# 1.3.4 Focal-plane status

Author(s): Jos de Bruijne, Gonzalo Gracia-Abril, Neil Cheek, Cian Crowley, Claus Fabricius, Juanma Martín-Fleitas, Alex Hutton, Alcione Mora

## Focal-plane assembly

During the time interval covering Gaia EDR3, all elements of the focal plane have been operating nominally. There have been no major issues with the PEM-CCD electronics-detector couples and all 106 devices are functioning within specification with only one device – the AF3 CCD in row 1 – slightly exceeding the readout-noise specification (Kohley et al. 2012; Crowley et al. 2016).

## Focal-plane temperature

Amongst the many different temperature sensors on-board the satellite, three are placed close to the detector array. They are distributed across the focal plane with one placed near the SM1 CCD on row 1, one placed near the AF5 CCD on row 4, and one placed near the RVS3 CCD on row 5 (see Figure 1.2). Figure 1.16 shows the readings obtained from these three sensors after smoothing the data by convolving them with a one-hour, running-average smoothing kernel. One can immediately note three temperature increases. These correspond to controlled payload heating events which have been carried out in order to decontaminate the telescope mirrors (Section 1.3.3 and Table 1.10).

After thermal equilibrium is re-established following each heating event, the focal-plane temperature is extremely stable. Large temperature changes can dramatically affect detector characteristics (such as charge-transfer efficiency or gain), so long-term stability is an important aspect of Gaia’s performance. There is a stable gradient of a few degrees from sensor-to-sensor over the focal plane and the operating temperatures of all devices are close to the target value of $-110$${}^{\circ}$C. The slow, periodic, long-term evolution of the temperature readings is caused by the variable distance of the spacecraft to the Sun that is primarily linked to the eccentricity of the Earth’s orbit around the Sun.

## CCD cosmetic defects

No CCDs can be manufactured without a (small) number of non-perfect, or defected, pixels. These can be manifest in a number of ways, but typically there will be a number of ‘cold’ pixels which are under-responsive in comparison to the surrounding population, and a number of ‘hot’ pixels that exhibit elevated signal levels, even in darkness. Due to the TDI operation of the Gaia CCDs (Section 1.1.3), such pixel defects will be smeared out in some form due to the fact that the signal is integrated over all $4500$ CCD image-area pixels in the column before reaching the serial register (see Figure 1.2 and Figure 1.4 for the adopted CCD nomenclature). Of course, for TDI-gated observations, the effective integration distance is shortened, so it may be that a defect pixel far from the serial register produces an effect in the non-TDI-gated data but not in (all or some of) the TDI-gated data. In addition to these fabrication (native) cosmetics, radiation-induced defects are created which can evolve in-orbit.

For Gaia, apart from the standard need to characterise CCD defects for the processing of the science data, it is also important to take them into account in order to ensure correct operation of the on-board detection (the SM CCDs) and confirmation (the AF1 CCDs) chains (Section 1.1.3): defects and evolution in dark-signal non-uniformity (DSNU) need to be tracked and updated on-board as/if needed. The locations and characteristics of CCD defects have been diagnosed pre-flight and stored in the VPU memories for the on-board processing of the SM, AF1, and RP windows (the RP-window data is used on-board to provide a red magnitude for each object which is used as input when deciding whether an RVS window should be assigned to an object or not; see Section 1.1.3). So far, only one on-board change has been required (and implemented on 16 December 2014, OBMT 1653.0) in order to update the DSNU table for one SM2 CCD (on row 4) which contains a weak ($\sim$15 ADU), radiation-induced hot column.

When analysing the whole focal plane using the analysis of defect pixel evolution as described in Section 3.3.3, it is found that the number of radiation-induced defects is low. The sensitivity of this analysis to weak defects will increase with time and more detailed statistics will become available with progressing mission duration (for further detail on defect pixel evolution, see Crowley et al. 2016).

## Prompt-particle events

As explained in Section 1.1.3 and in detail in de Bruijne et al. (2015), the object-detection algorithms running on-board Gaia scan the images coming from each sky-mapper (SM) CCD in search for local flux maxima. For each local maximum, spatial-frequency filters are applied to the two one-dimensional flux distributions within a so-called working window centred on it. Objects failing to meet star-like PSF criteria – i.e., that are too sharp or too smooth / extended – are rejected. Typically, rejected objects are either bright-star PSF ripples or prompt-particle events (PPEs; i.e., Solar protons, Galactic cosmic rays, etc.). A few example prompt-particle events are shown in Figure 1.17.

The filtering mechanism described above and taking place at the SM CCDs is followed by a confirmation stage in the AF1 CCDs. This stage removes those ‘objects’ that, having been detected in SM, do not re-appear in AF1 at the ‘right’ location with the ‘right’ flux. Non-confirmed objects are typically cosmic rays or Solar protons. Detection-process statistics (e.g., the number of rejected prompt-particle events) are telemetered to ground in the form of auxiliary-science-data (ASD) packets of type 4 (Section 1.1.3). These are only counters – in fact a ‘by-product’ of the detection chain – and neither energy nor nature of the impacting radiation particles are provided. Figure 1.18 shows the typical prompt-particle-event rates extracted from the auxiliary science telemetry. The prompt-particle-event time lines contain two main features, namely (i) a ‘background’ count rate (at $\sim$40 counts s${}^{-1}$) which is present at all times, on top of which are (ii) ‘peaks’ of counts (expectedly) concentrated around times of increased Solar activity.

The predicted background PPE rate $R_{\rm CCD}$ measured at a CCD (either SM or AF) produced by an incident, isotropic flux $F_{\rm CCD}$ of prompt-particle events is given by (see Sullivan 1971):

 $R_{\rm CCD}~{}[{\rm counts}~{}{\rm s}^{-1}]=\frac{F_{\rm CCD}\cdot A}{2},$ (1.4)

where $A$ is the effective surface area of the given detector in cm${}^{2}$ ($A=17.1$ cm${}^{2}$ for SM CCDs, taking the permanently-active TDI gate into account; Table 1.3) and $F$ is expressed in particles cm${}^{-2}$ s${}^{-1}$. For a typical particle background rate at L2 of $\sim$5 protons cm${}^{-2}$ s${}^{-1}$ (Barth et al. 1999; Catalano et al. 2014), the expected prompt-particle-event rate measured by Gaia at L2 is $R_{\rm CCD}\simeq 42.75$ counts s${}^{-1}$, which is in good agreement ($<$10% difference) with the actually measured rates. The peaks in the prompt-particle-event counters are clearly correlated with increased Solar activity. Figure 1.19 shows a selected, qualitative comparison of Gaia prompt-particle-event counters with data from dedicated radiation-monitoring instruments on-board other spacecraft, in particular ACE and GOES. The agreement is satisfactory, taking into account that the Gaia focal plane is shielded by the thermal-tent structure, the focal plane itself, and the service module of the spacecraft such that it should systematically measure reduced (low-energy) particle fluxes compared to the fully-exposed, dedicated radiation instruments on-board the other spacecraft.

Prompt-particle events impacting on a CCD can produce displacement radiation damage. The silicon-lattice defects (‘traps’) created in this process impact the charge transfer of the photo-electrons during the readout in TDI mode as charge is trapped and released, with typical time scales depending on the device temperature and the trap species (Section 1.3.4). The resulting degradation in the charge-transfer efficiency – or charge-transfer inefficiency (CTI) effect, with ${\rm CTE}=1-{\rm CTI}$ – introduces a loss of signal (flux) measured in the window as well as a bias in the estimated centroid position of each star with a dependency not only on the radiation dose but also on the (recent) illumination history of the pixels in question (Prod’homme et al. 2011, 2012; Holl et al. 2012). On-board measures have been taken to mitigate this effect, e.g., periodic activation of charge injections (CIs; Section 1.3.4) that help to keep existing (slow) traps filled. The flux loss and centroid bias, nonetheless, are significant, in particular when significant radiation levels have been experienced, and need to be corrected in the data processing (e.g., Short et al. 2013). Such calibrations have not yet been implemented for Gaia EDR3.

Gaia was launched in December 2013, just after the maximum of Solar cycle 24. Solar activity during the mission is therefore progressively decreasing towards Solar minimum (see Figure 1.20), resulting in a ‘benign’ situation from a CCD-radiation-damage perspective.

The Gaia prompt-particle-event counters are used to monitor the instantaneous radiation environment experienced by Gaia at L2 and its possible implications for radiation damage. For a quantitative assessment, however, first-pixel-response analysis of charge injections (Section 1.3.4) is used to study the level of radiation damage accumulated by the Gaia CCDs.

## Parallel charge-transfer inefficiency

The L2 environment offers many advantages for the operation of astronomical observatories, such as high thermal stability. Due to the large distance of L2 from Earth ($\sim$1.5 million km), the effect of geomagnetically-trapped charged particles is not a concern. However, it follows that L2 lacks the shielding of the Earth’s magnetosphere and is thus vulnerable to impacts from ionised particles from other sources. Indeed, the L2 radiation environment is dominated by particles from the effectively isotropic cosmic rays and the more directional Solar eruptive events (mostly protons). The cosmic-ray component is expected and is also observed to be rather steady throughout the mission, with the number of impacts expected and observed to vary smoothly and, as expected, anti-correlated with the Solar cycle (see Section 1.3.4). In contrast, the impacts of particles from Earth- / L2-directed Solar events are sporadic with more events expected and observed around the time of Solar maximum.

The most worrying effects of radiation on the Gaia science performances are expected to be due to performance degradations of the detectors over time. The most important effects are due to both ionising radiation damage and non-ionising radiation damage.

The effects of ionising radiation on the detectors are cumulative and occur when charge builds up, possibly resulting in threshold shifts, charge leakage, etc. A special calibration procedure is periodically run on-board to monitor the accumulation of charge in the oxide layers of the CCD detectors; this is further described in Section 1.3.4.

The non-ionising, lattice-displacement damage to the CCD silicon layers is the biggest radiation threat to the achievement of the Gaia science-performance goals. The generation of new energy levels between the valence and conduction bands in the silicon produces crystal defect sites that trap electrons from traversing charge packets, to be released from the trap at some later moment in time. This results in a smearing of the images that are readout from the CCD that cannot be calibrated in a simple manner (see Short et al. 2013). Indeed, charge-transfer inefficiency (CTI) of the CCDs was identified as a potential threat to the mission at an early stage. CTI in the transfer through the serial (readout) register is discussed in Section 1.3.4; however, Gaia is most sensitive to CTI in the image area (in the along-scan direction).

Besides shielding the detectors using (custom-placed) ‘surrounding’ masses, which is only partially effective given that the focal plane has to be an open structure in order to receive light from the telescopes, the most important hardware mitigation for along-scan CTI effects is the periodic injection of charge into the devices. This is carried out every $\sim$2 seconds for AF1 and AF2–9 devices and every $\sim$5 seconds for BP/RP devices. This scheme sees four contiguous lines of charge injected into the CCD and clocked out through the $4500$ TDI lines. This serves to periodically fill trapping sites with injected electrons, thus preventing the trapping of photo-electrons from astronomical sources. Indeed, for those traps with characteristic release time scales on the order of seconds, these traps will be kept filled over the period of the charge injection, thus keeping them nearly permanently filled and thus rendered effectively inactive. The regular presence of the injection features in the data stream also means that the CTI effects on the injections can be used to monitor the evolution of the trapping and release effects on the detector over time. One such diagnostic is what is known as the first-pixel response (FPR) and is measured by computing the number of electrons removed from the first injected pixel line through trapping (in practise, we also need to account for trapping from subsequent lines).

Shown in Figure 1.21 are the FPR results (displayed as fractional charge loss) averaged over all AF2–9 CCDs. The steady increase in CTI is apparent and is due to the low-energy tail of the Galactic cosmic-ray impacts. The step increases are correlated with Earth-directed Solar-proton events.

The almost linear degradation in the transfer efficiency is noticeable and rather similar for all devices. However, the effects of the Solar-proton events are heavily affected by the differential shielding across the focal plane. Therefore, the net CTI observed in each device is a convolution of the (low-level) initial pre-flight CTI due to traps generated during the manufacturing process, the effect of the cosmic-ray impacts, and the effects of the Solar protons. Shown in Figure 1.22 is the result of the extrapolation of in-flight FPR measurements to May 2020 (roughly one year after the end of Gaia’s nominal, five-year mission). The higher CTI FPR values for the RP CCDs are apparent and caused by the higher inherent along-scan transfer inefficiencies for these devices that were already measured pre-launch. Also apparent in this figure is a pattern showing higher FPR values towards the centre of the focal plane, which is a result of the shielding. However, the most important result that can be derived from these data is that the now-predicted end-of-mission CTI values are approximately an order of magnitude less than was feared before launch.

Furthermore, offline studies of the image shapes as a function of the time since charge injection have confirmed that the effects of CTI in the along-scan direction are low. A full analysis of the effects of CTI on the science data will become possible with time as more calibrations are carried out in the iterative data processing. Note that no periodic charge injection occurs in the SM and RVS devices, however. The CTI in the along-scan direction in these (and the other) devices is monitored through special calibration activities that take place every three to four months. The results of this monitoring are consistent with those presented here for the AF, BP, and RP CCDs. See Crowley et al. (2016) for an in-depth discussion of the effects of the radiation environment on the Gaia detectors over the first two years of the mission.

## Serial charge-transfer inefficiency

A common technique for monitoring the CTI in the serial (readout) registers of CCDs (sometimes abbreviated to sCTI) is to generate charge in the image area and then monitor any trailing into the post-scan samples (Figure 1.4) after the charge is transferred through the serial register. Since the Gaia VPUs are not capable of generating nominal data packets containing post-scan samples, a special calibration is run periodically where five different levels of charge are injected into each science device (apart from SM devices, where the definition of the CCD operating mode precludes the acquisition of post-scan pixels). At the same time, ‘virtual objects’ (VOs; Section 1.1.3) are placed over the end of the image area (in order to monitor the injected charge level) and are also used to acquire post-scan pixels. To circumvent the problem where post-scan samples cannot be stored in nominal packets, the VPU SIF functionality (Section 1.1.3) is used to gather the raw data, which is subsequently down-linked in a special data packet with high priority. Since radiation damage has been lower than expected (Section 1.3.4), it is currently sufficient to run this activity with a cadence of somewhere between three to four months.

During the time interval of Gaia EDR3, there have been nine runs of the serial-CTI calibration in-flight (Table 1.12). These runs allow the tracking of the evolution of the CTI in the serial register. However, it is found that the CTI is still dominated by the traps that have been present since device manufacture. Shown in Figure 1.23 are data derived from each run of the activity so far for the AF5 device on CCD row 5, and for the medium injection level. The co-added, high signal-to-noise cuts in the across-scan direction show the two pre-scan samples, followed by image-area samples showing the injected charge, and then the trailing into the post-scan samples due to the release of the injected electrons from traps in the serial register. In Figure 1.24, the very slow evolution of the serial CTI curve from run-to-run is readily apparent.

Similar to the situation for CTI in the along-scan direction in the image section (Section 1.3.4), the continued domination by pre-flight manufacturing traps, compared to the slow evolution of serial CTI in the readout register caused by radiation damage, is apparent all across the focal plane. In summary, CTI in the serial register is still dominated by the traps inherent to the manufacturing process and the radiation-induced degradation in CTI, whilst clearly measurable, is of the order of a few dozen percent of the pre-flight CTI ($\sim$$10$% for SM/AF/BP devices and $\sim$$20$% for the thicker, red devices used in RP/RVS – Section 1.1.3 – which had much lower native CTI to begin with).

## Flat-band voltage shift

Radiation-induced generation of electron-hole pairs in the CCDs cause a gradual accumulation over time of positive charge on the CCD-gate oxide layers. The end result is an effective change in the potentials under the gates of each of the CCDs, which is termed a ‘flat-band voltage shift’. The operating point of each device has been set to tolerate the maximum expected flat-band voltage shift of $\sim$0.5 V over $\sim$10 years, but the trend over time is being monitored to allow steps to be taken to avoid degradation in the performances of the devices. In principle, any shift can be compensated for by changing the operating voltages of the devices by an appropriate amount. A periodic run of a special calibration activity on-board is used to monitor the evolution.

During this activity, charge-injection data is acquired from every CCD and the potential under the injection drain ($V_{\rm ID}$) is successively increased from a low value of $12.5$ V in order to find the $V_{\rm ID}$ value at which charge stops being injected. This point, the turn-off voltage, corresponds to the point at which the $V_{\rm ID}$ equals the potential under the injection gate. The potential under the gate will evolve with increased charge accumulation in the oxide, but the potential under the drain should remain unaltered since it is connected directly to the silicon. Thus, the change in operating point over time diagnoses the flat-band voltage shift. This activity is being run throughout the mission with a cadence of $\sim$12 months. In the time range applicable to Gaia EDR3, the procedure has been run on-board two times, excluding two further dry-runs of the activity shortly after launch, during commissioning. Table 1.13 summarises the flat-band-voltage-shift-calibration operations that have been executed during the time interval covering Gaia EDR3.

The average and standard deviation for the on-ground-measured turn-off voltages for the 106 flight devices is $15.574\pm 0.258$ V. A comparison with in-orbit data collected so far (including data acquired in July 2019) shows no clear evidence for a measurable flat-band voltage shift beyond the measurement noise for any one device considered in isolation (the mean shift over all devices between the on-ground, pre-launch measurements and the July-2019 on-board measurements is also insignificant, at $+0.0035\pm 0.0212$ V). Therefore, it is expected that flat-band voltage shifts will not cause detector performance issues even before the end of an extended, ten-year mission.

## Bias non-uniformity calibration

As discussed in detail in Section 3.3.2 and in Hambly et al. (2018), the electronic readout of the Gaia CCDs suffers from bias-level (offset) variations within a TDI line, also referred to as (bias) non-uniformity or (offset) instability, which have a non-trivial dependency on the exact readout scheme. Since most CCDs are operated in a windowing mode (Table 1.2 and Figure 1.3), where the exact locations of the windows depend on the sky configuration as well as the continuously-changing superposition of both fields of view (Section 1.1.3), it follows that a continuous reconfiguration of the readout scheme is applied to most devices.

There exists a phenomenological model of the offset variation as a function of various readout parameters (Hambly et al. 2018) that needs to be calibrated in order to properly correct these offsets in the data processing on ground. Some details on the various effects are discussed in Section 3.3.2 and an extensive description of the model as well as its efficacy is given in Hambly et al. (2018); here, we discuss only the status of the special on-board calibration that is carried out to monitor bias non-uniformity.

During the non-uniformity calibration activity, the VPUs are, one at a time, taken out of operational mode and artificial ‘virtual object’ (VO; Section 1.1.3) detections are placed into the VPU algorithms (see Section 1.1.3). These detections result in empty (sky) windows to be read out from each device as the VOs are propagated across the various science CCDs in the row controlled by that VPU. The VO patterns are designed to sample different readout configuration timings and are used to calibrate the offset non-uniformity model (see Hambly et al. 2018 for details). On 31 August 2016 (OBMT 4150.6), the VO pattern on board was updated in view of the intra-TDI phase anomaly (again, see Hambly et al. 2018 for details). The nominal cadence of each calibration run has settled to be between three to four months and, in the Gaia EDR3 time range, the activity has been performed on-board nine times (Table 1.14).