Charge collection studies on heavily irradiated “spaghetti” diodes

Charge collection studies on heavily irradiated “spaghetti” diodes

G. Kramberger, V. Cindro, I. Mandić, M. MikužϮ, M. Milovanović, M. ZavrtanikJožef Stefan Institute, Ljubljana, SloveniaϮ also University of Ljubljana, Faculty of Physics and Mathematics

Many thanks to I. Tsurin , University of Liverpool, for mask design.

2

MotivationRD50 had/has a “multiplication wafer” run with Micron, which included special devices/diodes for studying impact of various parameters on charge collection:

implant diffusion time energy of implantation ions detector thickness

In addition it is interesting to explore: role of implant geometry : spaghetti vs. strip detectors long term annealing performance? operation at Feq~1017 cm-2?

15.11.2012 G. Kramberger, Charge collection studies on heavily irradiated "spaghetti" diodes, 8th Trento Workshop, 2013, Trento

How does charge collection differ over the strip – homogeneity of response? dependence of multiplication on hit position annealing effects

3

SamplesSpecial diodes-pad detectors were designed on that wafers which are particularly suitable for studies of charge collection:

DC coupled, guard ring structure – high breakdown voltage 80 mm pitch, 20 mm implant width (ATLAS geometry) 4x4 mm2 , 300 and 150 mm thick

Type 3 – fully metalized implant (CCE)


All strips connected together at one side: almost the same electric field

as in a strip detector much simpler handling (CCE,

CV-IV, TCT measurements) simpler analysis weighting field has same shape

as the electric field

Type 1 – partially metalized implant (TCT)

4

Irradiations and measurements

Sample treatment:Neutron irradiations: in steps up to 8∙1016 cm-2,80 min annealing at 60oC in betweenMeasurements done in the range [-20ºC, -25oC]some samples irradiated to a fix fluence - single stepsome standard p-type strip detectors from Micron were also used

Samples from different wafers:2935-2…9 – standard : 2e15 cm-2, 150 keV P, 5e15 cm-2, 80 keV B, 220 nm thermal oxide2935-10 - double diffusion : 1000oC for 3h2912-2,3 - double energy : 300 keV of P ions, doubly charged2885-5 – thin

Amplifier+shaper

90Sr/241 Am

water cooled heat sink

cold plate

Peltier cooler scintillator

pad detectorTex2440

thermal isolation


HV

Oscil.

CCE Setup (absolute charge) Scanning-TCT Setup(relative charge over strips)

1064 nm<100 ps8 mm FWHM

5

CCE and noise for non-irradiated samples

variation of ~10% for charge at V>Vfd – several samples were re-measured and reproducibility was found to be better than that

good agreement of Vfd determined from Q-V with that of C-V noise performance in accordance with expectations – not optimized

noise without detector


CCE Setup

G. Kramberger, Charge collection studies on heavily irradiated "spaghetti" diodes, 8th Trento Workshop, 2013, Trento 6

CCE comparison for all wafers

almost no difference in charge collection efficiency for different implants (but only limited parameter region investigated)

superior performance of thin detectors (black squares) at lower voltages very high CCE for thin detector (~10-11 ke for 3∙1015 cm-2). up to 1000 V thin are at least as good as thick only moderate increase of charge collection with high bias voltages for thin device –

why don’t we see larger increase of multiplication?

15.11.2012

cm-2 cm-2

80 min@60oC 80 min@60oC - 2 irr. steps

T=-23oC T=-23oC

CCE Setup

7

CCE of “standard” wafers for all fluences

Even at 8∙1016 cm-2 a signal of ~1200 e can be expected i.e. few mip sensitivity with present electronics

An interesting observation – at very high fluences (2,4,8∙1016 cm-2) no problems with micro discharges – very stable operation!


CCE Setup

8

Dependence of CCE on voltage


Power law dependence of QMVP on fluence (“an empirical formula”) surprisingly works over two orders of magnitude.

The relative difference in collected charge at given bias voltages is independent on fluence (as if combination of trapping and multiplication would perfectly compensate).

500 V: Q0=12600 800 V: Q0=20200 1000 V: Q0=26000

9

Leakage current performance


all scaled to -23oC

Leakage current does not increase linearly with fluence, but exhibits saturation. it falls below the expected generation current (the field is present in the entire volume) Possible carrier recombination (minority carrier lifetime decreases) limits the current

How much does it influence the CCE (trapping times << recombination times)? A very rough estimation of lower limit gives recombination lifetime of <1ns (8∙1016 cm-2).

expected generation current from entire bulk

10

Noise performance at high fluences


Main observations: there is no apparent increase of noise with leakage current noise is Gaussian

color corresponds to figure on left at the highestvoltage shown

Conclusions: Decrease of M with fluence? M increases with voltage

standard MicronProcess samples

11

Annealing of wafers with different implants

There is no large difference between different implants in terms of CCE during long term annealing (double diffusion seem to show less increase in multiplication at 1280 min)

Thin detector performs best also during long term annealing The slope of the charge rise with voltage increases with annealing – seen already

several times before (CERN, JSI, Glasgow…) CCE>1 for the thin device already at 600 V after 5120 min @ 60oC


1e16 cm-2 , annealed at 60oC

12

Annealing of wafers with different implants

The Q-V plot shows rapid rise of charge after 1280 min annealing. At low bias voltages the charge is smaller for longer annealing times (<400 V). Increase of noise is related to increase of charge, but a detector with smaller electrodes

would have larger S/N (series noise should dominate over shot noise). I-V has a similar shape as Q-V with a difference because of generation current annealing

(initial drop of current). Similar behavior seen for all investigated materials.


2912-3 2912-3 2912-3

13

Difference between spaghetti and strips


Two standard FZ-p strip detectors of same strip geometry from Micron were used for comparison: Electric field is almost the same, hence multiplication effects are the same. The difference is therefore due to effects of the weighting field “trapping induced charge

sharing” – residual charges on the neighbors. The QMPV dependence on bias voltage has the same shape.

p+ strips dioden+ strips

n+ - higher signal in hit electrode p+ - wider clusters

sign

al

Trapping induced charge sharing

for a track through the center of the strip!

usual threshold

G. Kramberger, Charge collection studies on heavily irradiated "spaghetti" diodes, 8th Trento Workshop, 2013, Trento

Uniformity of response over the strip

Non-irradiated detectorBias = 100 V

no metal

Guard ring

metal

Does charge collection depend on position of the e-h pair generation ? electrons may drift to different locations in the implant different path lengths result in different amount of trapping

This can be probed by TCT by scanning the inter-strip region.

15.11.2012 14


Φeq = 5·1015 n/cm2;Bias = 1000 V;

charge collection profile more non-uniform with annealing

direction of the scan


Φeq = 5·1015 n/cm2;Bias = 600 V;

TCTSetup

impl

ant

impl

ant

enhanced multiplication at the edge of implants for both bias voltages

The increase of collected charge after annealing is smaller than measured by 90Sr electrons?

Sam

ples

ann

eale

d in

the

setu

p

15.11.2012


2e15, annealed 5120 min 5e15, annealed 5120 min


15.11.2012 16

17

Conclusions Within the parameter space investigated in RD50 Micron Multiplication run:

the “double energy” of implantation ions and the “double diffusion time”

processed diodes perform equally after irradiation and also during long annealing to spaghetti diodes processed in a standard way.

Thin diodes perform better than standard ones for both fluences for bias <=1000 V.

Strong increase of charge during long term annealing, but also noise and leakage current.

As expected the spaghetti diodes perform slightly worse than strip detectors (due to trapping induced charge sharing) at given fluence.

Spaghetti diodes are still “alive” at 8∙1016 cm-2 – the charge of 1250 e can be expected (probably more for strip detectors) at 1000 V .

The uniformity of the signal over the strip grows with multiplication – highest signal at implant edges.


Documents

Charge collection studies on heavily irradiated “spaghetti” diodes