Single pixel performance of a 32\times32 Ti/Au TES array with broadband X-ray spectra
Matteo D'Andrea, Emanuele Taralli, Hiroki Akamatsu, Luciano Gottardi, Kenichiro Nagayoshi, Kevin Ravensberg, Marcel L. Ridder, Davide Vaccaro, Cor P. de Vries, Martin de Wit, Marcel P. Bruijn, Ruud W. M. Hoogeveen, Jian-Rong Gao
11 Single pixel performance of a 32 ×
32 Ti/Au TESarray with broadband X-ray spectra
Matteo D’Andrea, Emanuele Taralli, Hiroki Akamatsu, Luciano Gottardi, Kenichiro Nagayoshi,Kevin Ravensberg, Marcel L. Ridder, Davide Vaccaro, Cor P. de Vries, Martin de Wit, Marcel P. Bruijn,Ruud W. M. Hoogeveen and Jian-Rong Gao
Abstract —We are developing a kilo-pixels Ti/Au TES array asa backup option for Athena X-IFU. Here we report on single-pixelperformance of a 32 ×
32 array operated in a Frequency DivisionMultiplexing (FDM) readout system, with bias frequencies inthe range 1-5 MHz. We have tested the pixels response atseveral photon energies, by means of a Fe radioactive source(emitting Mn-K α at 5.9 keV) and a Modulated X-ray Source(MXS, providing Cr-K α at 5.4 keV and Cu-K α at 8.0 keV). First,we report the procedure used to perform the detector energy scalecalibration, usually achieving a calibration accuracy better than ∼ ∆ E FWHM = 2.40 ± ± ± ∼ α line at 1.5 keV, generatedby fluorescence inside the experimental setup. We analyzed thisline to obtain a first assessment of the single-pixel performancealso at low energy ( ∆ E FWHM = 1.91 eV ± Index Terms —X-Ray detectors, TES, MXS, X-IFU, EnergyResolution, Energy Calibration
I. I
NTRODUCTION A THENA is a large X-ray observatory (0.2-12 keV energyband), planned to be launched by ESA in the early 2030s.One of the two focal plane instruments is the X-ray IntegralField Unit (X-IFU) [1], which is a cryogenic spectrometerable to perform simultaneously detailed imaging (5” angularresolution over a Field of View of 5 arcmin diameter) andhigh resolution spectroscopy ( ∆ E FWHM < < ∆ E FWHM < Hot and Energetic Universe science theme,providing in particular: integral field spectroscopic mappingof the hot plasma in galaxy clusters; weak spectroscopicdetection of absorption and emission lines from the WarmHot Intergalactic Medium (WHIM); physical characterizationof the most energetic phenomena in the Universe (e.g. AGNwinds and outflows) [2]. The core of the instrument is a
M. D’Andrea is with INAF/IAPS Roma, Via del Fosso del Cavaliere 100,00133 Rome, Italy. During this research, he was with NWO-I/SRON Utrecht,3584 CA Utrecht, The Netherlands. e-mail: [email protected]. Taralli, H. Akamatsu, L. Gottardi, K. Nagayoshi, K. Ravensberg, M.L. Ridder, D. Vaccaro, C. P. de Vries, M. de Wit, M. P. Bruijn, R. W. M.Hoogeveen and J.-R. Gao are with NWO-I/ SRON Utrecht, 3584 CA Utrecht,The Netherlands.J.-R. Gao is also with the Faculty of Applied Science, Delft University ofTechnology, 2600 AA Delft, The Netherlands. large array of ∼ α calibration line generated by meansof a radioactive Fe source.In this work, we present single-pixel performance at differ-ent photon energies (i.e. 1.5 keV, 5.4 keV, 5.9 keV and 8.9keV), and we preliminary investigate the achievable energyscale calibration accuracy, the detector linearity and the count-rate effect on the performance.II. E
XPERIMENTAL S ETUP
The detector under test is a uniform 32 ×
32 pixels array,with 240 × µ m Au absorbers (2.35 µ m thickness). TheTESs have an aspect ratio (lenght × width) of 140 × µ m ,and consist of Ti/Au (35 nm/ 200 nm) bilayers with a criticaltemperature T C ∼
90 mK and a normal resistance R N ∼ Ω . The pixel heat capacity is C ∼ T C , and theexpected thermal conductance is G ∼
95 pW/K at T C . Moredetails about the detector design and fabrication can be foundin [9], while a deep characterization of the array has beenreported in [10] and [11].The detector has been operated in our FDM readout system[12] [13]. Under the FDM scheme, TES are coupled to high-Q LC filters [14] and AC biased at MHz frequencies. TheTES signals are then summed and amplified by a low-noisetwo-stage SQUID amplifier chain, and finally demodulated atroom temperature by digital electronics. The main parts of thecold stage setup are shown in Figure 1. We have used a 18channels LC-filter chip, with a coil inductance L = 2 µ H anda DC connector chip. The pixels bias frequencies are between1 and 5 MHz. The SQUID amplifiers are a VTT J3 (FrontEnd SQUID) and a VTT F5 (Amp SQUID).The setup has been installed in a dilution refrigerator, andthermoregulated at a temperature of 40 mK. It has been sus-pended with respect to the cryostat mixing chamber via kevlar a r X i v : . [ a s t r o - ph . I M ] F e b Fig. 1. Picture of the setup (cold stage) used to operate the kilo-pixel array.The labels indicate the main components. strings, in order to avoid microvibrations due to the pulse tubeoperation, and magnetically shielded by a superconductingaluminum shield (at cold), and by a mu-metal shield (outsidethe cryostat).The detector has been illuminated both by a radioactive Fe source (from the back side) and a Modulated X-raySource (MXS, from the front side). The MXS is a device ableto produce X-rays with an adjustable count rate, by hittingtarget materials with accelerated electron, which are emittedby a photocathode stimulated by UV leds. Details about theMXS design are reported in [15]. The MXS has been vacuumtight connected to an aperture in the cryostat Outer VacuumChamber. All the thermal shields inside the cryostat have beenprovided with apertures covered with 20 microns thick mylarcoated with 200 nm Al. The main purpose of such mylarwindows is to block the stray light and let the X-ray passthrough the cryostat.Figure 2 shows the comparison of the energy spectraacquired by one of the detector pixels (Px8, 2.7 MHz biasfrequency) with the MXS turned ON and OFF. The Mn-K α (5.9 keV) and Mn-K β (6.4 keV) lines are emitted by the Fe Fig. 2. Energy spectrum measured by Px8 (2.7 MHz bias frequency) withthe MXS turned ON (black curve) and OFF (red curve). source. The MXS provides the Cr-K α (5.4 keV), Cr-K β (5.9keV), Cuk-K α (8.0 keV) and Cu-K β (8.9 keV) lines, and thebremsstrahlung continuum. In addition, two low-energy linesare slightly visible in the spectra, generated by fluorescenceinside the setup: the Al-K α at 1.5 keV (generated in thealuminum foils used to attenuate the Fe source), and theSi-K α at 1.7 keV (generated in the detector substrate).III. A NALYSIS PROCEDURE & E
NERGY SCALECALIBRATION
We have operated the TES array in single-pixel modeconfiguration, biasing one pixel at a time, while all the othersare left in the superconducting state. We have configured theMXS in order to have a count-rate of ∼ α at 5.9 keV);3 - We have preliminarily fitted all the main K α and K β lines in the spectrum (Figure 3). The intrinsic line shapeshave been described according to the ASTRO-H SXSCalibration Database for Line Fit [16]; Fig. 3. Preliminary fit of the main K α and K β lines (generated by MXS andinternal Fe X-ray source) acquired by Px1 biased at 1.1 MHz.
Fig. 4. (Left) Measured energy shift of the detected lines as a function ofthe energy, as obtained by the preliminary fit (data from Px1 biased at 1.1MHz). The blue dashed lines is a 3th order polynomial fit of the data. (Right)Residuals of the 3th order polynomial fit. The fit accuracy is better than 0.5eV. ∼ α lines in the spectrumto assess the detector energy resolution at the differentenergies. Before this final fit procedure, we performeda simple bremsstrahlung subtraction from the lines ( ∼ NERGY RESOLUTION
The best measured energy resolutions at the different ener-gies are shown in Figure 5. The spectra have been acquiredusing the Px1 (1.1 MHz bias frequency) in single-pixel read-out, and they show ∆ E FWHM = (2.40 ± ± ± α , Mn-K α and Cu-K α comparing 3 pixels at different bias frequencies: Px1 at 1.1MHz, Px8 at 2.7 MHz and Px16 at 4.6 MHz (i.e. low,mid and high bias frequency). All the measurements havebeen performed in single-pixel readout. The dashed black linerepresents the final X-IFU requirement at instrument level, andit is overplotted just for reference. Note that, for each pixel,the energy resolution degradation between 5.9 and 8.0 keV( < . eV) is lower than the one requested by the X-IFUrequirements ( ∼ . eV). Fig. 6. Energy resolution as a function of the MXS emission line energies for3 pixels at different bias frequencies: Px1 at 1.1 MHz, Px8 at 2.7 MHz andPx16 at 4.6 MHz. The dashed black line represents the X-IFU requirementat instrument level.
V. L OW - ENERGY RESPONSE & LINEARITY
To investigate the detector behavior also at low energy, weperformed long measurements ( ∼ α complex acquired in a 20 houracquisition with Px 8 (2.7 MHz bias frequency). The measuredenergy resolution is ∆ E FWHM = (1.91 ± α Fig. 5. Best fit to data (red lines) for each detected K α line (MXS and internal 55Fe X-ray source) for Px1 biased at @ 1.1 MHz. The lines intrinsic shape(blue lines) has been described according to [16]. Fig. 7. Best fit (red line) of the detected Al K α line for Px8 biased at @2.7 MHz. The line intrinsic shape (blue line) has been described according to[16]. (1.49 keV) and the Si-K α (1.74 keV) lines. Because of thelow Si-K α count rate ( ∼
80 counts in 20 hours), it has beennot possible to assess the detector performance also at thatenergy.Considering the whole energy range we have explored sofar (from 1.5 to 8.9 keV), we can estimate the linearity of thedetector. Figure 8 shows the optimal filtering output parameter(i.e. the normalized area of the optimal filtered pulses, whichwe use as optimal energy estimator) obtained for each detectedemission lines as a function of the real lines energy. Thedashed blue line is the linear trend calibrated on the detectedlow energy lines. We observe a deviation from the linearity of8% at 5.9 keV and 15% at 8.9 keV.
Fig. 8. Detector linearity in the range from 1.5 to 9 keV. The optimal filteringoutput is shown as a function of the detected line energies (red points). Thedashed blue line is the linear trend calibrated on the low energy lines.
VI. C
OUNT RATE EFFECT
Finally, we have performed a preliminary investigation ofthe energy resolution trend as a function of the count rateon the pixels. The emission rate of the MXS can be indeedincreased by increasing the current in the UV-LEDs. Figure 9shows the measured energy resolution of the Cr-K α line (5.4 Fig. 9. Energy resolution of the Cr-K α line (5.4 keV) as a function of thecount rate. Red points are average values over the three pixels biased at 1.1,2.7 and 4.6 MHz. The blue points indicate the fraction of good events used tofit the spectra. Consider that on the X-IFU 30 cps/pixel roughly correspondto 1 Crab flux. keV) as a function of the count rate. The red points areaveraged values over the three pixels biased at 1.1, 2.7 and4.6 MHz, and operated in single-pixel readout. The blue linerepresent the fraction of good events used to fit the spectra. Weperformed a very simple pulse selection, in which we rejectevents where multiple pulses are detected within our optimalrecord lenght (26 ms). We observe that for up to around10 counts per second no significant performance degradationis visible, and that the fraction of high resolution events ishigher than 80%. Just for reference, consider that one of theX-IFU requirements is to observe the average flux of thesupernova remnant CasA ( ∼ >
80% high resolution events [17].Note that here, in single-pixel readout, the primary mech-anism of resolution degradation is the induced thermal in-stability on the pixel baseline. To proper evaluate the effectof the thermal crosstalk due to the neighbor pixels, wehave performed also a dedicated measurements in multiplexedreadout. This activity is reported in another paper in this issue[11]. VII. C
ONCLUSION
We are developing a large Ti/Au TES array as a backuptechnology for ATHENA X-IFU. In this work, we havepresented the single-pixel performance at different X-rayenergies of a uniform 32x32 pixels prototype, which has beenoperated in the FDM readout system. The achieved energyscale calibration accuracy is usually better than ∼ ∆ E FWHM = (1.91 ± ± ± ± ∼
10 count/s.These results represent a promising first reference for thekilo-pixel array behavior in a large energy band, as wellas conclude an extensive array characterization includingperformance and characteristics uniformity [10] and thermalcrosstalk minimization [11]. A further optimization of the pixel design, already showing significant performance improvement[7], could lead in the near future to demonstrate a full arrayinside the X-IFU requirements.A
CKNOWLEDGMENT
This work is partly funded by European Space Agency(ESA) and coordinated with other European efforts underESA CTP contract ITT AO/1-7947/14/NL/BW. It has alsoreceived funding from the European Union’s Horizon 2020Programme under the AHEAD (Activities for the High-EnergyAstrophysics Domain) project with grant agreement number654215. R
EFERENCES[1] Pajot, F., Barret, D., Lam-Trong, T. et al. The Athena X-ray Inte-gral Field Unit (X-IFU). J Low Temp Phys 193, 901–907 (2018).https://doi.org/10.1007/s10909-018-1904-5[2] D. Barret and the Science Advisory Team, 2019, Science withthe Athena X-ray Integral Field Unit, X-IFU Preliminary Re-quirements Review, http://x-ifu.irap.omp.eu/wp-content/uploads/2019/01/XIFU-SN-XI-09012019-IRAP.pdf, Visited 27/11/2020[3] Miniussi, A.R., Adams, J.S., Bandler, S.R. et al. Performance of an X-ray Microcalorimeter with a 240 µ m Absorber and a 50 µµ