www.sciencemag.org/content/362/6411/201
/suppl/DC1
Supp
lementary
Material
s
for
A hot and fast ultra
-
stripped supernova that likely formed a compact
neutron star
binary
K.
De
,
*
M. M.
Kasliwal
,
E. O.
Ofek
,
T. J.
Moriya
,
J.
Burke
,
Y.
Cao
,
S. B.
Cenko
,
G. B.
Doran
,
G. E.
Duggan
,
R. P.
Fender
,
C.
Fransson
,
A.
Gal
-
Yam
,
A.
Horesh
,
S. R.
Kulkarni
,
R. R.
Laher
,
R.
Lunnan
,
I.
Manulis
,
F.
Masci
,
P. A.
Mazzali
,
P. E.
Nugent
,
D. A.
Perley
,
T.
Petrushevska
,
A. L.
Piro
,
C.
Rumsey
,
J.
Sollerman
,
M.
Sullivan
,
F.
Taddia
*
Corresponding author.
Email: kde@astro.caltech.edu
Published 12 October
20
1
8
,
Science
362
,
201
(20
1
8
)
DOI:
10.1126/science.
aas8693
This PDF file
includes:
Materials and Methods
Supplementary Text
Caption for Data S1
Figs. S1 to S15
Tables S1 to S7
Reference
s
Other Supp
lementary
Material
for this manuscript
include
s
the following
:
(available at
www.sciencemag.org/content/362/6411/201
/suppl/DC1
)
Data S1 (
.zip
)
Materials and Methods
Observations
iPTF Discovery
iPTF 14gqr (SN 2014ft) was discovered by the intermediate Palomar Transient Factory (iPTF;
(
11,53
)) in data taken with the CFH12K 96-Megapixel camera (
54,55
) mounted on the 48 inch
Samuel Oschin Telescope at Palomar Observatory (P48), on 2014 October 14.18 [Modified
Julian Date (MJD) 56944.18; Coordinated Universal Times are used throughout this paper].
The source was discovered at coordinates right ascension (
↵
)
=
23
h
33
m
27.95
s
, declination
(
)
=33
38
0
46
.
1
00
(J2000 equinox), while it was not detected on 2014 October 13.32 (MJD
56943.32; 0.86 days before discovery) up to a limiting magnitude of
g
21.5. We nominally
adopt the average MJD 56943.75
±
0.43 as the explosion date, and calculate all phases with
reference to this epoch. However, the actual explosion could have taken place before the last
non-detection depending on the (unknown) behavior of the early emission. Hence, we allow the
explosion time to vary as a free parameter in our modeling, and discuss the last non-detection
individually in the context of the physical models.
Optical light curves
We obtained
g
band photometry of iPTF 14gqr with the P48 CFH12K camera, along with ad-
ditional follow-up photometry in the
Bgri
bands with the automated 60-inch telescope at Palo-
mar Observatory (P60; (
56
)). Point spread function (PSF) photometry was performed on the
P48 images using the Palomar Transient Factory Image Differencing and Extraction (PTFIDE)
pipeline (
11
), while the P60 images were reduced using an automated pipeline (
57
). The pho-
tometric evolution of the source is shown in Figure 2 and the data are presented in Table S1.
The data have been corrected for Galactic extinction (
58
) for
E
(
B
V
)=0
.
082
mag and
A
V
=0
.
255
mag. We do not expect any additional host extinction owing to the remote location
33
of the transient, which is consistent with the absence of Na I D absorption lines in all our spectra.
Additional follow-up photometry in
BV gri
bands was obtained with the Las Cumbres Ob-
servatory (LCO) 1-meter telescope located at the McDonald Observatory (
59
). These data were
processed using the tools available in the
lcogtsnpipe
package followed by PSF photom-
etry (
60
). Owing to the faint peak magnitude of the source (close to the sensitivity limit of
LCO), the photometry obtained from these observations are relatively noisy due to their low
signal-to-noise ratio. Since P60 obtained contemporaneous observations with LCO with much
higher signal to noise ratio, we chose not to include the LCO data in our analysis. Nonetheless,
the contemporaneous photometry from LCO and P60 are completely consistent with each other,
and we present the LCO photometry in Table S1 for completeness.
Swift
UV / X-ray observations
We triggered
Swift
follow-up of the source in the
V
,
B
,
UV W
1
and
UV W
2
bands with the
Swift
Ultra-Violet/Optical Telescope (UVOT; (
61
)) and X-ray follow-up with the
Swift
X-ray
telescope (XRT; (
62
)). We processed the data with the
HEAsoft
package (
63
) and detected the
transient in the
UV W
1
and
UV W
2
bands in the first three and two epochs of observation re-
spectively, while only non-detections were obtained at subsequent epochs. The UV photometric
evolution is shown in Figure 2 along with the optical light curves, and the data are presented in
Table S1, where we used empirical fits (
64
) to compute extinction coefficients for the
UV W
1
and
UV W
2
bands. Only upper limits were obtained in all epochs of
Swift
XRT observations.
The corresponding flux limits are summarized in Table S2.
We also stacked all of the cleaned event files obtained over a period of 21.9 days, amounting
to a total exposure time of 12.2 ks. We obtain a 5
flux upper limit of 1.24
⇥
10
3
counts s
1
corresponding to a 0.3 - 10 keV unabsorbed flux limit of 3.5
⇥
10
14
ergs cm
2
s
1
(assuming
34
a photon index
= 2). The Galactic neutral hydrogen column density along this line of sight is
5.7
⇥
10
20
cm
2
(
65
), yielding a corresponding X-ray luminosity limit of 3.4
⇥
10
41
ergs s
1
for the source at a distance of
D
= 284.5 Mpc.
Optical spectroscopy
We obtained a sequence of spectroscopic observations of the source starting from 4.3 hours
after first detection to +59 days after
r
band peak using the Dual Imaging Spectrograph (DIS)
mounted on the 3.5 m Astrophysical Research Consortium telescope at Apache Point Observa-
tory (APO), the auxiliary port camera (ACAM; (
66
)) on the 4.2 m William Herschel Telescope
(WHT), Andalusia Faint Object Spectrograph and Camera (ALFOSC) on the Nordic Optical
Telescope (NOT), the Gemini Multi-Object Spectrograph (GMOS; (
67
)) on the Gemini North
(N) telescope, the Low Resolution Imaging Spectrograph (LRIS; (
68
)) on the Keck I telescope
and the DEep Imaging Multi-Object Spectrograph (DEIMOS; (
69
)) on the Keck II telescope.
All spectra were reduced using standard tasks in IRAF and IDL, including wavelength calibra-
tion using arc lamps and flux calibration using standard stars.
The sequence of spectra obtained are shown in Figure 3, and the times of the spectra are
shown as dashed vertical lines in Figure 2. The spectroscopic observations are summarized in
Table S3. We were unable to obtain a high signal-to-noise ratio (SNR) spectrum of the transient
at epochs beyond
⇡
30 days from light curve peak. We also obtained a spectrum of the apparent
host galaxy nucleus with APO DIS on 2014 October 14 (shown in Figure S1) which was found
to exhibit narrow emission lines of H
↵
,H
, [SII], [NII], [OII] and [OIII]. Additionally, we
obtained one spectrum of the transient location
⇠
800 days after the explosion as a part of a
spectroscopic mask observation and did not detect any nebular emission features at the source
location. All spectra are available via the WISeREP repository (
70
).
35
NIR imaging
We observed the field of iPTF 14gqr using the Wide Field Infrared Camera (WIRC; Wilson et
al. 2003) on the Palomar 200-inch telescope on the night of 2014 Oct 19 (UT). We obtained 21
images (1
⇥
60 s each) using the
J
filter, 25 images (2
⇥
20 s each) using the
K
s
filter, and 24
images (2
⇥
20 s each) using the
H
filter, for a total integration time of 21, 16.7, and 16 minutes
respectively in each band.
The data were processed using a custom reduction pipeline including flat-fielding and sky
subtraction as well as special filtering steps to remove artifacts associated with the replacement
detector in use at the time. The source is well-detected in all three filters in the final stacks.
We performed aperture photometry within IDL and obtain magnitudes of
J
=19
.
76
±
0
.
08
,
H
=19
.
58
±
0
.
12
, and
K
s
=19
.
05
±
0
.
15
.
Radio observations
We observed iPTF 14gqr with the Very Large Array (VLA) radio telescope on both 2014 Oc-
tober 15.4 and 2014 October 25.0. Each observation was performed using C-band (centered
at 6.1 GHz) and K-band (centered at 22 GHz) in the C configuration. The Wideband Interfer-
ometric Digital Architecture (WIDAR) correlator was used in continuum mode with a 2 GHz
bandwidth in C-band and a 8 GHz bandwidth in K-band. We analyzed the data with standard
AIPS routines, using 3C 48 as the flux calibrator and NVSS J234029+264157 as the phase
calibrator. Our observations resulted in null detections in both bands at each epoch. The obser-
vational limits are
11
.
6
microJanskys (
μ
Jy) and
11
.
7
μ
Jy at C-band and K-band [measured as
the
1
root-mean-squared (RMS) noise of the reduced image], respectively, on 2014 October
15. On 2014 October 25, the observational limits are
13
.
0
μ
Jy and
15
.
0
μ
Jy, respectively. An
additional limit was obtained using the Arcminute Microkelvin Imager (AMI; (
71
)) Large Array
telescope. The AMI observation was undertaken on 2014 October 14.7 at a central frequency
36
of 15 GHz. The reduction of the AMI observation was conducted using the fully-automated
calibration and imaging pipeline AMIsurvey (
72
) and resulted in null detection with a
1
RMS
of
58
μ
Jy.
Host imaging and spectroscopy
Late-time imaging
We undertook deep imaging of the source region in the
g
-band and
R
-band filters with LRIS
on 2015 June 13 (MJD 57186.5) for a total exposure time of 960 s and 840 s respectively. The
data were reduced and processed with standard image reduction procedures in
lpipe
(
73
). No
source was detected at the transient location down to a 3
AB magnitude of
R>
26.1 mag
and
g>
26.5 mag (without extinction correction). This constrains the presence of any stellar
association at the location of the transient to
M
R
>
11
.
4
mag and
M
g
>
11
.
1
mag. Late-time
images of the host galaxy region are shown in Figure S14.
Host environment spectroscopy
iPTF 14gqr was discovered in the outskirts of an extended spiral galaxy showing clear signs of
tidal interactions with nearby companions. Since none of the apparent extended sources near
the transient region had measured redshifts in the NASA/IPAC Extragalactic Database (NED),
we undertook one spectroscopic mask observation on 2016 November 28 (
⇡
800 days after
discovery) of the region with LRIS on Keck I in order to confirm the interaction scenario.
Additionally, this would allow us to ascertain whether any of the other nearby fainter galaxies
could have potentially hosted the transient (i.e. was at a similar redshift) and if the spiral host
galaxy itself was a part of a galaxy group or cluster.
We selected a total of 32 extended sources classified as galaxies (including the apparent
spiral host) in the Sloan Digital Sky Survey (SDSS) within 5.4
0
of the transient location (out of
a total of 254 objects) to place the slits on the spectroscopic mask, along with one slit at the
37
location of the transient. The selection of sources for the slit mask was prioritized based on
the projected distance of the galaxy from the transient. The spectra were reduced with standard
routines in IRAF. Details of the spectroscopic mask observation are given in Table S3, while
Table S4 lists the redshifts of the galaxies identified from the mask observation. The positions
of the galaxies whose redshifts could be determined are shown in Figure S2 as circles while all
other cataloged SDSS galaxies are marked by crosses.
The faintest source placed in the mask had a SDSS magnitude of
r
⇡
22.51 mag, while the
faintest mask source within 100 kpc (
⇡
81
00
) of the transient had
r
⇡
21.89 mag. Amongst the
galaxies whose redshifts could be determined, the faintest source had SDSS magnitude of
r
⇡
22.11 mag, while the same for galaxies within 100 kpc of the transient was
r
⇡
21.60 mag. Our
redshift identification was complete for all sources brighter than
r
⇡
20.29 mag within 100 kpc
of the transient, corresponding to an absolute magnitude of
M
r
⇡
17
.
0
at the source redshift.
Photometric evolution
Basic properties
We summarize the basic photometric properties of the light curve in the
g
,
r
and
i
bands (where
we had sufficient coverage) in Table S5. We find the peak magnitudes, time of maximum and
corresponding rise time (between assumed explosion time and peak of light curve) in each
filter by fitting a low order polynomial to the photometry near peak. We characterize the post-
peak light curve in terms of the light curve decline rate (magnitudes per day). The absence of
photometric data points beyond
⇠
10 days after peak does not allow us to estimate the more
commonly used quantity
m
15
, the drop in magnitude in 15 days after light curve peak. The
uncertainties on these parameters were estimated by Monte Carlo sampling of 1000 realizations
of the photometric data points using their associated uncertainties.
The observed peak absolute magnitudes are on the low end of the distribution of SN Ic
38
peak magnitudes reported in (
18
), where SNe Ic were found to have peak
h
M
R
i
=
18
.
3
±
0
.
6
(uncertainties are 1
), while they are within the typical peak magnitudes found in the sample
of (
20
) (who find
h
M
r
i
=
17
.
66
±
0
.
21
and
h
M
g
i
=
17
.
28
±
0
.
24
). The rise times of the
light curves are shorter in the bluer bands as typically observed in Type Ib/c SNe (
74
). When
compared to other Type Ic SNe, the
r
band light curve rise time for iPTF 14gqr falls on the
extreme low end of the distribution observed in Type Ic and Ic-BL SNe. For example, (
74
)
find
h
t
rise
i
of 11.5
±
0.5 days and 14.7
±
0.2 days for Type Ic and Ic-BL SNe respectively
based on light curve from the SDSS-II supernova survey. If the decline rates are converted to an
equivalent
m
15
, we find the estimated
m
15
(
⇡
2.5 mag in
g
band and
⇡
1.3 mag in
r
band)
to be much higher than that observed for all normal Type Ib/c SNe (
18,20
), but similar to those
observed in the fast Type I events SNe 2002bj (
75
), 2010X (
22
) and 2005ek (
21
).
Comparison with other sources
We compare the multi-color light curves of iPTF 14gqr to other relatively faint and fast evolving
Type I SNe from the literature in Figure S3. These include the core-collapse Type Ic SN 1994I
(
76
), the Ca-rich transient PTF 10iuv (
77
) and the rapidly evolving transients SNe 2002bj (
75
),
2010X (
22
) and 2005ek (
21
). We plot the photometric evolution normalized to peak magnitude
in the upper panel and on an absolute scale in the lower panel. Owing to the lack of observations
in identical filters, we are constrained to compare the light curve evolution of these transients
in corresponding pairs of
R
/
r
,
V
/
g
and
I
/
i
bands (which in general we refer to as
r
,
g
and
i
bands respectively). We first focus on a comparison of the second peak of the light curve of
iPTF 14gqr to that of these events.
Figure S3 shows that the light curve shape and timescales (for the second peak) of iPTF 14gqr
are broadly similar to several of the events in this sample. In general, the light curves are faster
than those of PTF 10iuv but slower than the very fast decay exhibited by SN 2005ek. Overall,
39
SN 1994I exhibits light curves most similar to that of iPTF 14gqr near peak light, although the
rise time to peak for iPTF 14gqr is shorter than that of SN 1994I. For SN 2005ek, the best upper
limit for the rise time was at
⇡
9 days before
R
band maximum while the same for SN 2002bj
was 7 days. On the other hand, iPTF 14gqr attained a peak absolute magnitude (see Table S5)
fainter than that of SN 2002bj and SN 1994I, but similar to that of SN 2005ek and SN 2010X.
We compare the color curves of these transients to that of iPTF 14gqr in Figure S4, in
corresponding pairs of
V
/
g
R
/
r
and
R
/
r
I
/
i
colors. The sharp color jump after the first
data point arises from the rapid blue first peak. Subsequently, iPTF 14gqr has one of the fastest
color
g
r
color evolution among this set of transients, displaying a progression from a very blue
transient (
g
r
⇡
0
.
4
mag) at early times to a relatively red transient (
g
r
⇡
0
.
5
) within
10 days of explosion. This fast reddening is indicative of rapid cooling of the ejecta since
the spectra at these phases are broadly consistent with featureless continua with weak broad
features. We conclude that the multi-color light curves of the main peak of iPTF 14gqr exhibit
several similarities (light curve shape and timescales) as well as unique differences (short rise
time) in this sample of transients.
The rapid first peak of the light curve is perhaps the most distinguishing feature of iPTF 14gqr
when compared to this sample of transients, and hence we compare this first peak to that
of other known SNe exhibiting double peaked light curves in Figure S5. These include the
Type Ic iPTF 15dtg (
78
), the GRB associated broad-lined Type Ic SN 2006aj (
79
), the Type Ib
SN 2008D (
80
) and the Type IIb SN 2011dh (
81
). We show the iPTF 14gqr light curve in
g
band
since it has the best coverage. As shown, iPTF 14gqr has one of the fastest rise times (
0
.
5
days, as constrained ) and decay rates of the first peak in this sample. The width of the first
peak of iPTF 14gqr is most similar to that of the Type Ic-BL SN 2006aj, while the peak absolute
magnitude is similar to that that of iPTF 15dtg.
40
Optical / UV SEDs
We performed blackbody fitting of the multi-color photometry at all epochs for which we had
photometric detections in 3 or more filters. In particular, we have two epochs with photometric
data from all optical / UV bands, and the resulting blackbody SEDs are shown in Figure 3.
The first epoch was within the first peak of the light curve (at
⇡
14 hours after explosion),
where the UV / optical photometry is consistent with a blackbody of temperature
>
30,000
K. For comparison, we plot a blackbody fit of a spectrum obtained within an hour before this
epoch of photometry. This spectrum is also well described by a blackbody consistent with the
photometric fit within the uncertainties. We also plot a blackbody fit to the second epoch where
we had complete multi-color coverage, which was near the main peak of the light curve (at
⇡
5.3
d after explosion). In this case, the black dashed line represents a blackbody fit to the optical
photometry only (
eff
>
4000
̊
A). The UV photometric points are found to be significantly
fainter than the optical blackbody fit at this epoch (with T
⇠
10000
K), which is indicative
of significant line blanketing at UV wavelengths (as expected from Fe group elements in the
ejecta). The NIR photometric magnitudes obtained near this epoch (
⇡
1
day earlier) are also
consistent with the optical blackbody fit.
Bolometric light curve
We construct a bolometric light curve of iPTF 14gqr using three methods. We first fit a Planck
blackbody function to the observed photometry at all epochs where we have detections in 3
or more filters to obtain a best-fitting blackbody and corresponding temperature, radius and
luminosity. The relatively featureless optical continua of the source allows us to obtain good
blackbody fits at all epochs where multi-color photometry is available. We constrain the black-
body luminosity before the first detection by calculating the fraction of total flux within
g
band
at first detection (
⇡
2.6%), and use it to estimate the luminosity upper limit with the latest pre-
41
discovery limit in
g
band. The resulting bolometric luminosity curve is shown in Figure 4 as
black filled circles. The corresponding radius and temperature evolution is also shown in Figure
4. Although the blackbody approximation is valid within the first peak (as indicated by the best
fitting SEDs), the fit overestimates the luminosity during the second peak of the light curve due
to line blanketing in the UV.
We therefore compute a pseduo-bolometric light curve of the source by performing trape-
zoidal integration of the
gri
photometric fluxes (from the lower wavelength edge of the
g
band
to the higher wavelength edge of the
i
band) at all available epochs. Our peak photospheric
spectra indicate the presence of additional flux at wavelengths
4000
̊
A (the lower edge of
the
g
band) as well as at
8000
̊
A (the upper edge of the
i
band), which are not accounted
for by a simple trapezoidal integration. Hence, we estimate the contribution of these fluxes by
integrating the optical spectra from 4000
̊
A to 8000
̊
A and comparing them to the integrated
value over 3000
̊
A to 10000
̊
A (the full range of our optical spectra). Performing this procedure
on multiple spectra around peak light, we find that the simple trapezoidal integration underesti-
mates the total optical flux by a factor ranging from 1.42 to 1.56 over multiple epochs. Since we
do not have simultaneous spectroscopy with all epochs of multi-band photometry, we choose
to scale the fluxes obtained from a trapezoidal integration by an average factor of 1.48, while
conservatively adding an uncertainty of 10% to account for the possible errors on this fraction.
The pseudo-bolometric light curve is shown in Figure 4, which is found to be consistently
smaller than the blackbody luminosity as expected. The true bolometric luminosity for the
second peak lies in between these two estimates, and is likely to be very close to the pseudo-
bolometric luminosity we estimate. Hence, we use the pseudo-bolometric luminosities for mod-
eling the properties of the second light curve peak. We also compute a
g
band luminosity
F
for comparison, and show it as black empty circles in Figure 4.
Since our observations did not capture the rise to the first peak, we can only constrain the
42
peak luminosity of this phase to be greater than
⇠
2
⇥
10
43
ergs s
1
(i.e. the luminosity near first
detection), while the corresponding temperature at the first peak to be higher than
⇠
32
,
000
K.
The second peak reaches a peak (pseudo-bolometric) luminosity of
⇠
2
.
2
⇥
10
42
ergs s
1
with
a peak temperature of
⇠
10,000 K. The total integrated energy emitted within the second peak
(starting from
⇡
2 days to
⇡
14 days after explosion) of the pseudo-bolometric light curve
is
⇠
1.8
⇥
10
48
ergs. The blackbody temperatures exhibit rapid cooling from
>
32
,
000
K to
⇠
10
,
000
K at initial phases (
<
2 days after explosion), followed by a slower cooling phase at
later epochs. Similarly, the blackbody radius exhibits an initial fast increase with time, followed
by a slower increase at later phases. A linear fit to the photospheric radius evolution after peak
light (between 8 and 15 days after explosion) produces an expansion velocity of
⇠
11
,
400
km
s
1
, while the same for the early light curve (
<
2 days from explosion) gives an expansion
velocity of
⇠
33
,
500
km s
1
.
Spectroscopic evolution
First peak
The spectroscopic sequence for iPTF 14gqr is shown in Figure 3. The earliest spectra were
obtained within a day from explosion, and exhibit intermediate width emission lines of He
II
,
C
III
and C
IV
which evolve rapidly with time. In particular, the earliest +13.9 h spectrum
exhibits emission lines of He
II
(
4686) and C
IV
(
5801, 5812) with FWHMs of
⇠
5000 km s
1
and
⇠
2000 km s
1
respectively, although the He
II
line may be contaminated with emission
from a nearby C
III
4650 line. The presence of C
III
is confirmed from the spectrum taken at
+25.2 hours where the C
III
5696 line becomes prominent.
The intermediate width He
II
line weakens in the +30.5 h spectrum while exhibiting sig-
natures of an extended red wing, before disappearing in a spectrum taken about 3 hours later.
We highlight the rapid evolution of the He
II
line in Figure 3, where we plot F
as function of
43
velocity shift from the He
II
line for all spectra taken within the first day of detection. In Figure
S6, we compare our early spectra to those of the SNe iPTF 13ast (
13
) and iPTF 13dqy (
15
),
where early spectroscopy had also revealed significant temporal evolution of the flash ionized
spectra. Both these events were of Type II and hence exhibit prominent H emission lines which
are absent in iPTF 14gqr. The He
II
4686 line is a common prominent feature of the flash
ionized spectra of these events, and the C
III
4650 and C
IV
5801 lines were also observed
in iPTF 13ast. Note that the flash ionized spectra of these events exhibit significantly narrower
emission lines (with central FWHM of
⇠
100 km s
1
superimposed on broad Lorentzian wings)
compared to those of iPTF 14gqr.
The spectral evolution of the C high ionization lines in the early spectra is very similar to
that seen in the WC sub-type evolution of galactic Wolf-Rayet stars (
82, 83
). In particular, the
C
III
5696 / C
IV
5801 ratio increases in the later and cooler sub-types (WC7 - WC9) of this
class, consistent with the increasing ratio observed in this source with decreasing photospheric
temperature. On the other hand, the C
III
4650 line decreases in strength with decreasing tem-
perature in the later WC stars as the C
III
5696 line becomes stronger (
83
). We can confirm
the presence of He
II
in the spectra by noting that the
4686 emission feature is continuously
present from +13.9 h to +30.5 h, even after the C
III
5696 disappears in the +30.5 h spectrum
(this C
III
feature is expected to become stronger than the
4650 feature at lower tempera-
tures (
82,83
)), suggesting that this emission feature has a dominant contribution from a species
different from C
III
(i.e. He
II
). Nonetheless, the the
4686 feature in the earlier spectra (at and
before 25 h) may have a contribution from C
III
4650, as is observed in the earlier sub-types
of WC stars (
83
).
44
Photospheric phase
Spectra taken about a week after explosion show characteristic absorption features of Type Ic
SNe, including lines of Fe
II
, Ca
II
,O
I
and Ti
II
. We compare the photospheric phase spectra
of iPTF 14gqr to those of other fast and normal Type Ic SNe in Figure S7. The comparison
candidates include the fast Type Ic events SN 2010X (
22
) and SN 2005ek (
21
), as well as the
spectroscopically normal events SN 1994I (
84
) and SN 2004aw (
85
). The phases of the spec-
tra indicated in this section are relative to
r
/
R
-band peak because the explosion times for the
literature events are not well constrained. The comparison clearly shows that the photospheric
spectra of iPTF 14gqr remain relatively blue and featureless compared to those of the normal
Type Ic SNe (SN 1994I and SN 2004aw) at similar phases.
The only prominent features in the
1
d and
+0
d spectra are those of Fe
II
and Ca
II
,
which are also seen in the spectra of the other events. Also apparent are weaker features of
C
II
at 6300
̊
A and 6900
̊
A. The spectrum obtained at +4 days shows a progressively cooling
continuum as P-Cygni absorption features of O
I
and Ca
II
become prominent in the red part
of the spectrum. A possible weak absorption feature of Si
II
also appears near 6200
̊
A. The
continuum shape as well as several spectral features (Fe
II
and Ca
II
on the blue side in partic-
ular) at this phase are best matched to that of SN 2005ek near peak light. However, SN 2005ek
also displays prominent absorption features of Si
II
and C
II
near 6500
̊
A, which are relatively
weaker in iPTF 14gqr.
We measure the photospheric velocities from the spectra obtained at
+0
d and
+4
d from
r
band maximum. We are restricted to measuring only the Fe
II
5169 velocity since we do not
see a prominent Si
II
6355 line in any of our photospheric spectra. We measure the velocity of
the Fe
II
line by fitting a parabola to the absorption minimum. The resulting fits give a velocity
of
⇡
10
,
500
km s
1
and
9
,
600
km s
1
for the
+0
d and
+4
d spectra respectively. A similar
fitting procedure performed on the O
I
7773 P-Cygni profile in our +4 d spectrum gives a
45
velocity of
⇡
9
,
100
km s
1
.
Type Ic SNe generally also exhibit the nearby Fe
II
lines of
5018 and
4924, which are
blended into a single blue-shifted feature with respect to the
5169 line (see for example, the
spectra of SN 1994I and SN 2004aw in Figure S7). The three features are blended into a sin-
gle broad absorption component in the case of the high velocity Type Ic-BL SNe, and can
potentially cause errors in a velocity measurement if this effect is not taken into account (
86
).
Although we do not separately detect these features in our spectra, we do see a broad Fe II
absorption feature, so followed the methods given in (
86
) to ascertain if the absence of these
features could be explained by velocity broadening. The best fitting models produce absorption
features that have markedly different shapes, in particular, that are symmetric with respect to
the minimum, unlike the shape observed here, and hence, a high velocity broadening is unlikely
to be present. Thus, we conclude that the
5018 and
4924 features are present (owing to the
presence of an asymmetric Fe feature) but not prominent enough to create a separate absorption
feature in these spectra.
Early nebular phase
Our final spectrum (with good SNR) was obtained
⇡
34 days after explosion, and show that
the source was transitioning very early into the nebular phase. Prominent features in this spec-
trum include the Ca
II
IR triplet, [Ca
II
]
7291, 7324, and a weak feature of [O
I
]
6300,
6364. In particular, the early appearance of [Ca
II
]
7291, 7324 and the apparent high ratio of
[Ca
II
]/[O
I
] is similar to that seen in the class of Ca-rich gap transients (
77,87
). We compare the
only nebular spectrum of iPTF 14gqr to the nebular spectra of other Type I SNe which exhibited
an early nebular transition at similar phases in Figure S8. This sample includes the Type Ic SNe
SN 1987M (
88
) and SN 1994I (
84
), the Ca-rich transients PTF 10iuv (
77
) and SN 2012hn (
89
),
SN 2010X (
22
) and SN 2005ek (
21
).
46
Amongst the normal Type Ic SNe, the nebular spectrum of iPTF 14gqr is perhaps closest
to that of SN 1987M in terms of the prominent nebular lines ([Ca
II
]
7291, 7324 and the
Ca
II
near-IR triplet), albeit at a later phase (
⇡
+60 d). SN 1994I exhibits a stronger [O
I
]
feature compared to iPTF 14gqr. Both SN 2010X and SN 2005ek also show a strong Ca
II
near-
IR feature but still show prominent P-Cygni profiles of O
I
near 7700
̊
A. The best spectral
match to iPTF 14gqr in this sample are to that of the Ca-rich gap transients PTF 10iuv and
SN 2012hn. Both these transients subsequently evolved to exhibit a strong [Ca
II
] 7300
̊
A
feature at later phases (with a high [Ca
II
]/[O
I
] ratio) characteristic of Ca-rich transients. We
could not obtain a good SNR spectrum of iPTF 14gqr at later epochs to trace the evolution of the
Ca features, although the association of iPTF 14gqr to the class of Ca-rich transients is unlikely
(see supplementary text).
Modeling
Arnett Model for the main peak
Type I SNe which do not show signs of interaction (such as iPTF 14gqr) are predominantly
powered by energy released in the radioactive decay chain of
56
Ni to
56
Co to
56
Fe. Since the
peak photospheric spectra of iPTF 14gqr display a number of similarities to those of fast and
normal SNe Ic, we prefer the scenario where this peak is powered by
56
Ni decay as in normal
SNe Ic, and derive explosion parameters using a simple Arnett model (
17
). We use the analytic
light curve expressions for a
56
Ni powered photospheric phase light curve given in (
19
) and (
90
)
to fit the pseudo-bolometric light curve of iPTF 14gqr. We exclude the data points within the
first peak (
<
2 days from assumed explosion) for this modeling.
The only parameters of this model are the nickel mass
M
Ni
and the photon diffusion timescale
⌧
M
. The photon diffusion timescale is related to the ejecta mass
M
ej
and explosion kinetic en-
ergy
E
k
by equation (3) in (
19
), or equivalently to the peak photospheric velocity
v
ph
and
M
ej
47
by their equation (1). We fit the observed bolometric light curve to this model using the Markov
Chain Monte Carlo (MCMC) method in the Python
emcee
package (
91
). Keeping the explo-
sion time
t
0
as an additional free parameter, we obtain a best-fit model (shown in Figure 4)
with
M
Ni
=0
.
051
+0
.
002
0
.
002
M
,
⌧
M
=
4
.
57
+0
.
77
0
.
62
days and
t
0
=
0
.
94
+0
.
49
0
.
62
days (i.e. the explo-
sion occurred 0.94 days before the assumed explosion time). The uncertainties indicate 68%
confidence intervals estimated from the MCMC simulations although these are likely more con-
servative because we adopted conservative uncertainties when calculating the bolometric light
curve. We caution that the Arnett model has several simplistic assumptions (see below) which
likely affect the estimation of the uncertainty intervals.
The best-fitting explosion time is earlier than our last non-detection (0.43 days before as-
sumed explosion), although the predicted flux from the Arnett model would be below our detec-
tion threshold. However, this explosion time could still be inconsistent with the pre-discovery
limits depending on the exact evolution of the first peak. We show the degeneracies between
the various model parameters in Figure S9, where the degeneracy between
t
0
and
⌧
M
is par-
ticularly prominent, since
⌧
M
controls the width of the light curve. For a photospheric ve-
locity of
10
4
km s
1
and optical opacity of
opt
= 0.07 cm
2
g
1
relevant for stripped enve-
lope SNe (
19, 20, 92
), we derive
M
ej
=0
.
23
+0
.
08
0
.
06
M
and explosion kinetic energy of
E
K
=1
.
38
+0
.
51
0
.
35
⇥
10
50
ergs, where the uncertainty intervals (given at 68% confidence) may again
be affected by the assumptions of the model. The lack of late-time photometric coverage does
not allow us to put any constraints on this model at late times.
The several assumptions of the Arnett model may also affect our estimates. These include
assumptions of homologous expansion, spherical symmetry, completely centralized location of
56
Ni and of optically thick ejecta (as in the photospheric phase). The rise and early decay of the
bolometric light curve depends sensitively on the extent of mixing of
56
Ni in the outer layers
of the ejecta. A comparison of parameters derived from the Arnett model to hydrodynamic
48
simulations by (
93
) suggest that the Arnett model can over-estimate the
56
Ni mass by about
50%. Additionally, they find the assumption of constant opacity to be a major limitation of this
model since the derived parameters depend sensitively on the assumed opacity. For instance, if
we adopt
opt
=0
.
1
cm
2
g
1
(relevant for material with one electron per four nucleons, e.g.,
singly ionized He), our ejecta mass estimate would change to 0.16 M
.
Nonetheless, we find our estimate of the extremely low ejecta mass (lower than previously
known core-collapse SNe) and low explosion energy to be fairly robust. One potential caveat in
the estimation of the ejecta mass is if the ejecta contain He that is effectively transparent. (
94
)
discuss the possibility that some stripped envelope SNe may have transparent He in their outer
layers that does not contribute to the ejecta opacity, since He is effectively transparent in the
continuum at low temperatures. They show that some stripped envelope SNe indeed show low
photospheric temperatures (
<
10
4
K) simultaneously with low photospheric velocities (
<
8000
km s
1
), suggesting that some of the outer (and faster) He layers may be transparent. However,
the photospheric temperatures in iPTF 14gqr are higher than
10
4
K up to peak bolometric light
(near the epochs of the peak photospheric spectra) while the velocities are also high (
⇠
10
,
000
km s
1
). Thus, the properties of this source are different from the events where He could have
been hidden, suggesting that this effect is not prominent in this source.
Shock cooling model for the first light curve peak
The collapse of the core in a SN explosion produces a radiation mediated shock that travels
outward through the stellar envelope, accelerating and heating material along the way. When
the optical depth to the shock becomes low enough, the shock breaks out while the shock-heated
envelope subsequently cools, producing early optical / UV emission. (
23
) show that the shock
cooling emission arising in normal progenitors (i.e. progenitors with no extended envelopes)
can create double peaked light curves in the blue and UV bands, but not in the redder
R
and
49
I
bands. The presence of a double peaked structure in the redder bands requires the presence
of a low mass extended envelope, where the low mass allows for a short photon diffusion time
(and hence a rapid peak and decline) while the extended structure prevents large adiabatic losses
at initial times. This corresponds to a case where the shock breaks out from the surrounding
circumstellar material (also known as a ‘CSM breakout’; e.g. (
95,96
)).
In the case of iPTF 14gqr, we observe a double-peaked structure in
r
band as well, and
hence we refer to the extended progenitor models of (
23
). The envelope mass can be derived
using the velocity of the external envelope
v
ext
, the rise time to peak
t
p
and the assumed opacity
. While early spectroscopy near the first peak should allow a measurement of photospheric
velocity at this phase, the relatively featureless continuum of our spectra do not allow us to
measure this velocity. Instead, we use equation (13) in (
23
) to estimate the radius of the external
envelope
R
ext
from our first temperature measurement in the bolometric light curve (which
is lower than the temperature at bolometric peak
T
obs,peak
). For
⇡
0.2 cm
2
g
1
(relevant
for electron scattering dominated opacity in a hydrogen free atmosphere),
t
p
0.5 days and
T
obs,peak
30000
K, we get
R
ext
1.5
⇥
10
12
cm.
(
25
) developed analytic light curves for the shock cooling emission from such an extended
envelope, and we use them to compare to the early multi-color light curves of iPTF 14gqr. The
model uses a simple one-zone treatment of the extended envelope, where all of the mass
M
e
is assumed to be concentrated around a radius
R
e
. The only other parameters are the velocity
of the extended material
v
e
and the mass of the core
M
c
, where we set
M
c
=0
.
23
M
the
explosion energy to
E
=1
.
38
⇥
10
50
ergs to derive
v
e
. We also let the explosion time
t
0
vary
such that the model is consistent with the last non-detection. Comparisons between the model
and the data are done by performing synthetic photometry on the blackbody spectra predicted
by the model.
We fitted the model parameters with the observed data using the Markov Chain Monte Carlo
50
method in the Python
emcee
package (
91
). We only use photometric detections obtained
within 1.1 days from the assumed explosion date since we find later times to be affected by
the rising portion of the
56
Ni light curve. We find a best-fitting model with
M
e
=8
.
8
+0
.
8
0
.
7
⇥
10
3
M
,
R
e
=3
.
0
+0
.
3
0
.
3
⇥
10
13
cm (
430
+43
43
R
), and
t
0
=
0
.
58
+0
.
04
0
.
04
days (i.e. the explosion
occurred 0.58 days before the assumed explosion time). The resulting light curves for this set
of parameters is shown in Figure 2. The error bars indicate the 68% confidence interval for the
derived parameters, as estimated from the MCMC simulations.
The shock cooling models for an extended progenitor envelope reproduce the optical light
curves (Figure 2), although there are discrepancies in the UV light curve. Such discrepancies
were also noted by (
26
) who suggest that the acceleration of the shock near the lower density
edge of the envelope can cause a stronger temperature evolution than predicted by the one zone
model, leading to a poorer match at shorter wavelengths. The optical light curves (particularly
in the redder bands) start rising around 1.5 days after explosion, suggesting that the underlying
56
Ni light curve becomes important at this phase. As in the case of the Arnett model, we expect
the simple assumptions of this model to produce values which are only approximately correct.
We thus conclude that the early shock cooling emission was produced by an extended envelope
with a mass of
⇠
0
.
01
M
and located at a radius of
⇠
500
R
.
We show the degeneracies between the model parameters in Figure S10. Although there are
degeneracies between the parameters, the range of values occupy a relatively small phase space
around the best-fitting value. We also consider how the assumed ejecta mass and explosion
energy affect the derived parameters by considering the range derived from the Arnett modeling.
Adopting the lower end of ejecta mass and explosion energy, we find a best-fitting model with
M
e
=8
.
2
+0
.
8
0
.
6
⇥
10
3
M
,
R
e
=3
.
6
+0
.
3
0
.
3
⇥
10
13
cm, and
t
0
=
0
.
53
+0
.
04
0
.
05
days. On the other
hand, adopting the higher end of the ejecta mass and explosion energy distribution, we find
M
e
=10
+1
0
.
9
⇥
10
3
M
,
R
e
=2
.
4
+0
.
2
0
.
2
⇥
10
13
cm, and
t
0
=
0
.
64
+0
.
04
0
.
04
days. Thus, our
51
estimates of the envelope properties appear to be fairly insensitive to the adopted ejecta mass
and explosion energy.
Interaction with a companion
An alternative explanation of the early excess emission in iPTF 14gqr might be the interaction
of the SN ejecta with a non-degenerate companion. Such interaction signatures have been
previously observed in some Type Ia SNe ( (
97–99
); see (
16
) for a review of SN classification)
where comparison of the data to theoretical models (
100
) allows the inference of the orbital
separation of the binary system. The models show that the presence of a companion can produce
a void in the expanding ejecta, producing a reverse shock that powers excess luminosity at early
times when viewed close to the direction of the void. In such a scenario, we might expect high
ionization emission lines in the spectra arising from recombination of the companion wind,
although the relatively large widths of the lines would suggest unusually high wind velocities
for the companion. However, despite the high signal-to-noise ratio of our early spectra, we see
no evidence for the presence of broad lines, as expected from the reverse shock produced in the
ejecta-companion interaction.
We consider the photometric properties of the first peak in a companion interaction scenario.
As shown in (
100
), the early luminosity evolution depends on the viewing angle of the observer,
where the excess flux is most prominent along the direction of the companion and relatively
weak along directions perpendicular to or oriented away from the companion. In the case of
iPTF 14gqr, we observe a very rapid decline of the bolometric luminosity at early times (
⇠
a factor of 10 in
<
1 day), ruling out viewing angles away from the companion ( (
100
), their
figure 2). For viewing angles along the companion direction, the bolometric luminosity is well
approximated by the analytic equations presented in (
100
).
We attempted to fit the early multi-color photometry with the models presented in (
100
),
52
but were unable to obtain a good fit to the data. This is primarily because the luminosity
from the interaction is relatively long-lived (a few days) when the peak luminosity is large
(at least
⇠
2
⇥
10
43
ergs s
1
), while the predicted color evolution is very different. This is
readily apparent when comparing the luminosity and temperature evolution of the model with
the observations – the analytical model predicts a luminosity evolution scaling with time as
t
1
/
2
, which is almost the same as the color temperature evolution, which scales as
t
37
/
72
.
Our early observations clearly show that the luminosity drops by a factor of 10, while the
temperature drops by only a factor of 3. Thus, we find the companion interaction scenario to be
inconsistent with the properties of the early peak.
Analysis of early spectra
The early spectra of iPTF 14gqr exhibit prominent emission features of highly ionized He and
C that are broader (FWHM
⇠
3000
km s
1
) than that typically observed in the flash ionized
spectra of other core-collapse events. We consider two possible scenarios for the origin of these
features – one where they arise from recombination of material surrounding the progenitor after
ionization during the SN shock breakout, and one where they are powered by interaction of the
SN ejecta with a dense CSM. While the intermediate width features seen in the early spectra
are somewhat similar in width to those seen in interacting SNe at and before peak light (
101
),
the observed recombination of highly ionized species is better explained by shock cooling of
extended material. In particular, we observe clear evidence of recombination of C
IV
in our
earliest spectra followed by C
III
in later spectra, that is more consistent with shock cooling
emission rather than continued energy supply from interaction. We thus favor shock cooling
emission as the origin of the early excess emission as well as the intermediate width emission
lines.
We analyze the properties of the flash ionized spectra of iPTF 14gqr to estimate properties
53
of the emitting material surrounding the progenitor, where the optically thin flash ionized lines
originate. Following the methods outlined in (
96
) and (
15
) where they estimate the mass loss
rate from the luminosity of the H
↵
line at 6563
̊
A, we can relate the mass of the doubly ionized
helium region generating the He II
4686 line to the line luminosity using,
L
He II
,
4686
⇠
A
m
He
Z
r
r
1
n
e
4
⇡
r
2
⇢
(
r
)
dr
(S1)
where
L
He II
,
4686
is the luminosity of the
4686 line,
A
=
4
⇡
j
4686
n
e
n
He
++
,
j
4686
is the emission
coefficient for the
4686 transition,
m
He
is the mass of a He nucleus,
n
He
++
is the number
density of the doubly ionized He and
n
e
is the number density of electrons. The order of
magnitude approximation holds because there may be unknown collisional excitation and de-
excitation processes operating, in addition to the recombination processes considered here.
We use
A
⇡
1
.
3
⇥
10
24
ergs cm
3
s
1
(
102
) for a temperature of
10
4
K, electron density of
⇠
10
10
cm
3
(the value of
A
is fairly insensitive to the assumed electron density) and Case B
recombination. We consider a density profile with the density
⇢
varying as
⇢
=
Kr
2
, where
K
is the mass loading parameter. The total mass in a region between
r
1
and
r
is given by (
96
),
M
He
=
Z
r
r
1
4
⇡
r
2
⇢
(
r
)
dr
=4
⇡
K
r
(S2)
where
=
(
r
r
1
)
/r
is assumed to be of the order unity and
M
He
is the total mass of the
emitting material. Assuming
n
e
⇡
2
n
He
++
, and using the density profile, we write the
4686
luminosity as,
L
He II
,
4686
⇠
8
⇡
A
m
2
He
K
2
r
1
(S3)
We first attempted to measure the luminosity of the
4686 line by subtracting the best-
fitting blackbody continuum from the spectrum and fitting a Gaussian profile to the
4686
54
line in the +13.9 h spectrum. With the absence of simultaneous photometry for the +25.2 h
spectrum, we are unable to accurately measure the luminosity of the line in this spectrum. We
thus only use the +13.9 h line luminosity for our calculations. Since the
4686 feature likely
has a non-negligible contamination from C
III
4650 feature at this epoch, we find that the
simple Gaussian fit has a blue-shifted peak (with respect to
4686) and a red shoulder excess
(consistent with
4686).
Using the +25.2 h spectrum which has a higher signal to noise ratio to only better estimate
the relative contributions, we are able to get a better fit using a simple two component Gaussian
profile, which suggests that the He
II
line has a FWHM of
⇠
4500
km s
1
and contributes 60%
of the flux in the line, while the C
III
feature has a FWHM of
⇠
2000
km s
1
and contributes
the remaining 40% of the flux. The width of the C
III
feature is similar to that measured for the
C
IV
5801 feature in the first spectrum. We are unable to better constrain the contributions of
He
II
4686 and C
III
4650 in the +13.9 h spectrum than adopting the above measured ratio
of 3:2 for the contribution of the He
II
and C
III
lines. This suggests that the He
II
line has a
luminosity of
⇠
1
.
1
⇥
10
40
ergs.
The location of the emitting region can be constrained by requiring that the Thompson
optical depth in the region must be small for the lines to escape. The Thompson optical depth
produced by this ionized material can be written as,
⌧
=
2
T
m
He
K
r
1
(S4)
where
⌧
must be small for the lines to escape. This gives a lower limit on the location of the
line forming region at,
r
⇠
L
He II
,
4686
2
T
2
⇡
A
⌧
2
(S5)
Taking the +13.9 h luminosity measured for the
4686 line, we get
r
⇠
6
⇥
10
14
⌧
2
cm,
K
⇠
2
.
9
⇥
10
15
⌧
1
g cm
1
and
M
He
⇠
0
.
01
2
⌧
3
M
where we take
⇠
1
. These
55
estimates can be affected if the CSM was not characterized by a wind-like
r
2
profile, if the
emitting region was confined to a thin shell (
⌧
1
), if the emitting region was not spherically
symmetric (e.g. if it was clumped) or if our estimate of the C III
4650 contamination in the
4686 emission feature was incorrect by a factor of a few. Given these caveats, we take the
calculated values to be accurate to an order of an magnitude only. The inferred radius of the
optically thin material is larger than the envelope producing the early shock cooling emission
as seen in the light curve, and suggests that the highly ionized lines likely arise from a lower
density extension of the same envelope.
An additional consistency check for the location of the emitting material can be found by
noting that the line emission produced due to a short ionizing burst of radiation would be visible
for at least the light crossing time between the part of the shell pointing towards the Earth and
the part perpendicular to this line of sight (i.e.
t
cross
=
r/c
). This is because the inferred
recombination times at these high densities (
n
e
=2
Kr
2
/m
He
⇠
10
9
10
10
cm
3
) are short
(
⇠
10 minutes) and hence the emitting material recombines promptly after shock breakout.
While there is some uncertainty in the time of explosion, the light curve models for the first
peak, as well as the Arnett model, favor earlier explosion times than the one assumed. We thus
constrain the first spectrum to have been taken between
⇡
+14 h (for the assumed explosion
time) and
⇡
+36 h after explosion (for an explosion at 0.92 days before the assumed time).
Similar constraints for the last featureless spectrum within the first day of detection suggest that
it was taken between
⇡
+33 h and
⇡
+55 h after explosion. This constrains the emitting region
to
r
6
⇥
10
15
cm, noting that this estimate is relatively insensitive (within a factor of
⇡
2
) to
the exact time (between +33 h and +55 h after explosion) when the lines disappear.
Given the large inferred size of the emitting envelope (
⇠
6
⇥
10
14
cm, corresponding to a
light travel time of
⇠
5 hours), we now consider potential effects of light travel time (LTT) on
the observed spectrum. A study of the flash-ionized spectrum of SN 2013cu (
13
) by (
27
) suggest
56
that including LTT effects decreases the expected luminosity of the spectral lines and affects the
observed line profiles depending on the location of the line forming region in the envelope. The
lower luminosity has the equivalent effect of increasing the inferred mass of the envelope, as
was inferred in the modeling of SN 2013cu (
27
). While the emitting mass in iPTF 14gqr may
also be higher than our estimates, the rapid rise time of the early peak constrains the mass to be
⇠
0
.
01
M
.
(
27
) also show that lines formed in the outer region of the envelope are substantially blue
shifted for high wind velocities, while lines formed in the central regions are less affected.
Comparing to our observations, we note that the blue-shifted peak of the
4686 feature and
C
III
5696 lines in the +13.9 h and +25.2 h spectra can potentially be explained by LTT effects.
The blue-shifted profile of the
4686 feature may not require contamination by C
III
4650 in
such a case. The observed blue-shift corresponds to a velocity of
⇠
2200 km s
1
, similar to the
velocity inferred from the widths of the emission lines.
Assuming that the electron density in the envelope is dominated by He ionization (
n
e
⇠
10
10
cm
3
), we try to estimate the amount of C in the envelope. Using the C
IV
line in the +13.9
spectrum, we measure a C
IV
5801, 5812 luminosity of
⇠
6
.
1
⇥
10
39
ergs. Using a C
IV
5801,
5812 recombination coefficient of
A
=1
.
4
⇥
10
24
ergs cm
3
s
1
(
103
), we determine a C mass
of
⇠
4
⇥
10
3
M
. We can use our early spectra to place constraints on the amount of hydrogen
in the envelope using the non-detection of the H
↵
emission. We use the +13.9 h spectrum to
place a 3
upper limit on the H
↵
luminosity of
4
.
5
⇥
10
39
ergs s
1
. Taking an electron density
of
n
e
⇠
10
10
cm
3
in the emitting region and recombination coefficient of
A
=2
.
6
⇥
10
25
ergs cm
3
s
1
, we constrain the hydrogen mass of the envelope to a 3
upper limit of
⇠
10
3
M
.
57