Performance of LGADs and AC-LGADs towards 4D tracking · Performance of LGADs and AC-LGADs towards 4D tracking IntroductionDataGeant4 SimulationCaliforniumAC-LGADsConclusions Low

Performance of LGADs and AC-LGADs towards 4D trackingIntroduction Data Geant4 Simulation Californium AC-LGADs Conclusions

Performance of LGADs and AC-LGADstowards 4D tracking

G. D’Amen1, W. Chen1, G. Giacomini1, L. Lavitola2,S. Ramshanker3, A. Tricoli11Brookhaven National Laboratory (US)2Universita’ degli studi Federico II (IT)3Oxford University (UK)

9 December 2019

CPAD INSTRUMENTATIONFRONTIER WORKSHOP 2019

University of Wisconsin-Madison

1 / 21


OutlineTime resolution - LGADI. Introduction to LGADs

II. LGAD response to 90Sr β−

III. Response to DT fast neutronsIV. Comparison with Geant4 simulationV. Response to 252Cf fast neutrons

Space & Time - AC-LGADVI. The AC-LGAD concept

VII. Characterization with IR laser and 90Sr

Conclusions and Future activities

LGAD wafer (BNL)

AC-LGAD matrix (BNL)2 / 21


Low Gain Avalanche DiodeIntroduction

Low Gain Avalanche Diode (LGAD): highlydoped layer of p-implant (Gain layer) near p-njunction creates a high electric field thataccelerates electrons enough to startmultiplication.

I Electric Field: ∼300 kV/cm in Gain LayerI Silicon-based technology with low,

adjustable gain (2 - 100)I Breakdown Voltage ∝ Gain parameters

(dose, energy)I High Signal/Noise ratioI Ability to achieve fast-timing O(20-30) ps

in high radiation environments

Efield

Questions to be answered:I MIPs detection capabilities already proven,

fast neutron response to be characterizedI How fast is the response to fast neutrons?I What are out limits of detectable neutron

energy?3 / 21


LGAD structure

Wafer structure (W1836,W1837,W1840)

I 1×1 mm2 sensor sizeI 50 µm 28Si p epitaxial layer, 10B and 11B doped

(7×1013cm−3)I Different doping concentrations (3, 3.25 and 2.7×1013cm−3) and gain layer thickness

I 500 µm substrateI Aluminum thin layerI Silicon Oxide SiO2

I n++ layer, 31P dopedI Gain p+ layer, 11B doped

4 / 21


90Sr interactionsSignal waveforms

Waveforms from β− 90Sr signals

> W1836, W1837, W1840 show narrowpeaks with widths O(1 ns)

> Sensors Gain for β− compatible to that ofX-rays

> σj = 〈σnoise

(dVdt

)−1〉 ∼ 20 ps

Sensor Gain (X-Ray):W1836: ∼ 15W1837: ∼ 20W1840: ∼ 25

5 / 21


Deuterium-Tritium neutron generator

BNL Thermo-Fisher MP 320 Neutron Generator (prototype)3T +2 D →4 He+ n(14.1 MeV ) (1)

Neutron energy spectrum very narrow σE = O(10−2 MeV) and isotropic, with estimated neutronproduction of 6×107 neutrons/sec, with a flux of 7×104 neutrons/(cm2 sec) at sensor position

6 / 21


Fast Neutron interactionsSignal waveforms

Waveforms from neutron signals (Vtrig = 10mV)

> W1836, W1837, W1840 show narrowpeaks with widths O(1 ns)

> Sensor Gain for neutrons compatible to theone measured with X-rays

> σj = 〈σnoise

(dVdt

)−1〉 ∼ 20 ps

Sensor Gain (X-Ray):W1836: ∼ 15W1837: ∼ 20W1840: ∼ 25

7 / 21


Fast Neutron interactionsDeposited Energy distributions

Energy deposited by the neutron interactioncomputed as integral of each signal:

Edep [eV ] =3.6 [eV ]

Gn Rfb qe

∫wf

Adt

Sensitive Range in deposited energy (∝ (Gn)),limited by trigger voltage and maximumsignal amplitude in oscilloscope window.

For a 10 mV trigger level and Gn = 15,sensitivity to neutron signals with depositedenergy as low as ∼ 30 keV.

8 / 21


Generated energy spectrum

Distribution of energy deposited by DT neutroninteractions as simulated by Geant4 shows goodagreement with experimental data from W1836in the sensor sensitive range Edep = [30, 450] keV

Superimposing Edep distributions generated byneutrons with different energies can give us anestimate of minimum neutron energy sensitivity

9 / 21


Neutron energy sensitivity

Extrapolation of sensitivity to various neutron energies based on 14.1 MeV data

W1836 sensitivity (according to 14.1 MeVdeposited E distribution) to 300- and 500- keV

neutrons

W1836 sensitivity (according to 14.1 MeVdeposited E distribution) to 20 MeV neutrons

10 / 21


Californium 252 Decays

252Cf decay scheme:

- ∼ 96% Alpha decay

- ∼ 3% Spontaneous Fission (SF) (n, γ)

- < 1% rare decays

Energy spectrum (SF):

> Neutrons: Landau(µ = 2 MeV, σ = 0.5 MeV)

> Photons: Landau(µ = 400 keV, σ = 100 keV)

> α: either 6.076 MeV or 6.118 MeV, entirelyabsorbed

11 / 21


LGAD sensitivity to 252CfUnshielded sensor (Geant4 simulation)

Spontaneous Fission photon flux ∼ 8/3 neutronflux. Lead shielding should decrease γpopulation.

Lead shielding (2.5 cm) (Geant4 simulation)

• Edep < 80/90 keV Photon dominated

• Edep = 90 - 200 keV Photon/Neutronpopulation

• Edep > 200 keV Neutron dominated12 / 21


LGAD sensitivity to 252Cf

Distribution of energy deposited by 252Cfneutron and photon interactions assimulated by Geant4 shows goodagreement with experimental data fromW1840 in the sensor sensitive range Edep

= [15, 140] keV (photon dominated)

Jitter from Cf signals ∼ 20 ps, compatibleto DT and MIPs.

Additional data covering mixed- andneutron- dominated regions are beingcollected as we speak.

13 / 21


4D detectors: AC-LGAD tests with IR laser and 90Sr

> The AC-LGAD concept> LGAD vs AC-LGAD comparison> Cross-Talk studies> Timing performance

14 / 21


AC-LGADconcept

LGAD limits:I Dead volume (local gain ∼ 1)

within the implanted region ofthe gain layer

I Pixels/strips (pitch ∼ 100 mm)with gain layer below the implanthave a Fill Factor «100%

I Good for timing, hardly for 4Dreconstruction

AC-LGAD goals:I ∼ 100% Fill Factor and fast timing information at a

per-pixel level achievedI Signal generated by drift of multiplied holes into the

substrate but AC-coupled through dielectricI Electrons collect at the resistive n+ and then slowly

flow to a ohmic contact at the edge.

15 / 21


AC-LGADSignal comparison with LGADs

I Sensor wire-bonded to 16 channel Trans-impedanceAmplifier board by FermiLab

I AC-LGAD: 3×3 pixel matrix, 200µm × 200µmAC-coupled pads bonded to TAs

I LGAD: same AC-LGAD device where the n++ isread-out by the TA (same bias conditions and gain)

I Comparison of pulse amplitudes of betas from 90Sr.I Essentially equal distribution (same gain) for LGAD

and AC-LGAD AmplitudesI Is this signal well spatially localized? Need to

estimate Cross-Talk between pixels/strips16 / 21


Cross-talkStrip Map

Cross-talk measured as ratio between signal amplitudepeaks in different strips

Crosstalkratio A2/A1 100%ratio A3/A1 13%ratio A4/A1 6%ratio A6/A1 4%

Response of a single strip asa function of shining positionof IR or red laser (TCTscan).

Border effect: n++ is a lowresistance path that couplesthe signals back to the stripunder measure.

17 / 21


Cross-talkPixel Map

Cross-talk measured as ratio between signal amplitudepeaks in different pixels

Dose n+ 1/100 Dose n+ 1/10ratio A5/A1 7% 9%ratio A9/A1 11% 16%

Response of a single pixel asa function of shining positionof IR or red laser (TCTscan).

Border effect: n++ is a lowresistance path that couplesthe signals back to the pixelunder measure.

18 / 21


Timing Resolution

I AC-LAGDs and LGADs show similar response(waveforms)→ expected ∼ same timingperformance

I Using beta signals from a 90Sr source on AC-LGADslead to estimated σjitter ∼20 ps

I NEXT: Measuring timing resolution in coincidenceswith a trigger sensor, using 3D-printed Beta Scopesetup ready with ∼ 180 MBq 90Sr source

I Developed a setup such that our probe station canoperate both at room temperature and at -30◦Cwhich will be used for pre/post irradiation IV andCV scans

19 / 21


Conclusions

LGADs can be used to detect neutrons in the 100s keV - MeV (and beyond?)energy range in high flux conditions for applications where fast time (∼20 - 30ps) measurements are needed

Fast timing for fast neutrons ensured by jitter measurement of O(20) ps

Good agreement between data and G4 simulation; extrapolations from Geant4simulations shows potential sensitivity to neutrons with energies <100 keV

By changing a few photolithographic masks and tuning process flow parameters,AC-LGADs have been fabricated as well

Precision space resolution (50-100 µm) available with AC-LGAD technology

Cross-talk and time resolution tested with mips and TCT, leading to positiveresults 20 / 21


Additional info/links

I G. Giacomini, W. Chen, F. Lanni, and A. Tricoli, Development of a technologyfor the fabrication of Low-Gain Avalanche Diodes at BNL

I G. Giacomini, W. Chen, G. D’Amen, A. Tricoli, Fabrication and performanceof AC-coupled LGADs

21 / 21

https://arxiv.org/abs/1811.04152




Performance of LGADs and AC-LGADs towards 4D tracking

Backup

22 / 21


Motivation

Low-Gain Avalanche Diodes (LGAD) are gathering interest in thePhysics community thanks to fast-timing and radiation-hardness:

I HEP: ATLAS (HGTD) and CMS (MTD) timing detectors atthe HL-LHC

I NASA: neutron flux studiesI Medical Imaging: PET scansI Quantum information, Nuclear and forward physics,

etc...

MIPs detection capabilities already proven, investigating theresponse to neutrons in the O(MeV) region (fast neutrons)

Wafer of LGADs produced at BNL

23 / 21


Sensor gain computationSignals max amplitude Max amplitude scaled by Gain (normalized)

Distributions of maximum signal amplitude (left) aredivided by the sensor gain Gn (right), as obtainedfrom X-ray measurements.

• Sensor Gain:W1836: ∼ 15W1837: ∼ 20W1840: ∼ 25

24 / 21


Slew Rate

Average signal NoiseI W1836: (0.39±0.54) mVI W1837: (0.10±0.43) mVI W1840: (0.19±0.5) mVI W1849: (-0.11±0.42) mV

25 / 21


Sensitive range

Full width at half maximum (normalized)

Sensitive region limited by trigger voltage(10 mV for W1836, W1837, W1840, 3.5 mVfor W1849) and maximum signal amplitudein oscilloscope window.

Energy distributions constrained in regionbetween:

Ith =√2π Vth

〈FWHM〉2.355

with V minth = trigger level and V max

th = maxwindow amplitude

26 / 21


Signal waveforms

Waveforms acquired withTektronix MSO64 mixed-signalsoscilloscope;

W1836, W1837, W1840 (50 µm)show narrow peaks with widthsO(1 ns), while W1849 (300 µm)produces longer (∼ 8 times)signals.

27 / 21


Sensor gain computationSignals max amplitude Max amplitude scaled by Gain (normalized)

Distributions of maximum signal amplitude (left) aredivided by the sensor gain Gn (right), as obtainedfrom X-ray measurements.

• 50 µm Gain:W1836: ∼ 15W1837: ∼ 20W1840: ∼ 25

• 300 µm Gain:

W1849: ∼ 10

28 / 21


Jitter measurementJitter is an important component of the timeresolution of the sensor and is computed as ratiobetween the noise (∼0.5 mV for all the sensors)and slew rate (dV/dt):

σj = 〈σnoise

(dV

dt

)−1

〉

Sensor Gain Jitter [ps]W1836: ∼15 14.8 ± 3.6

W1837: ∼20 17.5 ± 4.3

W1840: ∼25 21.3 ± 4.3

W1849: ∼10 222.4 ± 42.7

Slew rate (normalized)

29 / 21


Deposited Energy distributions300 µm sensor comparison

W1849 (300µm) has been compared to the 50µm sensors:

I Compatible shape in the sensitive rangeafter gain correction

I Higher detection efficiency (×54 timesvolume)

I Different minimum threshold of sensitiverange:Emin

dep =∼ 30keV (50µm) vs ∼ 200keV(300µm)

30 / 21


Characterization of neutron processes

I Neutron Elastic interactionsignificant for 14 MeV neutroninteractions with depositedenergy up to ∼ 1.85 MeV

I Neutron Inelastic interactiondominant contribution for highdeposited energies

I In the range Edep = [30, 450] keVminimal contributions fromphotons and electronselectromagnetic processes(ionization, Compton effect,photoelectric effect) and decays

31 / 21


Scan of neutron energy sensitivity

Distributions of deposited energyfor neutrons with:

I K = 10/100 keV(top-left)

I K = 200/300 keV(top-right)

I K = 500/700 keV(bottom-left)

I K = 1 MeV(bottom-right)

for Trigger threshold 10 mV andGn = 15, expected sensitivity to300 keV neutrons

32 / 21


LGAD structure

Wafer structure (W1836,W1837,W1840)

I 1×1 mm2 sensor sizeI 50 µm 28Si p epitaxial layer, 10B and 11B

doped (7×1013cm−3)I Different doping concentrations (3, 3.25 and

2.7 ×1013cm−3) and gain layer thickness

I 500 µm substrateI Aluminum thin layer, thickness 0.5 µmI Silicon Oxide SiO2, thickness 0.3 - 0.5 µmI n++ layer, 31P doped, thickness 0.5 µmI Gain p+ layer, 11B doped, thickness 0.5 µm

33 / 21


Geant4 simulationIntroduction

Sensor response modelled with Geant4 10.4MonteCarlo simulation software

Simulation parameters:I QGSP_BIC_HP physics list used for

high precision simulation of neutrons ≤ 20MeV

I 10 million 14.1 MeV neutrons generatedeach simulation run with randomized initialdirection

I 1.6 mm of 27Aluminum interposed betweenneutron generator and sensor, to simulatethe Deuterium-Tritium generator casing

34 / 21


AC-LGAD characterizationIV-curve

35 / 21


AC-LGADFabrication at BNL

Process:I Process starts from a Std (DC-) LGAD PadI Change METAL (Aluminum) and thus ContactsI n++ runs at the periphery only; replaced by resistive

n+ in the active area with 10/100 less doseI Thin insulator (100 nm SiN ) over the n+

Std-LGAD Pad:

AC-LGAD Pixels:

AC-LGAD Strips:

36 / 21


Near future plans

• Lower trigger threshold from10 mV to 2 mV (×4 averagenoise); expected sensitivity toEn < 100 keV:

• Edep th @10mVW1836: ∼30 keVW1837: ∼20 keVW1840: ∼22 keV

• Edep th @2mVW1836: ∼6 keVW1837: ∼4 keVW1840: ∼4 keV

37 / 21


Limits of LGADs

Lateral dimensions of Gain layer must be much larger than thickness of substrate, to createuniform multiplication.Dead volume (local gain ∼ 1) extends within the implanted region of the gain layer:

I Pixels/strips (pitch ∼ 100 mm) with gain layer below the implant have a Fill Factor «100%(Voltage dependent)

I Large pads (∼ 1 mm) are preferred (e.g. ATLAS HGTD or CMS MTD)I Good for timing, hardly for 4D reconstructionI Various possible ways to overcome this issue with different geometries

38 / 21


AC-LGADconcept

Main differences w/r to LGADs:

1. One large low-doped high-ρ n+

implant running overall the activearea, instead of a high-dopedlow-ρ n++

2. Thin insulator over the n+, wherefine-pitch electrodes are placed,patterned over the insulator

Expected Results:I ∼ 100% Fill Factor and fast timing information at a

per-pixel level achievedI Signal generated by drift of multiplied holes into the

substrate but AC-coupled through dielectricI Electrons collect at the resistive n+ and then slowly

flow to a ohmic contact at the edge.

39 / 21


Fast Neutron interactionsJitter measurement

Jitter is an important component of the timeresolution of the sensor; computed as ratiobetween the noise (∼0.5 mV for all the sensors)and slew rate (dV/dt):

σj = 〈σnoise

(dV

dt

)−1

〉 ∼ 20 ps

Slew rate (normalized)

40 / 21


AC-LGADSignal Sr90

41 / 21

Performance of LGADs and AC-LGADs towards 4D tracking · Performance of LGADs and AC-LGADs towards 4D tracking IntroductionDataGeant4 SimulationCaliforniumAC-LGADsConclusions Low

Documents