Imaging the positron annihilation line with 20 years of INTEGRAL/SPI observations

TL;DR

The paper tackles the origin and distribution of Galactic positrons by producing the most sensitive all-sky map of the 511 keV annihilation line using 20 years of INTEGRAL/SPI data. It deploys Richardson–Lucy image deconvolution with joint background optimization and bootstrap-based uncertainty estimation, accelerated by GPU computing. The reconstructed map confirms a bright central component, a broad bulge, and a disk aligned with the Galactic plane, while revealing marginal hints of asymmetric bulge features and potential emission from massive-star regions such as Sco–Centaurus. If confirmed by future MeV missions like COSI, these features could illuminate low-energy positron production and propagation, offering a path to resolving the long-standing positron puzzle in the Milky Way.

Abstract

The e$^+$ e$^-$ annihilation line at 511 keV provides a unique probe for studying the distribution and origin of positrons in our Galaxy. The SPI spectrometer on INTEGRAL has observed this gamma-ray line for two decades. We analyze 20 years of INTEGRAL/SPI observations to produce the most sensitive all-sky map of the 511 keV line emission to date, aiming to reveal new features and provide refined measurements of known sources. We perform image deconvolution using the RL algorithm and employ bootstrap analysis to evaluate statistical uncertainties of fluxes from regions of interest. Systematic uncertainties in parameter choices are also considered. We utilize GPU acceleration to enable this computationally intensive analysis. The reconstructed image successfully recovers the basic morphological features reported in model-fitting studies: a bright central component, a broad bulge, and an elongated disk component along the Galactic plane. We also report hints of new spatial features in the reconstructed image, including an asymmetric structure in the broad bulge emission and 511 keV emission potentially associated with massive stars from the Sco-Cen and other OB associations. While the significance of these new features is marginal ($\sim 2σ$), they are spatially consistent with $^{26}$Al emission from massive stars in that region, suggesting that this 511 keV emission originates from its $β^{+}$ decay. Our 20-year dataset provides the most detailed 511 keV emission map to date, reproducing global structures suggested in model-fitting approach while revealing hints of new spatial features. These findings provide insights into the origin of Galactic positrons and propagation of low-energy positrons in the interstellar medium. Future MeV gamma-ray observations, such as COSI, are expected to confirm the reported features and shed further light on the nature of positrons in our Galaxy.

Figures (20)

• Figure 1: Exposure map for the 511 keV line observations with INTEGRAL/SPI over 20 years in the Galactic coordinates.
• Figure 2: Map of the narrow-line 3$\sigma$ point-source sensitivity at 511 keV derived from the exposure map. The contour lines show the sensitivity of $2\times10^{-5}$, $3\times10^{-5}$, $5\times10^{-5}$, $1\times10^{-4}$, and $2\times10^{-4}~\mathrm{cm^{-2}~s^{-1}}$ from purple to red.
• Figure 3: Reconstructed image of the 511 keV line emission from the 20-year INTEGRAL/SPI data. The image is shown in Galactic coordinates. We note that several dark spots appearing around the Galactic center, such as the one at $(l, b) \approx (353^{\circ}, -4^{\circ})$, are artifacts often seen in the image deconvolution.
• Figure 4: Same as Fig. \ref{['fig_best_image']} but with annotations showing the regions used for the flux measurements in Table \ref{['tab_global_flux']}.
• Figure 5: Longitude profile of the 511 keV line emission obtained by integrating the flux from $b = -20$ to $+20$ degrees in 4-degree longitude bins. The red region shows 1$\sigma$ statistical uncertainties derived from bootstrap samples. The dashed lines show systematic uncertainties estimated by varying the parameters in the image reconstruction algorithm. The blue band represents the 95% detection sensitivity derived from background-only bootstrap samples.