© The Authors 2023

1. Introduction

On 2022 October 9 at 13:16:59.988 UT, the Fermi Gamma-ray Burst Monitor (GBM) detected the exceptionally bright long gamma-ray burst (GRB) GRB 221009A (Veres et al. 2022; Lesage et al. 2022, 2023). The burst was also observed by the Fermi Large Area Telescope (LAT) up to the energy of 100 GeV (Pillera et al. 2022). Potentially remarkable detections of over 5000 very high-energy photons with energies up to 18 TeV were reported by the Large High Altitude Air Shower Observatory (LHAASO; Huang et al. 2022), and a possible 251 TeV photon was reported by Carpet-2 (Dzhappuev et al. 2022), triggering the interest of the broader physics community.

The burst was localised by the Neil Gehrels Swift Observatory’s Burst Alert Telescope (Dichiara et al. 2022) and followed up by the Very Large Telescope (VLT) X-shooter instrument (de Ugarte Postigo et al. 2022; Malesani et al. 2023), which determined that it occurred at a redshift of 0.151 and belongs to very near long GRBs (Oates 2023). It was also detected by a multitude of other instruments: AGILE/GRID (Piano et al. 2022), AGILE/MCAL (Ursi et al. 2022), BepiColombo/MGNS (Kozyrev et al. 2022), Insight-HXMT and SATech-01/GECAM-C (HEBS; An et al. 2023), INTEGRAL/SPI-ACS (Gotz et al. 2022), Konus-WIND & SRG/ART-XC (Frederiks et al. 2023), MAXI and NICER (Williams et al. 2023), Solar Orbiter/STIX (Xiao et al. 2022), STPSat-6/SIRI-2 (Mitchell et al. 2022), and XMM-Newton (Tiengo et al. 2023).

This brightest ever recorded GRB (Burns et al. 2023; O’Connor et al. 2023) saturated many of the GRB detectors in orbit, hampering the efforts to determine its peak luminosity. In this Letter, we present the peak flux and peak isotropic-equivalent luminosity of this extraordinary transient as measured by the GRBAlpha nano-satellite.

2. GRBAlpha

GRBAlpha (Pál et al. 2020) is a 1U CubeSat carrying a GRB detector as a technology demonstration for an envisioned future CubeSat constellation (Werner et al. 2018; Mészáros et al. 2022). It was launched on 2021 March 22 into a Sun-synchronous polar orbit at an altitude of ∼550 km and thus became the smallest astrophysical space observatory. About a third of the polar orbit is affected by high particle background, and the duty cycle of the detector is around 67%. GRBAlpha’s detector consists of a 75 × 75 × 5 mm³ CsI(Tl) scintillator read out by an array of Silicon PhotoMultipliers (SiPMs), called multi-pixel photon counters (MPPCs), by Hamamatsu. The SiPM detectors are protected from proton damage by a 2.5 mm thick lead (PbSb3 alloy) shield, and their degradation is being monitored. The on-board data acquisition software stack is being continuously upgraded to increase the duty cycle and data downlink rate. The ground segment is also supported by the radio amateur community, and it takes advantage of the Satellite Networked Open Ground Station (SatNOGS)¹.

Following a commissioning phase, GRBAlpha started collecting data, monitoring the particle and photon background environment on a low-Earth polar orbit, and detecting transients (see Řípa et al. 2022a). When the satellite operates continuously, on average it detects a transient every 5–6 days². GRBAlpha detected GRB 221009A (Řípa et al. 2022b) when traversing the northern polar regions, and during the peak brightness of the burst its detectors were not saturated; however, the measurement was influenced by pile-up.

The pile-up effect occurs due to the chance coincidence of the arrival and interaction of two or more gamma-ray photons in the detector’s scintillator within the detector’s inherent resolving time. In that case, two or more events are registered as one event, and the energies of the events within the same resolving time window are added (Knoll 2000).

3. Data analysis

GRBAlpha observed a peak count rate of 22 000 cnt s⁻¹ in the ∼80 − 950 keV energy band at 13:20:51.5 UTC on 2022 October 9. The duration of the GRB was > 250 s (Řípa et al. 2022b). During the detection, the satellite was flying above the northern polar region with elevated background levels. The end part of the GRB was recorded while passing the outer Van Allen radiation belt; therefore, we can only report the lower limit on the duration and cannot determine the T₉₀ duration of the GRB. The GRBAlpha data for this event are composed of binned light curves with 4 s temporal resolution recorded in 16 energy bands. However, the three lowest-energy bands are below the set on-board low-energy threshold limited by the noise peak of the MPPCs, making only 13 energy bands suitable for spectral analysis. The recorded raw count-rate curves in multiple energy bands are presented in Fig. 1. We note that the energy calibration was performed by using radionuclides in the laboratory and by observing activation lines in orbit.

Fig. 1. Raw light curve of GRB 221009A as observed by GRBAlpha in multiple energy bands with 4 s cadence. The bottom-rightmost panel shows the light curve in the whole sensitivity range. Times tb, ts, and tp mark the beginning of the used background interval, the source spectral data, and the peak time, respectively. About 30 s after the peak at tp, the GRB shows a second strong peak. Time tr marks the approximate moment when the satellite entered the outer Van Allen radiation belt, which resulted in the final part of the GRB prompt emission being flooded by the particle background. The sharp increase just after tr is due to this background.

We performed spectral analysis to determine the peak flux of the GRB and its isotropic-equivalent peak luminosity. The mass model of GRBAlpha used for generating detector response matrices employed in the spectral analysis is displayed in Fig. 2. It contains: the detector made of a 75 × 75 × 5 mm³ CsI(Tl) scintillator in a 1.5 mm thick Al casing; a 2.5 mm thick PbSb3 radiation shield; a standard 1U CubeSat Al platform with four stacks of empty printed circuit boards (PCBs; glass epoxy FR4) and two LiFePO₄ batteries.

Fig. 2. Mass model of GRBAlpha used for generating detector response matrices, with zenith and azimuth angles marked. The CsI(Tl) detector in its aluminium casing is shown in red and the PbSb3 radiation shield in green. The aluminium CubeSat platform, with four stacks of PCBs and two LiFePO4 batteries, is shown in grey.

A full Monte Carlo simulation based on the GEometry ANd Tracking (Geant4) toolkit³ (Allison & Amako 2016) was carried out to simulate incident photons with a flat energy distribution in the 5 − 1000 keV band, covering the entire satellite structure. The deposited energy as a function of input photon energy was calculated by scanning more than 200 incident Θ (zenith) and Φ (azimuth) angles. The effective area versus incident photon energy for a few zenith angles, Θ, and a constant azimuth angle Φ = 270° is shown in Fig. 3. The drop over 900 keV is an artefact because we simulated photon energies only up to 1000 keV and the photo-absorption peak (with broadening of the energy resolution) of such high-energy photons is not accounted for in this response. Therefore, in the spectral analysis, we conservatively only used data up to ∼810 keV.

Fig. 3. Total response efficiency (effective area) vs. incident photon energy for a few zenith angles, Θ, and a constant azimuth angle, Φ = 270°.

To check for pile-up, we plotted the fraction of the detected counts in each spectral band versus the total counts over the whole energy range for GRB 221009A (see Fig. 4). The figure shows that above ∼10 000 cnt bin⁻¹ = 2500 cnt s⁻¹ the fraction of counts in the lower-energy bands decreases while the fraction of counts in the higher-energy bands increases, which may indicate the presence of pile-up. We note that below ∼10 000 cnt bin⁻¹ the decrease in the rates at high energies as a function of total counts in each bin is due to very low count numbers, and the diagonal lines correspond to 1, 2, 3… cnt bin⁻¹. At the peak of this transient, GRBAlpha measured a count rate of m = 22 000 cnt s⁻¹ in the full band. The shaping parameter of the analogue electronics is τ₁ ≈ 10 μs. During the signal digitisation, the pulse processing is done within the resolution time of τ₂ = 15 μs. We can calculate the probability that a pile-up appears as (Knoll 2000)

$$\begin{array}{c}\hfill P=1-e^{-n{\tau }_2}.\end{array}$$(1)

Fig. 4. Fraction of the detected counts in each spectral band vs the total counts over the whole energy range for GRB 221009A and for the background as it passed the north polar region. The counts were detected in 4 s time bins. The figure shows that above ∼10 000 cnt bin = 2500 cnt s−1 the fraction of counts in low-energy bands decreases while the fraction of counts in high-energy bands increases, which can indicate a pile-up effect.

Here n is the true event rate, which can be calculated from the detected count rate, m, as

$$\begin{array}{c}\hfill n=-\frac{\ln (1-\mathrm{m}\tau )}{\tau },\end{array}$$(2)

where for m = 22 000 cnt s⁻¹ we obtain n = 26 700 event s⁻¹. For the observed peak count rate we obtain a pile-up probability of ∼33%, whereas the probability drops to ∼3.7% for a count rate of m = 2500 cnt s⁻¹. Therefore, the observed spectrum at the peak time is affected by pile-up, and in the following spectral analysis we avoid fitting the peak spectrum (which results in a flatter power-law fit) and instead fit the GRB spectrum before the peak time, when the pile-up is negligible.

The spectral fitting was performed by the X-ray spectral fitting package XSPEC⁴ (Arnaud 1996). The source spectrum was fitted from time t_s = 13 : 20 : 33.5 UTC to t_s + 4 s, and the background spectrum was taken from time t_b = 13 : 17 : 33.5 UTC to t_b + 4 s. By applying this background, we obtained a background-subtracted spectrum without any substantial bump or dip around 500 keV, where the GRBAlpha background shows a bump. The average background rate in the whole sensitivity range in this part of the orbit is 340 cnt s⁻¹, and the rate in the chosen 4 s background bin is 345 cnt s⁻¹. We set a systematic error of 2% to be added when evaluating χ². The peak time in the measured light curve in the whole sensitivity range of ∼80 − 950 keV was measured to be from t_p = 13 : 20 : 49.5 UTC to t_p + 4 s. Figure 5 displays the detected count spectra at the time bins t_s, t_b, and t_p. At an approximate time t_r  =   ∼ 13 : 24 : 42.0 UTC, the GRBAlpha nano-satellite entered the outer Van Allen radiation belt, and the signal from the GRB became flooded by particle background. These times are marked in Fig. 1.

Fig. 5. Detected count spectra of GRB 221009A and the background spectrum. The count spectrum of the GRB at time ts minus the background spectrum at time tb is shown in black. The background spectrum at time tb is shown in red. The count spectrum of the GRB at the peak time tp = 13 : 20 : 49.5 UT on 2022 October 9, minus the background spectrum at time tb, is shown in green and it reveals a spectral hardening at the highest energies present around the GRB peak.

GRBAlpha does not have an active attitude control system, and the attitude knowledge was not being recorded at the time of GRB 221009A. The background variation had a periodicity of ∼47 s when the satellite was near the north pole. From the communication with the satellite on the same day, we observed in the radio waterfall a periodicity of ∼23 s in the received radio power. The Ultra High Frequency (UHF) antenna on GRBAlpha is a half-wave dipole, and thus a periodicity in the waterfall corresponds to half of the rotation or precession period of the satellite. Hence, the satellite completed more than one revolution during the duration of the GRB. We therefore performed the spectral analysis for many detector response matrices simulated for various Θ and Φ incident angles. We fitted the measured spectra each time in order to obtain the best-fit χ² map and to obtain the most likely direction of the GRB with respect to the satellite’s coordinate system.

We performed the spectral fitting with the power-law model, a power law with an exponential cutoff model (CPL), and with the Band function. The lowest reduced χ² was obtained for the CPL model, as described in the following section. The reduced χ² maps show a similar trend; therefore, in the next section, we report results obtained with the CPL model only.

After we determined the most likely incidence angle of the GRB with respect to the detector, we calculated the flux and the isotropic-equivalent luminosity of the GRB using the best-fit spectral model. Scaling up the flux and the luminosity obtained from the spectral fit at time t_s allowed us to determine the flux and the luminosity at the peak time, t_p. The scaling factor is approximately the ratio of the total detected counts at the time bin t_p over the total detected counts at the time bin t_s. The results are described in detail in the following section.

4. Results

The best fit for the source spectrum at the time bin t_s and after subtraction of the background at the time bin t_b was obtained for the on-axis direction Θ = 180° and Φ = 0° by the CPL model:

$$\begin{array}{c}\hfill N(E)=A{\left(\frac{E}{1\mathrm{keV}}\right)}^{-\alpha }\exp (-\frac{E}{E_0})\end{array}$$(3)

(see Fig. 6). The fit was performed in the range 76 − 809 keV, giving χ²/D.o.F. = 13.4/8 = 1.68. The best-fit parameters are: the photon index α = 0.7 ± 0.1; the roll-off energy $E_0={750}_{-200}^{+410}$ keV; and the normalisation $A=8_{-4}^{+6}$ ph keV⁻¹ cm⁻² s⁻¹ at 1 keV. The relatively large uncertainty of the normalisation is due to its correlation with the photon index alpha and the roll-off energy. The 68% confidence interval (CI) parameter uncertainties were calculated by XSPEC from the fit covariance matrix. The peak energy is E_p = E₀(2 − α) = 980 ± 410 keV.

Fig. 6. Spectrum of the GRB 221009A best fit by the CPL model fitted from ts = 13 : 20 : 33.5 UTC on 2022 October 9 to ts + 4 s in the range 76 − 809 keV for Θ = 180° and Φ = 0°.

The flux in the energy range of 80 − 800 keV in the observer frame for the best-fit model for the same time interval t_s and the same Θ and Φ angles was derived to be $F_{\mathrm{ph}}^{\mathrm{s}}=68\pm 1$ ph cm⁻² s⁻¹, or $F_{\mathrm{erg}}^{\mathrm{s}}=3.3\pm 0.1\times {10}^{-5}$ erg cm⁻² s⁻¹ (68% CI). For the isotropic-equivalent luminosity in the energy range 92 − 920 keV in the rest frame for the redshift z = 0.151 (de Ugarte Postigo et al. 2022; Malesani et al. 2023), the cosmological parameter Ω_Λ = 0.685, a flat universe, and a Hubble constant of H₀ = 67.4 km s⁻¹ Mpc⁻¹ (Planck Collaboration VI 2020), we obtain $L_{\mathrm{iso}}^{\mathrm{s}}=2.2_{-0.0}^{+0.1}\times {10}^{51}$ erg s⁻¹ (68% CI). The uncertainties in flux and luminosity were derived from the Markov chain Monte Carlo method (Hastings 1970) with following parameters: a proposal Gaussian fit; the covariance matrix used in the proposal distribution re-scaled by a factor of 0.125; a chain length of 50 000; the first 1000 steps discarded prior to storing the chain; and a Goodman-Weare chain type with 16 walkers.

The approximate scaling factor f = (C_p − C_b)/(C_s − C_b) – the ratio between the total detected counts C_p at the peak time bin t_p minus mean background counts C_b = 1360 cnt bin = 340 cnt s⁻¹ and the total detected counts C_s at the time bin t_s minus the mean background counts – gives f = (87 950 − 1360)/(8689 − 1360) = 11.8. We scaled the flux and the luminosity obtained from the spectral fit at time t_s by this factor and by the pile-up correction factor of 1.21 in order to determine the flux and the luminosity at the peak time, t_p. By applying this scaling to the flux and the isotropic-equivalent luminosity derived above for the time bin t_s, we obtain for the peak flux $F_{\mathrm{ph}}^{\mathrm{p}}=970\pm 20$ ph cm⁻² s⁻¹, or $F_{\mathrm{erg}}^{\mathrm{p}}=4.6\pm 0.2\times {10}^{-4}$ erg cm⁻² s⁻¹, and for the peak luminosity $L_{\mathrm{iso}}^{\mathrm{p}}=3.1\pm 0.1\times {10}^{52}$ erg s⁻¹.

We have so far not considered the systematic uncertainty due to the uncertain attitude of the satellite (detector). Although we derived the most likely orientation of the satellite with respect to the GRB from the best-fit χ² map, the difference in reduced χ² near the on-axis direction (Θ = 180°) and near the backside direction (Θ = 0°) is relatively low, implying that the on-axis direction is not guaranteed. Therefore, we also calculated the flux and the luminosity using a more conservative approach.

We took the distribution of fluxes and luminosities obtained by fitting the spectra by the CPL model at the same time bin t_s with the same background time bin, t_b, as described above, and used 192 different Θ and Φ directions, which isotropically sample the sphere around the satellite following the Hierarchical Equal Area isoLatitude Pixelization (HEALPix)⁵ tessellation (Górski et al. 2005). The best-fit χ² was obtained for each direction. The reduced ${\chi }_{\text{r}}^2$ map is shown in Fig. 7. We removed six cases that have reduced ${\chi }_{\text{r}}^2$ > 4 because these we consider as not acceptable fits. They correspond to the direction towards the MPPCs’ lead shield. The resulting distribution contains 186 fluxes (luminosities). The median flux in the 80 − 800 keV range (observer frame) and the isotropic-equivalent luminosity in the 92 − 920 keV range (rest frame) at the same time bin t_s are: $F_{\mathrm{ph}}^{\mathrm{s}}={94}_{-17}^{+85}$ ph cm⁻² s⁻¹, or $F_{\mathrm{erg}}^{\mathrm{s}}=4.0_{-0.5}^{+2.6}\times {10}^{-5}$ erg cm⁻² s⁻¹; and $L_{\mathrm{iso}}^{\mathrm{s}}=2.6_{-0.3}^{+1.7}\times {10}^{51}$ erg s⁻¹.

Fig. 7. Reduced χ2 map of the best fits with the CPL model for different angles of Θ (vertical) and Φ (horizontal) in degrees.

By applying the scaling factor f = (C_p − C_b)/(C_s − C_b) = 11.8 and a pile-up correction factor of 1.21, we obtain for the peak flux $F_{\mathrm{ph}}^{\mathrm{p}}={1300}_{-200}^{+1200}$ ph cm⁻²s⁻¹, or $F_{\mathrm{erg}}^{\mathrm{p}}=5.7_{-0.7}^{+3.7}\times {10}^{-4}$ erg cm⁻² s⁻¹, and for the isotropic-equivalent peak luminosity $L_{\mathrm{iso}}^{\mathrm{p}}=3.7_{-0.5}^{+2.5}\times {10}^{52}$ erg s⁻¹. When we repeated this analysis for the fluxes integrated over a larger extrapolated energy range of 0.9 − 8690 keV, we obtain a peak flux $F_{\mathrm{erg}}^{\mathrm{p},\mathrm{bol}}=1.3_{-0.2}^{+0.4}\times {10}^{-3}$ erg cm⁻²s⁻¹ and a bolometric isotropic-equivalent peak luminosity $L_{\mathrm{iso}}^{\mathrm{p},\mathrm{bol}}=8.4_{-1.5}^{+2.5}\times {10}^{52}$ erg s⁻¹ (4 s scale) in the 1 − 10 000 keV range (rest frame). Results for the on-axis direction and those obtained by varying different incident directions are summarised in Table 1.

Knowing the peak flux, $F_{\mathrm{erg}}^{\mathrm{p}}$, we can calculate the approximate fluence, S, of the GRB as

$$\begin{array}{c}\hfill S=\frac{C-k{C}_{\mathrm{b}}}{C_{\mathrm{p}}-C_{\mathrm{b}}}{F}_{\mathrm{erg}}^{\mathrm{p}}\mathrm{\Delta }t,\end{array}$$(4)

where C is the total number of detected counts during the GRB, C_p is the number of detected counts at the peak time bin (t_p) of duration Δt = 4 s, C_b is the mean counts per bin, and k is the number of bins over which we calculate the fluence. We obtain the fluence $S=2.2_{-0.3}^{+1.4}\times {10}^{-2}$ erg cm⁻² in the 80 − 800 keV range of the first GRB episode, which lasted 300 s and was observable by GRBAlpha from time 13:19:29.5 to 13:24:29.5 UTC on 2022 October 9. In the same way, but using the wider extrapolated energy range 0.9 − 8690 keV, we obtain the fluence $S^{\mathrm{bol}}=4.9_{-0.5}^{+0.8}\times {10}^{-2}$ erg cm⁻² and the isotropic-equivalent released energy $E_{\mathrm{iso}}^{\mathrm{bol}}=2.8_{-0.5}^{+0.8}\times {10}^{54}$ erg in the 1 − 10 000 keV range (rest frame).

5. Discussion

The highest peak flux and fluence of all GRBs recorded with CGRO/BATSE (Paciesas et al. 1999) and Fermi/GBM (von Kienlin et al. 2020) are 496 ph cm⁻² s⁻¹ (50−300 keV) and 8.1 × 10⁻⁴ erg cm⁻² (50−300 keV) for GRB 130427A. The measured peak flux and fluence of GRB 221009A are ${1020}_{-160}^{+950}$ ph cm⁻² s⁻¹ and $8.3_{-1.1}^{+5.3}\times {10}^{-3}$ erg cm⁻² in the 50−300 keV band. The duration of GRB 221009A is longer than 100 s, compared to the 10−20 s for GRB 130427A, resulting in a relatively large fluence ratio. This indicates that GRB 221009A is indeed exceptionally bright. While the size of the CsI scintillator on GRBAlpha is not small (∼55 cm² face-on), the low-energy threshold of the detector is relatively high at around 80 keV due to the radiation damage of the SiPM detectors. This allowed GRBAlpha to avoid saturation of the count rate in the digital processing, in turn allowing us to trace the light curve. We note that the brightest gamma-ray flux from a celestial object was observed during the giant flare from the magnetar SGR 1806-20 in 2004; its peak flux was around 10–20 erg s⁻¹ in the 20–10 000 keV band (Mazets et al. 2005; Terasawa et al. 2005), which is 4–5 orders of magnitude higher than that of GRB 221009A.

This exceptionally bright GRB has one of the highest isotropic-equivalent peak luminosities ever reported (Burns et al. 2023). Gamma-ray bursts with higher peak luminosities have been observed in the past, but at significantly larger redshifts (see e.g., Yonetoku et al. 2010). In their Fig. 3, Yonetoku et al. 2010 provide the correlation between the rest-frame spectral peak energy and 1-s peak luminosity of short and long GRBs. Our measurements of $L_{\mathrm{iso}}^{\mathrm{p},\mathrm{bol}}=8.4_{-1.5}^{+2.5}\times {10}^{52}$ erg s⁻¹ and $E_{\text{p}}^{\ast }=E_{\text{p}}(1+z)=1120\pm 470$ keV are consistent with this relation within the systematic errors.

In Fig. 1 one can see that at low energies, mainly at 80 − 120 keV, the background shows a wavy pattern. This is most likely due to the periodic motion of GRBAlpha that exposes a different cross-section of its detector to the cosmic X-ray background and the secondary gamma-rays induced by cosmic rays in the Earth’s atmosphere (Galgóczi et al. 2021). On the same day that we detected this GRB, we also observed a periodicity in the received radio power in the waterfalls while downloading data from the satellite; this indicates that the satellite completed more then one revolution during the duration of the GRB. Therefore, a change in the satellite’s attitude between times t_s and t_p can result in a systematic error of the reported peak luminosity and total emitted energy. As shown in Table 1, the luminosity derived for the edge-on direction is 3.5 times higher than the median value. We cannot exclude that the detector on board GRBAlpha was oriented edge-on towards the GRB at the time of the peak, and therefore, we cannot exclude that the obtained median peak flux and luminosity are underestimated.

The time-resolved spectral analysis of GRB 221009A by Konus-WIND (Frederiks et al. 2023) reveals a spectral evolution during the burst with a higher photon index, α, and the peak energy, E_p, around the peak time reaching E_p ≈ 3 MeV. Hence, our method based on scaling the flux from the spectrum at time t_s to peak time t_p may underestimate the resultant peak flux and fluence. It is possible that the true peak flux and fluence are higher.

The highest-energy bands around 800 − 950 keV, as shown in Fig. 1, have an increased number of counts during the GRB peak compared to lower-energy bands around 700 − 800 keV. This is also evident in the raw count spectrum at the peak time, t_p, presented in Fig. 5. This could be due to an instrumental effect because of the high incident photon rate.

6. Conclusions

Our conclusions can be summarised as follows:

• GRBAlpha, a low-cost nano-satellite in a low-Earth orbit, detected the exceptionally bright long gamma-ray burst GRB 221009A without saturation, but affected by pile-up, providing light curves of the prompt emission in 13 energy bands from 80 keV to 950 keV.

• The peak flux in the 80 − 800 keV range (observer frame) is measured to be $F_{\mathrm{ph}}^{\mathrm{p}}={1300}_{-200}^{+1200}$ ph cm⁻² s⁻¹, or $F_{\mathrm{erg}}^{\mathrm{p}}=5.7_{-0.7}^{+3.7}\times {10}^{-4}$ erg cm⁻² s⁻¹, and for the wider extrapolated energy range of 0.9 − 8, 690 keV we obtain $F_{\mathrm{erg}}^{\mathrm{p},\mathrm{bol}}=1.3_{-0.2}^{+0.4}\times {10}^{-3}$ erg cm⁻² s⁻¹.

• The peak isotropic-equivalent luminosity in the 92 − 920 keV range (rest frame) was $L_{\mathrm{iso}}^{\mathrm{p}}=3.7_{-0.5}^{+2.5}\times {10}^{52}$ erg s⁻¹, and the bolometric peak isotropic-equivalent luminosity in the 1 − 10 000 keV range (rest frame) was $L_{\mathrm{iso}}^{\mathrm{p},\mathrm{bol}}=8.4_{-1.5}^{+2.5}\times {10}^{52}$ erg s⁻¹ (4 s scale). Our measurement of $L_{\mathrm{iso}}^{\mathrm{p},\mathrm{bol}}$ is consistent with the Yonetoku relation between the rest-frame spectral peak energy and the peak luminosity of GRBs.

• The fluence in the 80 − 800 keV range of the first GRB episode, which lasted 300 s and was observable by GRBAlpha from 13:19:29.5 to 13:24:29.5 UTC on 2022 October 9, was measured to be $S=2.2_{-0.3}^{+1.4}\times {10}^{-2}$ erg cm⁻² and $S^{\mathrm{bol}}=4.9_{-0.5}^{+0.8}\times {10}^{-2}$ erg cm⁻² for the extrapolated range of 0.9 − 8690 keV. We infer from this extrapolated fluence that the isotropic-equivalent released energy of the first GRB episode was $E_{\mathrm{iso}}^{\mathrm{bol}}=2.8_{-0.5}^{+0.8}\times {10}^{54}$ erg in the 1 − 10 000 keV band (rest frame).

• It is possible that, due to the spectral evolution during the peaks of this GRB, as reported by Konus-WIND (Frederiks et al. 2023), the true peak flux, fluence, L_iso, and E_iso may be higher than our obtained values. We cannot exclude that the GRBAlpha’s detector was oriented near edge-on with respect to the GRB 221009A at the peak time, which would also mean that the true values may be higher than our estimations.

Acknowledgments

We are thankful to Peter Veres and Stephen Lesage for discussions regarding the detection of this GRB by Fermi/GBM and other instruments and to Tomáš Plšek for discussions about MCMC and XSPEC. We are also thankful to Nozomu Kogiso, Maumu Yoneyama, Mizuki Moritaki, Tatsuya Kano and James W. Cutler for providing their SatNOGS stations for collecting GRBAlpha’s telemetry. We acknowledge support by the KEP-7/2018 and KEP2/2020 grants of the Hungarian Academy of Sciences and SA-40/2021 of the Eötvös Loránd Research Network for satellite components and payload developments and the grant IF-7/2020 for providing the financial support for ground infrastructure. This research has been supported by the European Union’s Horizon 2020 programme under the AHEAD2020 project (grant agreement n. 871158) and by the MUNI Award for Science and Humanities funded by the Grant Agency of Masaryk University. G.D. is supported by the Ghent University Special Research Funds (BOF) project BOF/STA/202009/040 and the Fonds Wetenschappelijk Onderzoek (FWO) iBOF project BOF20/IBF/124. This work was supported by the Internal Grant Agency of Brno University of Technology, project no. FEKT-S-20-6526. This research has been also supported by JSPS and HAS under Japan-Hungary Research Cooperative Program, JSPS KAKENHI Grant Number 17H06362, 19H01908, and 21KK0051. We are grateful to the Ministry of Education, Science, Research and Sport of the Slovak Republic for the support of the GRBAlpha mission.

References

All Tables

All Figures

Fig. 1. Raw light curve of GRB 221009A as observed by GRBAlpha in multiple energy bands with 4 s cadence. The bottom-rightmost panel shows the light curve in the whole sensitivity range. Times tb, ts, and tp mark the beginning of the used background interval, the source spectral data, and the peak time, respectively. About 30 s after the peak at tp, the GRB shows a second strong peak. Time tr marks the approximate moment when the satellite entered the outer Van Allen radiation belt, which resulted in the final part of the GRB prompt emission being flooded by the particle background. The sharp increase just after tr is due to this background.

In the text

	Fig. 2. Mass model of GRBAlpha used for generating detector response matrices, with zenith and azimuth angles marked. The CsI(Tl) detector in its aluminium casing is shown in red and the PbSb3 radiation shield in green. The aluminium CubeSat platform, with four stacks of PCBs and two LiFePO4 batteries, is shown in grey.
In the text

	Fig. 3. Total response efficiency (effective area) vs. incident photon energy for a few zenith angles, Θ, and a constant azimuth angle, Φ = 270°.
In the text

Fig. 4. Fraction of the detected counts in each spectral band vs the total counts over the whole energy range for GRB 221009A and for the background as it passed the north polar region. The counts were detected in 4 s time bins. The figure shows that above ∼10 000 cnt bin = 2500 cnt s−1 the fraction of counts in low-energy bands decreases while the fraction of counts in high-energy bands increases, which can indicate a pile-up effect.

In the text

Fig. 5. Detected count spectra of GRB 221009A and the background spectrum. The count spectrum of the GRB at time ts minus the background spectrum at time tb is shown in black. The background spectrum at time tb is shown in red. The count spectrum of the GRB at the peak time tp = 13 : 20 : 49.5 UT on 2022 October 9, minus the background spectrum at time tb, is shown in green and it reveals a spectral hardening at the highest energies present around the GRB peak.

In the text

	Fig. 6. Spectrum of the GRB 221009A best fit by the CPL model fitted from ts = 13 : 20 : 33.5 UTC on 2022 October 9 to ts + 4 s in the range 76 − 809 keV for Θ = 180° and Φ = 0°.
In the text

	Fig. 7. Reduced χ2 map of the best fits with the CPL model for different angles of Θ (vertical) and Φ (horizontal) in degrees.
In the text