---
title: 'Pantheon+: Precision SN Ia Cosmology'
url: https://www.emergentmind.com/topics/pantheon
type: topic
---

# Pantheon+: Precision SN Ia Cosmology

Pantheon+ is a Type Ia supernova compilation constructed to serve both as the low-redshift anchor for the SH0ES distance ladder and as a primary probe of cosmic acceleration and the dark-energy equation-of-state parameter \(w\). In its released form it contains 1701 light curves of 1550 spectroscopically confirmed SNe Ia drawn from 18 distinct surveys, spanning from the very nearest events (\(z\approx0.0008\)) to \(z\approx2.3\), and, unlike the original Pantheon compilation, explicitly includes SNe with \(z<0.01\) so that SN systematic covariance can be included in a joint measurement of \(H_0\) and \(w\) [2112.03863, 2202.04077].

## 1. Compilation, scope, and sample definitions

Pantheon+ extends the original Pantheon release in both size and low-redshift leverage. Compared to Pantheon, which contained 1048 SNe over \(0.01<z<2.3\), Pantheon+ increases the sample to 1701 light curves of 1550 distinct SNe Ia over \(0.001\le z\le2.26\), nearly doubles the low-\(z\) sample from \(\approx350\) to \(\approx800\) SNe with \(z<0.1\), adds new DES and Foundation data, and incorporates \(z<0.01\) events for the first time [2112.03863, 2202.04077]. The survey content spans nearby searches such as CfA1–4, LOSS1–2, CSP DR3, SOUSA, Foundation, and CNIa0.02; Hubble-flow samples such as SDSS, PS1MD, SNLS, and DES; and high-redshift HST programs including HDFN, GOODS/PANS, SCP, and CANDELS+CLASH [2112.03863].

The release also formalizes cross-survey redundancy. Pantheon+ uses the large sample to compare properties of 151 SNe Ia observed by multiple surveys and 12 pairs/triplets of “SN siblings” found in the same host galaxy, with additive “intrinsic” covariance between measurements of the same SN by different surveys set equal to \(\sigma_{\rm int}^2\), while covariance between siblings is found to be negligible [2112.03863].

Different downstream analyses apply distinct sample cuts. For example, the missing-covariance study of Bidenko et al. restricts the release to 1590 light-curve measurements of 1473 distinct Type Ia SNe after a low-\(z\) cut of \(z>0.01\), using distance moduli \(\mu_i\equiv m_{b,i}-M\) calibrated via the SH0ES three-rung ladder and the full \(1590\times1590\) covariance supplied by the release [2308.05157].

| Analysis setup | Sample definition | Stated use |
|---|---|---|
| Full Pantheon+ release | 1701 light curves of 1550 distinct SNe Ia, \(0.001\le z\le2.26\) | Joint \(H_0\), \(w\), and SN cosmology [2202.04077] |
| Light-curve release description | 1701 light curves of 1550 unique SNe Ia, \(z\approx0.0008\) to \(z\approx2.3\) | Unified SN + SH0ES framework [2112.03863] |
| \(z>0.01\) subset | 1590 light-curve measurements of 1473 distinct SNe, \(0.01<z<2.26\) | Tests for missing covariance [2308.05157] |

## 2. Light-curve standardization and calibration pipeline

Pantheon+ reprocesses the constituent surveys in a unified photometric and light-curve framework. All 18 surveys are re-calibrated in a single scheme, described as “Supercal + \(\delta zp\)” and also as the “Supercal–Fragilistic” approach, so that zeropoint offsets are tied to the latest CALSPEC network; in particular, the CfA3 natural vs. standard-system definitions and the mis-applied SDSS AB offsets in Pantheon are corrected [2112.03863, 2202.04077]. The light curves are fit with a revised SALT2 model, including SALT2-B22 in the cosmological analysis, retrained with nine realizations of zeropoints and filter shifts, with observed-frame filters restricted to rest-frame wavelengths in \(300\)–\(700\,{\rm nm}\) and epochs \(\pm15\) to \(45\) days from peak [2112.03863, 2202.04077].

Distance estimation follows the Tripp/BBC standardization. In the cosmological-constraints analysis the standardized distance modulus is written as
\[
\mu = m_B + \alpha x_1 - \beta c - M - \delta_{\rm bias}(z,x_1,c,M_\star) + \delta_{\rm host}(M_\star),
\]
with \(m_B=-2.5\log_{10}(x_0)\), \(x_1\) the stretch, \(c\) the color, and bias corrections derived from detailed SNANA simulations tuned to each survey’s cadence, depth, detection, and spectroscopic efficiency [2202.04077]. The light-curve release describes the same structure as
\[
\mu = m_B + \alpha x_1 - \beta c - M - \delta\mu_{\rm bias},
\]
with nominal \(\alpha=0.148\) and \(\beta=3.112\) [2112.03863].

Selection and measurement biases are treated with BEAMS with Bias Corrections (BBC), using per-survey simulations of cadence, depth, and efficiency, while intrinsic scatter is modeled following “It’s Dust” (Brout & Scolnic 2021), in which color-luminosity variations arise primarily from dust rather than exotic SN physics [2112.03863]. For nearby SNe, peculiar-velocity corrections are applied using the 2M++-based flow model of Carrick et al. (2015) and the techniques of Peterson et al. (2021), with a residual velocity uncertainty of \(\sim250\,{\rm km\,s^{-1}}\) folded into the error budget [2112.03863].

## 3. Covariance construction and inference framework

A defining feature of Pantheon+ is its explicit treatment of correlated systematics through an unbinned, per-SN covariance matrix. In the light-curve release the total covariance of Hubble residuals is
\[
C_{\rm tot}=C_{\rm fit}+C_{\rm sys}+{\rm diag}(\sigma_{\rm int}^2),
\]
while in the cosmological-constraints analysis the likelihood is written as
\[
-2\ln\mathcal L=\chi^2=\Delta D^T\,[C_{\rm stat}+C_{\rm syst}]^{-1}\,\Delta D,
\]
with \(\Delta D_i=\mu_i-\mu_{\rm model}(z_i)\) [2112.03863, 2202.04077]. The systematic block is propagated in the Conley et al. form
\[
C_{{\rm sys},ij}=\sum_k \left(\frac{\partial \mu_i}{\partial {\rm syst}_k}\right)\left(\frac{\partial \mu_j}{\partial {\rm syst}_k}\right)\sigma_k^2,
\]
with summaries referring to 40 systematic perturbations in the light-curve release and to \(\sim20\) sources in the cosmological-constraints analysis, including calibration, redshifts, dust, scatter, filter transmission, host-mass step, and SALT2 model errors [2112.03863, 2202.04077].

Sampling in the baseline cosmology analysis is performed with PolyChord within CosmoSIS and verified with CosmoMC and SNANA’s fast grid search [2202.04077]. The same covariance formalism underlies later methodological extensions. Bidenko et al. replace the total covariance by
\[
C_{\rm tot}=C_{\rm stat}+\Sigma_{\rm gp},
\]
and introduce a zero-mean Gaussian process in redshift \(z\) with either a stationary squared-exponential kernel,
\[
\Sigma_{{\rm gp},ij}=\sigma^2\exp\!\left[-\frac{(z_i-z_j)^2}{2\ell^2}\right],
\]
or a Matérn kernel with free \(\nu\), correlation length \(d\), and amplitude \(\sigma\), fitting the parameter vector \(\theta=\{H_0,\Omega_m,M,\sigma,d,\nu\}\) through
\[
\ln L(\theta)=-\tfrac12 \Delta\mu^T C_{\rm tot}^{-1}\Delta\mu-\tfrac12\ln|C_{\rm tot}|
\]
[2308.05157].

The same basic covariance formalism also supports hierarchical and alternative reductions. Lane et al. reconstruct a cosmology-independent covariance directly from the SALT2 parameter triplet \(\tilde\mu=(m_B,x_1,c)\), define \(C=C_{\rm stat}+C_{\rm syst}+C_{\rm dupl}\), and omit FLRW-dependent peculiar-velocity and \(\sigma_{\rm floor}\) contributions, thereby pushing the notion of model independence well beyond marginalization over \(w\)CDM parameters [2311.01438].

## 4. Baseline cosmological constraints

The canonical Pantheon+ cosmology analysis reports a factor of two improvement in cosmological constraining power relative to the original Pantheon sample [2202.04077]. For a flat \(\Lambda\)CDM model, SNe Ia alone yield
\[
\Omega_M = 0.334\pm0.018.
\]
For a flat \(w_0\)CDM model, SNe Ia alone give
\[
w_0=-0.90\pm0.14,
\]
and when Cepheid host distances and covariance are included the joint SN + SH0ES fit gives
\[
H_0=73.5\pm1.1\ {\rm km\,s^{-1}\,Mpc^{-1}}.
\]
For flat \(w_0w_a\)CDM, Pantheon+ alone gives
\[
w_a=-0.1^{+0.9}_{-2.0},
\]
while the SN + SH0ES fit gives
\[
H_0=73.3\pm1.1\ {\rm km\,s^{-1}\,Mpc^{-1}}.
\]
Combining the SN likelihood with CMB and BAO yields
\[
w_0=-0.978^{+0.024}_{-0.031},
\qquad
w_a=-0.65^{+0.28}_{-0.32},
\]
with both \(w_0\) determinations reported as consistent with a cosmological constant [2202.04077].

Pantheon+ is also used as a calibrated late-time anchor in model extensions. In \(f(T)\) cosmology, adding the SH0ES calibration breaks the \(M\)–\(H_0\) degeneracy and yields \(H_0\approx72\pm1\) independent of the specific \(f(T)\) model, while BAO pulls \(H_0\) slightly downward by \(\sim1\,{\rm km\,s^{-1}\,Mpc^{-1}}\) but still \(\sim3\sigma\) above the Planck 2018 \(\Lambda\)CDM value [2303.13840]. In the interacting-dark-energy and Hu–Sawicki \(f(R)\) study of Wang, Pantheon+ alone gives weak constraints, whereas the combination with CMB, BAO, and cosmic chronometers yields \(\epsilon=0.048\pm0.026\) and \(\log_{10}f_{R0}< -6.32\) at \(2\sigma\), together with a quintessence-like \(w(z)>-1\) signal beyond \(2\sigma\) in \(z\in[0.70,1.05]\) for the CMB + CC + Pantheon+ reconstruction [2207.07164].

Within the collaboration analysis, however, the principal conclusion is conservative: systematic uncertainties in the use of SNe Ia along the distance ladder comprise less than one third of the total uncertainty in the measurement of \(H_0\) and cannot explain the present Hubble tension between local measurements and early-Universe predictions from the cosmological model [2202.04077].

## 5. Sensitivity analyses, consistency tests, and internal debates

Pantheon+ has been the subject of unusually detailed sensitivity studies because its formal covariance treatment is central to both \(H_0\) and dark-energy inference. Brownsberger et al. vary the gray photometric zeropoint of each survey through free offsets \(\Delta\mu_S\), constrained by Gaussian priors, and show that the joint SH0ES + Pantheon+ \(H_0\) measurement is robust against inter-survey photometric miscalibration: the additional uncertainty in \(H_0\) is no larger than \(0.2\,{\rm km\,s^{-1}\,Mpc^{-1}}\), and even unbounded survey-offset priors increase \(\sigma(H_0)\) by at most \(0.15\,{\rm km\,s^{-1}\,Mpc^{-1}}\). By contrast, the best-fit values of \(\Omega_M\) and \(w\) slip, to first order, by \(0.04\) and \(-0.17\) per \(25\,{\rm mmag}\) of inter-survey calibration uncertainty, and allowing \(\sigma_{P,S}=25\,{\rm mmag}\) inflates the \(\Omega_M\times w\) contour area by \(\sim50\%\) [2110.03486].

A separate line of criticism concerns whether the released covariance is too large rather than too small. Keeley, Shafieloo, and L’Huillier analyze 1580 Pantheon+-like objects, find a best-fit flat-\(\Lambda\)CDM value \(\chi^2=1387.10\) for 1580 data points, and show that the normalized residuals have observed standard deviation \(\sigma_{\rm resid}\approx0.93\) instead of 1.0, with no evidence for heavy tails or non-Gaussian outliers. They interpret this as a \(\sim7\%\) overestimation of errors on SN distance moduli, note that subtracting \(\delta^2=0.002\) from each diagonal element or equivalently scaling \(C\rightarrow0.93^2C\) restores \(\chi^2\approx1580\), and conclude that flat \(\Lambda\)CDM remains consistent with the corrected data [2212.07917].

Bidenko et al. test the opposite possibility: unmodeled redshift-correlated covariance not already present in the release. Their GP-marginalized analysis finds no statistically significant evidence for extra \(\Sigma_{\rm gp}\), constrains the GP amplitude to
\[
\sigma<0.031\ {\rm mag}\qquad(95\%\,{\rm C.L.}),
\]
and reports that the baseline Pantheon+ values
\[
\Omega_m=0.332\pm0.018,\qquad H_0=73.55\pm1.02\,{\rm km\,s^{-1}\,Mpc^{-1}}
\]
shift only to
\[
\Omega_m=0.333\pm0.021,\qquad H_0=73.21\pm1.17\,{\rm km\,s^{-1}\,Mpc^{-1}}
\]
after GP marginalization, with the Hubble-tension significance reduced from \(5.3\sigma\) to \(4.5\sigma\) in the most extreme case they consider [2308.05157]. This distinction suggests that the “over-estimated covariance” and “missing covariance” arguments probe different failure modes of the likelihood.

Another internal debate concerns the homogeneity of the standardized SN Ia absolute magnitude. Perivolaropoulos and Skara replace the single-\(M\) likelihood with a piecewise \((M_<,M_>,d_{\rm crit})\) model and obtain
\[
M_< = -19.392 \pm0.050,\quad
M_> = -19.205 \pm0.030,\quad
d_{\rm crit}=19.95\pm0.10\,{\rm Mpc},
\]
with \(\Delta\chi^2=-19.6\) relative to the single-\(M\) baseline and a transition significance exceeding \(3\sigma\). After removing the \(z<0.01\) Hubble-flow points, however, the remaining signal drops to \(\sim2\sigma\), which they interpret as evidence that the fit improvement mixes the known volumetric redshift scatter bias at \(z<0.01\) with a milder possible luminosity transition near \(20\,{\rm Mpc}\) [2301.01024].

## 6. Extensions, local structure, and contested phenomenology

Pantheon+ has become a general-purpose late-universe laboratory rather than only a dark-energy Hubble diagram. At very low redshift, Wang’s generalized Hubble-law test,
\[
(cz)^\alpha = H_0 D_L(z),
\]
finds
\[
\alpha=1.00118\pm0.00044
\]
for the \(z<0.03\) subset of 468 SNe Ia, thereby confirming the validity of the linear Hubble law with \(0.04\%\) precision and identifying a transition redshift \(z_t=0.03\) and luminosity distance \(D_{L,t}=123.13\pm1.75\,{\rm Mpc}\) beyond which higher-order cosmographic terms must be included [2208.07271].

Tomographic analyses have reached less uniform conclusions. Wang’s 10-bin and equal-number slicing of Pantheon+ finds no obvious evidence of evolution of \(H_0\) and \(\Omega_m\) at the \(2\sigma\) confidence level, and emphasizes that the SH0ES calibration significantly compresses the allowed parameter space in each bin [2207.10927]. By contrast, Dainotti et al. report a high-\(z\) split at \(z>1\) with best-fit
\[
H_0^{\rm (high)}=34.37\,{\rm km\,s^{-1}\,Mpc^{-1}},\qquad
\Omega_m^{\rm (high)}=3.391,
\]
together with a profile-distribution exclusion of the full-sample \((H_0,\Omega_m)\) at \(>95\%\) confidence, although their mock-catalog significance for the joint \(H_0\downarrow/\Omega_m\uparrow\) trend remains in the \(1\)–\(2\sigma\) range and the AIC comparison marginally favors vanilla \(\Lambda\)CDM [2301.12725]. A later semi-parametric reconstruction of \(w(z)\) from Pantheon+SH0ES, based on analytic ansätze for the dimensionless comoving distance \(D(z)\) and its derivatives, finds results consistent with \(w=-1\) [2606.25315].

The isotropy literature is similarly mixed. Two dipole-modulated \(\Lambda\)CDM studies conclude that the full Pantheon+ sample is consistent with a null dipole, while a low-redshift subsample shows a stable \(\sim10^{-3}\) dipole at roughly \(2\sigma\): for \(z_c=0.1\), Lin et al. find \(D=1.0^{+0.4}_{-0.4}\times10^{-3}\) toward \((l,b)=(334.5^\circ,16.0^\circ)\), about \(65^\circ\) from the CMB dipole, and for the low-\(z\) subsample Zhao et al. find \(A_D=0.952^{+0.454}_{-0.403}\times10^{-3}\) toward \((149.77^\circ,-12.20^\circ)\), with the signal traced mainly to surveys 5, 56, 63, and 150 or to the highly inhomogeneous SNLS footprint rather than to intrinsic cosmic anisotropy [2309.11320, 2606.22302]. In contrast, Sah et al. employ maximum-likelihood dipole fits in the redshift shell \(0.023<z<0.15\) and report a Hubble-rate dipole exceeding \(1.5\,{\rm km\,s^{-1}\,Mpc^{-1}}\) in all frames, together with a deceleration dipole at \(>5\sigma\), which they interpret through a tilted-flow picture rather than a cosmological-constant signal [2411.10838].

Cosmographic analyses reinforce the view that Pantheon+ is sensitive to local structure. Sorrenti et al. fit a third-order luminosity-distance expansion with a dipole and a monopole redshift correction, find a significant local infall attributed to an overdensity out to \(R\simeq120\,h^{-1}\,{\rm Mpc}\), and report that adding the monopole lowers \(\chi^2\) by \(\Delta\chi^2\simeq7.3\) while yielding \(v_{\rm bulk}\simeq300\,{\rm km\,s^{-1}}\) and \(v_{\rm in}(R)=120\pm45\,{\rm km\,s^{-1}}\) in the robust determination that uses all SNe to \(z\le0.8\) and applies infall only for \(z\le0.04\) [2407.07002]. A related multipole analysis of the Pantheon+SH0ES data finds, in addition to a dipole-only bulk flow \(|v^{\rm (bulk)}|=318\pm40\,{\rm km\,s^{-1}}\), a monopole \(\gamma=-106\pm40\,{\rm km\,s^{-1}}\) and a quadrupole amplitude \(\lambda=190\pm40\,{\rm km\,s^{-1}}\), with the monopole significant only at very low redshift and the quadrupole increasing with redshift [2403.17741].

Alternative-cosmology reanalyses have also used Pantheon+ to probe the meaning of “model independence.” Lane et al. construct a covariance matrix directly from \((m_B,x_1,c)\), introduce an empirical SN-based scale of statistical homogeneity \(z_{{\rm SHS}_\alpha}=0.075^{+0.007}_{-0.009}\), and report that the Bayes factor between timescape and \(\Lambda\)CDM varies strongly with the low-\(z\) cut: \(\ln B\approx2.9\) in favor of timescape as \(z_{\min}\to0\), \(\ln B\approx-2.8\) near \(z_{\min}\approx0.04\), and \(\ln B\approx0\) for \(z_{\min}>{\rm SHS}_\alpha\) [2311.01438]. Whatever the preferred interpretation, a consistent conclusion across these extensions is that Pantheon+ is simultaneously a precision supernova Hubble diagram and a sensitive diagnostic of low-redshift calibration, covariance construction, survey geometry, and local velocity structure.

Source: https://www.emergentmind.com/topics/pantheon