Charge collection properties in an irradiated pixel sensor built in a thick-film HV-SOI process

1 MB·

by  B. Hiti

PreparedforsubmissiontoJINST
Charge collection properties in an irradiated pixel
sensor built in a thick-film HV-SOI process
7 B.Hitia,1,V.Cindroa,A.Gorišeka,T.Hemperekc,T.Kishishitac,2,G.Krambergera,
1
H.Krügerc,I.Mandic´a,M.Mikuža,b,N.Wermesc,M.Zavrtanika
0
2
aJožefStefanInstitute,

r Jamova39,Ljubljana,Slovenia
a
M bUniversityofLjubljana,FacultyofMathematicsandPhysics,
Jadranska19,Ljubljana,Slovenia
8
cUniversityofBonn,PhysikalischesInstitut,

] Nußallee12,Bonn,Germany
t
e
d E-mail: [email protected]
-
s
n
Abstract: InvestigationofHV-CMOSsensorsforuseasatrackingdetectorintheATLASexper-
i
.
s iment at the upgraded LHC (HL-LHC) has recently been an active field of research. A potential
c
i candidateforapixeldetectorbuiltinSilicon-On-Insulator(SOI)technologyhasalreadybeenchar-
s
y acterizedintermsofradiationhardnesstoTID(TotalIonizingDose)andchargecollectionaftera
h
moderateneutronirradiation. Inthisarticlewepresentresultsofanextensiveirradiationhardness
p
[ study with neutrons up to a fluence of 1×1016n /cm2. Charge collection in a passive pixelated
eq

2 structurewasmeasuredbyEdgeTransientCurrentTechnique(E-TCT).Theevolutionoftheeffec-
v tivespacechargeconcentrationwasfoundtobecompliantwiththeacceptorremovalmodel,with
4
2 the minimum of the space charge concentration being reached after 5×1014neq/cm2. An inves-
3 tigation of the in-pixel uniformity of the detector response revealed parasitic charge collection by
6
0 theepitaxialsiliconlayercharacteristicfortheSOIdesign. Theresultswerebackedbyanumerical
1. simulationofchargecollectioninanequivalentdetectorlayout.
0
7
1 Keywords: Charge induction, Radiation-hard detectors, Solid state detectors, Particle tracking
:
v detectors(Solid-statedetectors),DetectormodelingandsimulationsII
i
X
r ArXivePrint: 1701.06324
a
1Correspondingauthor.
2NowInstituteofParticleandNuclearScience,KEKHighEnergyAcceleratorResearchOrganization.
Contents
1 Introduction 1
2 Sampleandexperimentaltechnique 2
3 Evolutionofsensitivedepthwithirradiation 4
4 Effectiveacceptorremovalparameters 5
5 Uniformityofchargecollectionefficiency 7
6 SimulationofchargecollectionwithKDetSim 10
7 Conclusions 12
1 Introduction
Intensive investigations of possibility to produce particle tracking detectors for experiments at
upgraded LHC (HL-LHC) [1] using technology for commercial integrated CMOS circuits [2]
are currently ongoing at several research institutions throughout the world. The potential of the
CMOStechnologyoffersproductionoffullymonolithicparticledetectors[3],whichwouldenable
asmallerpixelsize,simplerassemblywithoutcostlyinterconnectionsbetweensensorsandreadout
electronics,andconsequentlylessmaterialinthetrackingvolume[4–7]. Manufacturingdetectors
in commercial fabrication plants should also result in faster production, larger number of vendors
andconsequentlylowercost.
Monolithic particle detectors have been used already for several years [8–10] but they were
notsuitableforuseintheATLASexperimentattheLHCbecauseoftheirrelativelyslowspeedand
insufficientradiationhardness,sincethedominantchargecollectionmechanisminthesedetectors
is diffusion [11]. Recently new technologies were developed permitting usage of high voltages
and full CMOS circuitry on the same chip. This opened the possibility of designing active pixel
detectorswithsufficientdepletedthicknessforfastcollectionofchargedriftingintheelectricfield
[12]. Severalotherdevelopmentsfollowed[13–15]commonlyreferredtoasdepletedCMOSpixels
[16]. Prototype test structures from various producers were tested recently and their performance
afterirradiationwasinvestigated[17–21]. OneveryinterestingtechnologicaloptionisSiliconOn
Insulator (SOI), where Buried OXide (BOX) isolates the bulk from the top layer where electronic
circuitry exploiting full CMOS possibilities can be implemented. Since BOX is protecting the
electronics, high voltage can be applied to form a significant depleted layer in the bulk for fast
chargecollection. AschematicdrawingofapixelinvestigatedinthisworkisshowninFigure1b.
DetectorprototypesdesignedbytheUniversityofBonnwereproducedin180nmSOICMOS
in the XFAB process [22]. XFAB uses the thick film SOI process where double well structures
–1–
(a) (b)
Figure 1: (a) Microscope image of the array 2A on the XTB02 chip with indicated pads for con-
necting the single and peripheral pixels and the beam direction in E-TCT. (b) Cross section of a
pixel in the structure 2A. Shown are the deep n-well collecting electrode with the corresponding
dimensions,theburiedoxide(BOX),theLOGICsectiononthep-typeepitaxiallayer(white)with
p-wellsfortheactivecircuitry,andtheP-FIELDimplantforinter-pixelinsulation. XTB02isread
outbyconnectingtheamplifierdirectlytothedeepn-well(figuretakenfrom[25]). Thecoordinate
systemthefiguresistheoneusedinE-TCTmeasurements. Beamdirectionisalongthez-axis.
shield FET transistors from the charge trapped in the BOX after irradiation. For this reason it
is immune to the so called back gate effect [23] and can, unlike other SOI detectors, withstand
high ionizing doses of over 700Mrad [24]. Samples produced by XFAB were also tested with
radioactive sources [25] and in a test-beam experiment [26] with good results before irradiation.
InthispaperresultsofEdgeTransientCurrentTechnique(E-TCT)measurementswithXFABtest
structuresirradiatedupto1×1016n /cm2 arepresented.
eq
2 Sampleandexperimentaltechnique
ThechipinvestigatedinthisworkiscalledXTB02. Thesubstrateisp-typesiliconwitharesistivity
of≈ 100Ω·cm. Thechipthicknessis700µm. TheXTB02chiphousesseveralteststructureswith
different design parameters [14, 25]. The focus of this study is the test structure 2A, a 4×4 pixel
matrixwithpixeldimensionof110×100µm2 showninFigure1a. Crosssectionofapixelcanbe
seeninFigure1b. Thechargeiscollectedbydeepn-wellelectrodeswithdimensions40×50µm2
placedundertheBOXlayer. AbovetheBOXthereisafewµmthickepitaxiallayerInthep-well
area above each deep n-well, called LOGIC, active CMOS circuitry can be placed. The epitaxial
layer also contains an implant structure called P-FIELD around individual pixels. Its aim is to
modifytheelectricalfieldbellowtheBOXtobreaktheconductivechannelformedinthebulkdue
toBOXspacechargeformedafterirradiation[24,25]. XTB02devicesarededicatedforstudying
the properties of the silicon bulk, so there is no readout circuitry actually implemented in the
LOGIC.Thechargecollectionelectrodeisconnecteddirectlytotheexternalamplifierasshownin
Figure1b. Thesubstrateisbiasedviatheoutermostguardring,ap-typeimplantringsurrounding
the test structures. LOGIC and P-FIELD can be connected via separate bond pads. The deep
–2–
Peltiercontroller
beamdiameter
opticalfiber&
focusingsystem insilicon C possibleshorting
widtFhWoHfMligh≈t1p0u�lsmesdet troppusdeloo elbayt y x HV(periHphVe(rcael)ntral) scoppe-ring
≈300ps,repetition z (GND)
rate500Hz
Beamdirection
ztable xtable
Bias-T
Substrate(p-type) Pixels(n-well)
1GHzBW
fastcurrent oscilloscope
amplifier x
Laser
detectorHV z
Laserdriver
y
triggerline
(a) (b)
Figure 2: Schematics of the E-TCT measurement setup (a) and the detector connection scheme
(b).
n-well of one of the four central pixels is routed to an independent bonding pad while the other
pixels are connected together and routed to another pad (seen in Figure 1a). This layout allows
measurementsoneitherasinglepixelorontheentirearrayof16pixels,dependingonwhichpad
is connected to the amplifier. In measurements described here the sample was always biased with
positive high voltageapplied to the deep n-wells whilethe substrate was kept atground potential.
Unless otherwise noted, LOGIC was biased to the potential of the deep n-well and P-FIELD was
leftfloating. Themaximalallowablebiasvoltagewas300V.
The schematics of the system for E-TCT measurements and the detector connection are pre-
sented in Figure 2. The sample is placed with its edge in a focused beam of a pulsing infrared
laser(λ = 1064nm–absorptionlengthinSi≈ 1mm,pulsewidth≤ 300ps,repetitionrate500Hz,
FWHM of beam profile in focal point ≤ 10µm). The light pulses generate electron-hole pairs
alongthebeampathinthesample. Theamountofgeneratedchargeisnotcalibrated,butthelaser
pulsepoweriskeptconstantto≈ 5%duringindividualmeasurement. Betweendifferentmeasure-
ments the laser power was varied to some extent and different laser diodes were used, therefore
the amount of injected charge cannot be directly compared between separate runs. The position
ofthesampleinthelaserbeamiscontrolledbyasetofpositioningstageswithsub-µmprecision,
whichallowbeampositioninginthe xy-planewithaprecisionbetterthanthebeamwidth. Dueto
thelargeabsorptionlengthoftheinfraredlightthedepositedchargeisroughlyconstantalongthe
z-direction, meaning that the measurement has no resolution in this direction. The charge carriers
generatedbythelaserpulsestarttodriftinelectricfield,inducinganelectriccurrentonthereadout
electrodes. This current is amplified by a 1GHz bandwidth current amplifier and digitized by a
1GHz bandwidth oscilloscope. Waveforms of 50 pulses are averaged by the scope and stored to
the computer. All measurements are carried out at room temperature, since there were no occur-
rences of the thermal runaway of the sensor at any irradiation stage. A detailed description of the
E-TCTmethodisavailablein[27]. MeasurementsreportedhereweremadewithanE-TCTsystem
producedbyParticulars[28].
–3–
The sample was irradiated with neutrons at Jožef Stefan Institute’s research reactor [29]. A
single sample was available for the study and it was therefore irradiated in steps. The cumulative
fluence after each step was 2×1014, 5×1014, 1×1015, 2×1015, 5×1015 and 1×1016n /cm2
eq
witha10%errormargin. Duringirradiationwithneutronssamplesalsoreceiveanionizingdoseof
about 1kGy (in SiO ) per each 1e14n /cm2. The dose was estimated using RadFETs, dedicated
2 eq
pMOS transistors, where TID is estimated from the change of threshold voltage [30]. After each
irradiationstepthesamplewasannealed(80minat60◦C)beforeE-TCTmeasurementsweremade.
Thesamplewaskeptinthefreezerat−17◦Canditwaswarmedtoroomtemperatureonlyforfew
hoursduringmeasurements.
3 Evolutionofsensitivedepthwithirradiation
The collected charge in E-TCT is defined as the integral of the induced current pulse over 25ns
afterthebeginningofthepulse. Thecollectedchargewasmeasuredwithasinglepixelconnected
to readout. Laser was directed to different depths y (see Figure 2 for definition of the coordinate
system), while the horizontal beam position x was fixed to the centre of a pixel. The measured
chargeasafunctionofthebeampositioniscalledchargecollectionprofile. Chargecollectionpro-
filesmeasuredbeforeirradiationandaftereachfluencestepareshowninFigure3. Measurements
weretakenatthehighestbiasvoltageof300V. Thecurveswerenormalizedtothesamemaximal
value. The transition at the rising edge of the charge collection profile (charge close to the chip
surface, at y ≈ 20µm) corresponds to the laser beam gradually entering the sample. The transi-
tion takes place over ≈ 10µm which corresponds to the laser beam diameter. Once the transition
at the chip surface is finished, the beam is for a while fully contained within the depletion zone.
Thisresultsinaplateauinthechargecollectionprofile. Theslopeofthetopoftheprofilemaybe
related to charge trapping as the electric field and thus the carrier velocity is falling with distance
from the surface. The transition on the falling edge of the charge collection profile is slower than
on the surface side — the collected charge typically reduces from 90% to 10% of the maximum
value over a depth of approximately 50µm, — which is due to the form of the depletion zone
deviating from the abrupt junction approximation. Figure 3 shows that the width of the charge
collectionprofileincreaseswithirradiation,reachingamaximumatafluenceof5×1014n /cm2.
eq
The width reduces at higher fluences, however it is still significant even at the highest fluence of
1×1016n /cm2. Thisbehaviourisconsistentwithradiationinducedremovalofinitialacceptors
eq
[21]. Uponfirstirradiationstepstheeffectivespacechargeconcentrationisreducingasmoreinitial
acceptors are removed than new are created by irradiation. After the removal process is finished
thenegativespacechargeconcentrationincreaseswithincreasingfluence.
Thewidthofthesensitiveregionwasquantifiedbyevaluatingthefullwidthathalfmaximum
of the charge collection profiles. Charge collection width (i.e. profile width) as a function of bias
voltage is shown in Figure 4 for all fluences. The width of the depletion region in a sensor with
(cid:113)
planarelectrodegeometryiscalculatedinthemodelofconstantspacechargeasW = 2(cid:15) V,
depl q0Neff
with (cid:15) the electric permittivity of the bulk, q0 the elementary charge, Neff the effective acceptor
concentration in the space charge region and V the applied bias voltage. With the rest of the
quantities known, one can use this dependence to extract the values of Neff for different fluences.
For the measurement before irradiation a fit could be made, yielding the value of Neff ≈ 1.3 ×
–4–
)
.
b
r 0.8
a
( Φ= 0
ge 0.7 Φ= 2e14 neq cm-2
ar 0.6 Φ= 5e14 neq cm-2
h Φ= 1e15 neq cm-2
c
0.5 Φ= 2e15 neq cm-2
Φ= 5e15 neq cm-2
0.4
Φ= 1e16 neq cm-2
0.3
V = 300 V
0.2 bias
0.1
0
0 50 100 150 200 250 300 350
y (µm)
Figure 3: Normalized charge collection profiles along the pixel centre at 300V bias for different
neutronfluences. Arrowsindicatethesequenceofirradiationsteps.
1014cm−3, consistent with the initial resistivity of the substrate. After irradiation the width of the
sensitive region grows faster than the square root function of voltage and the fit is therefore not
possible. InFigure4anunusualbehaviourat1×1015n /cm2andhigherfluencescanbeobserved
eq
at low bias voltages, where one can notice very low charge collection width for the initial bias
voltages followed by a relatively fast increase, which looks like a certain threshold voltage has to
bereachedbeforechargecollectionstarts. Themechanismbehindthisbehaviourisnotunderstood.
4 Effectiveacceptorremovalparameters
Irradiation of silicon with neutrons introduces defects into its crystal structure. Interaction of ini-
tial dopants with these defects may turn them electrically neutral so that they do not contribute
to the effective space charge any more [31, 32]. The radiation induced defects also form nega-
tively charged localized energy levels which contribute to the effective acceptor concentration in
thedepletedlayer. Ifthenumberofneutralizedinitialacceptorsislargerthanthenumberofnewly
created acceptors the effective space charge decreases, therefore increasing depleted depth. The
increase of depleted depth with irradiation was clearly observed as shown in Figure 3 and the be-
haviour can be explained by the removal of initial acceptors. This has a beneficial impact on the
signalafterirradiation. Asmentionedintheprevioussectionthedependenceofdepleteddepthon
biasvoltagedoesnotfollowthesimplesquarerootbehaviorafterirradiation. However,togetsome
comparison of acceptor removal behavior with other measurements [21, 33] the following proce-
durewasmade. Neff wascalculatedfromthewidthofthechargecollectionprofilesbyevaluating
(cid:113)
theformulaW = 2(cid:15) V atthebiasvoltageof300V. Asystematicuncertaintyonthevalueof
depl q0Neff
–5–
Width of charge collection region at 50% max
180
)
m
µ 160
( Φ= 0
M
140 Φ= 2e14 neq cm-2
H Φ= 5e14 neq cm-2
W
120 Φ= 1e15 neq cm-2
F Φ= 2e15 neq cm-2
100 Φ= 5e15 neq cm-2
Φ= 1e16 neq cm-2
80
60
40
20
0
0 50 100 150 200 250 300
Bias voltage (V)
Figure4: Widthofchargecollectionprofilesvs. biasvoltagefordifferentneutronfluences. Arrows
indicatethesequenceofirradiationsteps.
Neff wasestimatedfromthespreadofthevalueswhentakingthewidthofchargecollectionprofiles
at (50±10)% of the maximum. The measured values of Neff at different fluences calculated at a
biasvoltageof300VareshowninFigure5. TheevolutionofNeff asafunctionoffluenceisgiven
by[17,21,32]:
Neff = Neff,0−NA(1−exp(−c·Φeq))+gCΦeq, (4.1)
where Neff,0 denotestheinitialeffectiveacceptorconcentrationofthesubstrate, NA theconcentra-
tionoftheeffectivelyremovedacceptors,ctheacceptorremovalconstant,Φ the1MeVneutron
eq
equivalent fluence and g the generation rate of stable deep acceptors [34]. The measured data
C
werefitwithfunction4.1with Neff,0, NA,candgC asfreeparameters. Resultsofthefitareshown
inFigure5. TheratioNA/Neff,0 = 1indicatesacompleteinitialacceptorremoval. Thevalueofthe
parameter c = 1.1×10−14cm2 is by a factor of 2–3 larger than the value measured with a similar
method published in [21]. But the initial resistivities of samples measured in [21] were 10 and
20Ω·cmsothevaluec = 1.1×10−14cm2 estimatedinthisworkisconsistentwiththeobservation
that the acceptor removal constant is smaller in silicon with a lower initial resistivity [35]. The
valueofcisalsoreflectedinthefluenceatwhichthemaximalchargecollectionwidthisreached.
IntheXTB02samplethemaximumisreachedat∼ 5×1014n /cm2whereasforsamplesin[21]it
eq
occursat∼ 2×1015n /cm2. Thevalueoftheparameterg = 0.036cm−1 islargerthanthevalue
eq C
0.02cm−1 usually observed for neutron irradiated samples [21]. But this is not surprising since
√
we know that the depleted depth does not follow the V behaviour, pointing to an inconsistency
withtheuniformspacechargeconcentrationandabruptjunctionapproximation. ByevaluatingNeff
frommeasurementsatabiasvoltageof210Vweforexampleobtainthevalueofg = 0.08cm−1,
C
whereas at 150V the value is g = 0.2cm−1. However, the other three fit parameters are stable
C
within 10% of the value at 300V, rising the confidence that the value of the acceptor removal
constantextractedfromthefitinFigure5isagoodestimateforthissubstratematerial.
–6–
)
3
-m
4.5
c
4
110 4 c ^2 / NDF = 0.7
(
eff3.5
N
3
2.5
N = 1.3e14 cm-3
eff,0
2
N
A = 1.0
1.5 N
eff,0
1 c = 1.1e14 cm-2
g = 0.036 cm-1
0.5 C
0
0 20 40 60 80 100 120
F (1014 neq cm-2)
eq
Figure 5: Evolution of Neff of the substrate with fluence. The fit of the data with the function 4.1
andtheextractedfitparametersarealsoshown.
5 Uniformityofchargecollectionefficiency
An important requirement for a pixel detector is the response uniformity within a pixel. The uni-
formity can be studied with a two dimensional E-TCT scan, where the position of the laser beam
isvariedalongthesampledepth(coordinatey)aswellasalongtheedgeofthesample(coordinate
x, see Figure 2). Figure 6 shows collected charge as a function of coordinates x and y after a flu-
enceof2×1014n /cm2. Inthemeasurementallsixteenpixelsoftheteststructureareconnected
eq
together and read out simultaneously. Each region with a high collected charge corresponds to a
columnoffourpixelsalongthez-axis,whichcannotbedistinguishedfromeachotherinanE-TCT
measurement. Distinct regions with no collected charge can be seen between the columns. Size
and spacing of these efficiency gaps roughly coincides with the spacing between deep n-wells of
neighbouring pixels. As can be seen in Figure 1b, the n-well does not extend over the entire area
ofthepixel.
Figures 6a and 6b show the induced current pulses from an efficient and an inefficient region
at y = 50µm. While the former is a unipolar pulse with a non-zero integral (with superimposed
oscillationduetonon-matchingimpedancesofthecableandtheamplifier),thelatterisbipolarwith
a vanishing integral. A similar magnitude of both pulses confirms a uniform strength of electric
field at a given sample depth. According to the Shockley-Ramo theorem of signal formation, a
bipolar pulse with zero integral is observed when the drift path of charge carriers does not end
on a readout electrode [36, 37]. The pulse 6b therefore indicates a presence of a parasitic charge
collectingelectrodewhichisnotconnectedtoreadout. Theidentityofthiselectrodecanbededuced
from Figure 1b. It can be seen that there are two biased structures present on the top of the pixel
— the deep n-well, which is read out, and LOGIC, which is biased separately to the potential of
the deep n-well but does not have a low impedance connection to the readout. Although LOGIC
ispositionedabovetheinsulatingBOXlayer,itstillinfluencestheelectricfieldinthebulk. Ifthis
–7–
m) b)
µ140 ar
y ( 0.4 e (
120 g
ar
h
100 0.3 C
80
0.2
60
a b
40 0.1
20
0
0
0 50 100 150 200 250 300 350 400 450 500
x (µm)
b.) b.)
ar 400 ar 400
I ( I (
200 200
0 0
- 200 - 200
- 400 - 400
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (ns) t (ns)
(a)Highefficiency (b)Lowefficiency
Figure 6: Two dimensional charge collection profile in a detector irradiated to 2×1014n /cm2
eq
at V = 200V with induced pulses in regions with high and low charge collection efficiency
bias
respectively.
effect is strong the field lines will be roughly perpendicular to the chip surface. When charge is
not injected directly underneath the collecting n-wells, its drift path will end on the BOX rather
than on the deep n-well, resulting in a low collected charge (Figure 9). LOGIC therefore acts as
an AC coupled parasitic electrode. This hypothesis was confirmed by switching the deep n-well
and LOGIC connections, so that LOGIC was read out. This yielded a complementary picture —
low efficiency for charge injection underneath the deep n-wells and high efficiency for injection
underneathLOGIC.
TheelectronsdriftingtowardsLOGICeventuallystoponthenon-conductiveBOXlayer. Since
charge cannot accumulate on BOX indefinitely, it has to be removed laterally towards the deep n-
wells, which are the only conductive connection through the BOX. The exact mechanism of the
currentflowontheBOXsurfacewasnotaddressedinthiswork. Toinvestigatethelateralcurrent
assumption, the time scale of the E-TCT measurements was prolonged from 25ns to 1µs, while
stillusingthestandardbiasingscheme(Figure7). Thelongtimescalepulsesfromanefficientand
aninefficientregionoftheteststructurearecomparedinFigure7a. Itcanbeobservedthatafterthe
initial chargecollection inthe first25nsis finished, theamplitude ofthe pulsefrom aninefficient
region is higher than from an efficient region. Both currents appear due to the charge collected
on the BOX slowly discharging to the n-wells. Even in an efficient region some sections are not
–8–
IInndduucceedd ppuullsseess 11 mm ss ttiimmee ssccaallee Collected charge vs. integration time
b.) 20 b.) 200
ar ar
I ( 0 ge ( 0
ar- 200
- 20 ch
- 400 xx == 115500 mmmm,, yy == 110000 mmmm -- ""iinneeffffiicciieenntt""
- 40 xx == 115500 mm mm,, yy == 110000 mm mm - 600 xx == 220000 mmmm,, yy == 110000 mmmm -- ""eeffffiicciieenntt""
xx == 220000 mm mm,, yy == 110000 mm mm DDiiffffeerreennccee
- 60 - 800
- 80 - 1000
0 200 400 600 800 0 200 400 600 800
t (ns) integration time (ns)
(a) (b)
IInntteeggrraattiioonn ttiimmee 2255 nnss IInntteeggrraattiioonn ttiimmee 880000 nnss
my (m)140 700e (arb) my (m)140 1500e (arb)
120 600harg 120 1000harg
100 500C 100 C
80 400 80
500
60 300 60
40 200 40 0
20 100 20
0 0 0 - 500
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
x (m m) x (m m)
(c)Integrationtime25ns (d)Integrationtime800ns
Figure 7: (a) Induced electrical pulses from different efficiency regions on a time scale of 1µs.
Thepulsesareverticallytruncatedontheimagearoundt = 0tostressthedifferenceatlargert. (b)
Dependence of the pulse integral (collected charge) from integration time. (c) Charge collection
profilewith25nsintegrationtime. (d)Chargecollectionprofilewith800nsintegrationtime.
coveredbyn-wellsduetothesegmentationalongz-direction. Thecumulativetimeintegralofboth
pulses is shown in Figure 7b. Two characteristic regimes can be observed. For a short integration
time of 25ns about 2 of the maximum charge is collected from an efficient region, whereas the
3
chargecollectedfromaninefficientregioniszero,asalreadyobservedinFigure6. However,foran
integrationtimeofseveralhundrednsthesmalldifferenceinpulseamplitudebecomesimportant.
At600−800nsintegrationtimethechargecollectedfrombothregionsbecomesthesame. Thetwo
dimensional charge collection profiles corresponding to 25ns and 800ns charge integration times
are shown in Figures 7c and 7d. While efficiency gaps are present for the 25ns integration time,
the collected charge is relatively uniform for the 800ns integration. These measurements confirm
that a parasitic charge collection occurs via the AC coupled LOGIC electrode. The part of the
chargewhichendsitsdriftontheBOXinterfaceiseffectivelylostforthechargecollection, since
themeasuredcollectiontimefromtheBOXisoftheorderofmagnitudeof1µs.
A further test of charge collection uniformity was carried out by biasing LOGIC to differ-
ent bias voltages. In each measurement the deep n-well was always biased to +200V and the
outerguardringwasgrounded. LOGICwasthenbiasedtoeither+200V(standardconfiguration),
+150Vorleftfloating. Two-dimensionalchargecollectionprofilesfordifferentbiasconfigurations
were then recorded and are shown in Figure 8. All tests were done with a sample irradiated to a
–9–