Introduction

Unique signatures of supernova explosions can come from long-lived radioisotopes (with half-lives of the order of a million years). These freshly synthesized isotopes are ejected by the supernova and live long enough to either travel all the way to the solar system or emit radiation that does. Two such radioisotopes are 60Fe and 26Al with half-lives on the order of a million years. They have been detected in a number of Earth and space-based measurements: presolar stardust grains1, cosmic rays2,3, γ-ray measurements of radioactivities in the galaxy4,5,6,7,8,9,10, material deposited in deep-ocean crusts11,12,13,14,15,16,17,18 and on the surface of the moon19, indirect signatures in meteorites of their presence during the formation of the solar system20,21, and more. In addition to individual detections, the 60Fe/26Al ratio has been identified as an excellent probe to study nucleosynthesis in massive stars10,22. This ratio, extracted from γ-ray observations, can constrain stellar mixing processes and rotation associated with slow neutron capture process nucleosynthesis, as well as explodability of supernova models10. The latter impacts predictions of the neutron star and black hole mass distributions probed by gravitational waves. The ratio has also been used as a constraint in the study of the origin of the solar system, its possible pollution by a nearby supernova10, and its place and evolution within the local super bubble23. While accurate measurements of this ratio exist in the Galaxy and the early solar system9, models tend to strongly overestimate it22,24 and this discrepancy has been an unresolved mystery for several decades.

Both isotopes are predominantly produced in massive stars10,22,24,25, with small contributions from other sites such as novae, asymptotic giant branch (AGB) stars, type Ia supernovae, and electron capture supernova26,27,28. However, the two isotopes are synthesized in different parts of the massive star. While both are produced in the inner layers of the star, which are only ejected during the supernova explosion, 26Al is also produced in H-burning in the outer layers of the star. H-burning material can be ejected earlier due to stellar winds, which adds to the 26Al yield. For these reasons, the 60Fe/26Al ratio is a powerful tool for understanding the evolution and explosion of massive stars: if both isotopes are produced by the same source(s), then the source distance, location, and number cancel out, giving direct access to the stellar yield ratio right after the supernova explosion10. The same arguments could be applied to other ratios of supernova products, however, what makes the 60Fe/26Al ratio unique is the fact that it involves isotopes and not elements. Ratios of elements would have contributions from multiple stellar processes, while the 60Fe/26Al ratio provides a direct connection to the evolution of the massive star, stellar winds, and supernova explosion mechanisms. In fact, because the contributions of the two isotopes come from different parts of the star, a robust model that can match the 60Fe/26Al ratio indicates a good description of the stellar environment across a wide range of stellar material, which would be a significant accomplishment for the field.

The importance of the 60Fe/26Al ratio has been discussed extensively in the literature. For example, parameters such as stellar rotation and explodability (the ability of the star to undergo explosion) have been shown to impact the final 60Fe/26Al ratio values10,29. In these studies, a common theme appears in the discussion of model uncertainties, namely the uncertain nuclear reaction networks that produce/destroy the two relevant isotopes10,22,24,30,31. The nuclear reaction uncertainties related to the synthesis of 26Al are less extensive and were discussed in detail by Diehl et al.10.

The nuclear reaction network that produces 60Fe is relatively small (Fig. 1(a) inset). It consists of a series of neutron-capture reactions starting at the stable isotope 58Fe, which compete with either β decays or (p,n) reactions, depending on the astrophysical conditions31. In addition, the 22Ne(α,n)25Mg reaction is the dominant source of neutrons in massive stars32. Most reactions are well constrained as discussed in the recent review by Diehl et al.10. The most impactful and simultaneously most uncertain reaction in this network is the 59Fe(n,γ)60Fe reaction. This reaction is the dominant 60Fe production mechanism, and the 60Fe yield was shown to scale linearly with the reaction cross section30. It is challenging to measure this neutron-capture directly in the lab due to the short half-life of the target nucleus 59Fe (44 days)33. Therefore, indirect techniques have been used in the past to provide experimental constraints34,35. Each of the previously used techniques has its own limitations and uncertainties, however, they both have a common blind spot, namely the low-energy behavior of the γ-ray strength function (gSF).

Fig. 1: Experimental results.
figure 1

a Maxwellian Averaged Cross Section (MACS) of the 59Fe(n,γ)60Fe reaction as a function of neutron energy. The hatched and light-green bands represent previous results from Uberseder et al.34 and Yan et al.35. The default MACS used in astrophysical calculations from the non-smoker reaction code is shown as a dashed blue line. The results of the present work are represented by the purple band. For comparison purposes, the lower limit of our results without including the low-energy enhancement in the gSF is shown as a solid purple line. a inset: reaction network for the production/destruction of 60Fe in a massive star. b γ-ray strength function showing the presence of a low-energy enhancement in 60Fe, compared to 56Fe from previous works38,39. The vertical lines crossing each data point represent the uncertainties of the measurements.

Full size image

The gSF represents the reduced probability of the nucleus to emit a γ ray of certain energy and multipolarity36. It is one of the most essential quantities used in calculating neutron-capture reaction cross-sections of heavy nuclei37. The gSF has been studied for many decades, both experimentally and theoretically, mostly for stable isotopes. During the last decade, measurements of the gSF at low energies revealed a new phenomenon38, the so-called “low-energy enhancement” or “upbend”. The low-energy enhancement was shown to have a dipole character39, but it is still unclear whether it is of electric (E1) or magnetic (M1) nature40,41. Theoretical and experimental investigations show a dependence of the low-energy enhancement on the underlying nuclear structure41,42, becoming more significant near closed nuclear shells and gradually being reduced in more deformed nuclei. The impact of the low-energy enhancement on neutron-capture reactions was also investigated43, however its effects vary strongly from reaction to reaction.

Here we present an experimental investigation of the gSF in 60Fe and its impact on the 59Fe(n,γ)60Fe reaction cross-section. With our measurement we show that the reaction cross section is significantly higher than previously thought, which leads to the conclusion of an enhanced 60Fe production in massive stars. Our result shows that the discrepancy between models and observations in the 60Fe/26Al ratio persists despite the stronger nuclear physics constraints.

Results

We performed an experiment to investigate the gSF in 60Fe, especially at low energies using the β-Oslo method44,45 (see Methods section). The resulting gSF is shown in Fig. 1b (red squares) together with the extrapolation of the previous 60Fe study34 (solid black line) and two measurements of the 56Fe isotope for comparison38,39. It can be seen that similar to 56Fe, our results show the presence of a significant low-energy enhancement. This, in turn, results in a significant increase of the 59Fe(n,γ)60Fe reaction Maxwellian Averaged Cross Section (MACS) (Fig. 1a) compared to the previous measurements34,35 and compared to the recommended theoretical calculations (non-smoker)46 used in astrophysical models. More specifically, our lower and upper limits are factors of 1.6 and 2.1 higher than the recommended value. As a consistency check, we performed the MACS calculation removing the low-energy enhancement from our gSF (Fig. 1a, purple solid line) and keeping all other parameters the same. This calculation enables a more comparison with the Uberseder et al. measurement. Although other parameters, like the NLD, are not identical, it can be seen in Fig. 1a that by removing the low-energy enhancement the two measurements are consistent. The main difference between our result and the one from Uberseder et al. comes from the fact that the latter measurement was not sensitive to the presence of low-energy enhancement. The study by Yan et al.35 also measured at higher energies and used theoretical models to extrapolate to the astrophysically relevant energy. In their gSF extrapolation they also assumed no low-energy enhancement, which could explain the lower cross-section compared to the present work.

Discussion

To investigate the impact of the new 59Fe(n,γ)60Fe reaction cross section on the production of 60Fe we ran astrophysical calculations using the lower and upper limits of the new rate (Fig. 2). We evolved solar metallicity stellar models with initial zero-age main-sequence mass of 15, 20, and 25\({M}_{\odot }\) using the one-dimensional stellar evolution toolkit, Modules for Experiments in Stellar Astrophysics (MESA)47,48. For comparison, we also evolved a baseline model using the default choice of the 59Fe(n,γ)60Fe reaction rate in MESA adopted from Rauscher & Thieleman46 using the non-smoker Hauser-Feshbach model. Figure 2 shows the mass fraction profiles from the resulting 15\({M}_{\odot }\) and 20\({M}_{\odot }\) MESA stellar models as a function of Lagrangian mass coordinates for 60Fe at the pre-supernova stage (at the start of iron core-collapse). In agreement with Jones et al.30 we find that the dominant pre-supernova regions for 60Fe production is primarily in the He-, C-, and Ne-burning shell regions. The results from the 25\({M}_{\odot }\) models are very similar.

Fig. 2: 60Fe production in massive star simulations.
figure 2

a Calculation of a 15\({M}_{\odot }\) star, and b of a 20\({M}_{\odot }\) star. The green solid line represents calculations done using the default MACS for the 59Fe(n,γ)60Fe reaction46, while the blue dot-dashed and yellow dashed lines show the results using the upper and lower limits from the present work.

Full size image

As shown in Fig. 2, in agreement with previous studies30, the production of 60Fe both during the massive star evolution and during the supernova depends strongly on the 59Fe(n,γ)60Fe reaction rate. In addition, based on the study by Jones et al.30, the increase in the 59Fe(n,γ)60Fe reaction rate is expected to propagate throughout the stellar evolution and supernova explosion and impact the final ejected 60Fe yield. Such a significant increase in the production and ejection of 60Fe in massive stars has crucial implications in the interpretation of the different observations of 60Fe mentioned earlier, such as in ocean sediments, the surface of the moon, stardust, and cosmic rays. Here we focus on the impact on the 60Fe/26Al ratio.

The presently accepted value of the 60Fe/26Al ratio within the galaxy, based on γ-ray observations9, is 0.184+/− 0.042. Astrophysical calculations predict a wide range of values for this ratio. Most massive star models10,22,24 tend to overpredict this ratio by approximately a factor of 3-10. The disagreement between models and observations is often attributed to uncertain nuclear physics input in the models9,22. A recent investigation of the 59Fe β-decay rate in stellar environments indicated a reduction in the 60Fe/26Al ratio, bringing the models closer to the observations31. However, our results point to an increased production of 60Fe, resulting in a higher value of the 60Fe/26Al ratio in models. The firmer constraints on the nuclear reaction network from the present work show that a nuclear solution to this problem is unlikely. This suggests that the solution to the puzzle should come from the description of the astrophysical environment and processes that affect the two isotopes.

The importance of 60Fe and 26Al can further be highlighted when considering that they are produced in different parts of the massive star. The yield of 26Al in stellar winds was shown to be affected by the mass of the star, the stellar rotation, and the presence of a companion10,49. On the other hand, since 60Fe is ejected only by the supernova, its yield is intricately connected to the explodability of the star, which in turn is connected to black hole formation mass distribution. If a massive star does not explode, then the contribution to 60Fe is practically zero. Supernova studies exhibit large variation in stellar explodability, which results in significant variation of the 60Fe/26Al ratio24,29. For example, Pleintinger et al.29 found that the 60Fe/26Al ratio is significantly reduced in the case where no supernova exploded, thus no material ejection, above 2 5M⊙50. Limongi and Chiefi24 also showed that for some of their models, the 60Fe/26Al ratio decreased for lower mass limits. Therefore, the increased production of 60Fe found in the present work can be balanced by assuming a lower mass limit for supernova explosions in the models.

A second stellar parameter found to affect the 60Fe/26Al ratio significantly is stellar rotation. Increased stellar rotation was found to increase 26Al yields due to increased mass loss50. However, stellar rotation was shown to have an even larger impact on 60Fe yields. This is because stellar rotation causes an enlarged C-burning shell, which in turn increases neutron production10. Since neutron-capture reactions are the dominant mechanism of producing 60Fe, stellar rotation causes an increase in 60Fe yield, larger than the increase found for 26Al. As a result, the calculated 60Fe/26Al ratio when including stellar rotation is higher than for non-rotating stars29. Therefore, the increased 60Fe production found in the present work, which causes an even higher 60Fe/26Al ratio than previously predicted, cannot be explained by introducing stellar rotation to the models.

Additional stellar parameters can affect the 60Fe/26Al ratio and further investigations are needed to shed light on this puzzle. One example that was shown to have a significant impact is the explosion energy40. Jones et al.40 showed that the dependence of the 60Fe/26Al is different for 15, 20, or 25\({M}_{\odot }\) stars, therefore additional studies are needed to understand the impact in the overall 60Fe/26Al ratio.

In the present work, we investigated the production of 60Fe in massive stars. Our results provide the most complete estimate of the 59Fe(n,γ)60Fe reaction, including the low-energy enhancement in the γ-ray strength function. We found a higher reaction rate compared to previous measurements and the recommended theoretical value used in astrophysical models. The increase in the reaction rate resulted in an increase in the production of 60Fe in massive stars by almost a factor of 2. We investigated the impact of our results on the 60Fe/26Al ratio, which was identified as a sensitive probe for exploring stellar evolution and supernova dynamics. Before our measurement, the discrepancy between observations and theoretical models in the value of the 60Fe/26Al ratio was attributed to uncertain nuclear physics. While uncertainties in the nuclear physics aspects still remain, our result removes one of the most significant uncertainties in the 60Fe production. However, the discrepancy persists and is even larger. The solution to the puzzle must come from stellar modeling by, for example, reducing stellar rotation, assuming smaller explodability mass limits for massive stars, or modifying other stellar parameters.

Methods

The goal of the experiment was to constrain the 59Fe(n,γ)60Fe reaction. In the absence of a direct measurement, experimental techniques focus on constraining the nuclear properties used by theory to calculate the reaction cross section using the Hauser-Feshbach statistical model51. These properties are the neutron-nucleus optical model potential (a description of the interaction between the neutron and the nucleus), the nuclear level density - NLD (the number of energy levels per unit energy as a function of excitation energy, spin, and parity), and the γ-ray strength function - gSF (the probability to emit a γ ray of a particular energy and multipolarity). For a recent review of indirect neutron capture constraints for astrophysical processes the reader is referred to Larsen et al.37. For neutron capture reactions near the valley of stability the optical model potential is relatively well constrained52,53. The other two quantities, however, namely the NLD and gSF are significantly more uncertain, reaching up to two orders of magnitude variation in their theoretical prediction45. Here we focus on constraining the NLD and gSF for 60Fe and in this way constrain the 59Fe(n,γ)60Fe reaction cross section.

In the present work we used a 60Mn radioactive beam to populate excited states in 60Fe via β decay and extract its NLD and gSF. To do this we applied the β-Oslo method44,45, a variation of the traditional Oslo method54,55,56, which extracts the NLD and gSF in nuclei populated using various nuclear reactions. The β-Oslo method populates the nucleus of interest using β decay, which was introduced ten years ago for constraining neutron-capture reactions on nuclei far from stability. Here, we use the β-Oslo method to extract the NLD and gSF of 60Fe and constrain the cross section of the 59Fe(n,γ)60Fe reaction.

The experiment took place at the National Superconducting Cyclotron Laboratory at Michigan State University. A 64Ni primary beam was accelerated through the Coupled Cyclotron Facility to an energy of 140 MeV/u and impinged on a 510 mg/cm2 thick Be target. The produced cocktail beam centered around 60Mn was separated using the A1900 fragment separator57 and delivered to the gas stopping facility58, where it was slowed down and purified. A pure 60Mn beam at an energy of 30 keV was implanted into a Si surface barrier detector. The 60Mn beam consisted of 59(1)% ground state (Jπ = 1+, Τ1/2 = 0.28 s) and 41(1)% isomeric state (Jπ = 4+, Τ1/2 = 1.77 s) (https://www.nndc.bnl.gov). The identification of the two different states was done based on unique γ-ray signatures and their different half-lives. Both states β decay into different excited states of 60Fe, and the 1+ state also decays directly to its ground state.

The β-decay electrons were detected in the Si detector, and the emitted γ rays were detected in the Summing NaI(Tl) (SuN) detector59. SuN is a cylindrical total absorption spectrometer segmented into eight optically isolated segments. The total energy deposited in SuN (recorded as the event-by-event sum of all segments) provides the excitation energy, Ex, of 60Fe populated in each β decay. The individual segments are sensitive to the energy of the individual γ rays, Eγ, emitted during the deexcitation of each Ex. For our analysis we use a two-dimensional (2D) matrix of Ex vs Eγ, which includes all γ rays detected in our experiment following the β decay of 60Mn into 60Fe (Fig. 3). The matrix in Fig. 3a shows strong population of discrete states up to Ex = 3 MeV. This region was not included in our analysis to avoid strong transitions between discrete levels as the method is not valid for non-statistical γ decay.

Fig. 3: Raw experimental data and NLD for 60Fe.
figure 3

a Two dimensional raw matrix showing the γ-ray detection from the β decay of 60Mn into 60Fe. b NLD of 60Fe extracted in the present work (red squares) compared to the known discrete levels and theoretical calculations. Vertical lines crossing each of the red squares represent the uncertainties of the present measurement. RIPL3 (grey dashed line) refers to the recommended NLD from the RIPL3 library63. RIPL3 corrected (dashed red line) corresponds to the recommended NLD, adjusted to match the experimental data. The blue dotted line corresponds to the NLD calculated using the Back Shifted Fermi Gas model (BSFG)60, while the cyan dot-dashed line shows the NLD calculated by Demetriou et al.61.

Full size image

The starting point for the β-Oslo method analysis was the 2D γ-ray matrix mentioned earlier (Ex vs Eγ). The matrix was first unfolded with the response of the SuN detector55, and then, following an iterative subtraction process54, the primary γ-ray distribution was extracted. Primary γ rays are the first γ rays emitted in the deexcitation of an Ex bin, and their distribution depends on the functional form of the NLD and gSF56. The final step in this analysis was the normalization of the NLD and gSF using external information in order to extract absolute values for these quantities. Here we used the broad range of known discrete levels, up to \(\approx\)4.0 MeV (Fig. 3) to normalize our NLD, which allowed us to fix the slope both for the NLD and the gSF. Because the 60Mn beam consisted of both the 1+ ground state and the 4+ isomeric state, the populated NLD corresponded to a broad spin distribution. Assuming allowed β decays from the two initial spins, followed by dipole γ-ray emission, we expect to populate states in 60Fe with spins 0 – 6 of both parities. This range overlaps completely with the spin range expected to be populated in the neutron-capture on the 3/2- ground state of 59Fe. It should be noted that the 1+ ground state of 60Mn feeds the 60Fe ground state with a significant probability. High energy electrons from this gs-gs decay create a background in the SuN detector and for this reason, the primary γ-ray component that feeds the 60Fe ground state was excluded from the analysis.

The resulting NLD is shown in Fig. 3b (red squares) together with the known discrete levels (black solid line). For comparison, three commonly used theoretical NLD models are shown as well. The Back Shifted Fermi Gas (BSFG) is a phenomenological model often used in global NLD calculations60. The BSFG model seems to be in reasonable agreement with the experimental results in the statistical region. Other commonly used models are two semi-microscopic models61,62, which are normalized to experimental NLD data where available, or to known discrete levels. In Fig. 3b the model labeled “Demetriou 2001”61 uses the Hartree-Fock-Bogoliubov plus Bardeen–Cooper–Schrieffer approach in a parity-independent way. The second model62 was adopted by the RIPL3 library63 (and is therefore labeled as such) and it uses a Hartree-Fock-Bogoliubov plus combinatorial approach. In Fig. 3a we show an optimized normalization of the RIPL3 model, which matches the experimental data. Similar discrepancies from the adopted normalization used in RIPL3 have been observed in other nuclei previously64. Since the experimental data do not extend all the way to the neutron-separation energy (Sn=8.8 MeV), all three models shown in Fig. 3 were used to extract the 59Fe(n,γ)60Fe reaction cross section and the results were included in the uncertainty band shown in Fig. 1a.

Fixing the NLD slope allowed for the gSF slope to also be fixed, since in the Oslo method the slopes of the two quantities are directly linked. The final step in our analysis was the absolute normalization of the gSF. This normalization was taken from Uberseder et al.34 in the energy region between 5.4 and 6.5 MeV. The resulting gSF is shown in Fig. 1b of the main article.

The extracted NLD and gSF were used as input in the TALYS1.95 statistical model code65 in order to extract the Maxwellian Averaged Cross Section (MACS). The MACS results are shown in Fig. 1a of the main article. The uncertainty band includes analysis and statistical uncertainties, as well as uncertainties from the NLD extrapolation mentioned above and the gSF normalization. The final MACS results for the relevant temperatures are shown in Table 1, while the final reaction rates are shown in Table 2.

Table 1 MACS upper and lower limits at relevant temperatures
Full size table
Table 2 Reaction rate upper and lower limits as a function of temperature
Full size table