TiAu TES 32\times32 pixel array: uniformity, thermal crosstalk and performance at different X-ray energies
E. Taralli, M. D'Andrea, L. Gottardi, K. Nagayoshi, M. Ridder, S. Visser, M. de Wit, D. Vaccaro, H. Akamatsu, K. Ravensberg, R. Hoogeveen, M. Bruijn, J.R. Gao
aa r X i v : . [ a s t r o - ph . I M ] F e b JOURNAL OF L A TEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 1
TiAu TES 32 ×
32 pixel array: uniformity, thermalcrosstalk and performance at different X-rayenergies
Emanuele Taralli, Matteo D’Andrea, Luciano Gottardi, Kenishiro Nagayoshi, Marcel Ridder, Sven Visser,Martin de Wit, Davide Vaccaro, Hiroki Akamatsu, Kevin Ravensberg, Ruud Hoogeveen, Marcel Bruijnand Jian-Rong Gao
Abstract —Large format arrays of transition edge sensor (TES)are crucial for the next generation of X-ray space observatories.Such arrays are required to achieve an energy resolution of ∆ E < ×
32 pixelarray with (length × width) 140 × µ m TiAu TESs, which havea 2.3 µ m thick Au absorber for X-ray photons. The pixels have atypical normal resistance R n = 121 m Ω and a critical temperature T c ∼
90 mK. We performed extensive measurements on 60 pixelsout of the array in order to show the uniformity of the array. Weobtained an energy resolutions between 2.4 and 2.6 eV (FWHM)at 5.9 keV, measured in a single-pixel mode at AC bias frequenciesranging from 1 to 5 MHz, with a frequency domain multiplexing(FDM) readout system, which is developed at SRON/VTT. Wealso present the detector energy resolution at X-ray with differentphoton energies generated by a modulated external X-ray sourcefrom 1.45 keV up to 8.9 keV. Multiplexing readout across severalpixels has also been performed to evaluate the impact of thethermal crosstalk to the instrument’s energy resolution budgetrequirement. This value results in a derived requirement, for thefirst neighbour, that is less than 1 × − when considering theratio between the amplitude of the crosstalk signal to an X-raypulse (for example at 5.9 keV). Index Terms —Crosstalk, Energy resolution, Modulated X-raysource, Superconducting devices, X-ray detectors.
I. I
NTRODUCTION A THENA [1] is the second ‘Large mission’ of ESA’sCosmic Vision-programme to study astrophysical phe-nomena near black holes and galaxy clusters. One of the twoscientific instruments on board is the X-ray Integral Field Unit(X-IFU) [2], a detector consisting of an array of over 3000Transition Edge Sensor (TES) calorimeters, sensitive in 0.2-12 keV energy range, with 2.5 eV energy resolution below7 keV. SRON is currently developing X-ray transition edgesensor (TES) microcalorimeter array as a backup technology
E. Taralli, L. Gottardi, K. Nagayoshi, M. Ridder, S. Visser, M. de Wit, D.Vaccaro, H. Akamatsu, K. Ravensberg, R. Hoogeveen, M. Bruijn and J.R.Gao are with the Netherlands Institute for Space Research NWO-I/SRON,Utrecht, 3584 CA, The Nederland e-mail: (see [email protected]).J.R. Gao are also with Faculty of Applied Science, Delft University ofTechnology, 2600 AA Delft, The Netherlands.M. D’andrea is with INAF/IAPS Roma, Via del Fosso del Cavaliere 100,00133 Roma, ItalyManuscript received April 19, 2005; revised August 26, 2015. for X-IFU.During the recent past, we have been focused in the designoptimisation of our TiAu bilayer TESs. Number of mixedand uniform arrays, with different pixel designs, have beensuccessfully fabricated [3], [4]. Square pixels [3], [5] and highaspect ratio devices [6], [7] have been fully characterised usingour frequency domain multiplexing (FDM) readout system anda preliminary comparison of the performance, between samedevices biased under AC and DC, have been also investigated[8]. Lately, uniformity of a 32 ×
32 pixels array have beenextensively studied and the results have been widely reported[9].However, besides the homogeneity, many other aspects play acrucial role in fabrication and in the delivery of large arrays.Proper thermalisation process, to avoid undesired phononsdiffusing through the substrate after an X-ray has been de-tected, can have a major impact in the final evaluation of theinstrument’s energy resolution budget. Moreover the detectorenergy resolution needs to remain excellent for a large rangeof the energy spectrum.In this work, we first resume the main results in terms ofuniformity of critical temperature and energy resolution at5.9 keV among the 60 pixels measured over the uniform32 ×
32 array. Afterwards we present thermal crosstalk ratiofor four closest neighbours and eventually we conclude withthe evaluation of the energy resolution at X-ray with differentphoton energy using a modulated X-ray source (MXS).II. E
XPERIMENTAL SETUP
We selected a total amount of 60 TESs in order to investi-gate the performance from one side to the other of a uniform32 ×
32 pixels array and we measured them in 4 differentmeasurement runs as highlighted by colours in Fig. 1a. Alldevices have the same aspect ratio (length × width) of 140 × µ m (Fig. 1c) and the same bilayer thickness of Ti (35nm) and Au (200 nm), which have been selected to show asheet resistance of 25 m Ω / (cid:3) with superconducting transitiontemperature T c ∼
90 mK. All absorbers have also the samesize (240 × µ m ) and the same thickness (2.35 µ m of Au).Each absorber is connected to the corners of the membrane bymeans of 4 supporting stems and the other two are thermallycoupled to the centre of the TES, as shown in the schematicof a single pixel in Fig. 1b. A 0.5 µ m thick SiN membrane OURNAL OF L A TEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 2
Fig. 1: (a) Top view of 32 ×
32 pixels array with the pixelsmeasured during the 4 measurement runs: purple (Run1), green(Run2), blue (Run3) and red (Run4). (b) Schematic of thesingle pixel (not to scale). Supporting stems hold the absorberabove the TES which lays on the SiN membrane supported bythe Si muntins coated with Cu. (c) Picture of a 140 × µ m TiAu TES .supports each pixel. They are located on a grid of 300 µ mthick and 80 µ m width Si muntins shaped by back-etchingunder each TES to ensure a weak thermal link to the bath.The final pitch size is 250 µ m. Gold wire-bonds have beenused to thermalize the whole chip to the bath connecting onlythe silicon area on the top part of the chip. To increase thethermal conductivity between the chip and the bath, 1 µ mthick coating of Cu with Au capping layer has been addedonly to the sides and bottom of the muntins in the Si gridregion of the pixels, as shown in Fig. 1. More details on thefabrication of this TES array can be found in [4].Frequency Domain Multiplexing (FDM) [10], [11] is usedto read out these detectors. Each detector is separated infrequency by placing it in series with LC resonators of specificfrequencies. FDM readout applies a set of sinusoidal ACcarriers in order to bias the TES detectors in their bias points.When a TES detector is hit by an X-ray photon, its own carrieris amplitude modulated. The frequency band assigned to eachdetector is selected in order to prevent the detectors frominteracting with each other. In this way we are able to readoutmultiple TES pixels by means of one amplifier channel, whichuses only one set of Superconducting QUantum InterferenceDevices (SQUID) current sensors. For this work we have usedan 18-channels FDM readout system with a yield of ∼ HARACTERIZATION
A. T c and ∆ E at 5.9 keV
Critical temperature T c and spectral energy resolution ∆ E are generally correlated by the following expression ∆ E = p kT C ∝ T / where k is the Boltzmann constant and C isthe detector heat capacity. Low T c would certainly mean highresolving power and equally, uniform T c over the whole arrayrather means uniform performance in detecting X-ray photon.We have determined the critical temperature T c over the arrayby measuring the current-voltage characteristics ( IV curves) ofthe pixels under test for different bath temperatures T bath (from50 mK up to 90 mK). Using these curves we calculate thedissipated electrical power P TES , for example at the minimumof the TES’s IV along the phase transition, at every bathtemperature. Balancing the electrical and the thermal powerdissipated by the TES, we are able to get and summariseall the T c in the top histogram of Fig. 2a. The two peaksrepresent the critical temperatures associated to the north andsouth quadrant of the array, respectively. It is worth notingthat all the pixels belonging to the same quadrant show adifference in the transition temperature less than 0.7 mK,while the total gradient from one side to the other of thearray is less than 1.5 mK. We can consider the averaged T c = . ± Fe source, providing Mn-K α X-rays at anenergy of 5.9 keV with a count rate of approximately 1 countper second (cps) per pixel. More information about the setupare reported in this recent work [9]. Typically, we collect about5000 X-ray events for each spectrum to get a statistical error ofabout 0.15 - 0.18 eV for the reported energy resolution. Fig. 2breports the histogram of all the energy resolutions measuredfor all the pixels under test in their best bias point, where theaveraged energy resolution is 2.49 ± ∆ E FWHM between 2.4 and 2.6 eV in 1.5mK variation of the critical temperature.
B. Thermal crosstalk
Accurate thermalisation of large format and high-densityarrays of microcalorimeters plays a crucial role in the min-imisation of the thermal crosstalk between nearby pixels [12].Facing this problem turns out to be crucial in order to meet theenergy resolution requirements for the specific project [13].For instance, Athena has allocated 0.2 eV inside the instru-ment’s energy resolution budget for the impact of the thermalcrosstalk. It derives that the ratio between the amplitude ofthe crosstalk signal to an X-ray pulse (for example at 5.9keV) is less than 1 × − (for the first neighbour), less than4 × − (for the diagonal neighbour) and less than 8 × − (for the second nearest neighbour) [13]. The main goal of suchcharacterisation is to record any thermal pulse detected fromneighbour pixels (named as victims) when a pixel (named asperpetrator) detects an X-ray photon. OURNAL OF L A TEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 3
Fig. 2: (a) Histogram of the T c measured. Two peaks cor-respond to the different T c associated to the north and southquadrant of the chip, respectively. (b) Histogram of the energyresolution at 5.9 keV of the pixels measured in their best biaspoint. The smoothed lines in both the histograms serve noother purpose than to guide the eye. (c) Combined histogramwhere the contours define the most populated areas wheredetectors have similar T c and ∆ E .In Run 4 (red pixels in Fig. 1a), we have performed thermalcrosstalk measurement considering two perpetrators, e.g., pixel4 and pixel 1 (from now on named as Px4 and Px1), placedin the middle of the 12 pixels matrix. To minimise electricalcrosstalk issues due to carrier leakage, all the detectors havebeen connected to LC resonators with bias frequency separatedeach other by at least 200 kHz. To further reduce this effect, weperformed the measurement reading out only two pixels (oneperpetrator and one victim) at the time. Fig. 3 shows thermalpulses recorded by the victims pixels in linear and logarithmicscale respectively, when X-ray is detected for instance fromPx4. We usually collect around 5000 X-ray counts from theperpetrator in order to get an averaged X-ray pulse. Eachof these counts triggers the recording of the trace of thevictim pixel. Neglecting the traces corresponding to pileupevent detected by the perpetrator, we are able to define anaveraged crosstalk pulse for the selected victim. Fig. 3 reportsthe averaged crosstalk pulses for all the victims that have beenconsidered. It also shows some pixels with a considerablenegative spike just after the trigger induced by the X-raydetection. This is imputed to electrical crosstalk caused bythe presence of common inductance both at the SQUID inputand in the bias circuit and by mutual inductance in the LC-filter chip. This causes a scattering in the starting point of thethermal pulses. As a result, the electrical crosstalk effectively Fig. 3: Thermal pulses detected by victim pixels when X-rayphoton (line) is absorbed by perpetrator pixel 4 in linear (a)and logarithmic scale (b). 1 st neigh. (dashed line), 1.41 neigh(dash-dot line), 2 nd neigh. (dotted line) and 2.24 neigh. (dash-dot-dotted line) listed in the legend are first, first diagonal,second and second diagonal neighbour of Px4, respectively.lower the peak amplitude of the thermal pulse adding an errorin its evaluation.In order to take into account the impact of this electricalcrosstalk in the estimation of the thermal crosstalk, the analysisconsiders two types of fit: 1) a double exponentials fit functionand 2) a single exponential fit function. The first one is usedto extrapolate the peak amplitude of the pulse (dashed blackline in Fig. 4). We have already said that this value is under-estimated because the electrical crosstalk at the beginning ofthe pulse, as a matter of fact, reduces the actual peak of thethermal pulse. This provides a lower limit in the estimationof the pulse amplitude (orange dotted line in Fig. 4). Sincewe do not have a reliable information on the thermal crosstalkpulse in its initial phase (as said, its amplitude is lowered),we use the second fit to extrapolate the peak value at thebeginning of the thermal pulse considering the intersectionof the fit function with the positive slope of the X-ray pulse(dash-dot green line in Fig. 4). This provides an upper limitin the estimation of the pulse amplitude (green dotted line inFig. 4).To assess the thermal crosstalk we finally consider the ratiobetween the amplitude of the thermal crosstalk signal from the OURNAL OF L A TEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 4
Fig. 4: Thermal pulse detected by the first neighbour whenX-ray photon is absorbed by Px4. Black dashed and greendot-dashed lines define the double and single exponentialfit functions respectively. Dotted lines define the three peakamplitudes used to evaluate the lower and upper limit of thethermal crosstalk.victim to the X-ray pulse from the perpetrator, crosstalk = h victim / h . In Fig. 5 we plot the thermal crosstalk ratiomeasured for the perpetrator Px4 (Fig. 5a) and Px1 (Fig. 5b)as a function of the neighbour pixels, respectively. The insetof Fig. 5a shows the layout of the victim pixels specifically forPx4 as perpetrator: first nearest neighbour pixel (distance 1),first diagonal neighbour pixel (distance 1.41), second nearestneighbour pixel (distance 2) and second diagonal neighbourpixel (distance 2.24). We have included open and closeddots as for the lower and upper limit of the crosstalk ratio,respectively. In this manner, we consider the impact of theelectrical crosstalk defining a fair region where we expect tofind the ultimate thermal crosstalk value. It is worth notingthat we are well below the X-IFU requirement (red point inFig. 5) for the first neighbour even considering the upper limit.Moreover, for reasons that still needs to be understood, thefirst diagonal neighbour (distance 1.41), in both the sequencesconsidered, seems to be much more effected by the electricalcrosstalk as indicated by the error bar reflecting a largerscattering in the peak amplitude estimation. C. ∆ E at different X-ray energy photons
Large array of X-ray microcalorimers require to deliverhigh energy resolution over a large energy range in order tomeet the instrument specification and consequently the sciencegoal linked to the mission. Fig. 6 shows the two spectra,detected by a single TES, used in this work to characterisethe performance of the array in terms of energy resolution.As already mentioned, the two Mn-k α and Mn-k β (greendashed line in Fig. 6) are obtained with an internal Fesource, while the k α and k β of Cr and Cu (black line inFig. 6) are obtained with an external modulated X-ray source(MXS) with a count rate of ∼ α and Si-k α ) can be observed in the lower part of the spectra.They become visible only after having performed very longX-ray acquisition (more than 24 h). These lines are relatedto photons generated by the interaction between the X-rayphotons coming from our internal Fe source with the Al partsof the setup (superconducting shield and foils to reduce thecount rate) and the Si substrate of the TESs, respectively. Moreinformation about this characterization and the data analysisare provided in another paper which will be published in thisspecial issue [14].In this section we would like to give an overview of theenergy resolution of the array as a function of the measuredenergy. We characterised three pixels Px1, Px8 and Px16 atdifferent bias frequency 1, 2.7 and 4.6 MHz, respectivelyand the corresponding averaged energy resolution at fourdifferent energy lines (Al-k α , Mn-k α , Cr-k α and Cu-k α ) arereported in Fig. 7. We would like to mention that the energyresolution can be properly evaluated only for the k α lines,the natural line shapes of which are well described [15] andhave a decent number of counts as well. For this reason Si-k α , Mn-k β , Cr-k β and Cu-k β have been not included in theplot. Notwithstanding a degradation of ∼
4% in the energyresolution between low and high bias frequency pixels [9],detector performance is very promising in the low (1.45 keV)and in the high (8 keV) part of the energy spectrum comperedto the X-IFU requirement (blue dashed line). Improvementsat 5.4 and 5.9 keV are foreseen with the new pixels designcurrently under test.IV. C
ONCLUSION AND NEXT STEPS
We are developing large uniform array of TiAu transitionedge sensor microcalorimeters as backup option for X-IFUinstrument on board of the Athena space mission. In this workwe have presented the uniformity and the performance of auniform 32 ×
32 pixels array, each of which is an TiAu TESwith dimension (length × width) 140 × µ m with a normalresistance R n = 121 m Ω . We have shown an averaged criticaltemperature T c = 89.5 ± ∆ T c <1.5 mK. The largemajority of the pixels have a ∆ E FWHM between 2.4 and 2.6eV providing an averaged ∆ E FWHM = 2.49 ± × − for the first neighbour asrequired by Athena/X-IFU. The next step of this work is toperform a template for the electrical crosstalk pulse in order OURNAL OF L A TEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 5
Fig. 5: Lower (open dots) and upper (closed dots) limit for the thermal crosstalk assessment for pixel 4 (Px4) (a) and pixel 1(Px1) (b). As an example, the inset in (a) shows the layout of the victim pixels around Px4. Grey area shows the region wherewe expect to find the ultimate thermal crosstalk of our kilo-pixel array. Red points refer to the X-IFU requirement.Fig. 6: Spectra of the two sources used in this work tocharacterise the performance of this array: internal Fe source(green dashed line) and modulated X-ray source (black line).The main energy lines are named in the graph. Fig. 7: Three pixels average energy resolution as a functionof energy lines Al-k α , Mn-k α , Cr-k α and Cu-k α (black dots)compared to the X-IFU requirement as reference (blue dashedline). Error bars refer to the standard deviation of the meanenergy resolution obtained from three pixels. OURNAL OF L A TEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 6 to remove its repercussion on the thermal pulse.Energy resolution at different energy lines have been alsoevaluated finding an ∆ E FWHM less than 3 eV for X-ray energiesup to 8 keV with a best ∆ E FWHM less then 2 eV at 1.45 keV. Anext step is to characterize the large uniform arrays based onthe new pixel design, currently under test, showing promisingimprovement at 5.9 keV in order to demonstrate a full arraywhich meet the X-IFU requirement.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] F. Pajot, D. Barret, T. Lam-Trong et al. , ‘The Athena X-rayIntegral Field Unit (X-IFU)’,
J. Low Temp. Phys , (2018).https://doi.org/10.1007/s10909-018-1904-5;[2] European Space Agency (ESA), ‘ATHENA Mission Summary’, Available:http://sci.esa.int/athena/59896-mission-summary.[3] P. Khosropanah, E. Taralli, L. Gottardi, et al , ‘Development of TiAu TESx-ray calorimeters for the X-IFU on ATHENA space observatory’, Proc.SPIE 10699, Space Telescopes and Instrumentation 2018: Ultraviolet toGamma Ray, 106991M (6 July 2018); doi: 10.1117/12.2313439.[4] K. Nagayoshi, M. L. Ridder, M. P. Bruijn et al. , ‘Development ofa Ti/Au TES Microcalorimeter Array as a Backup Sensor for theAthena/X-IFU Instrument’,
J. Low Temp. Phys. et al. , ‘Complex impedance ofTESs under AC bias using FDM readout system’,
AIP Advances et al. , ‘Characterization of HighAspect-Ratio TiAu TES X-ray Microcalorimeter Array Under AC Bias’,
J. Low Temp. Phys. , 199, pp 80-87 (2020). https://doi.org/10.1007/s10909-019-02254-y.[7] M. de Wit, L. Gottardi, E. Taralli et al. , ‘High Aspect Ratio TransitionEdge Sensors for X-ray Spectrometry’,
J. Appl. Phys. et al. , ‘AC/DC Characterizationof a Ti/Au TES with Au/Bi Absorber for X-ray Detection’,
J. LowTemp. Phys. , 199, pp 102-109 (2020). https://doi.org/10.1007/s10909-020-02390-w.[9] E. Taralli, M. D’Andrea, L. Gottardi et al. , ‘Performance anduniformity of a kilo-pixel array of TiAu transition-edge sen-sor microcalorimeters’,
Rev. Sci. Instrum. , 92, pp 023101 (2021);https://doi.org/10.1063/5.0027750.[10] H. Akamatsu, L. Gottardi, J. van der Kuur et al. , ‘Progress in theDevelopment of Frequency-Domain Multiplexing for the X-ray IntegralField Unit on Board the Athena Mission’,
J. Low Temp. Phys. , 199, pp737-744 (2020). https://doi.org/10.1007/s10909-020-02351-3.[11] D. Vaccaro, H. Akamatsu, J. van der Kuur et al. , ‘Application of aFrequency Shift Algorithm to a Frequency-Domain Multiplexing Readoutfor X-ray TES Microcalorimeters’,
Rev. Sci. Instrum. , 92, pp (2021).[12] A. R. Miniussi, J. S. Adams, S. R. Bandler et al. , ‘Ther-mal Crosstalk Measurements and Simulations for an X-ray Mi-crocalorimeter Array’,
J. Low Temp. Phys. , 199, pp 663-671 (2020).https://doi.org/10.1007/s10909-019-02312-5.[13] R. den Hartog, P. Peille, T. Dauser et al. , ‘The impact of crosstalk inthe X-IFU instrument on Athena science cases’, Proc. SPIE 9905, SpaceTelescopes and Instrumentation 2016: Ultraviolet to Gamma Ray, Proc.,99055T (26 July 2018); doi: 10.1117/12.2232098.[14] M. D’Andrea, E. Taralli, L. Gottardi et al. , ‘Single pixel performanceof a 32x32 Ti/Au TES array with broadband X-ray spectra’,