Electrical Induced Annealing technique for neutronradiation damage on SiPMs

M. Cordellia, E. Diociaiutia,d, R. Donghiaa,e, A. Ferrarif, S. Miscettia,S. Mullerf, G. Pezzulloc,∗, I. Sarraa,b,∗

aLaboratori Nazionali dell’INFN, via Enrico Fermi 40, 00044 Frascati, ItalybMarconi University, via Plinio, 44, 00193 Rome, Italy

cDepartment of Physics, Yale University, 56 Hillhouse, New Haven, CT-06511, USAdDipartimento di Fisica, Universita di Roma ”Tor Vergata”, Via della Ricerca Scientifica,

1, 00133 Rome, ItalyeDipartimento di Fisica, Universita degli studi Roma Tre, Largo S. Leonardo Murialdo, 1,

00146 Rome, ItalyfHelmholtz-Zentrum Dresden-Rossendorf (HZDR), Bautzner Landstraße 400, 01328

Dresden, Germany

Abstract

The use of Silicon Photo-Multipliers (SiPMs) has become popular in the design

of High Energy Physics experimental apparatus with a growing interest for their

application in detector area where a significant amount of non-ionising dose is

delivered. For these devices, the main effect caused by the neutron flux is a

linear increase of the leakage current.

In this paper, we present a technique that provides a partial recovery of the neu-

tron damage on SiPMs by means of an Electrical Induced Annealing. Tests were

performed on a sample of three SiPM arrays (2× 3) of 6 mm2 cells with 50 µm

pixel sizes: two from Hamamatsu [1] and one from SensL [2]. These SiPMs were

irradiated up to an integrated neutron flux up to 8 × 1011 n1MeV−eq/cm2. Our

techniques allowed to reduced the leakage current of a factor ranging between

15-20 depending on the overbias used and the SiPM vendor.

Keywords: SiPM, radiation damage, induced recovery

∗Corresponding authorEmail addresses: g.pezzullo@yale.edu (G. Pezzullo), ivano.sarra@lnf.infn.it (I.

Sarra)

Preprint submitted to Nuclear Instruments and Methods A September 8, 2018

arX

iv:1

804.

0979

2v2

[ph

ysic

s.in

s-de

t] 2

7 A

pr 2

018

1. Introduction

A Silicon Photo-Multiplier (SiPM) is a novel semiconductor photo-detector

composed by a matrix of pixels operating few volts above the breakdown voltage

(Geiger-Mode Avalanche Photo-Diode). Distinctive features of such technology

are: single photon detection capability, high gain ∼ 106 and good time resolution

(< 1 ns) [1]. In addition SiPMs present small size and customisable granular-

ity, insensitivity to magnetic fields and are relatively un-expensive. For these

reasons several current and planned High Energy Physics experiments employ

SiPMs in their experimental setup [3, 4]. A growing interest exists in their

application when a high level of non-ionizing dose is expected to be delivered to

the SiPMs.

Different studies showed a correlation between the bulk defects in the Silicon

structure due to the radiation damage and the deterioration of the photo-

detector performances [5, 6]. Hadrons and high energy leptons can produce

point defects as well as cluster related defects in the photo-detector active vol-

ume. In particular, neutrons traveling within the Silicon lattice induce many

displacements of Silicon atoms that, at the end of the path, form a disordered

agglomeration of atoms called cluster [7]. From the macroscopic point of view,

some of these defects act as charge carriers generator centers, producing an

increase of dark noise. In high energy physics experiments neutrons can be pro-

duced by photo-nuclear reactions of X and γ rays with the surrounding media.

The energy spectrum of the emitted neutrons is broadened from thermal ener-

gies up to a few MeV.

The current state of the art technology does not provide any method to fully

recover neutron induced damage in Silicon based photo-detectors. Thermal an-

nealing can partially recover the neutron induced damage, as shown in Ref. [8],

where after ∼ 2 months of thermal treatment, the dark current (Idark ) of SiPMs

exposed to ∼ 1012 n1MeV−eq/cm2 got reduced by a factor of 2. Our tests re-

ported in the next sections indicate that an important recovery of the SiPM

damage can be achieved by means of an electrical induced annealing. Two dif-

2

ferent methods have been tested: a direct and an inverse polarization of the

SiPM using a non conventional over voltage - larger than the usual 5 V limit

indicated by several vendors [1, 2]. In this way two effects are combined: the

thermal annealing induced by heating the Silicon device and the presence of a

strong electric field to re-order the atoms in the Silicon lattice displaced by the

neutron-induced bulk damage [9].

2. Experimental setup

The measurements described in this paper were performed at the Silicon

Detector Facility of the Fermi National accelerator Laboratory 1. Two different

set of SiPM arrays were studied: Hamamatsu [1] and SensL [2]. Each detector is

composed by a 3×3 array of 6 mm2 with a pixel size of 50 µm. Both SiPM arrays

have a custom design, developed for the electromagnetic calorimeter of the Mu2e

experiment [10]. To improve the thermal dissipation capabilities compared to

commercial designs2, they have been built with a thermal resistance of about

5 × 10−4m2 K/W . The experimental setup consisted of:

• light tight thermal chamber TestEquity Model 140 [11];

• power supply PLH250 [12] to bias the SiPMs and measure the current;

• 30×20 cm2 Cu plate, 3 mm thick, used as thermal support for the SiPMs;

• 3 PT1000 [13] thermistor used to measure the temperatures of the SiPM

and of the Cu support.

Figure 1 shows a picture of the Cu support with one PT1000 and one SiPM

array plugged in. This setup allowed us to monitor the temperature of the

SiPM active region and of the Cu support during all the measurements and

bias the SiPM with a current limitation of few hundreds mA. The third PT1000

1http://www.fnal.gov2Hamamastu and SensL, private communication

3

Figure 1: Cu support with one PT1000 and one SiPM array plugged in.

sensor was used to check that the temperature drop on the pin-side of the SiPM

was within 2 °C w.r.t. the one measured on the active region.

3. SiPM recovery measurements

Two different configurations for biasing the SiPMs were tested: direct and

inverse polarization using an over-voltage larger than 5 V.

We define ∆VIND as the induced annealing operating over voltage w.r.t. to the

breakdown voltage (Vbr) and IIND as the drawing current of the photo-detector

in this configuration. Instead, we refer to Idark as the current measured in the

standard reverse configuration of the SiPM with an operation voltage Vop = Vbr

+ 3 V.

3.1. Recovery in direct polarization mode

Figure 2 shows the results obtained with one cell of a Hamamatsu SiPM

array polarized in direct configuration at ∆VIND = 10 V. The Idark shows an

exponential trend with a decay constant of τ = 500 s and a constant current

term I0−dark = 5.2 mA. A reduction of ∼ 40% in Idark after ∼ 5 min of direct

polarization was achieved. In this configuration the SiPM was generating a IIND

4

0 100 200 300 400 500 600 700 800 900Exposure time [s]

6

7

8

9

10

11Co

, 20

op

I [m

A]

afte

r ex

po

sure

@ V

0I 0.04299± 5.194 τ 33.87± 500.4 0I 0.04299± 5.194 τ 33.87± 500.4

))τ(1+#exp(-t/0I

Figure 2: Idark as a function of the exposure time for a direct polaritation. Points were fit

with an exponential curve.

of ∼ 530 mA, corresponding to a power of ∼ 5 W.

3.2. Recovery in reverse bias mode

To quote the performance of the recovery method in inverse polarization

two different settings of over voltage were tested. Both Hamamatsu and SensL

SiPMs, exposed to the same neutron irradiation [8], have been checked in this

configuration. Figure 3 shows the Idark variation as a function of the exposure

time for a Hamamatsu SiPM. The first four measurements were taken at VIND

= Vbr +10 V, then we improved the recovery by increasing the over voltage up

to 14 V. After about 13 min at the described conditions, the SiPM Idark reached

about 1 mA that corresponds to a reduction of a factor ∼ 10 when compared

to the starting value. Then we tried to increase the over-voltage up to 17 V,

but at this value of VIND the SiPM got broken. The IIND measured during

this operation was 130 (186) mA for a ∆VIND of 10 (14) V, corresponding to a

dissipated power of about 8 (12) W. The temperature measured on the SiPM

active region was ∼ 150 °C.

After the learning phase, we tested another Hamamatsu SiPM cell of the same

device setting ∆VIND at 14 V. Figure 4 shows the results of the Idark measurement

5

hEntries 0Mean 0Std Dev 0Underflow 0Overflow 0Integral 0Skewness nan

Exposure time [s]0 100 200 300 400 500 600 700 800

CoI [

mA]

afte

r exp

osur

e @

Vop

, 20

0

2

4

6

8

10

12h

Entries 0Mean 0Std Dev 0Underflow 0Overflow 0Integral 0Skewness nan

Over-voltage = 10 V

Over-voltage = 14 V

Inverse Polarization

Figure 3: Idark as a function of the exposure time for a Hamamatsu SiPM cell for an inverse

polarization. The legend indicates two different settings for the over-voltage used during the

test.

operated every 2 min. After about 6 min of electrical annealing, a total reduc-

tion of a factor 15 on the leakage current was observed.

Exposure time [s]0 50 100 150 200 250 300 350

CoI [

mA]

afte

r exp

osur

e @

Vop

, 20

1

10Over-voltage = 14 V

Inverse Polarization

Figure 4: Idark as a function of the exposure time for a Hamamatsu SiPM cell applying an

over-voltage of 14 V during the recovery.

A similar test has been carried out also for the SensL device [2]. Figure 5

6

shows the measured Idark for two different setting of ∆VIND3: 5 and 8 V. After

22 min we measured on overall reduction factor ∼ 16 on Idark (from 40.2 to 2.5

mA). Unfortunately after 22 min the SiPM got unplugged from the Cu plate,

hEntries 0Mean 0Std Dev 0Underflow 0Overflow 0Integral 0Skewness nan

Exposure time [s]0 200 400 600 800 1000 1200 1400

CoI [

mA]

afte

r exp

osur

e @

Vop

, 20

1

10

210h

Entries 0Mean 0Std Dev 0Underflow 0Overflow 0Integral 0Skewness nan

Over-voltage = 5 V

Over-voltage = 8 V

Inverse Polarization

Figure 5: Idark as a function of the exposure time for a SensL SiPM cell. The legend indicates

two different settings for the overvoltage used during the test.

reached a temperature larger than 200 °C and got broken. Figure 6 shows the

localization of the heating induced by the over voltage condition on the single

cell under bias. During the annealing treatment, we have measured a IIND

current of 200 (400) mA for ∆VIND of 5 (8) V, corresponding to a dissipated

power of about 6 (13) W. The temperature measured on the SiPM active region

was ∼ 150 °C.

We repeated the measurement on another cell of the same device at ∆VIND

of 8 V. Figure 7 shows the Idark variation as a function of the exposure time.

After ∼ 17 min the Idark was reduced from 42.9 to 2.7 mA that is compatible

with the reduction observed in the previous test.

3The breakdown voltage of the SensL SiPM at 20 °C is 24.87 V [2].

7

Figure 6: SiPM array on the Cu support at the end of the tests.

Exposure time [s]0 200 400 600 800 1000

CoI [

mA]

afte

r exp

osur

e @

Vop

, 20

10

Over-voltage = 8 V

Inverse Polarization

Figure 7: Idark as a function of the exposure time for a SensL SiPM cell applying an over-

voltage of 8 V during the recovery.

3.3. Thermal induced annealing

To understand if the observed reduction of Idark was due only to the related

thermal annealing, we exposed the SiPM to high temperatures at Vop4. For this

test we used another cell of the Hamamatsu SiPM array. We started keeping the

SiPM at 80 °C for 19 min, then increasing the temperature up to 120 °C. Results

4The breakdown voltage was adjusted w.r.t. the temperature of the test.

8

SiPM T [oC] Exposure time [s] Idark [mA]

20 ± 0.5 0 12.33 ± 0.01

80 ± 0.5 1120 9.93 ± 0.01

120 ± 0.5 600 9.50 ± 0.01

Table 1: Idark measured @ 54.7 V, 20oC after exposition of the SiPM at different thermal

conditions.

are summarised in Table 1. The thermal annealing alone does not provide in

30 min the same Idark reduction observed with the electrical induced annealing

technique in a shorter amount of time.

4. Conclusion

We have presented an electrical induced annealing technique that allows

to partially recover neutron damage of SiPM. Tests of such a technique were

performed on two different sets of SiPM arrays from Hamamatsu and SensL.

The benefit of the observed method is that it allows to partially recover (up

to a factor 15-20) the bulk damage in the SiPM exposed to a neutron flux up

to ∼ 1012n1MeV−eq/cm2. We also proved that the proposed technique improves

the results one may get from conventional Thermal annealing. However, as

the drawing current during the proposed annealing process can reach hundreds

of mA, the sensors need to be operated in a condition that provides thermal

dissipation in order to avoid the sensor heating up to the breaking temperatures

(200 °C).

Acknowledgments

This work was supported by the EU Horizon 2020 Research and Innovation

Program under the Marie Sklodowska-Curie Grant Agreement No. 690385. The

authors are grateful to many people for the successful realisation of the tests.

In particular, we thank Dr. Adam Para, for providing us a necessary and well

equipped facility at SiDet, Fermilab.

9

References

[1] Hamamatsu, TSV MPPC datasheet, http://www.nuclear.gla.ac.uk/

~jrma/A2/Tagger/NewFPD/SiPMT/KSX-I50014-E_S12642%20Series%20%

283mm%E2%96%A1-TSV%E3%83%BCArray%29.pdf, Accessed: 2018-03-12.

[2] SensL, SiPM datasheet, http://www.eqphotonics.de/cms/cms/upload/

datasheets/PB-SensL%20Product%20Summary_IEEE.pdf, Accessed: 2017-

09-2.

[3] L. Bartoszek et al., Mu2e Technical Design Report, 3rd ed. arXiv:1501.05241

[physics.ins-det], 2014.

[4] N. Strobbe, The upgrade of the CMS hadron calorimeter with silicon photo-

multipliers, JINST, 12, 01, c01080, 2017.

[5] Yu. Musienko et al., Radiation damage in silicon photomultipliers exposed

to neutron radiation, JINST, 12, 07, c07030, 2017.

[6] J.R. Srour, C.J. Marshall and P.W. Marshall, Review of Displacement Dam-

age Effects in Silicon Devices, IEEE Transactions on Nuclear Science, 50,

653-670, 2003.

[7] I. Pintiliea, G. Lindstroem, A. Junkesand and E. Fretwurs, Radiation In-

duced Point and Cluster-Related Defects with Strong Impact to Damage

Properties of Silicon Detectors,

[8] M. Cordelli and others, Neutron irradiation test of Hamamatsu, SensL and

AdvanSiD UV-extended SiPMs, arXiv:1710.06239, physics.ins-det, 2017.

[9] Claude Leroy and Pier-Giorgio Rancoita, Particle interaction and displace-

ment damage in silicon devices operated in radiation environments, Reports

on Progress in Physics, 70, 4, 493, 2007.

[10] N. Atanov et al., Design and status of the Mu2e electromagnetic calorime-

ter, NIM A 824, 695 - 698, 2016.

10

[11] TESTEQUITY web page, https://www.testequity.com/static/49/,

Accessed: 2018-03-12.

[12] PLH250 power supply datasheet, http://resources.aimtti.com/

datasheets/AIM-PLH+PLH-P_series_DC_power_supplies_data_

sheet-Iss1A.pdf, Accessed: 2018-03-12.

[13] PT1000 datasheet, http://www.keb.com.tw/Files/Other/PDF/Data%

20Sheet-PT1000-RW.pdf, Accessed: 2018-03-12.

11