schenectady, NY 12308 • Maura A. p. 9 McLaughlin Department of Physics and Astronomy, West Virginia University, P. p. 9O. Box 6315, Morgantown, WV 26506 • Patrick M. p. 9 Meyers ETH Zurich, Institute for Particle Physics and Astrophysics, Wolfgang-Pauli-Strasse 27, 8093 Zurich, Switzerland • Matthew T. p. 9 Miles Department of Physics and Astronomy, Vanderbilt University, 2301 Vanderbilt Place, Nashville, TN 37235

Appendix B: NANOGrav Collaboration Author List

• Chiara M. F.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI can perform Bayesian inference on pulsar timing data to set upper limits on continuous spin graviton spin scale rho g, as demonstrated by finding "the 95% credible level limit to be rho g < 1.23 x 10-23 eV."

  2. Improved AI can model and analyze timing residuals using a modified, rho g-dependent, marginalized likelihood which incorporates the spin scale parameter into the noise covariance matrix C.

  3. Improved AI can compute the posterior distribution of continuous spin graviton spin scale rho g by using MCMC sampling techniques such as PTArcade to sample parameter points based on priors defined in Table I.

  4. Improved AI can determine if a non-zero rho g is statistically significant by comparing the 1d marginalized posterior distribution on rho g against the prior limits, noting that the peak at rho g ∼ 10-23 eV marks the region where CSG-induced deviations slightly improve the fit to the NANOGrav 15-year data set, although the effect is not statistically significant.

  5. Improved AI can predict potential future constraints by applying the framework to upcoming datasets, expecting that the longest data set with evidence for a GWB will provide the most stringent limits on rho g.

Sources

Related papers