Traitement des glitches

Le traitement est effectu´e anneau par anneau [Planck Collaboration 2014d]. Le processus est it´eratif :

– Une estimation du ciel est faite en utilisant la redondance de la mesure sur un anneau (environ 60 cercles). On interpole le signal en utilisant des splines pour ´eviter les effets sub-pixel, et on soustrait ce ciel estim´e `a la TOI.

– On ouvre une fenˆetre de 1000 ´echantillons, et on estime la moyenne quadratique du bruit.

– Pour chaque ´ev´enement au dessus du seuil de d´etection (le seuil final est fix´e `a 3.2 σ du bruit), on fait un ajustement joint des mod`eles de glitches longs et courts1_.

– Dans le cas de glitches longs (et pour les glitches lents), les mod`eles ajust´es sont soustraits et la partie o`u le mod`ele est `a plus de 1 σ n’est pas utilis´ee pour les cartes. Dans le cas des glitches courts, les donn´ees pour lesquelles le mod`ele correspondant est `a plus de 0.1 σ ne sont pas utilis´ees, aucune estimation n’est soustraite.

– Une fois ce nettoyage effectu´e, on applique un filtre adapt´e pour une exponentielle d´ecroissante `a constante de temps de 2 s au signal restant, pour d´etecter les ´ev´enements ayant ´echapp´e au nettoyage.

– On fait une nouvelle estimation du signal du ciel, et on baisse le seuil de d´etection des glitches, puis on recommence la proc´edure de nettoyage.

A chaque it´eration, le seuil de d´etection diminue d’un facteur constant, jusqu’au seuil de 3.2 σ (cf. figure7.2). Une fois ce nettoyage termin´e, 8% `a 20% des donn´ees sont rejet´ees, selon les bolom`etres.

Rayons cosmiques sous le seuil de d´etection

La sensibilit´e des d´etecteurs de Planck-HFI est telle que presque tous les rayons cos- miques sont d´etect´es [Planck Collaboration 2013b]. En effet, pour certains bolom`etres, on observe une coupure `a basse ´energie dans la distribution cumulative des glitches longs (cf. figure 7.5), qui correspond au minimum d’´energie d´epos´ee par des protons ´energ´etiques sur le support en silicone (140 keV en incidence normale). Cette coupure est une des preuves les plus flagrantes que les glitches longs sont dˆus `a l’interaction des rayons cosmiques avec le silicone. Elle est significativement d´etect´ee pour tous les bolom`etres, et sert `a calibrer l’´energie d´epos´ee par les rayons cosmiques.

Planck HFI Core Team: Energetic Particle E↵ects in Planck Characterization, Removal and Simulations

electrons hitting the second grid and depositing about 10% of the energy deposited in the first grid. We also observe that some events are hitting both grids. Those appears in the cloud of points with the same amplitude in both PSB-a and b. Those represent about 2% of the events. We observe a lower coincidence rate for events at higher amplitudes corresponding to the bump. This is explained with the hypothesis that those events correspond to di- rect impact on the thermistor, which is not aligned with the grid of the other bolometer. Nevertheless, we have seen that all those high energy events are in coincidences with a small amplitude event in the second bolometer. This is attributed to the cross–

talk between bolometer in a pair, as discussed in Section3.4.

The non–linearity, appearing with the curved shape in the coin- cidences, can be entirely explained by the saturation of the high- est amplitude glitches. Nevertheless, we can evaluate the level of cross–talk as between 0.01 to 0.44% depending of the bolometer pair, for ⇡ 3000 sigmas events for which the e↵ect of saturation is small. The high amplitude events correspond to an energy of ⇡ 15 keV left on the grid+thermistor by a particle, this is very close to the expected energy deposited by a 1 GeV proton on the thermistor.

3.2. Interaction with the wafer

We identify the long glitches as produced by cosmic rays hit- ting the silicon die. This was first indicated by the ground tests (Catalano et al. 2013), showing that the NTD thermometer is sensitive to a temperature change of the silicon die. The HFI ground-based calibration show a rate of events compatible with the cosmic rays flux at sea level over the silicon die surface and also that almost all these events are in coincidence between PSB- a and PSB-b. The understanding is the following: phonons gen- erated by the event impact in the silicon die and produce fast ris- ing time of the Germanium temperature, which decays with the bolometer time constant. The slow part is the thermal response of the entire silicon die temperature rising and then falling as

the heat conducts out from the die to the heat sink (Catalano

et al. 2013). This hypothesis is reinforced by the comparison of cumulative counts N(>E) of long glitches from bolometer

to bolometer which is shown in Figure19. To make this com-

parison, we have normalized the counts such that we evalu- ated the number of event per unit of time and per unit surface of the silicon wafer (as the total surface varies from bolome- ter to bolometer). Also, as for short glitches, we have intercal- ibrated the energies of events between bolometers by match- ing the di↵erent counts at measured energies around 0.05 keV, which correspond to the deposited energy on the silicon die of

⇡ 103_{keV after absolute calibration. This absolute calibration}

is performed such that the observed faint–end break of counts, which is clearly observed in the figure, match the expected min- imum deposited energy by ⇠ GeV protons on the silicon die, which is ⇡ 140 keV, for normal incidence, as described later. Energy calibration cannot be performed accurately from mea- surement due to important uncertainties of heat capacities of the

silicon die. Figure19shows the cumulative counts for all HFI

bolometers used for the scientific analysis. We clearly see that they all match in shape and amplitude with very small scatter. For energies below 20 sigmas of the noise, we have also com- puted the cumulative counts for all events without distinction on families (green curves in the figure), but we have already seen that long events dominate at low energy, so the total counts are representative of counts of long glitches. The faint–end break in the counts is detected without ambiguity. This is an important result since it shows that there is a limited number of low en-

10 -4 10 -2 10 0 10 2 102 103 104 105 106 Amplitude (KeV)

Event number N(>E) / hour / mm

2 _Total

Long Toy model protons

Toy model He

Fig. 19. Cumulative distribution N(>E) of long glitches per unit surface of the wafer for all bolometers used for the scien- tific analysis. The black solid lines correspond to selected long glitches. The green lines correspond to total glitches, but are rep- resentative of long glitches since they dominate at low energy. A relative calibration of the energy, which indicated in the x- axis, has been for all bolometers such that the cumulative counts

match at ⇡ 103_{keV. The absolute calibration of the energy is}

performed such that the faint–end break of the counts match the expected deposited energy of ⇡ 140 keV in the silicon die. This calibration on the model is necessary since we only measure the energy in the grid+thermistor system, and the heat capacity and link conductivities are not known with a sufficient accuracy. Also, we use the values of peak of glitches to record the glitch amplitudes which are not representative of thermal processes, as the fast part of long glitches are expected to result from ballis- tic phonon e↵ects. Nevertheless, we observe that the shape of the cumulative counts match for all detectors with a low disper- sion. The break of the counts at faint–end is very significantly detected. This is our best indication from flight data that long glitches result from particles hitting the bolometer silicon die. Dot dash and dash lines correspond to the predicted power law distributions of deposited energy from protons, and helium, re-

spectively, from the toy model developed in3.3for sensibly en-

ergies higher than the stopping power energies. Predicted min- imum deposited energies for helium and protons are indicated with doted lines. We interpret the second apparent bump at en- ergies ⇠ 20 times the energy of the break as the signature of He nuclei. The fact that we observe an excess in the counts at the expected energy of the minimum stopping power for helium reinforces our hypothesis for the origin of long glitches. ergy glitches in data which are undetected, and would a↵ect the cosmological results. From this limit, we measure that the total number of events penetrating the shielding of the satellite around the bolometer is about 4.5 per second and square centimetre. This is in very good agreement with the expected value in the

bolometer environment computed in Section3.1of ⇡ 5 s 1_cm 2_.

The slope and amplitude of the distribution can be predicted with a simple model of interaction of particle with the silicon die as

shown by the toy model described in Section3.3. Predictions are

also represented in Figure19for primary galactic protons and

Helium nuclei. We observe a very good match of the model with the data showing that particles detected by Planck at the ener- gies above the faint–end break are primary galactic protons. The barely apparent second bump in cumulative counts at deposited energies ⇠ 3000 keV probably correspond to the contribution

10

Figure 7.5 – Distribution cumulative des glitches de type long par unit´e de surface

du support en silicone N(>E). Les lignes pointill´ees verticales repr´esentent les minima

d’´energie d´epos´ee par les protons (∼ 140 keV) et par l’h´elium (∼ 3000 keV) sur le

support en silicone. Figure dePlanck Collaboration [2013b].

Pour les d´etecteurs ayant un fort taux de glitches d´etect´e, dˆu `a un lien thermique plus important entre la grille et le support en silicone, la coupure est au del`a du seuil de d´etection de la proc´edure de nettoyage. Ainsi, ces bolom`etres d´etectent presque tous les ´ev´enements. Pour les autres d´etecteurs, une partie des glitches n’est pas d´etect´ee.

Comme on a pu l’observer, les paires de bolom`etres PSB sur un mˆeme module d´etectent la plupart des glitches longs en co¨ıncidence. Les spectres de puissance crois´es entre les TOI nettoy´ees des PSB-a et PSB-b sont montr´es figure7.6. On voit que les paires ´etant sur un mˆeme module (courbes bleues) sont bien plus corr´el´ees que les PSB sur des modules diff´erents (courbes rouges). Ces corr´elations sont dues aux ´ev´enements sous le

seuil de d´etection. Plus les d´etecteurs ont un taux de glitch ´elev´e, et donc un nombre plus faible de glitches non d´etect´es, plus la corr´elation est faible. Cela permet de quantifier le nombre d’´ev´enements sous le seuil.

10 -4 10 -2 10 0 10 2 10-3 10-2 10-1 100 101 Frequency (Hz)

Noise power spectra (arbitrary units)

Figure 7.6 – Spectres de puissance crois´es pour quelques TOI `a 100 GHz. L’estimation

du ciel, ainsi que les queues des glitches d´etect´es ont ´et´e retir´ees des TOI, c’est la partie non projet´ee sur les cartes qui est montr´ee. En noir, les auto-spectres calibr´es, en bleu, les spectres crois´es pour les PSB d’un mˆeme module, et en rouge pour des PSB de diff´erents modules. Pour les courbes bleues, les paires de bolom`etres ayant un taux de

glitches plus faible sont ceux ayant la plus forte corr´elation. Le pic `a 10 Hz est dˆu aux

vibrations du refroidisseur `a 4 K. Figure dePlanck Collaboration [2013b].

Il a ´et´e estim´e que les glitches, apr`es soustraction, ont un faible impact sur l’analyse cos- mologique au niveau du spectre de puissance. Cependant, le signal cr´e´e par les glitches ´etant bien au del`a de la statistique gaussienne (cf. sections suivantes), le nettoyage impar- fait des donn´ees pourrait polluer l’analyse des non-gaussianit´es. Les glitches sous le seuil de d´etection ajoutent un bruit syst´ematique non-gaussien (positif, piqu´e et asym´etrique).

Ce nettoyage imparfait peut aussi polluer la r´ealisation des cartes, en particulier le destriping des cartes (cf. section4.3.6), ce qui peut avoir un effet sur les non-gaussianit´es [Donzelli et al. 2009; Efstathiou 2005]. Les erreurs sur les offsets pourraient ˆetre non- gaussiennes `a cause des glitches sous le seuil ou des queues r´esiduelles.