Type Ia Supernova Distances at Redshift >1.5 from the Hubble Space Telescope Multi-cycle Treasury Programs: The Early Expansion Rate

, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , and

Published 2018 January 31 © 2018. The American Astronomical Society. All rights reserved.
, , Citation Adam G. Riess et al 2018 ApJ 853 126 DOI 10.3847/1538-4357/aaa5a9

Download Article PDF
DownloadArticle ePub

You need an eReader or compatible software to experience the benefits of the ePub3 file format.

0004-637X/853/2/126

Abstract

We present an analysis of 15 Type Ia supernovae (SNe Ia) at redshift $z\gt 1$ (9 at $1.5\lt z\lt 2.3$) recently discovered in the CANDELS and CLASH Multi-Cycle Treasury programs using WFC3 on the Hubble Space Telescope. We combine these SNe Ia with a new compilation of ∼1050 SNe Ia, jointly calibrated and corrected for simulated survey biases to produce accurate distance measurements. We present unbiased constraints on the expansion rate at six redshifts in the range $0.07\lt z\lt 1.5$ based only on this combined SN Ia sample. The added leverage of our new sample at $z\gt 1.5$ leads to a factor of ∼3 improvement in the determination of the expansion rate at z = 1.5, reducing its uncertainty to ∼20%, a measurement of $H(z=1.5)/{H}_{0}\,=\,{2.69}_{-0.52}^{+0.86}$. We then demonstrate that these six derived expansion rate measurements alone provide a nearly identical characterization of dark energy as the full SN sample, making them an efficient compression of the SN Ia data. The new sample of SNe Ia at $z\gt 1.5$ usefully distinguishes between alternative cosmological models and unmodeled evolution of the SN Ia distance indicators, placing empirical limits on the latter. Finally, employing a realistic simulation of a potential Wide-Field Infrared Survey Telescope SN survey observing strategy, we forecast optimistic future constraints on the expansion rate from SNe Ia.

Export citation and abstract BibTeX RIS

1. Introduction

Type Ia supernovae (SNe Ia) at redshift $z\gt 1$ offer unique leverage on investigations relating to the nature of their progenitors, their accuracy as distance indicators, and the parameters of the cosmological model. Unfortunately, ground-based facilities are extremely challenged to produce reliable discoveries of SNe Ia at $z\gt 1$, a task demanding significant and repeatable detections and robust classifications at $I\sim 26\,\mathrm{mag}$.

Thus, for the past two decades, the Hubble Space Telescope (HST) has offered the best perch from which to harvest these objects, with the rate of collection limited only by its relatively modest field of view. The first robust, multi-object sample of SNe Ia at $z\gt 1$ came from searching the GOODS fields with the HST Advanced Camera for Surveys (ACS) and its z-band filter, with crucial near-infrared follow-up observations of the rest-frame optical light obtained using NICMOS and confirming spectroscopy from the ACS grism. The first sample of 7 SNe Ia at $z\gt 1.25$ provided a crucial check that dimming from astrophysical effects was not mimicking cosmic acceleration (Riess et al. 2004). A follow-up program increased the sample of reliable SNe Ia at $z\gt 1$ to 18 (Riess et al. 2007) followed by another 12 from targeting cluster fields (Barbary et al. 2012; Suzuki et al. 2012). This sample of ∼30 successfully extended the SN Ia measurement of expansion to the matter-dominated era to break degeneracies between dark energy and dark matter.

Still, clues available only at $z\gt 1.5$ beckoned. Owing to the red-limit of HST CCDs and the roughly Gyr delay between progenitor formation and SN Ia explosion (Rodney et al. 2014), only ∼3 moderately constrained SNe Ia at $z\gt 1.5$ were previously discovered with HST: SN 1997ff at z = 1.755, SN 2003ak at z = 1.551, and SCP0401 at z = 1.713 (Gilliland et al. 1999; Riess et al. 2001, 2004; Rubin et al. 2013). An effective program to find SNe Ia at $z\gt 1.5$ required WFC3-IR, the first wide-area (greater than an arcminute) infrared HgCdTe detector on HST, installed in 2009, which extended the red cutoff to 1.6 μm. Two of the initial three Multi-Cycle Treasury (MCT) programs with WFC3, CANDELS (PI: Faber and Ferguson, Grogin et al. 2011; Koekemoer et al. 2011) and CLASH (PI: Postman, Postman et al. 2012) were selected to enable the discovery of SNe Ia at $z\gt 1.5$ with an additional program of coordinated SN follow-up observations (PI: Riess, Graur et al. 2014; Rodney et al. 2014). These MCT programs were three-year extragalactic imaging campaigns initiated in HST Cycle 18, beginning 2010 October. Both MCT programs employed ACS and WFC3-IR with cadences of ∼50 days between epochs, chosen to match the risetime of SNe Ia time-dilated to $1.5\lt z\lt 2.0$. (Rodney et al. 2014, hereafter R14) comprehensively described the SN search component of the CANDELS program and measured the volumetric SN Ia rate from the complete CANDELS sample of 65 SNe out to z = 2.5. (Graur et al. 2014, hereafter G14) presented the SN Ia rates analysis from the CLASH program, using a sample of 27 SNe detected in the HST parallel fields (∼6' from the galaxy clusters that make up the primary targets for CLASH). For full details of the survey design and observations, we refer the reader to R14 and G14.

These programs together identified 15 SNe Ia at $z\gt 1$, 9 of which (7 at $z\gt 1.5$) are sufficiently well-measured to derive reliable distance estimates. Detailed studies of the first two such events were presented by Rodney et al. (2012) and Jones et al. (2013), and a novel approach to SN classification via medium-band infrared imaging was presented for two others by Rodney et al. (2015).

Here for the first time we derive a set of distance estimates for this sample calibrated for a joint cosmological analysis with a compilation of SNe Ia from previous surveys (Scolnic et al. 2017). The most significant augmentation of the extant SN Ia sample is the set of SN Ia distances presented here at $z\gt 1.5$, which usefully extends the SN-based determination of the expansion rate of the universe to a higher redshift, $z\approx 1.5$, than previously possible. In Section 2, we present details of the SN sample, and in Section 3, we present constraints on the scale-free expansion history and carry out some related investigations. We summarize our conclusions in Section 4.

2. SN Ia Sample

From the total set of 92 CANDELS and CLASH SNe, we have identified 15 as likely SNe Ia at $z\gt 1$ with sufficient confidence for use as distance indicators. We present the coordinates of these objects in Table 1, their redshifts and classifications in Table 2, the properties of their host galaxies in Table 3, and their distance-related parameters in Table 4. The light-curve photometry is given in the Appendix. For inclusion in this subset, we require at least enough samplings of the light and color curves to exceed the number of free parameters in the light-curve fit. This effectively means that we require a minimum of four independent observation epochs, providing at least a modicum of constraint on the light-curve shape. We also require that the first epoch with $\gt 3\sigma $ detection must be no more than 10 days after the peak of the light curve in the rest-frame B band, consistent with the requirements used by Riess et al. (1996, 2007). Finally, we require that at least one of the epochs includes WFC3-IR observations in both the F125W and F160W bands, which provide a measurement of the SN color at rest-frame optical wavelengths for $1\lt z\lt 2.5$.

Table 1.  SNe Ia from CANDELS + CLASH at $z\gt 1$

SN ID Nickname Survey Field α(J2000) δ(J2000)
CLA10Cal Caligula CLASH Abell 383 IR par 02:48:25.74 −03:33:08.8
CLF11Ves Vespasian CLASH MACS2129 ACS par 21:29:42.60 −07:41:47.7
CLH11Tra Trajan CLASH MS2137 ACS par 21:39:46.05 −23:38:34.8
CLP12Get Geta CLASH RXJ2129 IR par 21:29:23.89 +00:08:24.8
COS12Car Carter CANDELS COSMOS 10:00:14.72 +02:11:32.6
EGS11Oba Obama CANDELS EGS 14:20:32.66 +53:02:48.2
EGS13Rut Rutledge CANDELS EGS 14:20:48.11 +53:04:22.1
GND12Col Colfax CANDELS GOODS-N Deep 12:36:37.58 +62:18:33.1
GND13Cam Camille CANDELS GOODS-N Deep 12:37:07.37 +62:10:26.9
GND13Gar Garner CANDELS GOODS-N Deep 12:36:40.81 +62:11:14.2
GND13Jay Jay CANDELS GOODS-N Deep 12:36:41.38 +62:11:30.1
GND13Sto Stone CANDELS GOODS-N Deep 12:37:16.77 +62:16:41.4
GSD10Pri Primo CANDELS GOODS-S Deep 03:32:38.01 −27:46:39.1
GSD11Was Washington CANDELS GOODS-S Deep 03:32:20.85 −27:49:41.5
UDS10Wil Wilson CANDELS UDS 02:17:46.33 −05:15:24.0

Download table as:  ASCIITypeset image

Table 2.  Final Redshifts and Classifications

SN ID Redshifta Redshift Sourceb P(Ia)c Supporting Evidenced Confidencee Primary Referencef
CLA10Cal 1.800 ± 0.1 phot-z 0.95 bronze Graur et al. (2014)
CLF11Ves 1.206 ± 0.007 spec-z (HST+G800L) >0.99 spec, early-type host gold Graur et al. (2014)
CLH11Tra 1.520 ± 0.04g phot-z >0.99 early-type host gold Graur et al. (2014)
CLP12Get 1.700 ± 0.04 phot-z >0.99 early-type host gold Graur et al. (2014)
COS12Car 1.540 ± 0.04 SN spec-z (HST+G141) >0.99 spec gold Rodney et al. (2014)
EGS11Oba 1.409 ± 0.002 spec-z (Keck+LRIS,DEIMOS) 0.9 bronze Rodney et al. (2014)
EGS13Rut 1.614 ± 0.005 spec-z (HST+G141, single line) >0.99 silver Rodney et al. (2014)
GND12Col ${2.260}_{-0.10}^{+0.02}$ phot-z >0.99 med. band gold Rodney et al. (2015)
GND13Cam 1.222 ± 0.002 spec-z (AGHAST, HST+G141) >0.99 silver Rodney et al. (2014)
GND13Gar 1.070 ± 0.02 SN spec-z (HST+G800L) >0.99 spec gold Rodney et al. (2014)
GND13Jay 1.030 ± 0.01 spec-z (AGHAST, HST+G141) >0.99 silver Rodney et al. (2014)
GND13Sto 1.800 ± 0.02 spec-z >0.99 med. band gold Rodney et al. (2015)
GSD10Pri 1.550 ± 0.0001 spec-z >0.99 spec gold Rodney et al. (2012)
GSD11Was 1.330 ± 0.02 spec-z (HST+G141) >0.99 spec gold Rodney et al. (2014)
UDS10Wil 1.914 ± 0.001 spec-z >0.99 spec gold Jones et al. (2013)

Notes.

aFinal composite redshift, incorporating all evidence from SN and host. bAll phot-z and spec-z redshifts are principally constrained by the host galaxy, except where a SN spec-z is noted. cClassification probability from the SN light curve, including host redshift priors, using STARDUST (R14). dAdditional factors influencing the classification confidence. "spec": SN spectrum; "med. band": pseudocolors from medium-band infrared imaging; "early-type host": host galaxy is identified as an early-type galaxy, unlikely to host core-collapse SNe. eConfidence in the Type Ia SN classification. fPrimary reference for further information on discovery, redshift, and classification. gRevised from Graur et al. (2014).

Download table as:  ASCIITypeset image

Table 3.  SN Host Galaxy Data

SN ID Host α(J2000) Host δ(J2000) Host Redshifta Morphology Star Formation Redshift Source
CLA10Cal 02:48:25.74 −03:33:08.8 1.8 ± 0.1 spheroid/disk active phot-z
CLF11Ves 21:29:42.62 −07:41:47.5 1.206 ± 0.007 spheroid passive HST+ACS
CLH11Tra 21:39:46.04 −23:38:34.6 1.52 ± 0.04 spheroid passive phot-z
CLP12Get 21:29:23.92 +00:08:23.8 1.70 ± 0.04 spheroid passive phot-z
COS12Carb 10:00:14.72 +02:11:32.6 undetected undetected
EGS11Oba 14:20:32.67 +53:02:48.1 1.409 ± 0.002 disk/irregular active Keck+LRIS
EGS13Rut 14:20:48.11 +53:04:22.1 1.614 ± 0.005 disk active HST+WFC3
GND12Col 12:36:37.51 +62:18:32.6 ${2.260}_{-0.10}^{+0.02}$ spheroid active phot-z
GND13Cam 12:37:07.38 +62:10:27.2 1.222 ± 0.002 spheroid/disk starburst HST+WFC3
GND13Gar 12:36:40.80 +62:11:14.6 1.86 ± 0.77 undefined starburst phot-z
GND13Jay 12:36:41.37 +62:11:29.5 1.03 ± 0.01 disk active HST+WFC3
GND13Sto 02:37:16.59 +62:16:43.4 1.80 ± 0.02 undefined active phot-z
GSD10Pri 03:32:37.99 −27:46:38.7 1.550 ± 0.0001 irregular starburst VLT+X-Shooter
GSD11Was 03:32:20.86 −27:49:41.5 1.042 ± 0.23 disk starburst HST+WFC3
UDS10Wil 02:17:46.33 −05:15:23.9 1.914 ± 0.001 spheroid starburst VLT+X-Shooter

Notes.

aPhotometric redshifts are marked as "phot-z" and spectroscopic redshifts are labeled with the observatory and instrument employed. bNo plausible host galaxy was identified for SN COS12Car. The coordinates given are for the SN itself.

Download table as:  ASCIITypeset image

Table 4.  SALT2 Light-curve Fit Parameters

SN ID mB x1 c ${{\rm{\Delta }}}_{B}$ μ(mag) Notes
CLA10Cal poor light-curve fit
CLF11Ves 25.38 (0.091) −1.24 (0.60) −0.288 (0.101) +0.27 25.73 (0.34)  
CLH11Tra 25.30 (0.095) −3.35 (2.10) −0.272 (0.090) fails x1 cut (${x}_{1}\lt -3$, ${\sigma }_{{x}_{1}}\gt 1$)
CLP12Get 25.73 (0.088) +1.01 (0.95) −0.139 (0.098) +0.18 26.06 (0.28)  
COS12Car 26.14 (0.122) +2.35 (0.83) +0.152 (0.083) +0.07 25.91 (0.21)  
EGS11Oba poor light-curve fit
EGS13Rut 25.92 (0.071) +0.98 (1.08) +0.055 (0.046) −0.07 25.93 (0.20)  
GND12Col 26.81 (0.056) +0.02 (0.91) +0.128 (0.133) −0.50 26.88 (0.25)  
GND13Cam 25.91 (0.061) −1.35 (0.48) −0.083 (0.043) Hubble diagram outlier ($\gt 4\sigma $)
GND13Gar 25.42 (0.259) +0.02 (0.99) +0.310 (0.179) fails color cut ($c\gt 0.3$)
GND13Jay 24.56 (0.672) −2.04 (0.92) −0.373 (0.447) fails color cut ($c\lt -0.3$)
GND13Sto 26.15 (0.074) −0.48 (0.70) +0.000 (0.071) −0.17 26.20 (0.19)  
GSD10Pri 25.76 (0.089) −0.51 (0.41) −0.186 (0.078) +0.16 26.01 (0.19)  
GSD11Was 25.32 (0.057) +1.04 (0.67) −0.089 (0.039) +0.09 25.60 (0.15)  
UDS10Wil 26.28 (0.172) −1.64 (0.76) +0.082 (0.152) −0.43 26.15 (0.26)  

Download table as:  ASCIITypeset image

For the cosmological analysis presented here, we further subdivide this sample into three confidence categories: gold, silver, and bronze, following the convention of Strolger et al. (2004), Riess et al. (2004, 2007). The gold sample comprises those SNe with compelling classifications as Type Ia, while the silver label indicates a "very likely" Type Ia classification, and the bronze objects are those that are probably Type Ia, but have some nonnegligible probability of misclassification.

As detailed by R14 and G14, the classifications of these SNe at $z\gt 1$ sometimes rely on photometric evidence. Spectra are available for 6 of the 15 SNe Ia at $z\gt 1$ (3 at $z\gt 1.5$), while two others use medium bands to measure the strength of SN Ia spectral features. This mixture of classification methods is necessitated by the difficulty of achieving a purely spectroscopic classification for such high-redshift SNe (see, e.g., Frederiksen et al. 2012; Rodney et al. 2012; Jones et al. 2013; Rubin et al. 2013). Photometric classification of these SNe was performed using STARDUST,24 a Bayesian algorithm employing a comparison of multi-band light curves against 43 template-based models representing Type Ia and core-collapse SNe (R14). For inclusion in the gold and silver samples, we require a Type Ia classification probability $P(\mathrm{Ia})\gt 0.99;$ the two objects with $0.9\lt P(\mathrm{Ia})\lt 0.99$ were relegated to the bronze sample.

The gold objects are further distinguished by having at least one piece of corroborating evidence to support the Type Ia classification. For 6 objects, we have a spectroscopic observation that is well-matched by a SN Ia spectral template, presented by R14 and G14. Two more SNe have medium-band infrared imaging that provides evidence for Type Ia spectral features in medium-band minus broad-band pseudocolors (Rodney et al. 2015). Finally, three of the gold sample SNe have a host galaxy that is classified as "early type" based on morphology and colors, indicating an old stellar population that would be unlikely to host a core-collapse SN (Riess et al. 2001).

We discard the 2 bronze SNe whose classification is too uncertain and proceed with the analysis of the remaining 13 gold and silver SNe at $z\gt 1$ (8 at $z\gt 1.5$) from the CANDELS and CLASH programs. Assuming the 3 silver SNe in the sample are Type Ia with 99% confidence, there is a ∼97% chance that all of the SNe in the cosmological analysis are Type Ia. We combine this set (hereafter, the MCT set) with a uniformly calibrated compilation of ∼1050 spectroscopically classified SNe Ia, the Pantheon compilation (Scolnic et al. 2017). This compilation includes SNe from the Harvard-Smithsonian Center for Astrophysics SN surveys (CfA, Hicken et al. 2009), the Carnegie Supernova Project (CSP, Stritzinger et al. 2011), the Sloan Digital Sky Survey (SDSS, Kessler et al. 2009), the Pan-STARRS1 Medium-Deep Survey (PS1, Rest et al. 2014), and the Canada–France–Hawaii Telescope Supernova Legacy Survey (SNLS, Conley et al. 2011). The compilation includes all SNe from the Rest et al. (2014) sample and from the samples included in the joint light-curve analysis (JLA; Betoule et al. 2014), all uniformly calibrated as presented in the Supercal analysis (Scolnic et al. 2015). The Pantheon compilation also includes 12 equivalently high-confidence SNe Ia at $1\lt z\lt 1.4$ from past HST SN surveys (see Table 5), 9 from Riess et al. (2004) and Riess et al. (2007) and 3 from Suzuki et al. (2012), that meet the criteria given in Scolnic et al. (2017).

Table 5.  SNe Ia at $z\gt 1$ from Other Surveys

SN ID Nickname Surveya Confidenceb Redshift References
1997ff 1997ff HDFN Gold 1.755 Riess et al. (2001)
2002fw Aphrodite Higher-z GOODS Gold 1.30 Riess et al. (2004)
2002fx Athena Higher-z GOODS Silver 1.40 Riess et al. (2004)
2002hp Thoth Higher-z GOODS Gold 1.305 Riess et al. (2004)
2002ki Nanna Higher-z GOODS Gold 1.141 Riess et al. (2004)
2003aj Inanna Higher-z GOODS Silver 1.307 Riess et al. (2004)
2003ak Gilgamesh Higher-z GOODS Silver 1.551 Riess et al. (2004)
2003az Torngasak Higher-z GOODS Silver 1.265 Riess et al. (2004)
2003dy Borg Higher-z GOODS Gold 1.34 Riess et al. (2004)
HST04Eag Eagle Higher-z PANS Gold 1.019 Riess et al. (2007)
HST04Gre Greenburg Higher-z PANS Gold 1.14 Riess et al. (2007)
HST04Mcg Mcguire Higher-z PANS Gold 1.357 Riess et al. (2007)
HST04Sas Sasquatch Higher-z PANS Gold 1.39 Riess et al. (2007)
HST05Fer Ferguson Higher-z PANS Gold 1.02 Riess et al. (2007)
HST05Gab Gabi Higher-z PANS Gold 1.12 Riess et al. (2007)
HST05Koe Koekemoer Higher-z PANS Gold 1.23 Riess et al. (2007)
HST05Lan Lancaster Higher-z PANS Gold 1.235 Riess et al. (2007)
HST05Str Strolger Higher-z PANS Gold 1.027 Riess et al. (2007)
SCP0401 SCP0401 SCP GOODS Gold 1.713 Rubin et al. (2013)
SCP05D0 Frida SCP CSS Gold 1.014 Suzuki et al. (2012)
SCP05D6 Maggie SCP CSS Gold 1.315 Suzuki et al. (2012)
SCP06A4 Aki SCP CSS Silver 1.192 Suzuki et al. (2012)
SCP06C0 Noa SCP CSS Gold 1.092 Suzuki et al. (2012)
SCP06F12 Caleb SCP CSS Silver 1.110 Suzuki et al. (2012)
SCP06G4 Shaya SCP CSS Gold 1.35 Suzuki et al. (2012)
SCP06H5 Emma SCP CSS Gold 1.231 Suzuki et al. (2012)
SCP06K0 Tomo SCP CSS Gold 1.415 Suzuki et al. (2012)
SCP06K18 Alexander SCP CSS Silver 1.411 Suzuki et al. (2012)
SCP06N33 Naima SCP CSS Silver 1.188 Suzuki et al. (2012)
SCP06R12 Jennie SCP CSS Gold 1.212 Suzuki et al. (2012)
SCP06U4 Julia SCP CSS Gold 1.05 Suzuki et al. (2012)

Notes.

aHDFN: SN 1997ff was discovered in observations of the Hubble Deep Field North (HDFN; Gilliland et al. 1999; Dickinson et al. 2001). Higher-z GOODS/PANS: Discoveries by the Hubble Higher-z SN Search team, from the SN component of the Great Observatories Origins Deep Survey (GOODS, HST-GO-9728, HST-GO-9352, HST-GO-9583; Giavalisco et al. 2004; Strolger et al. 2004) and the successor program Probing Acceleration Now with Supernova (PANS, HST-GO-10339; Riess et al. 2007). SCP-GOODS/CSS: Discoveries by the Supernova Cosmology Project (SCP) on the GOODS fields (HST-GO-9727) or in the Cluster Supernova Search (CSS, HST-GO-9425; Dawson et al. 2009). bConfidence in the Type Ia classification, as reported by Riess et al. (2007) or Suzuki et al. (2012), where the latter have been translated from "secure/probable/plausible" to "gold/silver/bronze."

Download table as:  ASCIITypeset image

3. High-redshift Measurements of the Hubble Parameter

At $z\gtrsim 1$, dark energy is a small contribution to the energy budget (${\rho }_{{\rm{\Lambda }}}/\rho \approx 0.2$ at z = 1 and ≈0.1 at z = 1.5) and therefore has a small effect on dynamics. With abundant and better-measured SNe at lower redshifts, constraints on typical one-or-two-parameter dark energy models are only weakly improved by observations of SNe at $z\gt 1$ (see also Andersen & Hjorth 2017, regarding $z\gt 2$). This is especially true for combined constraints when precise distances from cosmic microwave background (CMB) and baryon acoustic oscillations (BAO) measurements are included.

Nevertheless, the new SNe at $z\gt 1.5$ presented here allow us to constrain the (dimensionless) Hubble parameter $E(z)\equiv H(z)/{H}_{0}$ at greater redshifts than previously possible. The quantity H(z) is particularly useful because it is both a direct probe of cosmology and still closely tied to the data. As a dynamical quantity, H(z) contains information about the expansion history without reference to any physical cosmological model. Also, at least for current SN Ia data, the inferred H(z) measurements are fairly local; that is, they are predominantly influenced by SNe at nearby redshifts. The quantity E(z), which contains similarly useful information but can be measured using SN Ia data alone, makes the results independent of uncertainties associated with the determination of the absolute distance scale of SNe Ia (Riess et al. 2016).

As a direct probe of the expansion rate ($H\equiv \dot{a}/a$), measurements of E(z) are particularly dense with cosmological information. They provide, for instance, a straightforward way to test or falsify a given cosmological model (Mortonson et al. 2009, 2010; Shafieloo & Clarkson 2010). Given current constraints on its parameters, the flat ΛCDM model already makes very precise predictions for such basic observables. Constraints on the matter density ${{\rm{\Omega }}}_{m}$ from combined probes (e.g., Ade et al. 2016) imply that E(z), defined to be exactly one at z = 0, is predicted to a precision ranging from ∼0.1% at z = 0.1 to ∼1% at z = 2. Therefore, any new, independent measurement of E(z), particularly in a new redshift range, is a direct and nontrivial test of the standard cosmological model. Given the present $\gt 3\sigma $ tension between H(z) calibrated at $z\approx 0$ (Riess et al. 2016) and at $z\approx 1100$ by the CMB (Ade et al. 2016), it is especially worthwhile to see if the expansion rate fails to match the standard ΛCDM model prediction anywhere along this redshift range.

Furthermore, as we will illustrate, accurate estimates of E(z) from SN Ia data are a convenient and efficient form of data compression, allowing one to obtain SN Ia constraints on dark energy and other cosmological parameters quickly and robustly using a very small and easily provided set of measurements. Such data compression techniques will be especially useful as SN Ia samples grow significantly in size in the coming decade. Some recent SN Ia analyses (e.g., Betoule et al. 2014) have included compressed versions of the SN data in the form of binned distance moduli, and it is worth investigating the extent to which E(z) measurements can serve a similar purpose.

Finally, quantifying SN Ia constraints on E(z) facilitates a more direct comparison with other cosmological probes of geometry, such as anisotropic fits of the BAO feature, which effectively constrain a dimensionless measure of the expansion rate, the product of the Hubble parameter and the sound horizon, where the latter is inferred precisely from CMB observations.

Our aim here is to employ a new, well-calibrated compilation of SNe Ia, featuring the final addition of 9 new SNe Ia at $z\gt 1$ from the CANDELS and CLASH programs, to obtain unbiased estimates of the Hubble parameter E(z) up to $z\approx 1.5$.

In what follows, we will briefly review some proposed methodologies for inferring E(z) from SN Ia data (Section 3.1) and then discuss our approach and how it overcomes some important limitations (Section 3.2). In Section 3.3, we present constraints on E(z) for the Pantheon SN compilation supplemented by the MCT SNe (i.e., Pantheon + MCT25 ). We illustrate how the handful of high-redshift SNe from CANDELS and CLASH significantly improves the determination of E(z) at $z\approx 1.5$. We also illustrate the effectiveness of the E(z) measurements in subsequent inference of cosmological parameters, and, in Section 3.4, the ability of the high-redshift SNe Ia to distinguish cosmology from SN Ia evolution. Finally, employing a realistic simulation of a potential Wide-Field Infrared Survey Telescope (WFIRST) SN survey observing strategy, we compare our current results with optimistic future constraints on E(z) (Section 3.5).

3.1. SN Ia Measurements of E(z)

SNe Ia measure distances most directly; roughly speaking, each SN provides an independent measurement of the luminosity distance to its redshift. For a flat universe, we have

Equation (1)

therefore the (inverse) Hubble parameter, the derivative of the comoving distance, must be inferred indirectly when starting from raw SN Ia data.

A variety of interrelated methods have been used for this purpose. Some analyses have focused on model-independent reconstruction of an analytical E(z) function or of other dynamical quantities like the deceleration parameter q(z) (Sahni & Starobinsky 2006; Shafieloo et al. 2006; Shafieloo 2007; Ishida & de Souza 2011). Such reconstructions are useful for understanding where the data are most constraining, and they can indicate whether the functional form for H(z) naturally preferred by the data is consistent with that of a physical model like ΛCDM. On the other hand, it is not possible, or at least not straightforward, to subsequently incorporate the reconstructions in a likelihood function, or otherwise in a statistical analysis, in order to constrain cosmological parameters.

Other methods focus on obtaining direct measurements of E(z) at several redshifts by smoothing and/or weighting the individual SNe and differentiating the distance-redshift relation (Tegmark 2002; Daly & Djorgovski 2003, 2004; Daly et al. 2008). One proposed method (Wang & Tegmark 2005), which has been employed in some subsequent analyses (Riess et al. 2007; Avgoustidis et al. 2009; Mortsell & Clarkson 2009), seeks direct, independent estimates of E(z) in redshift bins by first converting SN distance moduli into their corresponding comoving distances ri, then transforming these ri into noisy, but locally unbiased, estimates of $E{(z)}^{-1}$ between neighboring SNe. A specific weighted average then yields a minimum-variance estimate of $E{(z)}^{-1}$ over a wider redshift bin. We have verified numerically26 that this procedure is actually equivalent to the familiar weighted least-squares fit of a line to the ri versus zi data over the same wide redshift bin, where the slope corresponds to $E{(z)}^{-1}$. Both the least-squares estimator and that of Wang & Tegmark (2005) have been shown to be unbiased and have minimum variance, assuming SN redshifts are known exactly and E(z) is constant over the redshift bin, so it is not surprising that these estimators coincide.

While such an approach is attractive in that it directly transforms the SN distances into independent measurements of E(z) at different redshifts, it has notable problems that make it unsuitable in practice. The first step requires converting SN distance moduli into comoving distances, and one must therefore assume a value for the intercept of the Hubble diagram, which is unknown a priori. As this quantity is partially degenerate with E(z), particularly the lowest-redshift measurement, fixing the intercept to some best-fit value would artificially remove a degree of freedom from the fit, resulting in underestimated uncertainties. One could instead interpret the estimates as estimates of ${AE}{(z)}^{-1}$, where A is an arbitrary constant. In this case, though, properly extracting cosmological information from the E(z) measurements would require fully marginalizing over A in a fit to multiple measurements of ${AE}{(z)}^{-1}$.

Furthermore, an E(z) estimate using this method reflects some average of E(z) over the redshift bin, not necessarily the value at the bin's center. Unless E(z) is constant over the redshift bin, this will lead to a bias, and since only a handful of E(z) values can be constrained robustly with current data, one might expect the bias to be significant. Indeed, by simulating instances of our SN data (see Section 3.2), we have verified that biases in such E(z) estimates are typically a large fraction (∼0.5) of their uncertainty, making the measurements unsuitable for later cosmological inference.

3.2. Parametrized E(z) and Interpolation

We now describe a somewhat different approach for determining E(z) from SN Ia data. We explain how it avoids the problems discussed in Section 3.1 and provides more meaningful and robust E(z) measurements. We will assume that the true, underlying E(z) function is a continuous, smooth function of redshift, which is certainly the case for most physical and empirical models studied in the literature.

In our approach, we parametrize E(z) by its value at several specific redshifts and employ a basic interpolation scheme to define the complete E(z) function, which can then be numerically integrated to compute the luminosity distance and compare to the data. This allows us to constrain the E(z) parameters using the full SN data set in its raw form, as one would in a standard dark energy analysis. This way, any nuisance parameters associated with the SN data, notably the distance scale or Hubble diagram intercept, can be properly marginalized over in the fit.

While the total number of E(z) values to constrain is somewhat arbitrary, there are several considerations. Choosing too many E(z) parameters results in weaker constraints and posterior distributions that are less likely to be Gaussian. Choosing too few E(z) values increases the chance that the estimates will be biased, as the interpolating function will deviate from the functional form of the underlying cosmology. The specific redshifts, while also somewhat arbitrary, should reflect the redshift range and distribution of the SNe. Since the E(z) measurements, especially those at neighboring redshifts, will naturally be somewhat correlated, choosing too small a separation in redshift between a given pair will lead to undesirably large pairwise correlations in the estimates. We choose redshifts such that the resulting E(z) correlation coefficients are ≲0.5 to avoid such redundancy.

Overall, we find that employing a shape-preserving piecewise-cubic Hermite interpolating polynomial (implemented as pchip in MATLAB; see Kahaner et al. 1988) to interpolate (and extrapolate) the E(z) function works particularly well, though other interpolation schemes (various splines, simple linear interpolation) are also generally suitable. For any specified E(z) (any fiducial cosmology), it is straightforward to determine whether the E(z) estimates resulting from the interpolation and fitting procedure are unbiased. To check this, we repeatedly simulate instances of our SN Ia data; that is, we keep the same SN redshifts and covariance matrix as the real SN data, but repeatedly sample the distance moduli from a multivariate Gaussian centered on the fiducial cosmology. Of course, unbiased constraints for the fiducial cosmology do not guarantee unbiased results for other cosmologies. In principle, one could perform this check for each specific model of interest; however, there is reason to worry only when a model predicts E(z) to vary rapidly or have features too narrow to be captured by the widely spaced E(z) parameters. For the highest-redshift SNe Ia, a modest amount (∼25%) of the integral of $E{(z)}^{-1}$ must be evaluated via extrapolation beyond the last redshift anchor of the E(z) function. However, our simulations indicate that this does not bias this highest-redshift measurement of E(z). Indeed, we have verified that all of the E(z) measurements are biased by ≲10% of their individual statistical uncertainties.

In essence, our procedure trades the ability to make direct, independent measurements of E(z) at redshifts that are somewhat uncertain (and not randomly so) for the ability to obtain precise, unbiased E(z) estimates at specific redshifts. Only this latter type of estimate allows for accurate subsequent cosmological inference with the E(z).

3.3. SN Ia Constraints on E(z)

We now constrain E(z) for the Pantheon compilation of 1040 SNe Ia, which we will supplement with the high-redshift CANDELS and CLASH SNe. The Pantheon compilation (Scolnic et al. 2017) includes data from multiple surveys (CfA(1–4), CSP, SDSS, SNLS, Pan-STARRS1, HST) calibrated for a joint cosmological analysis. Below we summarize the key aspects of the Pantheon analysis, and we refer the reader to Scolnic et al. (2017) for additional details and a complete discussion.

The Pantheon analysis presents the full set of spectroscopically confirmed SNe Ia from the Pan-STARRS1 (PS1) Medium Deep Survey, building on the earlier analysis of the first 1.5 yr of PS1 (Rest et al. 2014; Scolnic et al. 2014). It relies on the Supercal cross-calibration procedure presented by Scolnic et al. (2015), which uses the relative consistency of the Pan-STARRS1 photometry over 3π steradians of the sky to tie together the photometric systems of the individual surveys. The Pantheon analysis also incorporates the BBC methodology of Kessler & Scolnic (2017) (see also Scolnic & Kessler 2016), which corrects for distance biases dependent on the light-curve properties of the SNe and the surveys from which they are selected.

The Pantheon analysis employs the SALT2 light-curve fitter (Guy et al. 2007; Betoule et al. 2014), which determines an overall normalization of the log-flux (mB), a shape parameter (x1), and a color (c) for each SN light curve, along with associated uncertainties. We standardize the SNe by modeling an individual SN Ia distance modulus as

Equation (2)

The ${{\rm{\Delta }}}_{M}$ term is an additional correction for the empirical host-mass step, where SNe in high-stellar-mass host galaxies (${\mathrm{log}}_{10}({M}_{* }/{{\rm{M}}}_{\odot })\gtrsim 10$) are ∼0.05 mag brighter on average, after light-curve standardization.27 The ${{\rm{\Delta }}}_{B}$ term represents the distance bias correction. Note that M, α, β, and the amplitude of the mass step (included in the ${{\rm{\Delta }}}_{M}$ term) are all nuisance parameters that must be determined by a fit to the data. In our analysis, only M (effectively, the Hubble diagram offset) is fit along with the cosmological parameters E(z). The other parameters are well determined independently of cosmology in the Pantheon analysis. The inferred values are $\alpha \approx 0.15\mbox{--}0.16$ and $\beta \approx 3.0\mbox{--}3.7$, where the results vary depending on the intrinsic scatter model.28 Finally, note that the distance modulus as predicted by the cosmological model is given by

Equation (3)

where dL is the luminosity distance, which is a function of redshift and also depends on the set of cosmological parameters ${\boldsymbol{p}}$.

The statistical uncertainties of SN distance moduli are modeled, in the standard way, as a combination of observational measurement uncertainty, intrinsic scatter, and additional scatter due to gravitational lensing, peculiar velocities, and redshift measurement uncertainty.29 The inferred value for the intrinsic scatter is ${\sigma }_{\mathrm{int}}\approx 0.1$, although, like α and β, the value depends on the intrinsic scatter model. After bulk-flow corrections are applied to the low-redshift SNe, we add a peculiar-velocity scatter of ${\sigma }_{v}=250$ km s−1. We assume a value ${\sigma }_{\mathrm{lens}}=0.055z$ for the lensing scatter (Jönsson et al. 2010). Note that the distribution of the shift in observed magnitude due to lensing is non-Gaussian (e.g., Jönsson et al. 2006), with a tail of strongly magnified SNe; however, by examination of foreground structures we have verified that none of our CANDELS or CLASH SNe are likely to fall in this tail, making the lensing scatter contribution to the distance uncertainty a good approximation. Note that there is also statistical uncertainty in the host-mass correction and the distance bias correction.

The Pantheon analysis also includes a rigorous analysis of systematic errors, adding terms to the covariance matrix of SN distances to account for uncertainties in photometric calibration (including terms for individual survey calibration, the Supercal cross-calibration procedure, and the SALT2 model itself), the intrinsic scatter model, survey selection functions, Milky Way dust extinction, β evolution, the host mass step and its evolution, and peculiar velocity coherent flow corrections.

Standard data-quality cuts were applied to remove SNe that are not expected to follow the empirical standardization relations. Specifically, we keep only SNe with $| {x}_{1}| \lt 3$, ${\sigma }_{{x}_{1}}\lt 1$, $| c| \lt 0.3$, a light-curve fit with ${\chi }^{2}/{N}_{\mathrm{dof}}\lt 3$, and an uncertainty in the time of peak brightness of less than 2 days. Similar cuts have been used in most recent SN Ia cosmological analyses (e.g., Betoule et al. 2014; Rest et al. 2014; Riess et al. 2016). These cuts eliminate 3 of the silver and gold MCT SNe (CLH11Tra, GND13Gar, GND13Jay; see Table 4). Finally, a $4\sigma $ outlier rejection from the best-fit Hubble diagram is applied and removes GND13Cam, leaving 9 HST MCT SNe Ia in the joint analysis.30 Note that here we do include EGS13Rut, which is on the edge of the ${\sigma }_{{x}_{1}}$ cut but has typical light-curve fit parameters. Although the final MCT addition of 9 SNe represents $\lt 1$% of the combined sample, the unusually high redshifts (7 with $z\gt 1.5$) provide unique leverage on E(z) at z = 1.5.

Following the methodology and discussion in Section 3.2, we parametrize $E{(z)}^{-1}$ by its value at six redshifts (chosen to best summarize the sample) and therefore have six free parameters to constrain. It is important to remember that the Hubble diagram offset is a free parameter as well, though we analytically marginalize over this offset, with a flat prior, in the likelihood. We assume a flat universe (${{\rm{\Omega }}}_{k}=0$) throughout, so the E(z) measurements are cosmological-model-dependent in this sense. To obtain the constraints, we sample the likelihood using a custom Markov chain Monte Carlo (MCMC) code employing the basic Metropolis–Hastings algorithm. We impose flat, hard-bound priors on the $E{(z)}^{-1}$ parameters wide enough that extending the bounds does not affect the resulting constraints. The final MCMC chains were inspected to verify convergence.

The resulting marginalized posterior likelihoods for $E{(z)}^{-1}$ are Gaussian to a good approximation, and the constraints are given in Table 6. In Figure 1, we convert the measurements of $E{(z)}^{-1}$ into E(z) measurements by reprocessing the MCMC chains and then compare the results with and without the MCT SNe. It is not surprising that the MCT SNe subsantially improve the measurement of E(z) at z = 1.5. They permit a ∼20% measurement of $E(z=1.5)$, roughly a factor of three improvement over the result without the MCT SNe. While the CANDELS and CLASH SNe mostly affect the measurement at z = 1.5, they also improve and shift some lower-redshift measurements, which are somewhat correlated (≈8% and 4% improvements at z = 0.9 and 0.55, respectively). By eye, the set of E(z) measurements may appear somewhat discrepant with the fiducial ΛCDM model, but the overall ${\chi }^{2}$, which includes the moderate correlations, is 6.7 for the 6 degrees of freedom.

Figure 1.

Figure 1. Constraints on $E(z)\equiv H(z)/{H}_{0}$, relative to E(z) for a fiducial ΛCDM model (${{\rm{\Omega }}}_{m}=0.3$). We compare the constraints with (blue points) and without (red points) the high-redshift CANDELS and CLASH (MCT) SNe Ia. Note that these E(z) measurements are correlated and have non-Gaussian distributions (the error bars enclose 68.3% of the likelihood). For comparison, we also show the three (correlated) measurements of E(z) from combined BOSS DR12 BAO data (Alam et al. 2017) after calibration with Planck ΛCDM constraints on ${H}_{0}\,{r}_{d}$ (green points).

Standard image High-resolution image

Table 6.  Pantheon + MCT SN Ia Measurements of E(z)

z $E{(z)}^{-1}$ a Correlation Matrix E(z) Distance Residual ${\rm{\Delta }}\mu /(0.01\,\mathrm{mag}$)b
0.07 1.007 ± 0.024 1.00           0.994 ± 0.023 −0.13 ± 0.99
0.2 0.898 ± 0.016 0.40 1.00         1.113 ± 0.020 −0.23 ± 1.26
0.35 0.893 ± 0.029 0.52 −0.13 1.00       1.122 ± 0.037 +0.23 ± 1.32
0.55 0.732 ± 0.033 0.35 0.35 −0.18 1.00     1.369 ± 0.063 +0.11 ± 1.97
0.9 0.652 ± 0.051 0.02 −0.08 0.19 −0.41 1.00   1.54 ± 0.12 +1.15 ± 2.85
1.5 0.337 ± 0.078 0.00 −0.06 −0.05 0.16 −0.21 1.00 ${2.69}_{-0.52}^{+0.86}$ −3.42 ± 6.78

Notes.

aMean and standard deviation of the marginalized likelihood, approximately Gaussian in all cases. bEffective distance moduli relative to those of a fiducial ΛCDM cosmology (${{\rm{\Omega }}}_{m}=0.3$), as determined by an interpolated fit to the residuals using the same redshift control points as the E(z) analysis.

Download table as:  ASCIITypeset image

In Figure 2, we scale E(z) by ${(1+z)}^{-1}$ to illustrate the constraints on the time derivative of the scale factor $\dot{a}(z)$, relative to its present value, for the same data shown in Figure 1. In this space, it is clear that the low-redshift and high-redshift E(z) measurements together provide evidence for both recent acceleration and earlier deceleration epochs, as predicted by standard cosmological models. In addition to the fiducial ΛCDM model, we show dynamical models with fixed deceleration parameter q0. The $\dot{a}(z)$ values track the ${q}_{0}=-0.5$ model at $z\lesssim 0.5$ (where the low-z behavior matches that of a ΛCDM model with ${{\rm{\Omega }}}_{m}\approx 0.3$) but show deceleration with respect to that curve at higher redshifts. The coasting cosmology (${q}_{0}=0$), pure acceleration cosmology (${q}_{0}=-0.5$), and pure deceleration cosmology (${q}_{0}=0.5$, equivalent to a flat CDM model with ${{\rm{\Omega }}}_{m}=1$) are strongly disfavored with ${\rm{\Delta }}{\chi }^{2}=78.9$, ${\rm{\Delta }}{\chi }^{2}=40.5$, and ${\rm{\Delta }}{\chi }^{2}=357.9$, respectively, for 6 degrees of freedom. The measurement at z = 1.5 alone, while consistent with the other models, disfavors ${q}_{0}=-0.5$ with ${\rm{\Delta }}{\chi }^{2}=14.2$.

Figure 2.

Figure 2. For the same data as in Figure 1, we show constraints on the time derivative of the scale factor $\dot{a}(z)$ relative to its present value, obtained by scaling the E(z) values by ${(1+z)}^{-1}$. We compare the fiducial ΛCDM model to alternative models with a constant deceleration parameter ${q}_{0}=0$ (coasting cosmology), ${q}_{0}=-0.5$ (pure acceleration), and ${q}_{0}=0.5$ (pure deceleration), all assuming a flat universe.

Standard image High-resolution image

As an illustration of the power of the E(z) measurements in constraining (spatially flat) cosmologies, we compare constraints on common dark energy parameterizations in Figure 3. Remarkably, the constraints are nearly identical whether the parameters are constrained with the SN Ia data directly or with the E(z) measurements in Table 6. It may not be too surprising that E(z) captures the constraining power of the SNe for simple one-or-two-parameter models. One would not expect the same for fits with many degrees of freedom (e.g., more complicated dark energy models); in practice, however, current and near-future SN Ia data can only meaningfully constrain 2–3 expansion parameters anyway. For models that assume a flat universe and predict fairly smooth, featureless H(z), the E(z) constraints will be an efficient summary of the present SN Ia data.

Figure 3.

Figure 3. Constraints on ${{\rm{\Omega }}}_{m}$ and a constant equation-of-state parameter w in a flat universe (left panel) and for the w0wa model (Chevallier & Polarski 2001; Linder 2003), marginalized over ${{\rm{\Omega }}}_{m}$ and also assuming a flat universe (right panel). We compare the constraints when using the full SN Ia likelihood with individual distance moduli (filled blue contours) with the constraints from the six moderately correlated E(z) measurements (open red contours). Contours contain 68.3%, 95.4%, and 99.7% of the likelihood, and for the w0wa constraints we have also included distance priors derived from Planck data (Ade et al. 2016).

Standard image High-resolution image

3.4. High-redshift SNe Ia and Evolution

The use of SNe Ia as standardizable candles across redshift relies on the understanding that their uncommonly homogeneous luminosities and colors follow from their nature as carbon–oxygen white dwarfs close to the Chandrasekhar mass. While uncertainty persists regarding how these degenerate stars approach that mass limit, either by accretion from a nondegenerate companion or through the tidal disruption followed by accretion of a degenerate companion, there has long been agreement about this model based on the well-understood physics of degenerate stars (Hoyle & Fowler 1960; Arnett 1969; Colgate & McKee 1969). The thermonuclear detonation of a Chandrasekhar-mass carbon–oxygen white dwarf yields a mass of radioactive nickel whose energy output matches that of a SN Ia (Arnett et al. 1985) and whose modeled nucleosynthesis matches its spectral elements (Nomoto et al. 1984). More recently, prediscovery observations of SN 2011fe, a prototypical SN Ia in M101, demonstrated that the progenitor did not exceed a radius of 2% solar, fully consistent with the expected white dwarf (Li et al. 2011; Nugent et al. 2011; Bloom et al. 2012). Yet the difficulty and low likelihood of ever directly observing a white dwarf system before it becomes a SN Ia leaves enough uncertainty and model freedom to support the consideration of redshift evolution of the standardized SN Ia luminosity.

From SN Ia observations spanning a wide range of redshifts and sampling the epochs when cosmic expansion accelerates and decelerates, it is possible to distinguish such evolution from the uncertain properties of dark energy (Riess & Livio 2006). As an illustration of the power of SNe Ia at $z\gt 1$ to separate evolution from cosmology, we briefly reconsider the analysis of Tutusaus et al. (2017), which shows that power-law cosmology, where the scale factor evolves as $a(t)\propto {t}^{n}$ for some exponent n, is an equally good fit to SN Ia data (primarily at $z\lt 1$) as the ΛCDM model (with ${{\rm{\Omega }}}_{m}$ free) when the standardized luminosity is also allowed to vary with redshift according to some simplistic, empirical models of SN Ia evolution. Although such models are not astrophysically motivated, they may be useful for exploring the separation of other SN distance-dependent effects (e.g., gray extinction) from cosmological parameters.

Here, as an illustration, we consider Model B (${\rm{\Delta }}M(z)=\epsilon {z}^{\delta }$) from Tutusaus et al. (2017) with fixed $\delta =0.3$. We separately fit both ΛCDM and power-law cosmology to our combined (Pantheon + MCT) data; in each case, we fit for the Hubble diagram intercept, a cosmological parameter (${{\rm{\Omega }}}_{m}$ or n), and the amplitude epsilon of the assumed intrinsic luminosity evolution. We compare these fits in Figure 4. Fitting only the SNe at $z\lt 1$, a power law with n = 1.1 is a slightly better fit to the SN Ia data than ΛCDM. Indeed, when analyzing the JLA compilation, which features only ∼5 SNe at $z\gt 1$, Tutusaus et al. (2017) claims a mild preference for the power law (note that their analysis also included BAO and H(z) information).

Figure 4.

Figure 4. Comparison of ΛCDM and power-law cosmology ($a(t)\propto {t}^{n}$) fits to our SN Ia data, where in each case we allow the intrinsic luminosity to evolve as ${\rm{\Delta }}M(z)=\epsilon {z}^{\delta }$, corresponding to Model B from Tutusaus et al. (2017), where we fix $\delta =0.3$. The SN data are binned for clarity, and ${\rm{\Delta }}{\chi }_{{\rm{\Lambda }}\mathrm{CDM}}^{2}\,\equiv {\chi }^{2}-{\chi }_{{\rm{\Lambda }}\mathrm{CDM}}^{2}$.

Standard image High-resolution image

In contrast, when we include our 24 SNe at $z\gt 1$, a nearly coasting (marginally accelerating) power-law cosmology (best fit n = 1.04) together with simplistic SN Ia evolution is no longer as good a fit as the ΛCDM model, with a relative probability of $\exp (-{\rm{\Delta }}{\chi }^{2}/2)\approx 20 \% $. Without invoking evolution (that is, fixing $\epsilon \equiv 0$), the ΛCDM model is a much better fit than the power-law model, with the latter strongly disfavored with ${\rm{\Delta }}{\chi }_{{\rm{\Lambda }}\mathrm{CDM}}^{2}=8.3$, a relative probability of 1.6%, when including the new SNe at $z\gt 1.5$. Meanwhile, assuming ΛCDM and fitting for the evolution amplitude epsilon yields a value consistent with zero, $\epsilon =0.08\pm 0.15$, so there is no motivation for including it based on astrophysical or empirical considerations. A more comprehensive investigation of SN Ia evolution and cosmology is underway (D. L. Shafer et al., 2018 in preparation).

We note the addition of the MCT SNe to the Pantheon compilation also further reduces the already-low likelihood of the "empty universe" solution where ${{\rm{\Omega }}}_{m}\approx 0$ and ${{\rm{\Omega }}}_{{\rm{\Lambda }}}\approx 0$ in an open ΛCDM universe, a location Nielsen et al. (2016) claimed to be marginally consistent ($\sim 3\sigma $) with SN data alone using unconventional priors on SN distributions, to the boundary of the 6σ contour.

3.5. E(z) with WFIRST

WFIRST was the top space-based recommendation of the 2010 U.S. astronomy and astrophysics decadal survey. The mission is still in formulation, but current plans specify a 2.4 m primary mirror and include a wide-field instrument for cosmology. The cosmology science objectives, as detailed in the most recent report from the Science Definition Team (Spergel et al. 2015), will be accomplished through a combination of SN Ia, galaxy, and weak-lensing surveys.

The WFIRST SN survey is anticipated to yield a large sample of thousands of SNe, many at $z\gt 1$ with precise distances. These SNe will vastly improve upon the high-redshift E(z) measurements available today, allowing nontrivial and precision tests of the ΛCDM model independent of the BAO and weak-lensing constraints in a redshift range that is currently not well constrained.

Here we wish to forecast realistic constraints on E(z) from WFIRST. Typical forecasts (e.g., for dark energy figures of merit) rely on Fisher matrix formalism, which is exact only for Gaussian posterior distributions and otherwise underestimates parameter uncertainties. For SN Ia forecasts, one typically assumes idealized, or roughly estimated, redshift distributions and makes simple assumptions about the measurement error. Here instead we employ a detailed simulation of one potential observing strategy for the WFIRST SN survey (Hounsell et al. 2017). We then constrain the E(z) parameters using the methodology of Section 3.2 that was employed in Section 3.3 for our current Pantheon + MCT data.

For our illustration, we consider the Imaging All-z strategy described by Hounsell et al. (2017). This particular strategy relies on multi-band imaging for classification and assumes follow-up spectroscopy will provide host-galaxy redshifts. Hounsell et al. (2017) also assumes a large external sample of 800 SNe at $z\lt 0.1$. As the size of future systematic uncertainties is hard to predict, Hounsell et al. (2017) simulates a range of scenarios, and here we opt for all-around optimistic assumptions about future systematic errors (for what this entails, see Hounsell et al. 2017). In this scenario, the contribution of systematic errors is not negligible but is subdominant in the error budget.

In Figure 5, we compare our current Pantheon + MCT constraints on E(z) with simulated constraints from the WFIRST Imaging All-z strategy. We find that we are able to constrain E(z) robustly, albeit with moderate pairwise correlations, at 9 redshifts in the range $0.07\lt z\lt 2.5$. In Table 7, we list the percent errors for E(z) corresponding to our forecast. Note that these results are negligibly changed whether we quote percent errors on E(z) or its inverse. We find that WFIRST allows 8 measurements of E(z) at the 1%–3% level, along with a robust but less precise measurement at $z\approx 2.5$. Notably, this is a constraint on the expansion rate at a redshift higher than any SN Ia has even been observed to date.

Figure 5.

Figure 5. Simulated WFIRST constraints on $E(z)\equiv H(z)/{H}_{0}$, relative to E(z) for a fiducial ΛCDM model (${{\rm{\Omega }}}_{m}=0.3$). We compare the constraints from current data (blue points) with simulated constraints from the WFIRST Imaging All-z observing strategy (green points). We overlay the same dark energy models as in Figure 1.

Standard image High-resolution image

Table 7.  Simulated SN Ia Measurements of E(z) from WFIRST

z E(z) Percent Errora
0.07 1.3
0.2 1.1
0.35 1.5
0.6 1.5
0.8 2.0
1.0 2.3
1.3 2.6
1.7 3.4
2.5 8.9

Note.

aNote that these measurements are not fully independent; there are moderate pairwise correlations among some in the set. They assume a flat universe.

Download table as:  ASCIITypeset image

4. Summary and Conclusions

In this study, we analyzed the set of 15 high-redshift SNe Ia from the CANDELS and CLASH HST MCT programs, 9 of which ultimately pass classification confidence and quality cuts and 7 of which are at $z\gt 1.5$ where the relative expansion rate is poorly constrained. These are the first distance estimates for these SNe that are suitable for a joint cosmological analysis with a large compilation of lower-redshift SNe (the Pantheon compilation). We have introduced and employed a procedure to obtain unbiased constraints on the scale-free Hubble parameter $E(z)\equiv H(z)/{H}_{0}$ using only this extended Pantheon + MCT sample of SNe Ia (Table 6, Figures 12). The CANDELS and CLASH SNe at $z\gtrsim 1.5$ extend the Hubble diagram and allow us to achieve a robust measurement of the expansion rate at z = 1.5 that efficiently summarizes the cosmological leverage of these new SNe. Our measurement of $E{(z=1.5)}^{-1}=0.337\pm 0.078$ (equivalently, $E(z=1.5)={2.69}_{-0.52}^{+0.86}$) assumes a flat universe and smooth expansion history but is otherwise model-independent.

We also have demonstrated that the set of E(z) measurements can serve as a form of SN Ia data compression, allowing us to summarize SN Ia constraints on spatially flat cosmological models that feature a smooth expansion history, which comprise the majority of the commonly studied dark energy models. The E(z) are very economical, accurately reproducing parameter posteriors (even when non-Gaussian) using just 6 measured quantities in place of $\gt 1000$ (Figure 3). The computation time for this E(z) likelihood, relative to that for the full SN Ia likelihood, is negligible.

Future large, high-quality samples of high-redshift SNe Ia, notably from WFIRST, will allow precision constraints on the dark energy equation-of-state parameter w, especially for dynamical dark energy featuring a time-varying value of w. Still, there are uses for such high-redshift SNe beyond direct dark energy constraints, inspiring us to perform two additional investigations.

First, using our combined Pantheon + MCT set of SNe Ia, we have briefly illustrated how the added leverage of our larger sample of SNe at $z\gt 1$, including 7 at $z\gt 1.5$, can help distinguish empirical SN Ia evolution and nonstandard cosmological models from the ΛCDM model (Figure 4). We have shown that, while a nearly coasting power-law model ($a(t)\propto {t}^{n}$ with $n\approx 1$) is as good a fit to the $z\lt 1$ data as ΛCDM (at least when certain forms of SN evolution are allowed), adding the $z\gt 1$ SNe disfavors the power law, indicating a relative probability of ∼20%, even when permitting the same SN evolution.

Second, we have used our E(z) procedure in conjunction with a realistic simulation of a potential WFIRST SN Ia observing strategy to forecast optimistic WFIRST constraints on E(z). We find that WFIRST will permit 8 measurements of E(z) at the 1%–3% level across a wide range of redshifts, along with a robust measurement at $z\approx 2.5$ (Figure 5, Table 7). Such measurements will constitute precise tests of our expectations from the ΛCDM model separately from BAO and other high-redshift distance probes.

We thank Rebekah Hounsell for providing the WFIRST SN Ia simulations and the anonymous referee for helpful comments. It is our pleasure to thank program coordinators Patricia Royle and Beth Perriello, as well as the entire Space Telescope Science Institute (STScI) scheduling team, for their tireless efforts that made the CANDELS survey and the SN follow-up program possible.

This work was principally based on observations made with the NASA/ESA Hubble Space Telescope, which is operated by the Association of Universities for Research in Astronomy (AURA), Inc., under NASA contract NAS5-26555. These observations are associated with program IDs 12060, 12061, 12062, 12442, 12443, 12444, 12445, 12099, 12461, and 13063. The analysis presented here made extensive use of the Mikulski Archive for Space Telescopes (MAST). STScI is operated by AURA, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data is provided by the NASA Office of Space Science via grant NNX13AC07G and by other grants and contracts. Some of the data presented herein were obtained at the W.M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California, and NASA; the Observatory was made possible by the generous financial support of the W.M. Keck Foundation.

Financial support was broadly provided by NASA through grants HST-GO-12060 and HST-GO-12099 from STScI, and to SAR through grant HST-HF-51312. AVF is also grateful for generous financial assistance from the Christopher R. Redlich Fund, the TABASGO Foundation, and the Miller Institute for Basic Research in Science (U.C. Berkeley). AM acknowledges the financial support of the Brazilian funding agency FAPESP (Postdoc fellowship, process number 2014/11806-9). OG is supported by an NSF Astronomy and Astrophysics Postdoctoral Fellowship under award AST-1602595. JH was supported by a VILLUM FONDEN Investigator grant (project number 16599). SJ was supported by JPL RSAs 143563, 1448524, 1460278, and 1473597.

Facility: HST (WFC3). -

Appendix: CANDELS + CLASH Light-Curve Photometry

The light-curve photometry for the 15 likely SNe Ia is given in Table 8.

Table 8.  CANDELS + CLASH Light-curve Photometry

MJD Filter Band Fluxa Flux Unc. Magb Mag Unc.
SN CLA10Cal (Caligula)
55599.42 F814W I 11.982 1.084 24.8 0.1
55538.409 F125W J −0.035 0.847 >26.5
55558.446 F125W J 1.359 0.596 >26.9
55579.348 F125W J 0.805 0.637 >26.8
55599.383 F125W J 0.791 0.77 >26.6
55538.418 F160W H 0.794 0.797 >26.6
55558.455 F160W H 0.412 1.265 >26.1
55579.357 F160W H 0.612 1.175 >26.1
55599.392 F160W H 0.397 2.311 >25.4
55613.163 F160W H 1.386 0.775 >26.6
SN CLF11Ves (Vespasian)
55715.81 F775W X 0.916 1.056 >26.2
55736.78 F775W X 10.157 1.05 24.98 0.11
55747.67 F775W X 8.074 0.97 25.23 0.13
55715.8 F850LP Z 2.228 0.976 >26.3
55736.77 F850LP Z 14.909 0.96 24.57 0.07
55746.82 F850LP Z 11.119 1.252 24.88 0.12
55762.2 F850LP Z 7.178 0.972 25.36 0.15
55776.58 F850LP Z 3.46 0.937 26.15 0.29
SN CLH11Tra (Trajan)
55833.68 F775W X 0.321 0.947 >26.4
55854.04 F775W X 5.673 1.576 25.62 0.3
55872.41 F775W X 6.192 1.022 25.52 0.18
55885.16 F775W X 2.68 0.983 >26.3
55833.67 F850LP Z −0.348 0.778 >26.6
55872.443 F125W J 13.465 1.011 24.68 0.08
55885.216 F125W J 8.339 0.999 25.2 0.12
55872.45 F160W H 12.796 1.416 24.73 0.11
55885.224 F160W H 8.097 1.191 25.23 0.15
SN CLP12Get (Geta)
56048.121 F125W J −0.073 0.575 >26.9
56078.249 F125W J 0.253 0.938 >26.4
56105.648 F125W J 3.516 1.104 26.13 0.31
56117.553 F125W J 6.741 1.124 25.43 0.17
56131.585 F125W J 10.363 1.234 24.96 0.12
56138.833 F125W J 10.624 1.153 24.93 0.11
56146.219 F125W J 12.151 1.225 24.79 0.1
56164.368 F125W J 8.263 1.215 25.21 0.15
56184.516 F125W J 4.321 1.071 25.91 0.25
56197.283 F125W J 3.601 1.186 26.11 0.33
56048.129 F160W H 1.058 1.65 >25.8
56078.257 F160W H 0.899 1.588 >25.8
56105.656 F160W H 2.499 0.937 >26.4
56117.561 F160W H 6.284 1.077 25.5 0.17
56131.591 F160W H 9.995 1.238 25.0 0.12
56138.839 F160W H 10.703 1.291 24.93 0.12
56146.227 F160W H 8.411 1.253 25.19 0.15
56164.376 F160W H 6.156 1.265 25.53 0.21
56184.524 F160W H 4.135 1.02 25.96 0.25
56197.291 F160W H 5.663 1.278 25.62 0.23
SN COS12Car (Carter)
55954.58 F606W V 0.195 0.126 >28.6
55967.35 F850LP Z 0.988 0.597 >26.9
55908.8 F125W J 2.037 0.478 26.73 0.23
55957.504 F125W J 7.909 0.578 25.25 0.07
55967.543 F125W J 6.309 0.684 25.5 0.11
55968.674 F125W J 5.923 0.851 25.57 0.14
55977.603 F125W J 6.116 0.537 25.53 0.09
55991.105 F125W J 3.063 0.521 26.28 0.17
56006.455 F125W J 3.003 0.4 26.31 0.13
56026.291 F125W J 1.335 0.335 27.19 0.25
55908.806 F160W H 2.256 0.539 26.62 0.24
55957.516 F160W H 7.088 0.601 25.37 0.08
55963.439 F160W H 6.32 0.758 25.5 0.12
55977.588 F160W H 4.707 0.739 25.82 0.16
55991.09 F160W H 5.313 0.756 25.69 0.14
56006.389 F160W H 4.784 0.531 25.8 0.11
56026.216 F160W H 4.259 0.518 25.93 0.12
SN EGS11Oba (Obama)
55659.793 F125W J 0.782 0.636 >26.8
55710.639 F125W J 7.038 0.764 25.38 0.11
55725.214 F125W J 7.377 0.758 25.33 0.1
55739.126 F125W J 4.125 0.691 25.96 0.17
55753.522 F125W J 3.757 0.715 26.06 0.19
55659.799 F160W H 0.724 0.646 >26.8
55710.645 F160W H 6.529 0.713 25.46 0.11
55725.221 F160W H 6.968 0.845 25.39 0.12
55739.118 F160W H 4.252 0.712 25.93 0.17
55753.647 F160W H 4.585 0.707 25.85 0.15
SN EGS13Rut (Rutledge)
56438.18 F606W V −0.04 0.103 >28.8
56438.24 F814W I 1.684 0.389 26.93 0.25
56447.42 F814W I 2.228 0.384 26.63 0.19
56437.624 F125W J 7.044 0.732 25.38 0.1
56447.139 F125W J 8.717 0.751 25.15 0.09
56460.64 F125W J 8.766 0.696 25.14 0.08
56482.351 F125W J 6.659 0.669 25.44 0.1
56437.631 F160W H 6.361 0.673 25.49 0.11
56447.253 F160W H 7.588 0.747 25.3 0.1
56460.647 F160W H 9.158 0.832 25.1 0.09
56482.359 F160W H 6.599 0.705 25.45 0.11
56501.147 F160W H 4.986 0.674 25.76 0.14
56522.425 F160W H 3.93 0.645 26.01 0.16
SN GND12Col (Colfax)
56069.98 F606W V −0.269 0.285 >27.7
56016.29 F814W I 0.192 0.24 >27.9
56074.61 F814W I 0.003 0.317 >27.6
56116.49 F814W I −0.071 0.096 >28.9
56128.42 F814W I −0.207 0.108 >28.7
56180.92 F814W I 0.166 0.154 >28.3
56186.47 F814W I −0.191 0.119 >28.6
56200.16 F814W I −0.105 0.129 >28.5
56152.07 F850LP Z 0.043 0.323 >27.5
56018.628 F125W J −0.016 0.226 >27.9
56074.305 F125W J 3.877 0.464 26.03 0.12
56084.685 F125W J 3.609 0.479 26.11 0.13
56103.172 F125W J 2.818 0.488 26.38 0.17
56129.454 F125W J 0.858 0.319 >27.5
56145.532 F125W J 0.715 0.333 >27.5
56183.129 F125W J −0.011 0.228 >27.9
56241.974 F125W J 0.0 0.246 >27.8
56297.69 F125W J 0.045 0.239 >27.9
56084.763 F127M O 5.363 0.623 25.68 0.12
56082.777 F139M P 5.61 0.852 25.63 0.15
56083.095 F140W N 5.085 0.485 25.73 0.1
56091.533 F140W N 4.863 0.748 25.78 0.15
56104.231 F140W N 3.548 0.785 26.12 0.22
56085.627 F153M Q 4.825 0.891 25.79 0.18
56018.635 F160W H −0.048 0.763 >26.6
56074.312 F160W H 5.735 0.723 25.6 0.13
56084.645 F160W H 4.242 0.647 25.93 0.15
56103.164 F160W H 3.769 0.649 26.06 0.17
56129.379 F160W H 1.694 0.515 26.93 0.3
56145.47 F160W H 2.391 0.628 26.55 0.26
56183.136 F160W H 1.177 0.623 >26.8
56241.981 F160W H 0.278 1.21 >26.1
56297.697 F160W H −0.098 0.922 >26.4
SN GND13Cam (Camille)
56356.28 F814W I 5.791 0.456 25.59 0.08
56391.79 F814W I 0.558 0.353 >27.4
56356.43 F814W I 8.852 1.719 25.13 0.21
56236.854 F125W J 0.0 0.119 >28.6
56296.624 F125W J 0.507 0.208 >28.0
56346.757 F125W J 4.808 0.302 25.8 0.06
56369.646 F125W J 6.461 0.32 25.47 0.05
56382.019 F125W J 5.118 0.298 25.73 0.06
56406.455 F125W J 2.898 0.247 26.34 0.09
56425.556 F125W J 1.334 0.284 27.19 0.21
56459.362 F125W J 0.604 0.234 >27.9
56514.194 F125W J 0.746 0.263 >27.8
56356.545 F140W N 6.401 0.524 25.48 0.08
56236.861 F160W H 0.243 0.286 >27.7
56296.633 F160W H 0.437 0.324 >27.5
56346.764 F160W H 4.367 0.417 25.9 0.1
56369.638 F160W H 4.734 0.39 25.81 0.08
56382.011 F160W H 3.097 0.371 26.27 0.12
56406.462 F160W H 2.323 0.336 26.59 0.14
56425.548 F160W H 2.295 0.37 26.6 0.16
56459.369 F160W H 1.082 0.359 27.41 0.33
56514.201 F160W H 0.416 0.307 >27.6
SN GND13Gar (Garner)
56577.2 F814W I 0.739 0.44 >27.2
56446.46 F850LP Z 5.836 0.844 25.58 0.16
56347.887 F125W J 0.177 0.304 >27.6
56406.654 F125W J 11.106 0.523 24.89 0.05
56425.671 F125W J 16.562 0.627 24.45 0.04
56459.229 F125W J 7.356 0.499 25.33 0.07
56467.226 F125W J 8.3 0.605 25.2 0.07
56511.203 F125W J 3.154 0.499 26.25 0.16
56347.895 F160W H −0.058 0.888 >26.4
56406.661 F160W H 6.989 0.717 25.39 0.1
56425.664 F160W H 8.138 0.73 25.22 0.09
56459.236 F160W H 5.696 0.656 25.61 0.12
56467.22 F160W H 6.386 0.846 25.49 0.13
56511.21 F160W H 1.843 0.625 >26.8
SN GND13Jay (Jay)
56446.46 F850LP Z 5.459 1.013 25.66 0.2
56391.73 F814W I 1.39 0.222 27.14 0.17
56577.2 F814W I 0.305 0.458 >27.2
56347.887 F125W J 0.138 0.267 >27.7
56406.654 F125W J 12.196 0.555 24.78 0.05
56425.671 F125W J 11.243 0.576 24.87 0.05
56459.229 F125W J 5.738 0.484 25.6 0.08
56467.226 F125W J 5.179 0.538 25.71 0.1
56511.203 F125W J 2.185 0.469 26.65 0.21
56347.895 F160W H 0.0 0.669 >26.7
56406.661 F160W H 9.322 0.748 25.08 0.08
56425.664 F160W H 6.675 0.756 25.44 0.11
56459.236 F160W H 5.514 0.609 25.65 0.11
56467.22 F160W H 5.071 0.724 25.74 0.14
56511.21 F160W H 2.77 0.641 26.39 0.23
SN GND13Sto (Stone)
56458.33 F814W I 0.632 0.366 >27.4
56403.398 F125W J 0.785 0.372 >27.4
56457.501 F125W J 3.587 0.504 26.11 0.14
56467.888 F125W J 6.815 0.52 25.42 0.08
56490.59 F125W J 7.489 0.614 25.31 0.08
56513.396 F125W J 4.06 0.532 25.98 0.13
56531.078 F125W J 1.599 0.614 >26.8
56474.38 F139M P 6.79 1.279 25.42 0.19
56474.51 F140W N 7.239 0.817 25.35 0.11
56474.912 F153M Q 6.733 1.018 25.43 0.15
56403.405 F160W H 0.685 0.625 >26.8
56457.508 F160W H 4.431 0.674 25.88 0.15
56467.896 F160W H 5.174 0.646 25.72 0.12
56474.519 F160W H 8.116 0.628 25.23 0.08
56490.584 F160W H 6.292 0.723 25.5 0.11
56513.404 F160W H 5.539 0.728 25.64 0.13
56531.057 F160W H 3.829 0.681 26.04 0.18
56551.177 F160W H 2.981 0.534 26.31 0.18
SN GSD10Pri (Primo)
55607.39 F814W I −0.148 0.911 >26.4
55620.78 F814W I 0.054 0.367 >27.4
55626.45 F814W I −0.393 0.36 >27.4
55647.07 F814W I −0.438 0.262 >27.8
55584.58 F850LP Z −1.143 0.534 >27.0
55598.62 F850LP Z 1.54 0.597 >26.9
55620.18 F850LP Z 0.147 0.495 >27.1
55412.131 F125W J 0.138 0.069 >29.2
55417.79 F125W J 0.027 0.063 >29.3
55423.131 F125W J 0.17 0.114 >28.7
55479.905 F125W J 9.27 0.627 25.08 0.07
55501.266 F125W J 8.604 0.879 25.16 0.1
55507.662 F125W J 6.66 0.423 25.44 0.06
55521.198 F125W J 4.387 0.421 25.89 0.1
55528.374 F125W J 4.437 0.546 25.88 0.12
55535.157 F125W J 2.875 0.33 26.35 0.11
55543.739 F125W J 1.891 0.243 26.81 0.13
55578.695 F125W J 1.205 0.424 >27.2
55625.62 F125W J 0.909 0.448 >27.2
55718.133 F125W J −0.014 0.19 >28.1
55818.755 F125W J 0.162 0.195 >28.1
55974.319 F125W J 0.029 0.244 >27.8
55415.859 F160W H 0.0 0.098 >28.8
55416.858 F160W H 0.0 0.071 >29.2
55424.047 F160W H 0.156 0.115 >28.7
55438.906 F160W H 0.198 0.111 >28.7
55453.798 F160W H 2.024 0.256 26.73 0.13
55479.899 F160W H 8.43 0.707 25.19 0.08
55494.169 F160W H 6.721 0.59 25.43 0.09
55507.679 F160W H 6.12 0.601 25.53 0.1
55522.904 F160W H 5.249 0.805 25.7 0.15
55528.367 F160W H 4.76 0.752 25.81 0.16
55535.291 F160W H 3.281 0.464 26.21 0.14
55578.687 F160W H 1.477 0.55 >27.0
55625.613 F160W H 0.474 0.663 >26.8
SN GSD11Was (Washington)
55621.05 F814W I 8.785 0.392 25.14 0.05
55647.34 F814W I 3.344 0.287 26.19 0.09
55656.4 F814W I 2.276 0.373 26.61 0.18
55576.43 F125W J −0.013 0.343 >27.5
55625.085 F125W J 12.234 0.616 24.78 0.05
55635.536 F125W J 10.003 0.515 25.0 0.05
55647.127 F125W J 7.605 0.572 25.3 0.08
55656.45 F125W J 6.862 0.634 25.41 0.09
55720.129 F125W J 1.122 0.552 >27.0
55770.903 F125W J 1.783 0.506 26.87 0.28
55816.692 F125W J 0.037 0.62 >26.8
55576.422 F160W H −0.061 0.832 >26.5
55625.077 F160W H 11.116 0.753 24.89 0.07
55635.338 F160W H 7.612 0.717 25.3 0.09
55647.119 F160W H 7.493 0.828 25.31 0.11
55656.442 F160W H 4.581 0.694 25.85 0.15
55720.122 F160W H 3.18 0.775 26.24 0.24
55770.91 F160W H 1.361 0.679 >26.7
55816.699 F160W H 0.0 0.737 >26.6
SN UDS10Wil (Wilson)
55558.33 F814W I −0.146 0.176 >28.2
55574.65 F850LP Z −0.019 0.502 >27.1
55511.201 F125W J 2.193 0.511 26.65 0.23
55560.786 F125W J 5.182 0.565 25.71 0.11
55573.221 F125W J 2.806 0.399 26.38 0.14
55584.405 F125W J 1.831 0.364 26.84 0.2
55596.135 F125W J 1.451 0.368 27.1 0.25
55608.24 F125W J 0.698 0.297 >27.6
55511.181 F160W H 2.724 0.566 26.41 0.21
55560.721 F160W H 5.901 0.652 25.57 0.11
55572.954 F160W H 4.729 0.592 25.81 0.13
55584.343 F160W H 4.19 0.545 25.94 0.13
55596.06 F160W H 2.964 0.566 26.32 0.19
55608.106 F160W H 1.419 0.498 >27.1

Notes.

aFlux density in FLUXCAL units used by SNANA, which are normalized to a zeropoint of 27.5. bMagnitudes are on the AB photometric system. For epochs where the measured flux is $\lt 3\sigma $ significance, we report the magnitude as a 3σ upper limit.

Download table as:  ASCIITypeset images: 1 2 3 4 5

Footnotes

  • 24 

    STARDUST: Supernova Taxonomy And Redshift Determination Using SNANA Templates.

  • 25 

    Note that the Pantheon compilation as defined in Scolnic et al. (2017) includes the MCT SNe presented here.

  • 26 

    Proving this analytically involves calculating the inverse of a symmetric tridiagonal matrix (the covariance matrix describing the individual noisy estimates of the slope). We were able to verify analytically for the simplified case of a fit over a bin with 3 SNe with equal distance uncertainties.

  • 27 

    It is difficult to estimate the masses of high-redshift hosts in a way that is consistent with lower-redshift estimates. Therefore, in the analysis, the MCT SNe were assumed to be in low-mass hosts but near the host-mass cut such that a systematic uncertainty estimated by shifting the mass cut effectively marginalizes over their unknown host masses.

  • 28 

    In the Pantheon analysis, two alternative models for the intrinsic scatter are separately used to derive distance bias corrections, which are then averaged, with half of the difference included in the systematic uncertainty budget.

  • 29 

    Separate from standard propagation of redshift uncertainty, the derived distance moduli themselves depend on the observed redshift. We have verified that, for SN GND12Col, which has a large redshift uncertainty with asymmetric errors, repeating the analysis with both its redshift and distance shifted by 1σ does not significantly affect the results.

  • 30 

    In the Pantheon analysis, additional cuts were applied to remove SNe without an observation at least 5 days after peak brightness and with light-curve parameters that do not fall in the simulated distribution from the BBC method (see Scolnic et al. 2017). These cuts do not remove any of the remaining MCT SNe.

Please wait… references are loading.
10.3847/1538-4357/aaa5a9