NUCLEOSYNTHESIS IN A PRIMORDIAL SUPERNOVA: CARBON AND OXYGEN ABUNDANCES IN SMSS J031300.36–670839.3^*

Keller et al. (2014) described the discovery of SMSS J031300.36–670839.3 (hereafter SM0313–6708) and the analysis of the high resolution spectrum taken with the Magellan Inamori Kyocera Echelle (MIKE; Bernstein et al. 2003) spectrograph on the 6.5 m Magellan Telescope at Las Campanas Observatory in Chile. The spectrum was remarkable for the complete absence of lines of iron or any other metal, apart from Mg i and Ca ii. The fact that no iron line is evident in the signal-to-noise ratio (S/N) ∼ 100 (at λ = 4000 Å) spectrum indicated that the iron abundance in SM0313–6708 was at most $[{\rm Fe}/{\rm H}]=-7.2$,¹³ making this star at least ∼40 times more iron-poor than HE 1327–2326 ($[{\rm Fe}/{\rm H}]=-5.6$; Frebel et al. 2005).

The only elements whose abundances could be measured were lithium, carbon, magnesium, and calcium with $A({\rm Li})=12+{\rm log} ({\rm Li}/{\rm H})=0.71\pm 0.10$, $[{\rm C}/{\rm H}]=-2.44\pm 0.10$, $[{\rm Mg}/{\rm H}]=-4.30\pm 0.10$, and $[{\rm Ca}/{\rm H}]=-7.2\pm 0.10$, respectively. The level of lithium is presumably related to the decline from the primordial Spite plateau level of $A({\rm Li})=2.2$ during early post-main-sequence stellar evolution (Lind et al. 2009) and this evidence, together with a ${{T}_{{\rm eff}}}=5125\;{\rm K}$ and ${\rm log} g=2.3$ derived from spectrophotometry, is consistent with a star ascending the giant branch for the first time. The C, Mg, and Ca abundances can be attributed to a low energy Pop III SNe, and measurement of the oxygen abundance, in particular, offers an important test of this scenario. To measure the oxygen abundance in ultra-metal-poor K stars, however, it is necessary to work in the ultraviolet and observe the OH lines that persist well after the [O i] 6300.3 Å line has become undetectable.

In order to measure the OH (and CH) lines between 3100 and 3200 Å and obtain better detection limits on N, Na, V, Co, Ni, Fe, and Cu, we used Ultraviolet and Visual Echelle Spectrograph (UVES; Dekker et al. 2000) on the VLT (ESO Very Large Telescope) at Paranal in Chile. Between 2013 and 2014 September, 31 exposures totalling 29.26 hr of on-source integration were obtained. The UVES UV spectrum extended from 3060 to 3860 Å and the two red spectra from 4786 to 6803 Å.

The data were reduced using the default settings of the UVES pipeline as implemented in the "reflex" reduction environment. There was a high order sinusoidal-like modulation of the continuum remaining in much of the UV spectrum. The modulation with an amplitude of about 20% and frequency about 50 Å was removed by hand, leaving a residual variation of amplitude of 2%–4% and spacing of about 1.5 Å that makes the measurement of shallow features somewhat uncertain. It was very clear, however, that there were many strong OH lines present between the observational cutoff around 3060 and 3200 Å, and that the OH lines were much stronger than the CH lines, except at the band head of the CH C-X band near 3144 Å. In Figure 1 we compare the UV spectra of SM0313–6708 ($[{\rm Fe}/{\rm H}]\leqslant -7.2$) and HE 0107–5240 ($[{\rm Fe}/{\rm H}]=-5.4$), a C-rich ultra-EMP star with almost identical effective temperature and gravity, and possibly at the same evolutionary stage. The total absence of metal lines in SM0313–6708 (except for the Mg i lines around 3830 Å) is very marked as is the different distribution of the forest of molecular lines below 3250 Å. The CH lines dominate the spectrum of HE 0107–5240 whereas OH lines dominate in SM0313–6708. The complete absence of absorption lines in SM0313–6708 between 3250 and 3800 Å enables more stringent limits to be placed on those species with strong lines in that region.

Zoom In Zoom Out Reset image size

We have carried out a spectrum synthesis of the CH, OH, and NH molecular lines in SM0313–6708 using Turbospectrum (Alvarez & Plez 1998; Plez 2012). We adopted an interpolated "standard" metallicity MARCS 1D LTE model atmosphere structure (Gustafsson et al. 2008), for ${{T}_{{\rm eff}}}=5125\;$ K, ${\rm log} g=2.30$, $[{\rm M}/{\rm H}]=-5.0$, but with actual abundances from this paper for the atomic and molecular equilibria calculations. We used the new ¹²CH and ¹³CH line list from Masseron et al. (2014), while for OH and NH we used extensive line lists recently recomputed by T. Masseron (2015, in preparation). The Masseron OH $gf$ values are very similar to those of Gillis et al. (2001), but the NH $gf$ values are significantly smaller by $0.3-0.7\;{\rm dex}$ than the commonly used values of Kurucz (2011). The solar composition was adopted from Asplund et al. (2009).

We refitted the CH A-X and B-X lines in the blue MIKE spectrum and determined a carbon abundance of $[{\rm C}/{\rm H}]=-2.41\pm 0.05$, very similar to the previously determined A-X value. The CH C-X band synthesis is in good agreement with the A-X and B-X bands. The synthesized spectrum well reproduces the observed CH lines in strength and position for all J values and for all branches observed.

The OH lines extend from 3060 to redward of 3200 Å. Because of the lower S/N and the uncertain continuum placement at the shortest wavelengths where the lines are strongest, we derived the oxygen abundance in SM0313–6708 by fitting the weaker OH lines between 3178 and 3200 Å. Figure 2 clearly shows that the oxygen abundance is well constrained at $[{\rm O}/{\rm H}]=-2.00\pm 0.10$. In HE 0107–5240, the similar temperature and gravity star with abundances $[{\rm M}/{\rm H}]\approx -5.4$, $[{\rm C}/{\rm H}]=-1.58$ (Christlieb et al. 2004) and $[{\rm O}/{\rm H}]=-3.0$ (Bessell et al. 2004), these longer wavelength regions could not be used for OH fitting as they are dominated by CH lines as seen in Figure 3. We note that the treatment of scattering (as done in Turbospectrum), is very important for the analysis of the UV lines in SM0313–6708 and metal-poor giants in general.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 3 shows the comparison between the observed and synthetic spectra at the shorter wavelengths in SM0313–6708 and HE 0107–5240. The panels on the left show the region that was used to measure the oxygen abundance in HE 0107–5240, and the panels on the right shows the region around the CH C-X band head. Note the significantly different relative strengths of OH and CH lines in the two stars.

There is no indication of any NH features centered on 3360 Å, but the higher S/N of the UVES spectrum enables us to reduce the upper limit to $[{\rm N}/{\rm H}]\lt -4.2$, at least two orders of magnitude lower than C and O.

3.1. 3D–1D Corrections to C and O Abundances

3.2. ¹²C/¹³C Ratio

3.3. Revised Upper Limits on the Abundances of Fe, Na, Si, Sc, V, Co, Ni, Cu

Two of the Fe i transitions with the largest ${\rm log} gf$ values in the observed spectral range are located within the highest S/N portion of the UV spectrum, at 3719.94 and 3737.13 Å. Sections around the position of these lines were extracted, summed and compared with the summed synthetic spectra computed for a range of Fe abundances. Again, no feature was evident in the observed spectrum, and we estimate a revised abundance limit for $[{\rm Fe}/{\rm H}]$ of about −7.5.

The non-detection of lines of Sc, Ti, V, Co, Ni, and Cu in the UVES spectrum in the range from 3100 to 3700 Å provided revised lower upper limits for many elements given in Table 1 (A. Frebel et al. 2015, in preparation). The non-detection of the Si i line at 3905.52 Å in the blue MIKE spectrum indicates an upper limit of $-5.6\;{\rm dex}$.

Table 1.  Chemical Abundances of SM0313–6708

Element	${{[{\rm X}/{\rm H}]}_{1{\rm D},{\rm LTE}}}$	${{[{\rm X}/{\rm H}]}_{1{\rm D},{\rm NLTE}}}$	${{[{\rm X}/{\rm H}]}_{\langle 3{\rm D}\rangle ,{\rm LTE}}}$	${{[{\rm X}/{\rm H}]}_{\langle 3{\rm D}\rangle ,{\rm NLTE}}}$	${{[{\rm X}/{\rm H}]}_{3{\rm D},{\rm LTE}}}$
Li i a	0.70 ± 0.10	0.73 ± 0.10	...	0.56 ± 0.10	$0.71\pm 0.10$
C (CH)b	−2.41 ± 0.05	...	...	...	$-2.55\pm 0.09$
N (NH)b	$\lt -4.2$	...	...	...	$\lt -4.70$
O (OH)b	−2.00 ± 0.10	...	...	...	$-2.53\pm 0.15$
Na i b	$\lt -5.7$	$\lt -5.65$	$\lt -5.79$	$\lt -5.66$	...
Mg i b	−4.45 ± 0.07	−4.03 ± 0.03	−4.53 ± 0.09	$-4.08\pm 0.03$	...
Al i b	$\lt -6.2$	...	...	...	...
Si i b	$\lt -5.6$	...	...	...	...
Ca ii	−7.46 ± 0.10	−7.27 ± 0.10	−7.60 ± 0.10	$-7.26\pm 0.10$	...
Sc ii b	$\lt -5.5$	...	...	...	...
Ti ii b	$\lt -7.1$	...	...	...	...
V ii b	$\lt -4.3$	...	...	...	...
Cr i	$\lt -6.3$	...	...	...	...
Mn i	$\lt -5.8$	...	...	...	...
Fe i b	$\lt -7.80$	$\lt -7.28$	$\lt -8.27$	$\lt -7.52$	⋯
Co i b	$\lt -5.4$	...	...	...	...
Ni i	$\lt -6.9$	...	...	...	...
Cu i b	$\lt -5.5$	...	...	...	...
Zn i	$\lt -3.4$	...	...	...	...
Sr ii	$\lt -6.9$	...	...	...	...
Ba ii	$\lt -6.1$	...	...	...	...
Eu ii	$\lt -2.9$	...	...	...	...

Notes. Estimated (1σ) observational uncertainties in the quoted upper limit abundances are $0.3\;{\rm dex}$. For Fe the uncertainty is 3σ. Bold print indicates our recommended values used in the fitting of explosive nucleosynthesis.

^aThe abundance of Li is expressed as $A({\rm Li})={\rm log} ({\rm N}({\rm Li})/{\rm N}({\rm H}))+12$. ^bAbundances derived in this paper.

Download table as:  ASCIITypeset image

The UVES red spectrum was in good agreement with the MIKE spectrum for the Mg i lines and the Li i line, but the better S/N and weaker telluric lines in the UVES red spectrum indicates that the Na limit could be lowered to about $-5.7\;{\rm dex}$. Table 1 lists the current best estimates for the chemical abundances of SM0313–6708.

Heger & Woosley (2010) have followed the evolution and explosion of metal-free stars with masses $10\;{{M}_{\odot }}$–$100\;{{M}_{\odot }}$ and determined their nucleosynthetic yields, light curves, and remnant masses. Such stars would have been the first to form after the Big Bang and could have left a distinctive imprint on the composition of the early universe. The evolution of these stars may be affected in several ways by their zero-metallicity primordial composition. Because these stars suffered little mass-loss during their pre-explosive lifetimes, their masses at their deaths were larger than more metal-rich stars. The models produces some carbon, but for masses less than about $40\;{{M}_{\odot }}$ primary production of nitrogen is suppressed. These stars ended their lives as compact blue supergiants making their envelope more tightly bound, with two interesting consequences: (i) the more compact envelope increases the amount of mass that falls back after the explosion, meaning that they were more likely to collapse to black holes at lower masses compared to modern stars, trapping most of the heavy elements; (ii) mixing by the Rayleigh–Taylor instabilities is suppressed in these compact stars, so it is mainly unmixed material that escapes in the explosion. Heger & Woosley (2010) fitted their predicted nucleosynthetic yields to several examples of metal-poor stars and found that the best fits were obtained from relatively low energy explosions of 0.6–1.8 B ($1\;{\rm B}={{10}^{51}}\;$ erg) as compared to hypernova models (Nakamura et al. 2001). Overall, the large sample of metal-poor stars in Cayrel (2004) was well fit with a normal Salpeter IMF including all stars from $10\;{{M}_{\odot }}$ to $100\;{{M}_{\odot }}$ (but with a preference for stars lighter than $15\;{{M}_{\odot }}$), a small amount of fallback and with reduced mixing of 1% of the helium core mass (for reference, the default mixing value adopted to solar metallicity stars is 10% of the helium core mass; Heger & Woosley 2010). For the two ultra-iron-poor, but high ${\rm O}/{\rm Fe}$, stars HE 0107–5240 (Bessell et al. 2004; Christlieb et al. 2004) and HE 1327–2326 (Aoki et al. 2006; Frebel et al. 2006), the best fits were for similar low energy explosions, a lower upper mass range ($10\;{{M}_{\odot }}$ to less than $30\;{{M}_{\odot }}$), small fallback and an even lower mixing of 0.025.

The abundances of SM0313–6708 are much more extreme than the Cayrel (2004) EMP stars or the two C-rich ultra-EMP stars. The biggest difference are the large carbon, oxygen and magnesium overabundances relative to calcium, and the non-detection of all other metals. Of particular significance is the oxygen to carbon ratio in SM0313–6708 of ${{[{\rm O}/{\rm C}]}_{3{\rm D}}}=+0.02$ compared to ${{[{\rm O}/{\rm C}]}_{3{\rm D}}}\approx -1.0$ (Collet et al. 2006) in HE 0107–5240.

Fitting the current SM0313–6708 abundances (Table 1) against models from Heger & Woosley (2010) using their Starfit algorithm¹⁴ finds good fits for progenitor masses in the range $40\;{{M}_{\odot }}-60\;{{M}_{\odot }}$ and energies of $1.5\;{\rm B}-1.8\;{\rm B}$ with low amounts of mixing, $\lesssim 2\%$. This low amount of mixing is consistent with Heger & Woosley (2010) and hydrodynamic models of Pop III SNe by Joggerst et al. (2009). The best matching mass is sensitive to the upper limits of the Na and Al abundances. Generally, lower values on the upper limits to the abundances would push the fit toward higher initial masses, stars that produce a stronger odd–even effect. The low limits on N in SM0313–6708 constrain the mass on the upper end.

The range of [O/C] abundances produced in a range of core collapse supernovae is shown in Figure 4 together with a $40\;{{M}_{\odot }}$ fit from Starfit. Significantly, the observed [O/C] ratio is lower than what Pop III pair-instability SNe of Heger & Woosley (2002) can produce or what we can produce for high explosion energies ($E\gtrsim 3\;{\rm B}$) in more massive stars ($M\gtrsim 20\;{{M}_{\odot }}$). The structure and pre-SN evolution of stars with mass above $15\;{{M}_{\odot }}$ is very sensitive to initial mass (e.g., Sukhbold & Woosley 2014) and these stars may experience chaotic nuclear burning. As a result, the fraction of the C and O layers that are ejected varies strongly for a given explosion energy and is rather sensitive to mass and explosion energy. Currently there is no good first principle calculations providing reliable predictions for explosion energies for a given stellar structure for masses above $15\;{{M}_{\odot }}$.

Zoom In Zoom Out Reset image size

Our new 3D LTE values provide definite constraints on the [O/C] ratio instead of only an upper limit. The new lower value of [Fe/H] derived here strengthens the constraint not to eject iron group elements from the inner core of the progenitor star. This limits the explosion energy and the magnitude of mixing due to Raleigh–Taylor instabilities during the SN explosion.

Ishigaki et al. (2014) have proposed an alternative scenario to explain the unique abundances in SM0313–6708. They investigated supernova yields of metal-free (Population III) stars and find that the high $[{\rm C}/{\rm Ca}]$ and $[{\rm C}/{\rm Mg}]$ ratios and upper limits of other elemental abundances are well reproduced with the yields of core-collapse supernovae (which have normal kinetic energies of explosion of $1\;{\rm B}$) and hypernovae ($10\;{\rm B}$) of Population III $25\;{{M}_{\odot }}$ or $40\;{{M}_{\odot }}$ stars. Their best-fit models assume that the explosions undergo extensive matter mixing and fallback, leaving behind a black hole remnant. In these models, Ca is produced by static/explosive O burning and incomplete Si burning in the Population III supernova/hypernova, in contrast to our suggestion that the Ca originates from the hot-CNO cycle during pre-supernova evolution, which is, remarkably, consistent with the present data.

Although Ishigaki et al. (2014) well fits the observed abundances, their modeling uses several independent free parameters. For example, their models require significantly more mixing than is consistent with hydrodynamical simulations of Population III stars (Joggerst et al. 2009) and the fallback is adjusted independently of the explosion energy. Jetted or asymmetric explosions may be a possibility; however, this is another free parameter. In the Heger & Woosley (2010) models the mixing is consistent with hydrodynamical simulations of Joggerst et al. (2009) and the fallback is computed self-consistently with the explosion energy. Fuller discussion of the abundances of SM0313–6708 and SN modeling will be presented in a future paper.

The ultraviolet spectrum of SM0313–6708 is dominated by the lines of the OH A-X band, which are noticeably stronger than the CH C-X lines. The new line lists of Masseron fit the OH and CH lines very well in both wavelength and strength. We could detect no lines of ¹³CH, which places a limit of more than 40 on $^{12}{\rm C}{{/}^{13}}{\rm C}$, and no lines of NH putting nitrogen at least two orders of magnitude less than C and O. This supports there having been no mixing from the H-burning shell to the surface within SM0313–6708 and there being little nitrogen relative to C and O in the material from which SM0313–6708 formed. The clear lack of any metal lines apart from Mg i in the UVES ultraviolet spectrum also enabled us to lower the abundance limits for those elements whose strong lines are normally seen in this region, including Fe i, for which we now derive an $[{\rm Fe}/{\rm H}]$ upper limit of $-7.52\;{\rm dex}$ ($3\sigma$) from $\langle 3D\rangle$ LTE analysis. We derived an [${\rm O}/{\rm C}$] ratio of $+0.02$ with a $1\sigma$ uncertainty of $0.175\;{\rm dex}$ for SM0313–6708, updated abundances measurements for Mg and Ca, and derived new upper limits for nitrogen, the Fe-group elements and some of the alpha-elements. These abundances remain consistent with a single massive Population III progenitor in the mass range $40\;{{M}_{\odot }}-60\;{{M}_{\odot }}$, low mixing, and modest explosion energies; the Ca production remains consistent with an origin from hydrostatic hydrogen burning. Improved upper limits on intermediate-mass elements, in particular Na and Al, will further help constrain the models in the future. The strong Fe ii lines in the vacuum UV between 2340 and 2625 Å should allow one to constrain—or detect—the iron abundance of SM0313–6708 down to as low as $[{\rm Fe}{\rm H}]=-9$.

A.H. was supported by an Australian Research Council (ARC) Future Fellowship (FT120100363). R.C. is supported by an ARC Discovery Early Career Researcher Award (DE120102940). A.C. and T.M. have been supported by the European Union FP7 programme through ERC grant number 320360. A.F. acknowledges support from NSF Career grant AST-1255160. This research been supported, in part, by ARC Discovery Project grants DP120101237 and DP150103294 and NSF grant PHY-1430152 (JINA-CEE).

Facility: ESO(UVES) -