A&A 469, 1005-1012 (2007)
DOI: 10.1051/0004-6361:20066982

SPI observations of the diffuse 60Fe emission in the Galaxy

W. Wang1 - M. J. Harris1 - R. Diehl1 - H. Halloin2 - B. Cordier4 - A. W. Strong1 - K. Kretschmer1 - J. Knödlseder3 - P. Jean3 - G. G. Lichti1 - J. P. Roques3 - S. Schanne4 - A. von Kienlin1 - G. Weidenspointner3 - C. Wunderer5

1 - Max-Planck-Institut für extraterrestrische Physik, Postfach 1603, 85740 Garching, Germany
2 - AstroParticule et Cosmologie (APC), 11 place Marcelin Berthelot, 75231 Paris, France
3 - Centre d' $\acute{\rm E}$tude Spatiale des Rayonnements, BP 4346, 31028 Toulouse Cedex 4, France
4 - DSM/DAPNIA/SAp, CEA Saclay, 91191 Gif-sur-Yvette, France
5 - Space Sciences Laboratory, UC Berkeley, CA 94720, USA

Received 20 December 2006 / Accepted 26 April 2007

Aims. Gamma-ray line emission from radioactive decay of 60Fe provides constraints on nucleosynthesis in massive stars and supernovae.
Methods. spectrometer SPI on board INTEGRAL has accumulated nearly three years of data on gamma-ray emission from the Galactic plane. We have analyzed these data with suitable instrumental-background models and sky distributions to produce high-resolution spectra of Galactic emission.
Results. We detect the $\gamma$-ray lines from 60Fe decay at 1173 and 1333 keV, obtaining an improvement over our earlier measurement of both lines with now 4.9$\sigma $ significance for the combination of the two lines. The average flux per line is (4.4 $\pm $ 0.9) $\times $ $10^{-5}~{\rm ph}~{\rm cm}^{-2}~{\rm s}^{-1}~{\rm rad}^{-1}$ for the inner Galaxy region. Deriving the Galactic 26Al gamma-ray line flux with using the same set of observations and analysis method, we determine the flux ratio of $^{60}{\rm Fe}/^{26}{\rm Al}$ gamma-rays as 0.148 $\pm $ 0.06.
Conclusions. The current theoretical predictions are consistent with our result. We discuss the implications of these results for the widely-held hypothesis that 60Fe is synthesized in core-collapse supernovae, and also for the closely-related question of the precise origin of 26Al in massive stars.

Key words: ISM: abundances - nuclear reactions, nucleosynthesis, abundances - gamma rays: observations

1 Introduction

The radioactive isotope 60Fe is believed to be synthesized through successive neutron captures on Fe isotopes (e.g. 56Fe) in a neutron-rich environment inside massive stars, before or during their final evolution to core collapse supernovae (CCSN). Due to its long decay time ( $\tau \simeq$ 2.2 My), 60Fe survives to be detected after the supernova ejected it into the interstellar medium, by $\beta$-decay via 60Co and $\gamma$ emission at 1173 keV and 1333 keV - like other radioactive isotopes: 44Ti, 56,57Co, and 26Al. These isotopes provide evidence that nucleosynthesis is ongoing in the Galaxy (with CCSN occurring at intervals $\sim$30-100 yr, The et al. 2006; Diehl et al. 2006). The discoveries of the above isotopes have fulfilled predictions of nucleosynthesis theory, at least in its general outline. The 44Ti and 56,57Co isotopes have relatively short half-lives ($\leq$100 years) and are thus detected as point-like sources, e.g. in young type II supernova remnants (SNRs) Cas A and 1987A, respectively (Iyudin et al. 1994, 1995; Matz et al. 1988; Kurfess et al. 1992). With its much longer half-life ($\sim$106 years), 26Al may propagate over significant distances of $\simeq$few hundred pc, and accumulates in the interstellar medium (ISM) from many supernovae, until injection and $\beta$-decay are in balance in the ISM, giving rise to a diffuse and even Galaxy-wide glow (Mahoney et al. 1982; Prantzos & Diehl 1996). The behavior of 60Fe from the same massive-star sources should follow that of 26Al, since its half-life is similar, $\sim$2.2 $\times $ 106 years. In addition, 60Fe could also be produced in substantial amounts by rare subtypes of SN Ia (Woosley 1997), which would then be point sources of 60Fe gamma-rays.

Although the detections of these isotopes are in agreement with the broad outlines of CCSN nucleosynthesis theory, there are discrepancies in the details. For example, 44Ti lives long enough so that it should have been detected from several recent Galactic supernovae if these occur at a rate of $\sim$2 per century; but with the exception of Cas A (Iyudin et al. 1994; Renaud et al. 2004, 2006), no other SNRs have been detected yet. Mapping of the diffuse 26Al emission (Knödlseder et al. 1999; Plüschke et al. 2001) confirms that it follows the overall Galactic massive star population, but concentrations of 26Al are seen in very young OB associations where even the most massive stars are not expected to have exploded (e.g. Cygnus OB associations). This has led to the suggestion that most 26Al production is associated with ejections in an earlier, pre-explosion phase of stellar evolution (Knödlseder et al. 1999).

Measurements of 60Fe promise to provide new information about these issues, specifically the massive star nucleosynthesis in the late pre-supernova stages. RHESSI reported observations of the gamma-ray lines from 60Fe with 2.6$\sigma $ significance, and an average flux of (3.6 $\pm $ 1.4) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}}$ (Smith 2004). The analysis of the first year of data from the SPI spectrometer on ESA's INTEGRAL spacecraft resulted in a similarly marginally-significant detection of these $\gamma$-ray lines from 60Fe ($\sim$$3 \sigma$, Harris et al. 2005), with an average line flux of (3.7 $\pm $ 1.1) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}}$. SPI on INTEGRAL has accumulated more data since then; here we analyzed 2.5 years of data, aiming at a consolidation of the INTEGRAL/SPI measurement of 60Fe gamma-rays.

In the following section, we describe our observations with SPI. In Sect. 3, we describe the steps and methods of data analysis. The scientific results, including the spectra and intensities of both 1173 keV and 1333 keV lines, and the flux ratio of 60Fe/26Al are presented in Sect. 4. Discussions and a summary follow in Sect. 5.

2 Observations

The INTEGRAL spacecraft was launched on October 17, 2002, into a high-inclination, high-eccentricity orbit intended to avoid the increased background from the Earth's trapped radiation belts. INTEGRAL's orbital period is 3 days. The spectrometer SPI consists of 19 Ge detectors actively shielded by a BGO anti-coincidence shield. It has a tungsten coded mask in its aperture which allows imaging at $\sim$$3^{\circ}$ resolution within a  $16^{\circ}$ $\times $ $16^{\circ}$ full coded field of view (imaging on INTEGRAL is mainly performed at lower energies by the IBIS telescope, with which SPI is co-aligned). The Ge detectors are sensitive to gamma-rays between 15 keV and 8 MeV, with a total effective area $\sim$70 cm2 at 1 MeV, and achieve an energy resolution of $\sim$2.5 keV at 1 MeV (Roques et al. 2003; Attié et al. 2003). However, cosmic-ray (CR) impacts degrade this resolution over time, and the instrument is periodically shut off for a few days while annealing (by heating from cryogenic temperatures to 100 $^{\circ}$C) is applied to the detectors to restore the energy resolution by thermal curing heating of the CR-induced defects (Roques et al. 2003; Leleux et al. 2003).

In space operations, INTEGRAL with its IBIS and SPI telescopes is pointed at predesignated targets, with a fixed orientation for intervals of typically $\sim$2000 s (referred to as pointings), which are successively arranged as a standard pattern of neighbouring pointings $\simeq$$2^\circ$ apart (dithering), and covering target region of interest for improved imaging (Courvoisier et al. 2003).

The interval covered by the observations analyzed here is December 3, 2002-August 22, 2005 (orbits 17-359). The instrument was in operation for most of this time, but important gaps were caused by five of the annealing episodes described above, and by the regular perigee passages with gaps due to the radiation belts. The live time was further reduced by the failure of two of the 19 detectors during this interval (December 2003 and July 2004) and by the occurrences of solar flares.

3 Data analysis

The basic measurement of SPI consists of event messages per photon triggering the Ge detector camera. We distinguish events which trigger a single Ge detector element only (hereafter single event, SE), and events which trigger two Ge detector elements nearly simultaneously (hereafter multiple event, ME), which is the case for $\simeq$$30\%$ of 60Fe gamma-ray photons.

The analysis steps are: (1) Assembly of data selected to be free of contamination by, e.g., solar-flare events. (2) Modelling the instrumental background. (3) Fitting the measured data in narrow energy bins with the background model and a model of celestial gamma-ray emission, folded through the instrumental response into the data space of the measurement; the fitted amplitude of the celestial model per energy bin then comprises the spectrum of observed sky emission. (4) Deriving line parameters of the celestial signal. From the two event types, and from the two gamma-ray lines of 60Fe, we obtain four independent measurements of 60Fe emission.

In the following, we discuss the analysis steps and their implementation in the corresponding software utilities within MPE's INTEGRAL data analysis system.

3.1 Data selection and assembly

We test data quality per each "science window'' (typically a time interval of $\sim$30 min corresponding to one pointing), applying selection limits to "science housekeeping'' parameters such as the count rates in several onboard radiation detectors, instrument status codes, data ownership, and orbit phase (utility  spiselectscw). We employ the INTEGRAL Radiation Environment Monitor (IREM; Hajdas et al. 2003), the SPI plastic scintillator anticoincidence counter (PSAC), and the rate of saturating events in SPI's Ge detectors (from events depositing >8 MeV in a detector; hereafter referred to as GEDSAT rates) to exclude solar-flare events and other erratic background increases. Regular perigee background increases are additionally eliminated through a 0.05-0.95 window on orbital phase. From the selected events, spectra are accumulated per each detector and pointing, and together with dead time and pointing information assembled into the analysis database.

We establish databases of spectra per pointing per detector for single and multiple events (SE and ME), in spectral ranges with 1 keV bin size in $\sim$20 keV bands around each of the 60Fe lines, and in adjacent energy bands for determination of instrumental background (see below). Typical such raw data spectra around two $\gamma$-ray lines are presented in Fig. 1, and show that instrumental background including strong instrumental lines dominates the raw spectra. The energy ranges around the two lines extend from 1153-1193 keV and 1313-1359 keV. For the background modelling from the adjacent continuum, we use 1163-1169 keV plus 1177-1184 keV, and 1318-1328 keV plus 1336-1349 keV energy bands, respectively. From orbits 17 to 359, a total observation time of $\sim$24 Ms is thus obtained, with a data set consisting of 14 623 pointings with one spectrum per each detector and event type.

\includegraphics[width=6.1cm,clip]{6982f1b.eps}\end{figure} Figure 1: Raw data spectra around the energies of two 60Fe lines in one-INTEGRAL-orbit observations (3 days), representing the instrumental lines and continuum background. For the 1173 keV case ( left), three strong instrumental lines are obvious: 44Sc (1157 keV), 60Co (1172.9 keV), 182Ta (189.4 keV), and the 60Co line blends with the 60Fe line (Weidenspointner et al. 2003). For the 1333 keV case ( right), the 60Co background line (1332.5 keV) and the other strong instrumental line of 69Ge (1336.8 keV) blend with the 60Fe line, an instrumental line at 1347 keV also comes from 69Ge electron captures.
Open with DEXTER

\includegraphics[width=5.65cm,clip]{6982f2b.eps}\end{figure} Figure 2: Background coefficients for the 60Co radioactivity build-up model around the 1173 keV ( left) and 1333 keV ( right) lines, respectively.
Open with DEXTER

3.2 Background modelling

SPI spectra are dominated by the intense background radiation characteristic of space platforms undergoing CR bombardment. Much of this radiation is prompt, resulting directly from CR impacts, whose variation with time ought to follow that of the incident CR flux. Other components arise from radioactive isotopes produced by the CR impacts, whose decay lifetimes are long compared to the coincidence window of the detector-anticoincidence electronics. Radioactive isotopes will increase in abundance until CR production balances $\beta$-decay. A time series of their $\gamma$-ray emission will be the convolution of the time behavior of the prompt CR source with an exponential decrease from decay. Local radioactivity in the spacecraft and instruments themselves thus will generate both broad continuum background emission and narrow gamma-ray lines from long-lived radio-isotopes. Varying with energy, background components may exhibit complex time variability due to their origins from more than one physical source.

We derive models for the background contribution per energy bin and pointing in each detector from comparisons to presumed "tracers'' of background; for these, we use independent contemporaneous measurements aboard INTEGRAL (see Halloin et al. 2007; Diehl et al. 2006; also background model studies by Sizun et al. 2007). Comparing different candidate tracers to observed background variations, we find the best tracer(s) of background for each energy band. Candidate tracers with high statistical precision are the rates measured in SPI's plastic scintillator anticoincidence detector, the rate of saturated events in the BGO shield, the detector-by-detector GEDSAT rates (from events depositing >8 MeV in a Ge detector), and Ge-detector rates integrated over rather wide energy bands (for better statistical precision). For our analysis, we make use of the GEDSAT rates tracing the prompt CR activation to the 60Fe lines (see above).

Each such tracer is regarded as a "template'' for time variability of background. By normalizing such a template to the set of counts per detector spectrum (i.e. per pointing), a background model can be constructed, which is applicable to the analysis data set of spectra. We construct our background model from two components, one derived from adjacent energy bands, and an additional component modelling instrumental line contributions.

\includegraphics[width=7.8cm,clip]{6982f3b.ps}\end{figure} Figure 3: Two plausible sky distribution models adopted for diffuse 60Fe emission from the Galaxy. Both have been derived from COMPTEL 26Al measurements (Plüschke et al. 2001; Knödlseder et al. 1999). The image obtained with the Maximum-entropy method ( left) shows more structure, while the image derived with the Multi-resolution Expectation Maximization Method (MREM) appears smooth from its noise-filtering, accepting only image structure which is enforced by the measurements.
Open with DEXTER

The most important radioactive background line component for our analysis is the emission from 60Co decay inside the instrument, whose two lines at 1173 and 1333 keV are actually part of the celestial 60Fe decay chain we want to measure. The convolution of the GEDSAT CR source tracer with an exponential function  ${\rm e}^{-t/\tau}$, in which the decay time $\tau$ is that of 60Co (7.6 yr), is expected to give a good background model for energy bins containing contributions from these lines, i.e. around 1173 and 1333 keV (see Fig. 2). Another possible radioactive contaminant in our spectral region of interest may be the strong 69Ge K-shell electron capture line (1337 keV, Weidenspointner et al. 2003) which blends into the 1333 keV 60Fe line; its lifetime is 2.35 days. Because of the short decay time, we just take the GEDSAT as a background component tracer for this strong instrumental line feature.

Thus, in a first step, the detector-by-detector count rates in the four continuum bands are fitted by the GEDSAT time series to construct an "adjacent-energies'' background template. Then, the set of detector-by-detector spectra per pointing in 1 keV bins covering the $\sim$20 keV intervals around and including the 60Fe lines are fitted to the sum of adjacent continuum template, plus the radioactivity template for 60Co, plus an additional GEDSAT template to capture any additional prompt background. Thus we generate a background model for our actual set of spectra, which is constrained in its variability by the various background tracers. Contributions of celestial gamma-ray events are small, their variability is intrinsically different due to the coded-mask modulation of the background model supplemented by dithering, and thus is expected only to affect global normalization, but not the variability with time.

We then orthogonalize the different background model components for an improved convergence of the fitting (also see Sect. 2.2 of Diehl et al. 2006; this step ensures that each higher-order background component carries new and independent information).

Representing our data sets with background-models only, we obtain reduced $\chi^2$ values of 1.185 and 1.194 for the 1173 and 1332 keV line bands (SE, 277824 d.o.f.), and 0.665 and 0.663 for the 1173 and 1333 keV line bands (ME, 615038 d.o.f.), respectively. Note that for the low number of counts in ME spectra, $\chi^2$ statistics does not apply. Therefore, in our model fitting approach (Sect. 3.3), we use Poissonian statistics and maximize the Poisson-likelihood function.

3.3 Model fitting and spectra

We combine above background models with a spatial model for sky emission (Fig. 3) to fit our data, allowing for adjustments of fit parameters for background and sky intensities. For this, we use a maximum-likelihood method (spimodfit, for more details of methods, see Strong et al. 2005) based on Poissonian statistics. We then derive per energy bin, the fitted parameter values with uncertainties, the covariance matrices, and the fitted model components. The counts per energy bin, per detector, and per pointing are fitted to the background model described in Sect. 3.2 and the assumed sky map of celestial emission (e.g., 26Al distribution obtained by COMPTEL, see Fig. 3) as convolved into the domain of spectra per detector, energy bin, and pointing through the pointing sequence and the instrumental response:

D_{\rm e,d,p}=\sum_{m,n} A_{\rm e,d,p}^{m,n}\beta_{\rm s} I...
...\sum_{i=1}^3\beta_{b,i} B^i_{\rm e,d,p} + \delta_{\rm e,d,p},
\end{displaymath} (1)

where ${\rm e,d,p}$ are indices for data space dimensions: energy, detector, pointing; m,n indices for the sky dimensions (galactic longitude, latitude); A is the instrument response matrix, I is the intensity per pixel on the sky. Coefficients  $\beta_{\rm s}$ for the sky map intensity (constant in time) and  $\beta_{b,1}$, $\beta_{b,2}$, $\beta_{b,3}$ for three background intensities are derived (see Sect. 3.2 and Fig. 2; different normalizations allowed for each camera configuration of 19/18/17 functional detector elements due to failure of two Ge detectors), and the amplitude  $\beta_{\rm s}$ comprises the resultant spectra of the signal from the sky. $\delta$ is the count residue after the fitting. Generally, a good fit will lead to residuals being statistically distributed around zero (see Fig. 4).

\par {\hspace*{2cm}\includegraphics[width=10.1cm,clip]{6982f4a.ep...
\includegraphics[width=6.6cm,clip]{6982f4c.eps}\end{figure} Figure 4: Residuals of counts after fitting, in different projections. The upper figure shows the residuals versus time (days of the Julian Date, starts at 1 Jan. 2000), where counts per pointing have been re-binned into 3-day intervals for clarity (here for single events SE in the 1163-1182 keV energy band). The lower two figures present the residuals projected with the energy (SE cases for the bands around the 1173 and 1333 keV lines separately). Two different time periods are shown (dashed: first half of database; solid: second half of database). For the 1333 keV line case, because of the strong instrumental 69Ge line at 1337 keV, there still exist the large residuals around 1337 keV. Residuals around zero generally confirm that our background models are adequate except for the instrumental line around 1337 keV, with  $\chi ^2/{\rm d.o.f.}$ in model fittings (Sect. 3.3): 1.182 (SE) and 0.670 (ME) for the 1173 keV line; 1.191 (SE) and 0.667 (ME) for the 1333 keV line.
Open with DEXTER

Separate fits are made for each of the Fe lines and for SE and ME, respectively. These produce four spectra of sky brightness amplitudes per energy bin. If background models are adequate, the continuum energy bands should show sky amplitudes statistically consistent with zero sky brightness. Non-zero sky brightness should show up in the form of lines with a shape conforming to the instrumental line width, possibly broader if the celestial line energy is broadened already (as had been discussed for 26Al, see Diehl et al. 2006). Due to the low intensity of expected 60Fe emission, we do not attempt to derive line shape information at all, and rather derive the 60Fe line amplitude by fitting a Gaussian with fixed instrumental width to the spimodfit result spectra, determining line centroids and amplitudes in this way. From strong nearby instrumental lines, we determine that a Gaussian with a width of 2.76 keV is an adequate model for narrow lines in the regime of the 60Fe lines; we use this shape to determine 60Fe line parameters from the spimodfit result spectra. For a consistent celestial signal from 60Fe, the lines should be around the 60Fe decay energies of 1173 and 1333 keV and equal (within errors) in amplitudes for all four data sets.

\par\includegraphics[width=14.7cm,clip]{6982f5.eps}\end{figure} Figure 5: The spectra of two gamma-ray lines of 60Fe from the inner Galaxy: 1173 keV and 1333 keV. We have shown the results both from SE and ME databases. The data points are fitted with Gaussian profiles of fixed instrumental width (2.76 keV), and fixed continuum slope (flat), with $\chi ^2/{\rm d.o.f.}$ of fits: 1.06 (SE) and 1.09 (ME) for the 1173 keV line and 1.16 (SE) and 1.11 (ME) for the 1333 keV line. For the SE database, we find a line flux of (4.2 $\pm $ 1.6) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$ for the 1173 keV line and (3.5 $\pm $ 1.5) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$ for the 1333 keV line. For the ME database, the line flux is (5.8 $\pm $ 1.9) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$ for the 1173 keV line and (5.2 $\pm $ 2.1) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$ for the 1333 keV line (also see Table 1).
Open with DEXTER

\includegraphics[width=12cm,clip]{6982f6.eps}\end{figure} Figure 6: The combined spectrum of the 60Fe signal in the inner Galaxy, superimposing the four spectra of Fig. 2. In the laboratory, the line energies are 1173.23 and 1332.49 keV; here superimposed bins are zero at 1173 and 1333 keV. We find a detection significance of 4.9$\sigma $. The solid line represents a fitted Gaussian profile of fixed instrumental width (2.76 keV), and a flat continuum. The average line flux is estimated as (4.4 $\pm $ 0.9) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$.
Open with DEXTER

\par\includegraphics[width=12cm,clip]{6982f7.eps}\end{figure} Figure 7: Flux ratio of the gamma-ray lines from the two long-lived radioactive isotopes 60Fe/26Al from several observations, including our SPI result (also see Table 2), with upper limits shown at 2$\sigma $ for all reported values, and comparison with the recent theoretical estimates (the upper hatched region from Prantzos 2004; the horizontal line taken from Timmes et al. 1995; the lower hatched region, see Limongi & Chieffi 2006). Our present work finds the line flux ratio to be (14.8 $\pm $ 6.0)$\%$. Note that the primary instrument results on 60Fe is compromised in the figure by different and uncertain 26Al fluxes used for the 60Fe/26Al ratio (see Table 2 for details).
Open with DEXTER

4 Results

4.1 SPI detection of 60Fe

From independently fitting the SE and ME data sets at 1 keV-wide energy bins in the energy ranges of the two lines of 60Fe emission (1173 keV and 1333 keV), we obtain the four spectra from spimodfit (Sect. 3.3) shown in Fig. 5. Our fitted model consists of the sky intensity distribution of 26Al from 9-year COMPTEL observations (Fig. 3, maximum entropy image as a standard input, Plüschke et al. 2001), together with the background model based on measurements in energy bands adjacent to the 60Fe lines and on background tracers (see above). We find excess amplitudes for the celestial component in each of the four spectra.

Residual background imperfections are small except for the instrumental line at 1333 keV. The continuum flux levels are close to zero within $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}~keV^{-1}}$, without apparent energy dependence, and adding a linear slope to our spectral fit function does not improve the fit[*]. The 69Ge line at 1337 keV has also been eliminated rather well, though not completely. Suppression is a factor $\sim$4 better than our previous work (Harris et al. 2005). However, this strong instrumental line still leads to a spectral signal which is recognized clearly as an artifact and thus not confused with the celestial line.

We then determine the 60Fe line parameters in these spectra fitting a continuum and Gaussians representing the SPI spectral response (Sect. 2). The width of the Gaussian profiles has been fixed at 2.76 keV, the value derived from simultaneously-measured adjacent instrumental-line shapes (e.g. the instrumental line at 1107 keV). We thus obtain 60Fe line positions and intensities from fitting the four spectra of Fig. 5, where the total line flux for each line in the inner Galaxy region (e.g. $-30^\circ
<l<30^\circ $ and $-10^\circ <b<10^\circ$) is determined from the intensity of our fitted Gaussians and from the normalization of the input sky map. We present line fluxes of each of the two 60Fe lines for the SE and ME data in Table 1, respectively. All line flux values are consistent within uncertainties.

A superposition of the four spectra of Fig. 5 is shown in Fig. 6. Line energies of the 60Fe lines in the laboratory are 1173.23 and 1332.49 keV. For this superposition, we therefore define the zero of the relative energy axis at 1173 and 1333 keV, to derive the summed spectrum of all 60Fe signals. The line flux estimated from the combined spectrum is (4.4 $\pm $ 0.9) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$. Our significance estimate for the combined spectrum is $\sim$$4.9\sigma$, adding uncertainties of the individual spectra in quadrature. This improves upon earlier 60Fe signal reports from RHESSI (Smith 2004) and the first year of SPI data (Harris et al. 2005).

The signal of celestial 60Fe is very weak, and marginally significant (<$3 \sigma$) in each of our four spectra. Therefore we cannot evaluate line shape information. The lines in our spectra appear well represented by Gaussians with instrumental widths (Figs. 5 and 6), suggesting that the celestial 60Fe lines are intrinsically-narrow lines. If we assume a line broadening of 1 keV, line fluxes would increase by $\sim$$6\%$.

4.2 60Fe and 26Al

It has been argued that 26Al and 60Fe share at least some of the same production sites, i.e. massive stars and supernovae (e.g., Timmes et al. 1995; Limongi & Chieffi 2006). In addition both are long-lived radioactive isotopes, so we have good reasons to believe their gamma-ray distributions are similar as well. Therefore we adopt the sky distribution of 26Al gamma-rays as our model for celestial 60Fe gamma-ray distribution.

From our model-fitting approach (Sect. 3.3), in principle, the spectral result depends on the input sky map (also see Eq. (1)). We do not know the real distribution of 60Fe sources in the Galaxy, but we need a realistic model for these sources to derive correct spectra and line intensity. As an alternative to the COMPTEL 26Al maximum entropy and MREM sky maps, we also tried a different sky map: an exponential-disk model with scale radius 4 kpc, and scale height 180 pc. This should also be a good first-order representation of 26Al emission from the inner Galaxy, and but avoids the fine structure inputs from the COMPTEL image measurements which may partly arise from instrumental or analysis imperfections. We find that different input sky maps do not change the line profiles and intensities significantly (variations of the line fluxes are within 10% of the quoted values, i.e. below the uncertainties). If we use very different and scientifically implausible sky distribution models, e.g., a point source at $l = \pm 20^\circ$, a bulge model (a Gaussian profile in the Galactic center), or a COMPTEL 26Al map zeroed in the inner region of $-40^\circ < l <40^\circ,\ -10^\circ<b<
10^\circ$, then no signals are found near the 60Fe line energies.

If we hypothesize a massive-star origin for 60Fe, then the measurement of the gamma-ray flux ratio 60Fe/26Al is important for discussions of the astrophysical nucleosynthesis origins of the two radioactive isotopes, and the nuclear physics involved in models for their production (addressing the uncertain nuclear reaction cross sections and half-lives). For this purpose, we apply the same analysis method on the 26Al data from the same observations, analyzing the 1785-1826 keV energy band (1 keV bins). We again generate a specific background model (from adjacent energy bands and the GEDSAT background intensity tracer), and apply the same input sky map (the 9-year COMPTEL 26Al maximum entropy map) in model fitting the SE data. This yields a flux ratio of 60Fe/26Al in a self-consistent way, with $F(^{60}{\rm Fe})/F(^{26}{\rm Al}) = (14.8 \pm 6.0)\%$. Here, the uncertainty has been estimated from the respective model-fitting uncertainties of the two SE databases. Alternatively, from the combined spectrum of 60Fe lines (Fig. 6), and adopting the 26Al intensity measured with SPI before ( $F(^{26}{\rm Al})=
(3.04\pm 0.31)$ $\times $ $10^{-4}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$for the inner Galaxy, see Fig. 5 of Diehl et al. 2006), we obtain $F(^{60}{\rm Fe})/F(^{26}{\rm Al})= (14.5 \pm 4.0)\%$.

Table 1: 60Fe intensity in the inner Galaxy.

4.3 The Cygnus and Vela regions

The Cygnus region is one of the most active nearby star formation regions in our Galaxy. It contains a large number of massive stars and rich OB associations at a distance of 1-2 kpc (e.g. Knödlseder 2000; Le Duigou & Knödlseder 2002), and has been recognized as a prominent source of 26Al with COMPTEL (Diehl et al. 1995a; Plüschke et al. 2001) and with INTEGRAL/SPI (Knödlseder et al. 2004). From these massive stars, 60Fe gamma-rays at the 1173 keV and 1333 keV decay lines may also be expected. Since the majority of Cygnus region star-clusters are young ($\sim$3 Myr), from population synthesis studies of the massive stars in the Cygnus region it has been suggested that the 60Fe production is low, consistent with the small number of recent supernova events inferred for this region ( $F(1173~{\rm keV})\sim 2$ $\times $ $10^{-6}~{\rm ph~cm^{-2}~s^{-1}}$, see Knödlseder et al. 2002).

The Vela region in the southern sky includes even more nearby massive stars, and prominent candidate 26Al sources such as the Wolf-Rayet star WR 11 in the binary system $\gamma^2$Vel at $d\sim 260$ pc (van der Hucht et al. 1997), and the core-collapse supernova remnant Vela SNR at $d\sim 250$ pc (Cha et al. 1999). 1809 keV line emission of 26Al from the Vela region has been reported from COMPTEL (Diehl et al. 1995b). And this emission is not attributed to the nearby $\gamma^2$Vel source (Oberlack et al. 2000) and may be either diffuse or due to supernovae in the region. Therefore the Vela region should be a good candidate for gamma-ray line emission from 60Fe.

Splitting our model sky map into independent components to fit the data, we do not see significant contributions from either region; our estimated upper limits for 60Fe gamma-rays from the Cygnus ( $68^\circ <l<92^\circ$, $-10^\circ < b<14^\circ$) and Vela ( $255^\circ <l<273^\circ$, $-9^\circ < b<9^\circ$) regions are $\sim$1.1 $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}}$ ($2\sigma$). More detailed studies are in progress (Schanne et al. 2007; Martin et al., in preparation).

5 Summary and discussions

We report the detection of 60Fe decay gamma-ray lines in the Galaxy from 2.5 years of SPI observations. With our new measurements we detect both the 1173 keV and 1333 keV line of 60Fe from SPI single and multiple-detector events. Combining our four spectra from independent model fits we obtain an 60Fe signal from the Galaxy with a significance of $4.9\sigma$. This improves upon our previous measurements (see Harris et al. 2005). The average 60Fe line flux from the inner Galaxy region is (4.4 $\pm $ 0.9) $\times $ $10^{-5}~{\rm ph~cm^{-2}~s^{-1}~rad^{-1}}$, assuming intrinsically narrow lines and a sky distribution equal to that of 26Al as measured by COMPTEL (Plüschke et al. 2001). From the same observations and analysis procedure applied to 26Al, we find a flux ratio of 60Fe/26Al of (14.8 $\pm $ 6.0)$\%$. The 60Fe signals are too weak to determine line shape details; it appears that a Gaussian with the width of the instrumental resolution can fit the data well. This would imply that broadening of 60Fe lines from astrophysical processes is not significant; most of the 60Fe may be distributed in a rather normal interstellar medium with turbulent velocities below $\sim$300 km s-1. In order to investigate the variations over the Galaxy, we search for 60Fe emission from the Cygnus and Vela regions, and do not find 60Fe signals.

Table 2: Different measurements of 60Fe flux in units of 10-5 ph cm-2 s-1 rad-1 from the inner Galaxy and 60Fe/26Al flux ratio.

Many experiments and efforts were made (see Table 2 and Fig. 7) to measure the 60Fe gamma-ray emission that was predicted by theory - we now provide the most significant detection to date. In Fig. 7, we show the previous constraints on the flux ratio of 60Fe/26Al together with the result of this work, and compare the observational results with different theoretical predictions. The earliest observational limit was given from HEAO-3, $F(^{60}{\rm Fe})/F(^{26}{\rm Al}) = 0.09$ $\pm $ 0.08, an upper limit being 0.27 (Mahoney et al. 1982) (in Fig. 7, we chose to give limits at 2$\sigma $ for all reported values below a significance of 3$\sigma $). Another limit was obtained with the SMM Gamma-Ray Spectrometer, a flux ratio of 0.1 $\pm $ 0.08, the upper limit being $\sim$0.27 (Leising & Share 1994). OSSE aboard the COMPTON Observatory gave a flux ratio of 0.21 $\pm $ 0.15, and the upper limit is $\sim$0.51 (Harris et al. 1994, 1997). COMPTEL aboard the COMPTON Observatory also found 60Fe gamma-rays, and reported a flux ratio value of 60Fe/26Al of 0.17 $\pm $ 0.135, which translates into an upper limit $\sim$0.44 (Diehl et al. 1997). The Gamma-Ray Imaging Spectrometer (GRIS) reported an upper limit for the ratio of <0.14 (2$\sigma $, Naya et al. 1998). RHESSI then reported the first detection of 60Fe gamma-ray lines, and gave a flux ratio 0.17 $\pm $ 0.13with two-year data (Smith 2004). The first year data of SPI gave a flux ratio 0.11 $\pm $ 0.07 (Harris et al. 2005), and the present analysis of the 2.5-year SPI data finds a flux ratio 0.148 $\pm $ 0.06.

Theoretical predictions of the ratio of 60Fe/26Al have undergone some changes since Timmes et al. (1995) published the first detailed theoretical prediction. In their paper, they combine a model for 26Al and 60Fe nucleosynthesis in supernova explosions with a model of chemical evolution, to predict that the steady production rates are (2.0 $\pm $ 1.0)  $M_\odot~{\rm Myr^{-1}}$ for 26Al, and (0.75 $\pm $ 0.4)  $M_\odot~{\rm Myr^{-1}}$ for 60Fe, which corresponds to a gamma-ray flux ratio $F(^{60}{\rm Fe})/F(^{26}{\rm Al})= 0.16$ $\pm $ 0.12. This prediction would be consistent with our present measurements. Since 2002, theoreticians have improved various aspects of the stellar-evolution models, including improved stellar wind models and the corresponding mass loss effects on stellar structure and evolution, of mixing effects from rotation, and also updated nuclear cross sections in the nucleosynthesis parts of the models. As a result, predicted flux ratios 60Fe/26Al rather fell into the range 0.8 $\pm $ 0.4 (see Prantzos 2004, based on, e.g. Rauscher et al. 2002; Limongi & Chieffi 2003) - such high values would be inconsistent with several observational limits and our SPI result (but see Woosley & Heger 2007 for comments on nuclear reaction rate updates). Recently, new calculations of 26Al and 60Fe yields from massive stars of solar metallicity have been presented (Limongi & Chieffi 2006; Heger & Woosley 2007). Limongi & Chieffi (2006) combined their individual yields, using a standard stellar-mass distribution function, to produce an estimate of the 60Fe/26Al gamma-ray flux ratio expected from massive stars. Their calculations yield again a lower prediction for the 60Fe/26Al flux ratio of 0.185 $\pm $ 0.0625, which is again consistent with the observational constraints (see Fig. 7).

In summary, uncertainties still exist, both in models and measurements of 60Fe. On the theory side, stellar evolution in late stages is complex, nuclear reactions include neutron capture on unstable Fe isotopes, and explosive nucleosynthesis adds yet another complex ingredient. On the experimental side, cosmic ray induced 60Co radioactivity in the instrument and spacecraft and the limitations of spatial resolutions and sensitivity are issues reflected in the substantial uncertainties in experimental values. With more INTEGRAL/SPI data to come, and also with the development of next-generation gamma-ray spectrometers/telescopes, gamma-ray observations hopefully can help with an independent view on the astrophysical model components.

We are grateful to the referee for the constructive comments. The INTEGRAL project is supported by government grants in all member states of the hardware teams. The SPI project has been completed under responsibility and leadership of CNES. We are grateful to ASI, CEA, CNES, DLR, ESA, INTA, NASA, and OSTC for support.



Copyright ESO 2007