Charge collection efficiency in back-illuminated Charge-Coupled Devices
Guillermo Fernandez-Moroni, Kevin Andersson, Ana Botti, Juan Estrada, Dario Rodrigues, Javier Tiffenberg
FFERMILAB-PUB-20-278-E
Charge collection efficiency in back-illuminatedCharge-Coupled Devices.
Guillermo Fernandez-Moroni a , Kevin Andersson a , b , Ana Botti b , Juan Estrada a , DarioRodrigues a , b and Javier Tiffenberg. a a Fermi National Accelerator Laboratory, PO Box 500, Batavia IL, 60510 b Department of Physics, FCEN, University of Buenos Aires and IFIBA, CONICET, Buenos Aires,ArgentinaE-mail: [email protected]
Abstract: Low noise CCDs fully-depleted up to 675 micrometers have been identified as a uniquetool for Dark Matter searches and low energy neutrino physics. The charge collection efficiency(CCE) for these detectors is a critical parameter for the performance of future experiments. Wepresent here a new technique to characterize CCE in back-illuminated CCDs based on soft X-rays.This technique is used to characterize two different detector designs. The results demonstrate theimportance of the backside processing for detection near threshold, showing that a recombinationlayer of a few microns significantly distorts the low energy spectrum. The studies demonstrate thatthe region of partial charge collection can be reduced to less than 1 µ m thickness with adequatebackside processing.Keywords: Back-illuminated CCD, Backside processed CCD, CCE a r X i v : . [ phy s i c s . i n s - d e t ] J u l ontents Fe source 4
Charged Coupled Devices (CCD) with low readout noise and large active volume have beenidentified among the most promising detector technologies for the low mass direct dark mattersearch experiments, probing electron and nuclear recoils from sub-GeV DM [1–5]. The recentdevelopment of the Skipper-CCD [6, 7] demonstrated the ability to measure ionization eventswith sub-electron noise extending the reach of this technology to unprecedented low energies.Experiments based on this technology are planned for the coming years with total CCD active massgoing from 100 grams to several kilograms [8, 9]. At the same time the low noise CCD technologyhas been implemented in low energy neutrino experiments [10, 11] and are planned for futuredevelopments[12].There are several key performance parameters for the CCD sensors in future developments thatare part of a significant R&D effort for future projects [8, 9, 12]. The most important performancerequirements are the pixel dark current [7], readout noise optimization [13], Fano factor [14] andcharge transport in the sensor [15].The Charge Collection Efficiency (CCE) is defined as the fraction of the total charge producedduring a ionization event that is collected in the CCD pixel for later readout. For a fully depleteddetector, with a large electric field, CCE is approximately 100% [16] for the full active volume.In regions of the detector with lower electric field, CCE could be less than 100% due to chargerecombination. Regions of partial CCE distort the measured spectrum of ionization events, affectingenergy calibration and particle identification.Back illuminated CCDs in astronomy are treated to have a thin entrance window for light, withlow reflectivity. This is specially important when detectors are used for wavelength shorter than500 nm [17–19]. A 500 nm photon has an absorption length of 1 µ m in Silicon, and any layer withpartial charge collection (PCC) on the back surface will degrade the detection efficiency for bluelight. The measurements presented in Ref.[20] compare the detection efficiency for visible photonswith the reflectivity. These studies show that all photons with wavelength longer than 500 nm are– 1 –ully detected, unless they are reflected on the back surface. These results show that the bulk of thedetector has 100% CCE, and that any recombination on these sensors occurs only on the first 1 µ mnear the back surface.For thick CCDs, as those used in dark matter [1–7] and neutrino experiments [10, 11], abackside ohmic contact is required in order to apply the needed substrate bias to fully deplete sensor[21]. At the same time, different processing techniques are used on the backside to reduce darkcurrent. The backside processing of these sensors determines the field shaping near the surface, andhas a large impact in the CCE for events in that region. We study here the CCE for back-illuminateddetectors with more than 200 µ m thickness. Ɛ ( Z ) ɵ ++ ++++++ q f (= q i ) ++ ++++++ q i +++ q f (< q i ) μ mSiO [0, ?]Sideadlayer Partial Charge Collection CCD bulk(full charge collection)~670 μ mSi0.1 +0.04+0.02 μ mSiO +ZrO +In [0, ?]Si ~200 μ mSiUnthinnedCCD-AThinnedCCD-B Figure 1 . Sketch of the CCD back illumination with an X-ray source. The photon penetrates into the CCDproducing a cloud of charge q i , some fraction ε ( Z ) of this charge gets collected depending on the depth Z .The region near the back of the CCD where 0< ε ( Z ) <1 is the PCC layer. X-rays can be used to characterize the CCE near the back surface of a CCD. Figure 1 showsa cartoon of X-ray setup together with the most important variables that participate in the analysis.Some important aspects and definitions – 2 – The source emits photons with uniform angular distribution covering a full hemisphere. Theangular distribution on the sensor depends on the geometry of the setup. We model theangular distribution by f Θ ( θ ) , where Θ is measured as the angle of the incidence of thephoton in the CCD compared to perpendicular direction to the back surface of the sensor.• The X-ray photons can reach the PCC layer and the bulk of the sensor volume. The interactiondepth Z in the sensor depends on the incident angle and its probability distribution function( pdf ) can be written as f Z ( z | θ ) = ( cos ( θ )/ λ ) exp (− z cos ( θ )/ λ ) , where λ is the attenuationlength of the photon.• X-rays produce an ionization charge packet with mean value q i = E i / (cid:15) , where E i is theenergy of the photon and (cid:15) = 3.75 eV is the mean ionizing energy [14]. For now, we assumethat the initial charge packet is the same for all photoelectric absorption events, we discusslater how the Fano noise affects the final results. The primary charge ionization is the samefor the PCC layer and the bulk of the sensor as represented in Fig. 1.• ε ( z ) is the CCE function in the backside of the detector. The function indicates the fractionof carriers that are collected by the pixel after drifting away from the PCC layer (carriers thatdo not recombine in the PCC layer). This function depends on the depth of the interaction.If the primary charge packet occurs deep in the PCC layer (far from the bulk of the sensor),carrier will have more time to recombine before they reach the bulk. Thus, ε ( z ) increasesmonotonically.• q f is the charge that escapes from the PCC layer and can be collected and measured by thesensor. As illustrated in Fig. 1, this will depend on the interaction depth of the photon. Wewill refer as Q f to the random variable accounting for the possible values of the X-ray eventswith pdf f Q f ( q f ) . The distribution of Q f is the observable in our data.From the previous definitions the measured charge can be expressed as Q f = q i ε ( Z ) . (2.1) The measured spectrum of events normalized by the total number of events ( N T ) is an estimationˆ f Q f ( q f ) of f Q f ( q f ) . We can then use it to estimate the cumulative distribution function ( cdf ) of Q f : P ( Q f ≤ q f ) ≈ ˆ F Q f ( q f ) = ∫ q f ˆ f Q f ( x ) dx . (2.2)Using Eq.(2.1) and due to the monotonically increasing ε ( Z ) , P ( Q f ≤ q f ) = P ( q i ε ( Z ) ≤ q f ) = F Z ( z ) = ∫ z ˆ f Z ( z ) dz (2.3)where z is such that ε ( z ) = q f / q i , and ˆ F Q f ( q f ) = F Z ( z ) . (2.4)– 3 – able 1 . Fe X-rays energies, Intensity in photons per 100 disintegrations and attenuation length in µ m[22]. Mean e-h pairs production using the mean ionization energy . X K Energy (keV) Mean e-h production ( q i ) Intensity Attenuation length ( λ α ) α α β cdf integrating away from low charge values,ˆ F ← Q f ( q f ) = ∫ q i q f ˆ f Q f ( x ) dx , and F ← Z ( z ) = ∫ ∞ z ˆ f Z ( z ) dz . (2.5)For each q f , we find z such that ˆ F ← Q f ( q f ) = F ← Z ( z ) where the efficiency is ε ( z ) = q f / q i .The method to calculate the CCE using one X-ray peak is summarized in the Table 2 of theAppendix. Fe source Fe X-ray source has an extensive use in the calibration of typical performance parameters of CCDsand other sensors [16]. In this article we extend its use to characterize the charge collection in thePCC layer using the methodology proposed in Section 2.1. The main characteristics of the threeX-rays emitted by Fe are summarized in Table 1. K α X-rays have similar energy and attenuationlength and therefore can be treated as a single X-ray line for the purpose of this analysis.Then, pdf for the interaction as a function of depth are f Z α ( z | θ ) = ( cos ( θ )/ λ α ) exp (− z cos ( θ )/ λ α ) f Θ ( θ ) and f Z β ( z | θ ) = ( cos ( θ )/ λ β ) exp (− z cos ( θ )/ λ β ) f Θ ( θ ) for the X K α and X K β , respectively. With the same angular distribution in both cases.Generalizing Eq. (2.2) and (2.4) for two X-ray energies α and β , the measured cdf for Q f isˆ F Q f ( q f ) = P ( Q f ≤ q f ) = p α P ( Z α ≤ z α, ) + p β P ( Z β ≤ z β, ) , (2.6)where ε ( z α, ) = q f / q i ,α and ε ( z β, ) = q f / q i ,β , such that the depth cdf equals the measuredcumulative distribution of events. p α and p β are the relative intensities determined by Table 1normalized by the number of desintegrations. Since we assume a monotonically increasing ε ( z ) function, then z α, ≥ z β, . Using a more condense notationˆ F Q f ( q f ) = p α F Z α ( z α, ) + p β F Z β ( z β, ) with z α, ≥ z β, . (2.7)A recursive nonlinear numeric solver is used to find z α, and z β, simultaneously. Three featuresof the Fe source can be used to simplify the problem.• Larger X K α -flux than X K β -flux, since p α / p β = .
47– 4 – F Z α ( z α, ) is always greater than F Z β ( z β, ) because of the difference in the attenuation length( λ α < λ β ) and the fact that z α, ≥ z β, .• As q f becomes smaller than q i ,α then q f / q i ,α becomes closer to q f / q i ,β , and therefore z α, becomes closer to z β, . In fact, q i ,β and q i ,α differ only 10%.Most of the signal is dominated by the X K α photon and a small effect is introduced by assuminga unique z = z α, = z β, . This assumption allows to follow the same procedure presented in Section2.1 to solve equation 2.7. Assuming z = z α, = z β, the true collection efficiency at z lays between q f / q i ,β and q f / q i ,α . A simple approximation is ε ( z ) = q f /( p α q i ,α + p β q i ,β ) . The full method foran Fe source is summarized in Table 3, in the Appendix.
We study here two different CCDs.CCD-A was designed by the LBNL Microsystems Laboratory [23] as part of the R&D effortfor low energy neutrino experiments [10] and low mass direct dark matter search [3]. This is arectangular CCD with 8 million square pixels of 15 µ m × µ m each. The CCD is fabricated inn-type substrate with a full thickness of 675 µ m. The resistivity of the substrate greater than 10000 Ω -cm. The CCD is operated with 40V bias voltage that fully depletes the high-resistivity substrateusing the method developed in Ref.[21]. In order to trap impurities that migrate during the sensorprocessing, a 1 µ m thick in-situ doped polysilicon (ISDP) layer is deposited on the backside of thedetector. This layer plays a critical role controlling the dark current of the detector. Additional layersof silicon nitride, phosphorous-doped polysilicon and silicon dioxide are added to the backside ( 2 µ m total thickness). Phosphorous can migrate into the high resistivity material producing a regionof a few microns where charge can recombine before drifting to the collecting gates of the detector.This region constitutes the PCC layer that we characterize with Fe X-rays, as shown in Figure 1.CCD-B is similar to CCD-A with a few important differences. The detector has 4 million pixels,with a thickness of 200 µ m. It is also fabricated in high resistivity n-type silicon. The backsideof the sensor has been processed for astronomical imaging. A backside ohmic contact is formedby low-pressure, chemical-vapor deposition in-situ doped polycrystalline silicon (ISDP). This layeris made thin for good blue response, typically 10-20 nm, and is robust to over-depleted operationthat is necessary to guarantee full depletion across the entire CCD. This detector is operated at biasvoltage of 40 V. Because of its backside treatment, this detector is not expected to have significantcharge recombination near the back surface. The detector is exposed to Fe X-rays on the backside,as shown in Figure 1.The Fe was located 3.55 cm away from the CCDs. The effective depth distribution ofinteracting photons was calculated using a Monte Carlo simulation, and the result is f Z ( z ) = I α ∗ exp (− z / τ α ) + I β ∗ exp (− z / τ β ) , (3.1)where I α ( I β ) represents the intensity and τ α ( τ β ) is the effective optical depth for the α ( β ) spectralline. I α = . I β = . τ α = . µ m, and τ β = . µ m.– 5 – .1 Results for CCD-A The spectrum of measured charge for CCD-A is shown in the top panel of Fig. 2, and comparedwith a Geant4[24] simulation assuming perfect CCE for the entire volume of the sensor ( ε ( z ) = α and K β peaks from Table 1 are evident. The excess of reconstructed events to the left ofthese peaks is attributed to the PCC layer, where charge recombination produces a measurementbelow the peak energy. The bump in the simulation around 1100 e − is an escape peak, as discussedin Ref.[25]. This data is used to measure the CCE function ε ( z ) following the prescription inSection 2.2, and the results are shown in the top panel of Fig.3. The depth scale is chosen such that ε ( z = ) = .
9. The shaded region corresponds to the energies between 5.4 keV and 7 keV wherethe events from K α and K β are dominant and systematic uncertainties are expected to be important.In this region the precise shape of ε curve is less reliable.
200 400 600 800 1000 1200 1400 1600 1800 (e-) f q - - - - - e v en t s pe c t r u m datasimulation
200 400 600 800 1000 1200 1400 1600 1800 (e-) f q - - - - - e v en t s pe c t r u m datasimulation Figure 2 . Event spectra for CCD-A (left) and CCD-B (right) calculated using bin size of 70 eV normalizedby the number of measured events in the K α peak. Blue: measured spectra. Magenta: Simulated spectra ofevents from Geant4. Left figure: spectra for CCD-A; 35195 events in the histogram; 26697 events in the K α peak. Right figure: event spectra for CCD-B; 5452 events in the histogram; 4482 events in the K α peak. Thedashed black line indicates the expected level of events if the partial charge collection layer on CCD-B wassame as the one measured on CCD-A. The spectrum of measured charge for CCD-B is shown in the bottom panel of Fig.2, and comparedwith a Geant4 [24] simulation with perfect CCE. As for CCD-A, the K α and K β spectral lines areevident, CCD-B has a different output stage producing higher resolution peaks [6]. The relative rateof events on the left of the peaks, are well below the rate observed for CCD-A and consistent withthe simulation. These events are produced mostly by low probability Compton scattering of X-rays.The CCE function ε ( z α ) is determined as discussed in Section 2.2 and the results are shown inFig.3 bottom panel, black circles. The measurement of ε ( z α ) is also performed after the predictedCompton spectrum is subtracted based on the simulation, the results are shown in bottom panel ofFig.3, magenta circles. As before, the horizontal axis is selected such that ε ( z = ) = . The results of CCD-A and CCD-B showed in Fig.3 demonstrate the large impact that the backsideprocessing could have in the CCE for back-illuminated detectors. When a layer of a few microns with– 6 – - - - - - - - - m depth (00.10.20.30.40.50.60.70.80.91 ( z ) e Figure 3 . Measured charge collection efficiency for a CCD-A (solid square markers) and CCD-B (opencircle markers). The black points show the results without considering the background events predicted bythe simulation. The magenta point shows the results after correcting the experimental spectra by subtractingthe events form the simulations. The shaded area indicates the region where the detailed shape of the X-raypeaks affect the measurement, introducing more uncertainty. charge recombination is present on the CCD, the spectrum for low energy X-rays gets significantlydistorted. The charge recombination generates a significant number of lower energy events in thespectrum. The backside processing performed in detectors optimized for astronomical instrumentseliminates this issue for the most part, as shown with CCD-B. The generation of low energy eventsconstitute a major concern for experiments looking for rare signals near the detector threshold [1–7, 10, 11].The results obtained here for CCD-B, optimized for astronomical imaging, are consistent withthe observations of detection efficiency and reflectivity in Ref.[20].A new technique was introduced here to characterize the CCE for back-illuminated CCDs, thistechnique can easily be generalized to other semiconductor detectors. The technique uses toolsthat are commonly available at the detector characterization laboratories. As shown here, the newmethod is capable of measuring a PCC layer of a few micrometers. The sensitivity to a very thinPCC layer is limited by the energy of the Fe X-rays, and the technique could be easily extended formuch thinner recombination layers using lower energy X-rays. This technique will be a powerfultool in the optimization of detectors for the next generation of low threshold experiments lookingfor rare events such as dark matter, or coherent neutrino nucleus scattering[9, 12].– 7 – ) Calculate angular distribution of incident photons:
Based on the geometry of the experiment evaluate f Θ ( θ ) .
2) Calculate depth distribution of events: f Z ( z | θ ) = ( cos ( θ )/ λ ) exp (− z cos ( θ )/ λ ) f Θ ( θ ) , where λ is the attenuation length of the photon.Then, calculate the cdf F Z ( z ) (or F ← Z ( z ) from Eq. (2.5)).
3) Make a spectrum of measured events:
Calculate the spectrum of events reconstructed from the data and normalize it by total numberof events ( N T ). This is the estimation ˆ f Q f ( q f ) .
4) Calculate integral of the measured spectrum up to a charge q f : Calculate cdf either ˆ F Q f ( q f ) (from Eq. 2.2), or ˆ F ← Q f ( q f ) (from Eq. 2.5).
5) Find z : Find z that equals the cdf of the interaction depth with the cumulative proportion of measuredevents. This is ˆ F Q f ( q f ) = F Z ( z ) , or ˆ F ← Q f ( q f ) = F ← Z ( z ) .
6) Calculate the efficiency at z : ε ( z ) = q f / q i .
7) Repeat steps 4, 5 and 6 for a different q f to complete ε ( z ) . Table 2 . Methodology to calculate the PCC efficiency function using one X-ray peak.
Appendix: Details of the method
The details of the method to measure the CCE in the backside of a back-illuminated CCD arepresented in Table 2. The details of method used with the Fe source having two X-ray lines ispresented in Table 3.
Acknowledgments
We thank the SiDet team at Fermilab for the support on the operations of CCDs and Skipper-CCDs,specially Kevin Kuk and Andrew Lathrop. We are grateful to Oscar von Uri for taking care ofno-solo-bar problem. This work was supported by Fermilab under DOE Contract No. DE-AC02-07CH11359. This manuscript has been authored by Fermi Research Alliance, LLC under ContractNo. DE-AC02-07CH11359 with the U.S. Department of Energy, Office of Science, Office of HighEnergy Physics. The United States Government retains and the publisher, by accepting the articlefor publication, acknowledges that the United States Government retains a non-exclusive, paid-up,irrevocable, world-wide license to publish or reproduce the published form of this manuscript, orallow others to do so, for United States Government purposes.
References [1] A. Aguilar-Arevalo et al. (DAMIC), Phys. Rev.
D94 , 082006 (2016), arXiv:1607.07410 [astro-ph.CO]. – 8 – ) Calculate angular distribution of incident photons:
Based on the geometry of the experiment evaluate f Θ ( θ ) .
2) Calculate depth distribution of events: f Z ( z | θ ) = ( p α ( cos ( θ )/ λ α ) exp (− z cos ( θ )/ λ α ) + p β ( cos ( θ )/ λ β ) exp (− z cos ( θ )/ λ β )) f Θ ( θ ) , where λ is the attenuation length of the photon. Then, calculate the cumulative distribution F Z ( z ) (or F ← Z ( z ) from equation 2.5).
3) Make an spectrum of measured events:
Calculate the spectrum of events reconstructed from the data and normalize it by total numberof events ( N T ). This is the estimation ˆ f Q f ( q f ) .
4) Calculate integral of the measured spectrum up to a charge q f : Calculate cumulative distributions either ˆ F Q f ( q f ) (from Eq. 2.2), or ˆ F ← Q f ( q f ) (from Eq. 2.5).
5) Find z : Find z that equals the cdf of the interaction depth with the cumulative proportion of measuredevents. This is ˆ F Q f ( q f ) = F Z ( z ) , or ˆ F ← Q f ( q f ) = F ← Z ( z ) .
6) Calculate the efficiency at z : ε ( z ) = q f /( p α q i ,α + p β q i ,β )
7) Repeat steps 4, 5 and 6 for a different q f to complete ε ( z ) . Table 3 . Methodology to calculate the partial charge collection efficiency function using Fe source.[2] A. Aguilar-Arevalo et al. (DAMIC), Phys. Rev. Lett. , 141803 (2017), arXiv:1611.03066[astro-ph.CO] .[3] A. Aguilar-Arevalo et al. (DAMIC), Phys. Rev. Lett. , 181802 (2019), arXiv:1907.12628[astro-ph.CO] .[4] M. Crisler, R. Essig, J. Estrada, G. Fernandez, J. Tiffenberg, M. Sofo haro, T. Volansky, and T.-T. Yu(SENSEI), Phys. Rev. Lett. , 061803 (2018), arXiv:1804.00088 [hep-ex] .[5] O. Abramoff et al. (SENSEI), Phys. Rev. Lett. , 161801 (2019), arXiv:1901.10478 [hep-ex] .[6] J. Tiffenberg, M. Sofo-Haro, A. Drlica-Wagner, R. Essig, Y. Guardincerri, S. Holland, T. Volansky,and T.-T. Yu, Phys. Rev. Lett. , 131802 (2017), arXiv:1706.00028 [physics.ins-det] .[7] L. Barak, I. M. Bloch, M. Cababie, G. Cancelo, L. Chaplinsky, F. Chierchie, M. Crisler,A. Drlica-Wagner, R. Essig, J. Estrada, E. Etzion, G. Fernandez Moroni, D. Gift, S. Munagavalasa,A. Orly, D. Rodrigues, A. Singal, M. Sofo Haro, L. Stefanazzi, J. Tiffenberg, S. Uemura, T. Volansky,and T.-T. Yu, arXiv e-prints , arXiv:2004.11378 (2020), arXiv:2004.11378 [astro-ph.CO] .[8] M. Settimo, arXiv e-prints , arXiv:2003.09497 (2020), arXiv:2003.09497 [hep-ex] .[9] The Oscura project is an R&D effort supported by Department of Energy to develop 10 kgskipper-CCD dark matter search.[10] A. Aguilar-Arevalo et al. (CONNIE Collaboration), Phys. Rev. D , 092005 (2019).[11] Connie Collaboration, A. Aguilar-Arevalo, X. Bertou, C. Bonifazi, G. Cancelo, B. A.Cervantes-Vergara, C. Chavez, J. C. D’Olivo, J. C. Dos Anjos, J. Estrada, A. R. Fernandes Neto,G. Fernandez-Moroni, A. Foguel, R. Ford, F. Izraelevitch, B. Kilminster, H. P. Lima, M. Makler, – 9 – . Molina, P. Mota, I. Nasteva, E. Paolini, C. Romero, Y. Sarkis, M. S. Haro, J. Tiffenberg, andC. Torres, Journal of High Energy Physics , 54 (2020), arXiv:1910.04951 [hep-ex] .[12] The Violeta Collaboration is planning a kg-scale Skipper-CCD experiment at a nuclear reactor facility.[13] G. Cancelo, C. Chavez, F. Chierchie, J. Estrada, G. Fernandez Moroni, E. E. Paolini, M. Sofo Haro,A. Soto, L. Stefanazzi, J. Tiffenberg, K. Treptow, N. Wilcer, and T. Zmuda, arXiv e-prints ,arXiv:2004.07599 (2020), arXiv:2004.07599 [astro-ph.IM] .[14] D. Rodrigues et al. , (2020), arXiv:2004.11499 [physics.ins-det] .[15] M. Sofo Haro, G. Fernandez Moroni, and J. Tiffenberg, arXiv e-prints , arXiv:1906.11379 (2019),arXiv:1906.11379 [physics.ins-det] .[16] J. R. Janesick,
Scientific charge-coupled devices , Vol. 83 (SPIE press, 2001).[17] S. Nikzad, M. E. Hoenk, P. J. Grunthaner, R. W. Terhune, F. J. Grunthaner, R. Winzenread, M. M.Fattahi, H.-F. Tseng, and M. P. Lesser, in
Proceedings of the SPIE , Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference Series, Vol. 2198, edited by D. L. Crawford and E. R.Craine (1994) pp. 907–915.[18] E. T. Hamden, A. D. Jewell, C. A. Shapiro, S. R. Cheng, T. M. Goodsall, J. Hennessy, M. Hoenk,T. Jones, S. Gordon, H. R. Ong, D. Schiminovich, D. C. Martin, and S. Nikzad, Journal ofAstronomical Telescopes, Instruments, and Systems , 036003 (2016), arXiv:1701.02733[astro-ph.IM] .[19] C. J. Bebek, J. H. Emes, D. E. Groom, S. Haque, S. E. Holland, P. N. Jelinsky, A. Karcher, W. F.Kolbe, J. S. Lee, N. P. Palaio, D. J. Schlegel, G. Wang, R. Groulx, R. Frost, J. Estrada, and M. Bonati,Journal of Instrumentation , C04018 (2017).[20] M. H. Fabricius, C. J. Bebek, D. E. Groom, A. Karcher, and N. A. Roe, in Proceedings of the SPIE ,Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 6068, edited byM. M. Blouke (2006) pp. 144–154.[21] S. E. Holland, D. E. Groom, N. P. Palaio, R. J. Stover, and M. Wei, IEEE Transactions on ElectronDevices , 225 (2003).[22] M.-M. Bé, V. Chisté, C. Dulieu, E. Browne, C. Baglin, V. Chechev, N. Kuzmenko, R. Helmer,F. Kondev, D. MacMahon, and K. Lee, Table of Radionuclides , Monographie BIPM-5, Vol. 3(Bureau International des Poids et Mesures, Pavillon de Breteuil, F-92310 Sèvres, France, 2006).[23] Https://engineering.lbl.gov/microsystems-laboratory/.[24] Https://geant4.web.cern.ch.[25] J. Jaeckel and S. Roy, Phys. Rev.
D82 , 125020 (2010), arXiv:1008.3536 [hep-ph] ., 125020 (2010), arXiv:1008.3536 [hep-ph] .