Dense-Matter Uncertainty and Inference Domains

Dense-matter constraints come from theories, nuclei, pulsar timing, X-ray pulse profiles, and gravitational waves, but none measures pressure at one sharply defined density. Each datum is filtered through a model, calibration, selection, and prior, and each responds to a broad density interval. A sound joint inference preserves those domains, propagates correlated systematics, and tests how conclusions change when one evidence family or equation-of-state prior is removed. The dataset inventory below is current through 2026-08-18.

Required background. Hilbert Stress–Energy from Metric Variation supplies the stress tensor encoded by an equation of state.

Helpful background. The Einstein Equation maps pressure and energy density to stellar observables; Relativistic Matter and Compact Objects gives the inference context; and Computational Relativity supplies forward models and likelihood calculations.

Density domains are response ranges

Nuclear masses, charge radii, dipole polarizability, and weak form factors mainly constrain finite nuclei and sub- to near-saturation properties. Their mapping to uniform neutron-rich matter depends on an energy-density functional and correlations among symmetry coefficients. The CREX weak-form-factor result is a calibrated electroweak observable of calcium, not a direct measurement of $L$; the model translation is described alongside its systematic budget in CREX Collaboration 2022, pp. 042501-4–042501-5, Tables II–III, and Figs. 2–4.

Chiral effective field theory supplies a theory ensemble for nucleonic matter near saturation and can be extended toward roughly twice saturation with growing truncation, regulator, and many-body uncertainty. It does not directly constrain exotic high-density phases. Drischler, Holt, and Wellenhofer delimit that model domain and uncertainty organization in Drischler, Holt, and Wellenhofer 2021, §§ 2–4.

Stellar observables are integrals through an equilibrium configuration. Radii and tidal deformabilities often respond most strongly to pressure across an intermediate-density band, while a maximum mass reaches the highest central densities, but the response shifts with stellar mass and the assumed equation-of-state family. Quoting one density for a radius is therefore a model-dependent summary, not a local measurement. The 2024 Living Reviews synthesis compares laboratory, timing, NICER, and gravitational-wave domains in Kumar et al. 2024, §§7–8 and Tables 9–11.

Versioned observational inputs

Radio timing. The PSR J0740+6620 Shapiro-delay analysis gives a high-mass anchor using a named timing baseline, telescope data, ephemeris, noise model, and relativistic binary likelihood. Fonseca et al. 2021, Results and Tables 1–2 reports the refined mass and geometry. Calibration includes time standards, dispersion and propagation corrections, backend offsets, and pulse-noise treatment. Targeted timing pulsars are not a complete population sample; a maximum-mass population inference needs a selection function and mass-distribution model.

NICER pulse profiles. The current page uses the PSR J0740+6620 dataset accumulated from 2018-09-21 through 2022-04-21, analyzed jointly with XMM–Newton in the 2024 release. The likelihood uses phase- and energy-resolved counts, instrument responses, backgrounds, a radio mass prior, atmosphere emission, relativistic ray tracing, and hotspot geometry. Salmi et al. 2024, §§ 2–4 and Tables 1–3 documents the event selection, calibration choices, priors, sampler tests, and independent-analysis comparison. Systematics include NICER–XMM cross-calibration, background/source separation, atmosphere composition, spot parameterization, distance, and sampling of multimodal posteriors.

Gravitational waves. For a reproducible baseline, use the machine-readable GWOSC GW170817-v1 event release, DOI 10.7935/K5B8566F, with H1, L1, and V1 strain, data-quality segments, calibration information, and released parameter-estimation samples Gravitational Wave Open Science Center 2017, GW170817-v1 data release, strain files, and parameter-estimation samples. The likelihood couples masses, spins, tidal deformabilities, waveform calibration, and detector noise. Priors on spin, mass ratio, a common equation of state, waveform family, and high-frequency cutoff affect the tidal posterior. Population use of a gravitational-wave catalog additionally requires the detection selection function; a single-event reanalysis must still state data-quality selections and calibration uncertainty.

Nuclear theory and laboratory records. Each interaction family, regulator, many-body order, and covariance model is a distinct theory version. Laboratory inputs require published covariance and calibration rather than only central values. PREX and CREX probe different nuclei and can prefer different functional correlations; treating them as duplicate measurements of one scalar $L$ conceals model discrepancy.

Joint likelihoods and hidden priors

Let $\theta$ parameterize an equation-of-state ensemble and $\phi_k$ the nuisance variables for evidence family $k$. A schematic posterior is

$$p(\theta,\{\phi_k\}\mid\{D_k\}) \propto p(\theta)\prod_k p(\phi_k\mid\theta) \mathcal L_k(D_k\mid\theta,\phi_k),$$

provided the datasets are conditionally independent given their parameters and the nuisance prior factorizes as written. A shared calibration, source mass, or population parameter must instead appear once in a joint model. A detected population also requires the corresponding selection normalization. The forward map sends each EOS through stellar structure to the parameterized pairs $(M,R)$ and $(M,\Lambda)$, then through timing, pulse-profile, or waveform models to measured data. Multiple stable branches can make $R(M)$ or $\Lambda(M)$ multivalued, requiring a branch or formation model rather than silently choosing one solution.

The prior over functions is scientifically consequential. Piecewise polytropes, spectral expansions, sound-speed models, Gaussian processes, and catalogs of microscopic candidates cover phase transitions and high-density stiffness differently. Enforcing stability, causality, a crust match, and a lower maximum mass changes the induced prior on radii and tidal deformabilities. Posterior agreement under one parameterization is not robustness across function spaces.

Published posterior samples cannot generally be multiplied directly. To use a posterior density as a likelihood, divide out its analysis prior on the same variables and within that prior’s support; marginalization has already averaged hidden nuisance parameters with their original priors. Those priors cannot then be changed using only the marginalized posterior. A flat prior in one coordinate also need not be flat in another. For example, $p(\log x)=\mathrm{constant}$ induces $p(x)\propto1/x$ for $x\gt0$. Ignoring the Jacobian would silently change the EOS prior. A prominent double counting error uses the radio mass measurement as the NICER mass prior and then multiplies the same radio likelihood again. Another treats gravitational and electromagnetic products from GW170817 as independent despite shared binary parameters and simulation calibrations.

Uncertainty budget and discrimination

Separate uncertainty into at least five classes: nuclear interaction and many-body truncation; equation-of-state interpolation and high-density extension; experimental calibration and functional mapping; observational instrument and source modeling; and population selection. Statistical credible intervals condition on all of these model choices. Model discrepancy should not be hidden by widening one convenient Gaussian nuisance parameter.

The current evidence supports pressure sufficient for reliably measured approximately two-solar-mass pulsars, disfavors some extremely soft or stiff model families, and narrows intermediate-density behavior when datasets are combined. It does not uniquely identify core composition or establish a first-order transition. The two independent 2024 NICER analyses of the same expanded J0740 dataset are broadly compatible but differ in prior and sampler details; GW170817 remains sensitive to waveform and spin assumptions; laboratory-to-star mapping remains functional dependent.

A discriminating analysis should report which data move which density-dependent quantities. Useful summaries include prior-to-posterior information gain by density, correlation of each observable with $p(\varepsilon)$, and posterior predictive checks in the original data space. A narrow radius interval produced primarily by a restrictive function prior is not equivalent to one produced by a sharply informative likelihood.

Reproduction and leave-one-family-out checks

A machine-readable reproduction bundle should pin every dataset version, URL and checksum; calibration or response files; event and energy selections; timing ephemeris; GW strain and power-spectral-density choices; waveform and atmosphere software versions; equation-of-state prior generator; random seeds; likelihood code; and posterior samples. Where a collaboration has not released sufficient likelihood products, that boundary must be stated rather than replaced by a digitized contour.

The core robustness matrix refits the same equation-of-state prior with: all evidence; theory removed; laboratory data removed; radio masses removed; NICER removed; gravitational waves removed; and each high-leverage source removed individually. Repeat with at least a phase-transition-capable and a smooth equation-of-state prior. Compare pressure bands, radii, maximum mass, marginal likelihood, and prior-to-posterior information. This page specifies that analysis but does not execute it or claim new inference results.

Review is triggered by a new public catalog or calibration, revised NICER response or hotspot analysis, a new securely measured high mass, a new binary-neutron-star tidal likelihood, updated chiral interactions, or a published correction or withdrawal. The evidence cutoff must advance only after the versioned inputs and their correlations are rechecked.

Exercises

A joint analysis uses the J0740 radio mass posterior as the NICER mass prior, then multiplies the radio mass posterior once more with the NICER posterior. What is wrong, and how should it be repaired?

Solution

The radio information enters twice. Recover a NICER likelihood by dividing the released joint posterior by every prior actually used, including the radio mass prior, while retaining mass–radius and nuisance correlations; then multiply by the radio likelihood once. Alternatively, reuse the released NICER-plus-radio posterior once under its original prior assumptions, without adding the radio data again. If the necessary prior density or correlations for a desired reweighting are unavailable, that separation cannot be reconstructed from a contour plot.

References

• CREX Collaboration. “Precision Determination of the Neutral Weak Form Factor of Calcium-48.” Physical Review Letters 129, 042501 (2022). DOI.
• Drischler, Christian, Jeremy W. Holt, and Christian Wellenhofer. “Chiral Effective Field Theory and the High-Density Nuclear Equation of State.” Annual Review of Nuclear and Particle Science 71, 403–432 (2021). Open preprint.
• Fonseca, Emmanuel, et al. “Refined Mass and Geometric Measurements of the High-Mass PSR J0740+6620.” The Astrophysical Journal Letters 915, L12 (2021). DOI.
• Gravitational Wave Open Science Center. “GW170817 Event Data Release,” Version 1 (2017). Dataset DOI.
• Kumar, Rajesh, et al., and the MUSES Collaboration. “Theoretical and Experimental Constraints for the Equation of State of Dense and Hot Matter.” Living Reviews in Relativity 27, 3 (2024). DOI.
• Salmi, Tuomo, et al. “The Radius of the High-Mass Pulsar PSR J0740+6620 with 3.6 yr of NICER Data.” The Astrophysical Journal 974, 294 (2024). DOI.