Precise Time Delays from Strongly Gravitationally Lensed Type Ia Supernovae with Chromatically Microlensed Images

In the present analysis, we use the same glSN Ia population model as that in Goldstein & Nugent (2017). We consider only elliptical galaxy lenses and model their mass distribution as a Singular Isothermal Ellipsoid (SIE; Kormann et al. 1994), which has shown excellent agreement with observations (e.g., Koopmans et al. 2009). The SIE convergence κ is given by

$$\begin{eqnarray}&&\kappa (x,y)=\displaystyle \frac{{\theta }_{E}}{2}\displaystyle \frac{\lambda (e)}{\sqrt{{(1-e)}^{-1}{x}^{2}+(1-e){y}^{2}}},\end{eqnarray} \tag{ 1 }$$

where

$$\begin{eqnarray}&&{\theta }_{E}=4\pi {\left(\displaystyle \frac{\sigma }{c}\right)}^{2}\displaystyle \frac{{D}_{\mathrm{ls}}}{{D}_{s}}.\end{eqnarray} \tag{ 2 }$$

In the above equations, σ is the velocity dispersion of the lens galaxy, e is its ellipticity, and λ(e) is its so-called "dynamical normalization," a parameter related to three-dimensional shape. Here we make the simplifying assumption that there are an equal number of oblate and prolate galaxies, which Chae (2003) showed implies λ(e) ≃ 1. As in Oguri et al. (2008), we assume that e follows a truncated normal distribution on the interval [0.0, 0.9], with μ_e = 0.3, σ_e = 0.16.

We also include external shear to account for the effect of the lens environment (e.g., Kochanek 1991; Keeton et al. 1997; Witt & Mao 1997). We assume ${\mathrm{log}}_{10}{\gamma }_{\mathrm{ext}}$ follows a normal distribution with mean −1.30 and scale 0.2, consistent with the level of external shear expected from ray tracing in N-body simulations (Holder & Schechter 2003). The orientation of the external shear is assumed to be random.

We model the velocity distribution of elliptical galaxies as a modified Schechter function (Sheth et al. 2003):

$$\begin{eqnarray}&&{dn}\,=\,\phi (\sigma )d\sigma ={\phi }_{* }{\left(\displaystyle \frac{\sigma }{{\sigma }_{* }}\right)}^{\alpha }\exp \left[-{\left(\displaystyle \frac{\sigma }{{\sigma }_{* }}\right)}^{\beta }\right]\displaystyle \frac{\beta }{{\rm{\Gamma }}(\alpha /\beta )}\displaystyle \frac{d\sigma }{\sigma },\end{eqnarray} \tag{ 3 }$$

where Γ is the gamma function, and dn is the differential number of galaxies per unit velocity dispersion per unit comoving volume. We use the parameter values of Choi et al. (2007) from the Sloan Digital Sky Survey (SDSS; York et al. 2000): (ϕ_*, σ_*, α, β) = (8 × 10⁻³ h³ Mpc⁻³, 161 $\mathrm{km}\,{{\rm{s}}}^{-1}$, 2.32, 2.67). We assume that the mass distribution and velocity function do not evolve with redshift, which is consistent with the results of Chae (2007), Oguri et al. (2008), and Bezanson et al. (2011).

To convert Equation (3) into a redshift distribution, we use the definition of the comoving volume element:

$$\begin{eqnarray}&&{{dV}}_{C}={D}_{H}\displaystyle \frac{{(1+z)}^{2}{D}_{A}^{2}}{E(z)}\,{dzd}{\rm{\Omega }},\end{eqnarray} \tag{ 4 }$$

where ${D}_{H}=c/{H}_{0}$ is the Hubble distance, $E(z)\,=\sqrt{{{\rm{\Omega }}}_{M}{(1+z)}^{3}+{{\rm{\Omega }}}_{{\rm{\Lambda }}}}$ in our assumed cosmology, and D_A is the angular diameter distance. Since ${dn}\,=\,{dN}/{{dV}}_{C}$, for the unnormalized all-sky $(d{\rm{\Omega }}=4\pi )$ galaxy distribution, we have

$$\begin{eqnarray}&&\displaystyle \frac{{dN}}{d\sigma {dz}}=4\pi {D}_{H}\displaystyle \frac{{(1+z)}^{2}{D}_{A}^{2}}{E(z)}\phi (\sigma ).\end{eqnarray} \tag{ 5 }$$

Integrating Equation (3) over 0 ≤ z ≤ 1 and ${10}^{1.7}\,\mathrm{km}\,{{\rm{s}}}^{-1}\,\leqslant \sigma \leqslant {10}^{2.6}\,\mathrm{km}\,{{\rm{s}}}^{-1}$, we find that there are ${N}_{\mathrm{gal}}\simeq 3.8\times {10}^{8}$ elliptical galaxies, all sky, that can act as strong lenses. This gives the joint probability density function for σ and z:

$$\begin{eqnarray}&&p(\sigma ,z)=\displaystyle \frac{1}{{N}_{\mathrm{gal}}}\displaystyle \frac{{dN}}{d\sigma {dz}}.\end{eqnarray} \tag{ 6 }$$

SNe Ia exhibit a redshift-dependent volumetric rate and an intrinsic dispersion in rest-frame M_B. In our model of the SN Ia population, we take the redshift-dependent SN Ia rate from Sullivan et al. (2000). We assume that the peak rest-frame M_B is normally distributed with μ_M = −19.3 and σ_M = 0.2. For simplicity, we neglect extinction.

The lens and source populations are realized in a Monte Carlo simulation. We generate 10⁵ lens galaxies with parameters drawn at random from their underlying distributions. For each lens galaxy, an effective lensing area of influence is estimated as a ${[8{\theta }_{E,{z}_{s}=\infty }]}^{2}$ box centered on the galaxy.⁷ We simulate 5 × 10⁴ years of SNe Ia, randomly distributed across the box, rejecting systems where ${z}_{s}\lt {z}_{l}$. For each remaining source, we solve the lens equation using glafic (Oguri 2010) to determine the macrolensing magnification, image multiplicity, and time delays. In total, we generated 37,100 multiply imaged systems containing a total of 78,184 images. Since our simulation only covers ${10}^{5}/{N}_{\mathrm{gal}}\approx 0.026 \%$ of the sky, this corresponds to a rate of 2675 systems, all sky, per year to z_s = 2.

For each image, we calculate a source-plane magnification pattern with microlens (Wambsganss 1990, 1999), an inverse ray-tracing code. In this scheme, stars modeled as point-mass deflectors are realized from a mass function at random locations in a two-dimensional field of the lens galaxy. The size of the pattern is characterized by the Einstein radius ${\bar{R}}_{E}$ of a deflector of mass $\bar{m}$ projected onto the source plane,

$$\begin{eqnarray}&&{\bar{R}}_{E}=\sqrt{\displaystyle \frac{4G\bar{m}}{{c}^{2}}\displaystyle \frac{{D}_{\mathrm{ls}}{D}_{s}}{{D}_{l}}},\end{eqnarray} \tag{ 7 }$$

where D_l is the angular diameter distance to the lens, D_s is the angular diameter distance to the source, and D_ls is the angular diameter distance between the lens and the source. Our magnification patterns are $10{\bar{R}}_{E}$ on a side, which for typical source and lens redshifts (${z}_{s}=1.2,{z}_{l}=0.6$) corresponds to an angular scale of $10{\bar{R}}_{E}/{D}_{s}\approx 1.5\times {10}^{-5}$ arcsec. At the same redshifts, ${\bar{R}}_{E}\approx 2.7\times {10}^{3}\,\mathrm{au}$, which is roughly 5 times larger than the extent of the supernova model near peak brightness.

The magnification patterns are specified by four parameters: (1) the local convergence, κ, (2) the local shear, γ, including the contributions of both the SIE and external potentials, (3) the fraction of the local convergence in stars, f_*,⁸ and (4) the dynamic range of the stellar mass function, $q={m}_{\min }/{m}_{\max }$. Supplying these parameters allows microlens to solve the general microlensing equation, ${\boldsymbol{\beta }}={\boldsymbol{\theta }}-{\boldsymbol{\alpha }}$, which resolves to

$$\begin{eqnarray}{\boldsymbol{\beta }}=\left(\begin{array}{cc}1-\gamma -{\kappa }_{c} & 0\\ 0 & 1+\gamma -{\kappa }_{c}\end{array}\right){\boldsymbol{\theta }}-\displaystyle \sum _{i=1}^{{N}_{* }}\displaystyle \frac{{M}_{i}({\boldsymbol{\theta }}-{{\boldsymbol{\theta }}}_{i})}{{({\boldsymbol{\theta }}-{{\boldsymbol{\theta }}}_{i})}^{2}},\end{eqnarray} \tag{ 8 }$$

where ${\kappa }_{c}=(1-{f}_{* })\kappa$ is the local convergence in continuously distributed matter, the two-dimensional vector ${\boldsymbol{\beta }}$ is the angular position of the source in the absence of lensing, ${\boldsymbol{\theta }}$ is the angular position of the observed macroimage, ${{\boldsymbol{\theta }}}_{i}$ is the angular position of the i'th star, and N_* is the number of stars in the field, determined from the local convergence in stars ${\kappa }_{* }={f}_{* }\kappa$ using the procedure of Schneider & Weiss (1987).

We use a Salpeter (1955) mass function, ${dn}/{dm}\propto {m}^{-2.35}$, to model the population of stars in our microlensing calculations. As we will show in Section 3, this choice has no effect on our results as the achromaticity of glSN Ia microlensing is driven entirely by the color evolution of SNe Ia and not by the properties of the microlensing magnification patterns. Following Dobler & Keeton (2006), we take the mean mass $\bar{m}=1\,{M}_{\odot }$. The microlensing parameters κ and γ are determined by evaluating the SIE and external shear lensing potentials at the location of each image. f_* is estimated following the method of Dobler & Keeton (2006), assuming a de Vaucouleurs stellar profile normalized so that the maximum ${f}_{* }=1$. For each image, the dynamic range parameter q is sampled uniformly at random from (0.1, 0.2, 0.5, 1.0), appropriate for the old stellar populations in elliptical galaxies. Figure 1 shows a random selection of nine of the maps, highlighting their morphological diversity. The detailed parameters of each map are given in Table 1. Figure 2 shows two-dimensional projections of the joint distributions of the macrolensing parameters σ, e, and ${\gamma }_{\mathrm{ext}}$, the microlensing parameters κ, γ, and f_*, and the time- and wavelength-averaged microlensing magnification ${\mu }_{\mathrm{ML}}$.

Table 1.  Properties of the Magnification Patterns in Figure 1

	κ	f*	γ	q	$\langle \mu \rangle$
a	1.30	0.84	1.30	0.20	0.63
b	0.24	0.35	0.25	0.10	1.93
c	1.25	0.83	1.26	0.20	0.65
d	0.87	0.72	0.93	1.00	1.19
e	0.75	0.80	0.76	1.00	1.96
f	0.30	0.38	0.27	1.00	2.39
g	0.36	0.49	0.33	1.00	3.30
h	0.84	0.72	0.79	0.10	1.66
i	0.37	0.50	0.31	1.00	3.29

Note. κ: local convergence. f_*: fraction of surface density in stars. γ: local shear. q: mass ratio ${m}_{\min }/{m}_{\max }$ of the stellar mass function. $\langle \mu \rangle$: magnification of the field from a smooth mass model without microlensing.

Download table as:  ASCIITypeset image

Zoom In Zoom Out Reset image size

We use the well-understood, spherically symmetric⁹ SN Ia atmosphere model W7 (Nomoto et al. 1984) to estimate the time- and wavelength-dependent magnification of glSNe Ia due to microlensing. This radiation transport model is the result of a one-dimensional explosion simulation in which a Chandrasekhar-mass carbon–oxygen white dwarf undergoes a deflagration. The explosion of the white dwarf completely unbinds the star and deposits the energy liberated by nuclear burning into the ejected mass. The deposited energy controls the velocity distribution of the ejecta and its density profile, which is assumed to reach homology seconds after the explosion. We use the time-dependent, Monte Carlo radiation transport code SEDONA (Kasen et al. 2006) to calculate the spectral time series of the model. Details of our SEDONA simulations appear in Appendix A.

The observed spectrum F_λ of the model at wavelength λ and time t is obtained by convolving its time-evolving specific intensity with the lensing amplification pattern over a plane normal to the observer's line of sight. Since the model is spherical, this integral takes the following form:

$$\begin{eqnarray}&&{F}_{\lambda }(\lambda ,t)={D}_{L}^{-2}{\int }_{0}^{2\pi }\,{\int }_{0}^{{P}_{m}}{I}_{\lambda }(P,\phi ,\lambda ,t)\mu (P,\phi )P\,{dP}\,d\phi ,\end{eqnarray} \tag{ 9 }$$

where ϕ and P are azimuthal and impact parameter coordinates on the plane, I_λ is the specific intensity of the model, μ is the lensing amplification,¹⁰ D_L is the luminosity distance to the supernova, and P_m is the maximum impact parameter of the model. For a derivation of Equation (9), see Appendix B. The time- and wavelength-dependent magnification of a given magnification pattern is obtained by dividing F_λ by the unlensed spectrum of the model (Equation (9) with μ = 1).

We interpolate each magnification pattern bilinearly and convolve it with the redshifted¹¹ specific intensities of the supernova model. The redshift configurations control the projected size of the supernova on the magnification pattern and thus the magnification experienced by each differential element of the projected supernova atmosphere. They also control θ_E and thus κ, γ, and f_* at the location of the image. We always place the supernova model at the center of the magnification pattern. We model the homologous expansion of the supernova behind the magnification pattern (i.e., the projected size of the supernova on the magnification pattern changes with time), but not relative motion between the supernova and the lens galaxy star field. In general, supernova atmospheres both expand and move with respect to the lens galaxy, but the characteristic expansion velocity of the atmosphere ($\sim {10}^{4}\,\mathrm{km}\,{{\rm{s}}}^{-1}$) is much larger than the characteristic relative velocity between the lens galaxy and the supernova ($\sim {10}^{2}\,\mathrm{km}\,{{\rm{s}}}^{-1}$), so here we model only the effects of expansion.

What is the physics responsible for the "achromatic" and "chromatic" phases of microlensing evident in Figures 4 and 5? Figure 6 shows the specific intensity profile ${I}_{\lambda }(v)$ of our unlensed model at 20 and 40 days after explosion, where v is the velocity of the shell, equivalent to a radial variable (i.e., P).¹² The left panel of the figure shows that near peak (20 days after explosion), the ratio ${I}_{{X}_{1}}(v)/{I}_{{X}_{2}}(v)$, where X₁ and X₂ are any two bands, is roughly constant over all v. Thus the supernova near peak has a specific intensity profile that is independent of v up to an overall normalization factor. As a result, any magnification pattern μ(P, ϕ) will not change its color.

Zoom In Zoom Out Reset image size

However, after peak, the supernova expands and cools enough for some part of the atmosphere to reach a temperature of 7000 K. Kasen (2006) and Kasen & Woosley (2007) note that this is the temperature at which Fe iii recombines to Fe ii, which presents a significantly higher opacity to blue and ultraviolet radiation than Fe iii. This line blanketing has the effect of enabling one to see redder emission from deeper in the supernova, while emission in the blue and the UV is pushed to larger radii. Additionally, a "fluorescent shell" of iron recombination, which causes a peak in the redder bands in the specific intensity profile of the supernova, develops near the onset of the secondary maximum. This shell is clearly visible in the righthand panel of Figure 6. These two effects, line blanketing and a fluorescent shell, make the supernova's specific intensity ratio no longer spatially constant. As a result, the supernova atmosphere becomes susceptible to chromatic fluctuations.

Our detection strategy rests on three observational facts. First, normal SNe Ia are the brightest type of supernovae that have ever been observed to occur in elliptical galaxies (Maoz et al. 2014). Second, the absolute magnitudes of normal SNe Ia in elliptical galaxies are remarkably homogenous, even without correcting for their colors or light-curve shapes $({\sigma }_{M}\sim 0.4\ \mathrm{mag})$, with a component of the population being underluminous (Li et al. 2011). Finally, due to the sharp 4000 Å break in their spectra, elliptical galaxies tend to provide accurate photometric redshifts from large-scale multicolor galaxy surveys such as SDSS.

A high-cadence, wide-field imaging survey can leverage these facts to systematically search for glSNe Ia in the following way. First, by spatially cross-matching its list of supernova candidates with a catalog of elliptical galaxies for which secure photometric redshifts have been obtained, supernovae that appear to be hosted by elliptical galaxies can be identified. The hypothesis that one of these supernovae actually resides in its apparent host can be tested by fitting its broadband light curves with an SN Ia spectral template (as SNe Ia are the only types of supernovae that occur in ellipticals) fixed to the photometric redshift of the galaxy and constrained to obey $-18.5\gt {M}_{B}\gt -20$, a liberal absolute magnitude range for SNe Ia, assuming a fiducial cosmology. If the transient is a lensed supernova at higher redshift, then the spectral template fit will fail catastrophically, as the supernova light curves will be strongly inconsistent with the redshift and brightness implied by the lens galaxy.

We use SALT2 (Guy et al. 2007), a parametrized SN Ia spectral template that is the de facto standard tool to place SNe Ia on the Hubble diagram, to test this method. The template possesses four parameters: t₀, x₀, x₁, and c, encoding a reference time, an overall SED normalization, a supernova "stretch," and a color-law coefficient, respectively. The flux of the template is given by

$$\begin{eqnarray}&&{F}_{\lambda }(\lambda ,t)={x}_{0}[{M}_{0}(\lambda ,t)+{x}_{1}{M}_{1}(\lambda ,t)]\exp [c\,\mathrm{CL}(\lambda )],\end{eqnarray} \tag{ 11 }$$

where M₀ and M₁ are eigenspectra derived from a training sample of measured SN Ia spectra and $\mathrm{CL}(\lambda )$ is the average color-correction law of the sample (see Guy et al. 2007, for details). The template aims to model the mean evolution of the SED sequence of SNe Ia and its variation with a few dominant components, including a time independent variation with color, whether it is intrinsic or due to extinction by dust in the host galaxy (or both).

We consider an LSST lensed supernova search in which the photometry is performed with a PSF that is artifically enlarged to blend the multiple images together into a single source. We randomly assign each of the 37,100 simulated glSNe Ia from Section 2.1 an LSST field from the nominal observing strategy (minion_1016; LSST Science Collaborations 2018, in preparation).¹³ We compute the phase- and wavelength- dependent magnification μ(λ, t) of each lensed image by placing its corresponding microlensed W7 SED into the rest frame, then dividing each by the unlensed spectral sequence of the model. We then generate the rest-frame spectral model for the image F(λ, t) according to

$$\begin{eqnarray}&&F(\lambda ,t)=\mu (\lambda ,t)H(\lambda ,t),\end{eqnarray} \tag{ 12 }$$

where H(λ, t) is the Hsiao et al. (2007) SN Ia spectral template. We employ a warped Hsiao template rather than the microlensed W7 SEDs to mitigate uncertainties in the radiation transport.¹⁴ We then place the templates of each image at their correct redshifts, and we rescale and time-shift them to account for the macrolensing and time delays. Finally, we draw a random time for the system (arbitrarily chosen to be the observer-frame time of rest-frame B-band maximum of the first image) from the 11 year period spanning 6 months before the beginning of the survey until 6 months following the end of the survey.

We realize broadband photometry of each blended glSN Ia (summing the flux of each microlensed image) using the sky brightnesses, FWHMs, exposure times, observation times, and limiting magnitudes of the assigned field provided by minion_1016, assuming the total area covered by the survey is 25,000 deg². Starting from the first observation of the SN Ia, we fit the light curve with SALT2, fixed to the redshift of the lens galaxy (assumed to be known either as a photometric or spectroscopic redshift) and fixed to obey $-18.5\gt {M}_{B}\gt -20$ at that redshift (effectively a constraint on x₀). Additionally, we enforce bounds of [−0.2, 0.2] on c and [−1, 1] on x₁, values characteristic of normal SNe Ia (Scalzo et al. 2014). We use the CERN minimization routine MIGRAD (James & Roos 1975) to fit the data. If the light curve has at least one data point that is at least 5σ discrepant from the best fit and at least four data points with S/N ≥ 5, then the object is marked "detected." If not, then the next observation is added and the process is repeated until the object is detected or all observations are added, resulting in a nondetection.

Figure 7 shows an example of this procedure being used to detect one of our simulated glSNe Ia at z_s = 1.91. The red data points show the "current" light curve, and the red line shows the best-fit model, fixed to z_l = 0.96. Although the model fits the data well in the bluer bands, the high redshift of the source makes the data much brighter in the infrared than the model expects given the redshift of the lens. Thus the object is detected shortly after peak due to 10σ discrepant points in y-band.

Zoom In Zoom Out Reset image size

Our spectral template fitting approach to glSN Ia identification delivers almost twice as many LSST glSNe Ia than the method of Goldstein & Nugent (2017). In total, LSST should find ∼925 microlensed glSNe Ia with the new method over the duration of its 10 year survey. This is almost identical to the case with no microlensing, which would yield 935 glSNe Ia over the same period, with a nearly identical redshift distribution (see Figure 9). This represents a major increase in the expected glSN Ia yield for LSST, comparable to the number of expected lensed quasars (Oguri & Marshall 2010).

Figure 8 shows the rest-frame phase distribution of discovered microlensed glSN Ia images (a phase of 0 corresponds to peak brightness in B). The 68% confidence interval of the image phase distribution is $-{1.01}_{-10.77}^{+10.24}$, so about half of the images should be discovered before peak brightness. 73% of the images and 64% of the image pairs should be discovered during the achromatic phase.

Zoom In Zoom Out Reset image size

Since many of the glSNe Ia that will be discovered by LSST will require higher-spatial resolution follow-up observations to extract time delays, either with ground-based adaptive optics or space-based imaging, we simulate "time-delay observations" with the Wide Field Camera 3 (WFC3) on the Hubble Space Telescope (HST). The redshift distribution in Figure 9 implies that most LSST glSNe Ia will be brightest in the IR, so we use the F814W, F125W, and F160W filters (roughly I, J, and H bands, respectively) on WFC3 for this simulation.

Zoom In Zoom Out Reset image size

For each of the "detected" microlensed glSN Ia systems from Section 4.3, we realize 45 photometric observations of each image in F814W, F125W, and F160W with infinite signal-to-noise, with a uniform temporal spacing, spanning the light curve. The spectral template for each image is the same as that used in Section 4 (see Equation (12)). Using MIGRAD, we fit the realized light curves of each pair of images using the redshifted, unlensed spectral template $H(\lambda ,t)$. For each fit, we estimate the fitted time delay as

$$\begin{eqnarray}&&{\rm{\Delta }}t={t}_{\mathrm{0,2}}-{t}_{\mathrm{0,1}},\end{eqnarray} \tag{ 13 }$$

where ${t}_{\mathrm{0,2}}$ and ${t}_{\mathrm{0,1}}$ are the fitted reference times of the second and first images, respectively. We then measure the error on the fitted time delay as

$$\begin{eqnarray}&&\epsilon =\left|\displaystyle \frac{{\rm{\Delta }}t-{\rm{\Delta }}t^{\prime} }{{\rm{\Delta }}t^{\prime} }\right|,\end{eqnarray} \tag{ 14 }$$

where ${\rm{\Delta }}t^{\prime}$ is the true time delay of the pair of images.¹⁵

Next, we prune the photometric observations to the achromatic phase, masking all observations more than five rest-frame days from the date of B-band maximum. We then synthesize F814W − F125W, F125W − F160W, and F814W − F160W color curves from the photometry.

We apply the same fitting procedure to the color curves and estimate the fitted time delays and uncertainties using Equations (13) and (14). Figure 10 shows an example of the procedure being applied to the light curves and achromatic-phase color curves of two images of a supernova. While the light curve data show microlensing-induced offsets near peak brightness and the secondary maximum that produce a large time delay error of ∼10%, the residuals show that these offsets are nearly the same in all bands during the achromatic phase, and thus the fits to the color curve have errors <1%.

Zoom In Zoom Out Reset image size

Figure 11 shows the joint distribution of time delays and microlensing-induced time delay uncertainties for all detected pairs of images in our simulation. We find that the median time delay of detected pairs of images is ∼10 days, and that the median microlensing-induced time delay uncertainty using light curve fits is 4%, comparable to the current uncertainties on mass modeling. However, this number drops down to 1% (∼2.5 hours on a 10 day time delay) when achromatic-phase color curves are used instead of light curves.

Zoom In Zoom Out Reset image size

Figure 12 shows the systematic microlensing bias on time delays from fitting light curves and color curves in the achromatic phase. The achromatic phase color curve fits are consistent with zero bias down to Δt = 1 day, while the light curve fits are consistent with zero bias down to Δt = ∼ a few days. This result indicates that time delay bias from microlensing will not be a major systematic for cosmography with glSNe Ia.

Zoom In Zoom Out Reset image size

iPTF16geu is the only glSN Ia with resolved images that has been discovered to date. Here we consider its potential time delay precision and cosmological impact. Before it faded, multiwavelength follow-up observations of the event were obtained with a variable cadence using the Washington C, g, r, wide I, Z, J, and H bands of WFC3 on HST (DD 14862, PI: Goobar). The bluer bands were only observed sporadically, but the redder bands were observed with a roughly 4-day cadence.¹⁶ Unfortunately, the observations were obtained well into the chromatic phase, and thus the light curves and color curves exhibit significant microlensing uncertainties. Final photometry of the event has not yet been produced, as not enough time has passed to take final reference images. The current best estimate of the time delay on this system is 35 hours (Goobar et al. 2017). Based on Figure 11, we estimate the time delay uncertainty due to microlensing from this event to be ∼40%. If the event were discovered earlier so that color curves during the achromatic phase could be constructed, then the microlensing time delay uncertainty would drop to ∼10%.