The Search for Supernova-Produced Radionuclides in Terrestrial Deep-Sea Archives

Keywords: Nuclear reactions, nucleosynthesis, abundances — ISM: bubbles — ISM: supernova remnants

Stars with masses larger than 8 M_⊙ end their life in a supernova (SN) explosion. The nuclides that were synthesized in the late phases of the star and during the explosion are ejected and entrained in the supernova shell, which expands rapidly into the surrounding interstellar medium. If such an event happens in the solar vicinity, the remnant may leave certain traces in terrestrial archives. Ejecta of a SN contain freshly produced long-lived radionuclides; eventually there might be a chance of finding such radionuclides in terrestrial archives.

Such a signal of possible SN origin has been detected by Knie et al. (2004) in the hydrogenetic ferromanganese crust 237KD, which was recovered from the equatorial Pacific. The radionuclide ⁶⁰Fe with t_1/2 = (2.62 ± 0.04) Myr (Rugel et al. 2009) was measured with accelerator mass spectrometry (AMS) utilizing the MP-tandem accelerator in Munich. The signature was detected again in the same crust sample in layers at a depth of 6–8 mm in a second measurement in the same laboratory, using different chemical separation schemes for AMS target preparation (Fitoussi et al. 2008).

The age of the deep ocean's crust was determined through ¹⁰Be dating. Based on the half-life of t_1/2(¹⁰Be) = (1.51 ± 0.06) Myr (Hofmann et al. 1987), Knie et al. (2004) estimated a time range of 2.4–3.2 Myr for the layer that contained the anomaly. Fitoussi et al. (2008) measured the ¹⁰Be content consistently and re-evaluated this result using t_1/2(¹⁰Be) = (1.36 ± 0.07) Myr (Nishiizumi et al. 2007). The half-life of ¹⁰Be was recently remeasured by Chmeleff et al. (2010) and Korschinek et al. (2010) and the suggested value is now (1.387 ± 0.012) Myr. Following Fitoussi et al. (2008), we calculate the time profile with this new value and obtain a time range of 1.74–2.61 Myr. Figure 1 shows the recalculated age profile of the sediment core, including the data of Knie et al. (2004) and Fitoussi et al. (2008).

Figure 1  Measured ⁶⁰Fe/Fe ratios versus the age of the layer calculated with t_1/2(¹⁰Be) = 1.387 Myr. The black circles show the measurements by Knie et al. (2004), the white squares the data of Fitoussi et al. (2008).

A hint as to recent SN activity in the solar vicinity is the existence of a thin hot cavity in the local interstellar medium (ISM), the Local Bubble (LB), which embeds our solar system. Its extensions are about 200 pc in the Galactic plane and 600 pc perpendicular to it. Superbubbles are formed by several SN explosions, creating blast waves that push away the surrounding medium. They leave thin hot gas inside, encircled by a shell consisting mainly of accumulated material but also of SN ejecta. The latest age calculations of the LB, based on comparison between observations and simulations of ion column density ratios, suggest a value of 13.6–13.9 Myr (Avillez & Breitschwerdt 2012). Fuchs et al. (2006) identified a stellar moving group as having crossed the solar neighborhood and determined about 14–20 SN explosions to form the LB, the last of these occurring approximately 0.5 Myr ago. The remaining stars belong today to the subgroups UCL (Upper Centaurus Lupus) and LCC (Lower Centaurus Crux) of the Sco–Cen association.

It is suggested that the stars that created the LB also contributed to the ⁶⁰Fe excess on Earth (Berghöfer & Breitschwerdt 2002). Neuhäuser et al. (2011) have searched for the birthplace of neutron stars and their runaway companions to corroborate this theory.

To obtain further evidence for recent SN explosions in the solar vicinity, we have identified two suitable marine sediments from the Indian Ocean, the Eltanin Piston Cores ELT 45-21 and ELT 49-53. Both sediment cores originate from a location south-west of Australia, 38°58.7′S, 103°43.1′E and 37°51.6′S, 100°01.7′E (Allison & Ledbetter 1982). They were recovered from depths of 4237 and 4248 m in the years 1970 and 1971. With accumulation rates of 3–4 mm kyr^–1, the cores grow much faster than the ferromanganese crust (~2 mm Myr^–1). Such a high resolution allows a more precise dating of SN signals compared with previous measurements and thus one may constrain the time period of the incoming shock wave. In contrast to a crust, the formation process of a sediment is well understood. The uptake efficiency, the amount of an element to be incorporated into the sediment, might be up to 100%; in a manganese crust this value is only vaguely determined. If the SN signal is detectable in these two sediments, it allows us to draw conclusions regarding the production rate of the measured radionuclides and transport mechanisms from the SN envelope to Earth. The long-lived radionuclides ²⁶Al, ⁵³Mn, ⁶⁰Fe, with half-lives of the order of the SN event ~2 Myr ago, and the pure r-process element ²⁴⁴Pu are particularly applicable in this case.

More than 15 years ago it was suggested by Ellis et al. (1996) and Korschinek et al. (1996) that there might be a chance of finding long-lived radionuclides in terrestrial archives that were originally produced during a supernova — a possible site for the r-process. The first successful measurement was carried out by Knie et al. (1999) in a ferromanganese crust from Mona Pihoa in the South Pacific. An enhanced ⁶⁰Fe concentration was found in two layers corresponding to a total time range of 0–5.9 Myr, which was later confirmed with improved time resolution in the crust 237KD by Knie et al. (2004) and Fitoussi et al. (2008). This signal was again observed in another crust (29DR-32) recovered from the North Pacific ocean (G. Korschinek, in preparation).

A measurement of ⁶⁰Fe of a marine sediment by Fitoussi et al. (2008) gave less promising results. The core taken from the North Atlantic ocean did not show a clear signal. Fitoussi et al. (2008) suggested an overestimation of the fluence in the crust coming from the uncertainty of the uptake factor. Samples from the Moon have been analyzed by Cook et al. (2009) in terms of ²⁶Al, ⁶⁰Fe, and ¹⁰Be. They state that one sample shows a signal that cannot be explained by the production of ⁶⁰Fe by solar or Galactic cosmic rays. However, the detected signal was one order of magnitude lower than expected.

The presence of another isotope, the r-process radionuclide ²⁴⁴Pu, was investigated at TU Munich in ferromanganese nodules (Wallner et al. 2000) and in a small piece of the same ferromanganese crust as was used for the ⁶⁰Fe measurements (Wallner et al. 2004). Independently, Paul et al. (2001, 2003, 2007) analyzed a marine sediment for possible r-process signatures. All these measurements did not result in significant evidence of SN-produced ²⁴⁴Pu in such terrestrial archives. It is not clear whether only a fraction of SN explosions produce r-process ²⁴⁴Pu at all. The condensation and transport of ²⁴⁴Pu within dust particles in the SN shell into the solar system is also not understood sufficiently.

Future AMS measurements of SN-produced ⁶⁰Fe are planned by Bishop & Egli (2011). By extracting ⁶⁰Fe from magnetite microfossils produced by magnetotactic bacteria in deep-sea sediments, the search for a possible SN signal is extended to biogenic archives.

The red clay piston cores ELT 49-53 and ELT 45-21 might be suitable for detecting possible SN signatures. They were sampled at great depths in the Indian Ocean to minimize contamination from continental input. The ages of the cores were already roughly determined through magnetostratigraphy and biostratigraphy (Allison & Ledbetter 1982). Our sample of ELT 49-53 reaches from a depth of 120–517 cm and that of ELT 45-21 from 398–697 cm below the surface level of the ocean, covering the time range 1.7–3.1 Myr, which includes the time range of the ⁶⁰Fe peak identified in the ferromanganese crust 237KD. Dating with ¹⁰Be gives further details on the age profile. The cores are split into sections of 1 cm length, corresponding to ~3 kyr, with gaps of 8 cm in between, making 71 samples in total.

Adequate chemical methods to extract Be, Al, Mn, Fe, and Pu material (each element separately) from the sediment samples are described in Bourlès et al. (1989) and Merchel & Herpers (1999).

The most sensitive technique to date for detecting long-lived radionuclides is AMS. With this method, the expected small amounts of such radionuclides in terrestrial archives can be quantified.

This isotope is mainly produced by spallation of oxygen and nitrogen by cosmic rays in the Earth's atmosphere. By assuming the cosmic-ray flux was constant over the past few million years, it can be used to date the sections of the core with (Fitoussi et al. 2008)

where T_d is the age of the layer with depth d, C₀ the concentration of ¹⁰Be at the surface or any reference layer with known age and C_d the concentration at depth d. An increase of ¹⁰Be by an enhanced cosmic-ray flux from a SN would be small compared with the cosmogenic production. The yield of the ejected ¹⁰Be mass in a SN explosion is at least a factor of 10³ lower than ⁶⁰Fe (Woosley & Weaver 1995) and thus the SN input 2 Myr ago is estimated to be a factor of 10³–10⁴ lower than the cosmogenic fraction in a piece of sediment of 1 cm thickness. Therefore, quantification of any direct ¹⁰Be in the sediment cores above terrestrial cosmogenic ¹⁰Be is improbable.

The radionuclide ²⁶Al is dominantly produced by the ²⁵Mg(p, γ)²⁶Al reaction in the core H-burning phase of main sequence stars. A contribution in the later stages of massive stars occurs during C and Ne shell burning and explosive Ne burning. Its half-life is t_1/2 = (0.717 ± 0.017) Myr, which is the weighted mean with standard deviation of the values reported in Samworth et al. (1972), Middleton et al. (1983), Norris et al. (1983), and Thomas et al. (1984), as stated in Auer et al. (2009). ⁶⁰Fe is synthesized mainly by neutron capture on ⁵⁹Fe in He and C shell burning and during the explosion, when the blast wave moves through the star, roughly in the same region in which ²⁶Al is produced (Limongi & Chieff 2006).

The ²⁶Al and ⁶⁰Fe signals in a deep-sea sediment core can be estimated by scaling from the ⁶⁰Fe fluence estimation in the ferromanganese crust. The average ratio measured in the crust by Knie et al. (2004) was ⁶⁰Fe/Fe = 1.9 × 10^–15 in layers at a depth of 6–8 mm. This corresponds to a ⁶⁰Fe fluence of 8.2 × 10⁵ atoms cm^–2 after background correction, which was an averaged ⁶⁰Fe/Fe ratio of 2.4 × 10^–16. Two parameters remain uncertain. One is the uptake factor U, as discussed above, the other is the ratio of ⁶⁰Fe/²⁶Al as it enters the solar system in the supernova remnant (SNR). Knie et al. (2004) calculated an uptake efficiency of 0.6%. However, a much higher value of 50–100% is suggested (G. Korschinek, in preparation). The ⁶⁰Fe/²⁶Al flux ratio varies from 0.11–0.17 in observations from the Reuven Ramaty High Energy Solar Spectroscopic Imager (RHESSI) and SPectrometer for INTEGRAL (SPI) and from 0.16–1 in theoretical models (Limongi & Chieff 2006). With

the ²⁶Al fluence is calculated, where (⁶⁰Fe/²⁶Al)_m is the ratio of the fluences one will measure today. Table 1 shows the ²⁶Al fluence values as expected in the sediments for different parameters of U and the ⁶⁰Fe/²⁶Al flux ratio, and Table 2 the corresponding fluence values of ⁶⁰Fe.

Table 1.  Estimated ²⁶Al fluence F₂₆ (atoms cm^–2) in ELT 45-21 and ELT 49-53 with varying uptake factor U and ⁶⁰Fe/²⁶Al flux ratio

Table 2.  Estimated ⁶⁰Fe fluence F₆₀ (atoms cm^–2) in ELT 45-21 and ELT 49-53 with varying uptake factor U

The number of atoms per cm² in the sediment cores is calculated with

where w is the mass fraction, A the mass number of the isotope, and N_A Avogadro’s constant. The density of the cores was estimated to be ρ = 1.5 g cm^–3 and the thickness of the layer is h = 1 cm. Such a layer contains a time interval of ~3 kyr. We assume the sediments contain ~10 wt% of stable Al and ~5 wt% of stable Fe, thus with equation (3) we obtain 3.3 × 10²¹ atoms cm^–2 Al and 8.1 × 10²⁰ atoms cm^–2 of Fe. Therefore, we expect the ⁶⁰Fe/Fe ratio of the signal to be in the range 10^–15–10^–13 and the ²⁶Al/Al ratio between 5 × 10^–17 and 10^–13. The detection limit for measuring ²⁶Al/Al and ⁶⁰Fe/Fe with AMS is ~10^–16, making the most pessimistic values hard to measure. By applying a dedicated chemical sample preparation technique, it is possible to reduce the fraction of stable Al and achieve an isotope ratio 3–10 times higher (Fitoussi & Raisbeck 2007).

In contrast to ⁶⁰Fe, ²⁶Al has a much higher natural background. It is produced either by spallation from cosmic rays on Ar in the Earth's atmosphere or in situ in the sediment. The atmospheric production can on the one hand be obtained by scaling the ¹⁰Be concentration in the ferromanganese crust. Using an initial atmospheric ratio of ²⁶Al/¹⁰Be = 1.89 × 10^–3 (Auer et al. 2009), the background is estimated to be ~5 × 10³ atoms cm^–2 in a section of the core of 1 cm height. Since the uptake of ¹⁰Be is also not well known, this can be considered as a lower limit. On the other hand we use the atmospheric production rate of ¹⁰Be computed by Masarik & Beer (2009) of 0.0209 atoms cm^–2 s^–1 and derive 5 × 10⁵ atoms cm^–2 as an upper limit. A source of in situ produced ²⁶Al is the ²³Na(α, n)²⁶Al reaction. The α-particles are generated in the natural decay chains of ²³²Th, ²³⁵U, and ²³⁸U. If the Na, U, and Th content in the cores is similar to that in the ferromanganese crust, then the production will be in the order of 10⁴ atoms cm^–2, which is an order of magnitude lower than the estimated upper limit of the atmospheric production. The contribution from micrometeorites can be neglected, as was pointed out by Fitoussi & Raisbeck (2007).

The yield of ⁵³Mn ejected by a SN explosion is similar to or up to an order of magnitude higher than that of ⁶⁰Fe (Woosley & Weaver 1995; Rauscher et al. 2002) in the mass range of 11–25 M_⊙ stars (if we neglect the 13-M_⊙ star from Woosley & Weaver 1995). If we take equation (2), replace ²⁶Al with ⁵³Mn and apply the half-life of ⁵³Mn with t_1/2 = (3.7 ± 0.4) Myr (Honda & Imamura 1971), we obtain the results shown in Table 3.

Table 3.  Estimated ⁵³Mn fluence F₅₃ (atoms cm^–2) in ELT 45-21 and ELT 49-53 with varying uptake factor U and ⁶⁰Fe/⁵³Mn flux ratio

⁵³Mn is also produced in cosmic dust by solar cosmic rays. An extraterrestrial flux was calculated by Auer (2008) with ~200 atoms cm^–2 yr^–1 of ⁵³Mn, which sums up to a fluence of 6 × 10⁵ atoms cm^–2 in 1 cm of our sediment cores.

The long-lived radionuclide ²⁴⁴Pu (t_1/2 = (81.1 ± 0.3)Myr, calculated from the α decay half-life of t_1/2 = (81.2 ± 0.3) Myr (Aggarwal 2006) and spontaneous fission: t_1/2 = (66 ± 2) Gyr (Holden & Hoffman 2000) by Lachner et al. (2012)) is a product of the pure r-process in massive stars during a SN explosion. The yield in core collapse supernovae is estimated to be of the order of M_ej = 10^–8 M_⊙ for ²⁴⁴Pu (Lingenfelter et al. 2003). The number of atoms per cm² that reach the Earth is now calculated by spreading the ejected mass over a spherical shell with radius r, which is the distance to Earth (Fields et al. 2005):

with A the atomic mass and m_u the atomic mass unit. Choosing a very rough estimation for the distance of a SN explosion of r = 100 pc, we expect F₂₄₄ ≈ 10⁴ cm^–2. This value is small compared with the expected ⁶⁰Fe fluence, but with the most sensitive AMS facilities for measuring ²⁴⁴Pu it might be possible to detect a SN signal.

If ²⁴⁴Pu is found in the sediment cores, the question remains as to whether it is directly deposited from a single supernova. With its long half-life, ²⁴⁴Pu should be abundant in the ISM through the contribution of multiple nucleosynthesis events. The expanding SN shell accumulates the surrounding medium, which is the low-density medium of the Local Bubble. The hot temperature of the LB of ~10⁶ K (Berghöfer & Breitschwerdt 2002) prohibits the formation of molecules and dust grains. Therefore, it is still an open question as to how ²⁴⁴Pu is collected from the ISM and whether it is condensed into dust grains to be deposited on Earth.

The deposition time of SN ejecta on Earth is calculated using the thin-shell approximation. If we assume that the entire mass is accumulated in the shell, then its thickness d can be estimated with d/r ~ 1/12 for a strong shock into a monoatomic gas (e.g. Clarke & Carswell 2007). With a distance r of the SN explosion of the order of 100 pc, the shell has a thickness of ~8 pc. The SNR expands freely into the ambient medium until the ejected mass is comparable to the swept-up mass of the ISM; subsequently the velocity can be calculated with the Sedov–Taylor solution (e.g. Clarke & Carswell 2007):

The energy of a SN explosion is typically E = 10⁵¹ erg and ρ is the mass density of the LB plasma. This results in a value of 5 × 10⁷ cm s^–1 at 100 pc, which is still a Mach-3 shock. Since the medium inside the LB has a low particle density of n = 5 × 10^–3 cm^–3 and a temperature of T = 10⁶ K (Berghöfer & Breitschwerdt 2002), the pressure is non-negligible. The solution only holds if the kinetic energy of the remnant exceeds the thermal energy of the medium. With pressure P = 2nkT in an ionized gas, where k is the Boltzmann constant, and the adiabatic index γ = 5/3 (e.g. Clarke & Carswell 2007),

Therefore, we obtain for the deposition time . The thin-shell approximation refers to the ISM, which is compressed into a shell by the expanding shock wave. The SN ejecta is accumulated behind this shell, but because M_ej ≪ M_sw, where M_ej is the ejected mass and M_sw the accumulated ISM mass, we expect the time at which the SN ejecta overruns the solar system to be much lower than 15 kyr. It is not clear how the dust grains are transported into the solar system and finally deposited on Earth. These mechanisms might lead to the broadening of a possible signal. Fields et al. (2005) suggest a deposition time of the order of 10 kyr.

We have obtained two sediment cores from the Indian ocean and estimated a possible signal from the SN-produced radionuclides ²⁶Al, ⁵³Mn, ⁶⁰Fe, and ²⁴⁴Pu. In addition, the long-lived radionuclide ¹⁰Be is used to date the sediments. The calculations show that it is possible to measure ²⁶Al, ⁵³Mn, ⁶⁰Fe, and ²⁴⁴Pu with accelerator mass spectrometry, but the cosmogenic background of ²⁶Al and ⁵³Mn might be challenging to distinguish clearly from a SN signal. Because the sediment cores are split into sections of 1 cm length, which corresponds to a time range of ~3 kyr with gaps of 8 cm in between, a chance is given that the signal could be located in one of those gaps, corresponding to ~24 kyr. The combination of two independent sediments lowers this probability.

We will continue the search for SN-produced radionuclides in terrestrial archives by analyzing the two deep-sea sediment cores with respect to the aforementioned SN-produced radionuclides. AMS measurements are planned at the Vienna Environmental Research Accelerator (VERA) facility (University of Vienna), Australian National University (ANU) Canberra, Australia, Technische Universität München, Germany, Helmholtz-Zentrum Dresden–Rossendorf (HZDR) Dresden, Germany, and Eidgenössische Technische Hochschule (ETH) Zürich, Switzerland.

²⁶Al measurements with high time resolution will be performed; with a half-life of 0.717 Myr it is, like ⁶⁰Fe, sensitive to single SN events. A positive signal will allow comparison of its intensity with the observed ⁶⁰Fe signal in the ferromanganese crust (Knie et al. 2004). It provides a check of late stages in massive stars and a comparison with space-borne observations of γ-ray intensities in our Galaxy (see e.g. Diehl et al. 2010).

In parallel to ²⁶Al, the SN-produced radionuclides ¹⁰Be, ⁵³Mn, and ⁶⁰Fe will be measured in the same sections. ¹⁰Be is not sufficiently produced and ejected in a SN explosion, but it is constantly produced in the Earth's atmosphere and can be used to date the sections of the sediment cores.

The search for pure r-process ²⁴⁴Pu will be continued with new developments that improve the overall efficiency in AMS measurements, which was the prime limitation on the significance of the previous ²⁴⁴Pu data. While still only upper limits can be given for SN-produced ²⁴⁴Pu on Earth, the goal would be an unambiguous detection of extraterrestrial ²⁴⁴Pu, i.e. direct evidence that r-processes indeed happened within the last ~100 Myr and can be identified. The ²⁴⁴Pu signal was calculated for a SN occurring at a distance of 100 pc. It is very likely that the ⁶⁰Fe signal identified by Knie et al. (2004) and Fitoussi et al. (2008) originated from a SN closer than 100 pc to the solar system (Neuhäuser et al. 2011); therefore we might be able to find a higher signal than estimated here.

If all of the above-mentioned elements, which are all ejected in SN explosions, are detected, a clear link to their supernova origin will be provided. The time resolution of the sediment cores is 1000 times higher than for the ferromanganese crust, which will help to narrow the time range of the explosion in the solar vicinity.

It is not yet clear how dust is formed in SN ejecta and how the grains are transported to Earth. The radionuclides are only able to overcome the ram pressure of the solar wind and the interplanetary magnetic field if they are condensed into dust particles (Athanassiadou & Fields 2011). The dust condensation efficiency in SNRs is a critical value and not very well known yet. It ranges from 10^–4 in observations to 1 in theoretical simulations (Ouellette et al. 2010). Comparing our measurements with theoretical SN models should give an insight into the formation of dust particles in SNRs and nucleosynthesis scenarios in massive stars and during the explosion.