STIX observes the Sun in X-rays from 4 to 150 keV using 32 detectors, each divided into 12 pixels, and reconstructs images of flares from patterns cast by 30 pairs of grids. The instrument enjoys a privileged position: orbiting the Sun in interplanetary space rather than around Earth, it largely avoids the radiation belts and the South Atlantic Anomaly that plague Earth-orbiting X-ray missions, where geomagnetic storms can completely obscure solar signals. But STIX carries its own inescapable source of contamination.

Four tiny drops of the radioactive isotope barium-133, each with an activity of roughly 20 becquerel, sit at the corners of every detector just millimetres from the detector surface, providing about 80 becquerel of total activity in each detector's immediate vicinity. These calibration sources emit characteristic X-ray lines used to monitor detector health, but they also flood every measurement with a constant glow that can mask flares weaker than GOES class B3 as seen from a distance of one astronomical unit. The new analysis rests on 256 so-called background files recorded by STIX between January 2022 and July 2023.

These files are accumulated over thousands of seconds during periods when no flaring emission was recognised, and the team carefully discarded files that, despite being labelled quiet, were contaminated by genuine solar emission, particularly near perihelion passages. Their central assumption is that during most of the analysed period, the barium-133 calibration sources dominate the background across the full energy range. Supporting evidence is strong: the team estimated that the cosmic X-ray background is roughly 400 times weaker than the instrumental background in the 40 to 50 keV range, where the isotope's emission is already faint, and that even at perihelion, at 0.28 astronomical units, the instrumental background more than doubles the signal expected from a GOES B1-class flare.

Because barium-133 decays radioactively with a half-life of 10.52 years, the background should decline in a precisely predictable way. The researchers exploited this by fitting an exponential decay, with the slope fixed to the isotope's decay constant of minus 2.09 times ten to the minus nine per second, to the logarithm of the observed counts in each of the 384 detector pixels and in each of the science energy bins. Only the intercept of the fit had to be determined, which was done with the MpFit routine, with measurement errors propagated through the procedure.

Exponentiating the fit yields the background for any pixel at any time, together with an uncertainty derived from the range of fitted intercepts. An important practical improvement was the use of Energy Look-Up Table corrections calculated nearly daily rather than at the multi-month intervals typical of standard STIX processing, since the actual widths of the recorded energy bands drift over time. Building the pipeline also required computational ingenuity.

Processing the files through the standard OSPEX package took about 2.8 hours per file, meaning a full month of computation for the 256 files. The team instead wrote routines that reproduce OSPEX's data reduction independently, verified to give identical results to six decimal places, and cut the total processing time by a factor of 30 to under a day. The analysis revealed that the background is far from uniform across the instrument.

In the 28 to 32 keV band, which contains the strongest barium-133 line at about 30.85 keV, the small pixels record systematically lower fluxes than the large ones. This is partly a consequence of the single-event counting mode: photons that interact across two channels, for example through fluorescence escape or Compton scattering between neighbouring pixels, are discarded, and such cross-channel events are more likely in the small pixels because of the detector geometry. Individual detectors also show distinctive levels, and the distributions for the upper and lower rows of large pixels differ from each other because slightly different amounts of isotope were deposited in each source.

These pixel-to-pixel variations matter because they imprint a fixed pattern on the detector face that can interfere with the delicate moiré patterns used for imaging. The imaging consequences were tested on a sample of 414 compact flares with a single X-ray source, reduced to 405 events after two badly overestimated background files were excluded. The team compared source sizes, defined by the region containing 50 percent of the total signal, across several reconstruction algorithms.

The Richardson-Lucy type algorithms, Expectation Maximisation and MARLIN, told a different story: their reconstructions without background subtraction diverged markedly from the Forward Fit benchmark, but including background brought them into line, with improved correlation, standard deviation and mean absolute deviation for the flares exceeding 30,000 counts. When the interpolated background was used in MARLIN instead of background files, the agreement was even tighter, with a correlation close to one across all 405 flares, and reconstructed sources emerged slightly larger and cleaner, with less scattered signal. The spectral test delivered the most dramatic demonstration of why background selection matters.

Fitting spectra of the same 405 flares with either the file-based or interpolated background produced correlations of 0.99 for both the thermal temperature and emission measure, confirming broad consistency. For the non-thermal component, agreement remained high for the electron flux, at about 0.99, and somewhat lower for the power-law index, at about 0.84, reflecting the large uncertainties in weak flares where many fitted values had errors exceeding 50 percent. The clearest lesson came from one of the excluded events, a C1.8-class GOES flare on 31 March 2022 with an estimated STIX class of B7.

Using the overestimated background file, the fitted non-thermal parameters were physically unrealistic and the thermal uncertainties were several orders of magnitude larger than the parameters themselves, because the excessive subtraction stripped away real signal. With the interpolated background, the same spectrum fitted cleanly with a combined thermal and non-thermal model, and the reconstructed source diameter of 18 arcseconds agreed much better with the independently expected 20 to 30 arcseconds than the 9 arcseconds obtained with the flawed file. Beyond flares, the interpolated background solves a subtler problem with the instrument's attenuator, a device that reduces the solar signal at low energies.

Pre-flare or elevated background files contain a substantial solar component that is strongly suppressed when the attenuator is inserted, so subtracting them over-corrects the low-energy data. The purely instrumental interpolated background avoids this over-subtraction entirely. The model also permits a forward look: extrapolated to July 2030, when solar minimum is predicted, the declining barium-133 activity should lower the background substantially, potentially opening the quiet Sun's faint high-temperature emission to direct STIX observation.

The team's conclusion is measured but consequential. The interpolated background reproduces what the standard files deliver when those files are good, yet it eliminates the risk of adopting an overestimated background, works at any moment of the mission without waiting for a quiet interval, and enables a higher degree of automation in the analysis. That matters most for microflares and other weak events, whose statistics are limited precisely because the signal sits near the instrumental floor.

Lower background means narrower credibility intervals, fewer false detections, and the ability to reconstruct images from fewer counts. For a mission racing through the most active phase of the solar cycle while repeatedly diving inside Earth's orbit, knowing exactly how much glow comes from the instrument itself may prove as important as the flares it was built to catch.

Subject of Research: Analytical modelling of the STIX instrumental background on Solar Orbiter for solar flare X-ray spectral and imaging analysis

Article Title: Analytical modelling of STIX instrumental background and its application in the spectral and imaging analysis

Article References: Kułaga, K., Mrozek, T., Sylwester, J., Awasthi, A. K., Litwicka, M., Ścisłowski, D., & Limousin, O. (2026). Analytical modelling of STIX instrumental background and its application in the spectral and imaging analysis. Experimental Astronomy, 62(3), Article 24. https://doi.org/10.1007/s10686-026-10084-6

Image Credits: AI Generated

DOI: 10.1007/s10686-026-10084-6

Keywords: STIX, Solar Orbiter, solar flares, X-ray background, barium-133, image reconstruction, spectral analysis, microflares, hard X-rays, instrumental background, Experimental Astronomy, Analytical