SIMULATIONS OF THE SILICON MICROSTRIP DETECTOR - SCIPP
Transcript of SIMULATIONS OF THE SILICON MICROSTRIP DETECTOR - SCIPP
SCIPP-00/48
SIMULATIONS OF THE SILICON MICROSTRIPDETECTOR FOR THE GLAST EXPERIMENT.
ClaudioPiemonte
INFN, Sezione di Trieste, Via A. Valerio 2, I-34127 Trieste, Italy
Abstract
Devicesimulationsof thesiliconmicrostripdetectorsthatwill equiptheTrackingSystemof theGLAST experimenthave beenperformed.Theinfluenceon theelectriccharacter-isticsof bothstrip pitch andwidth have beenevaluated.First, a DC calculationis doneto predicttherisk of breakdown within theworkingbiasconditions.Then,thebackplaneandinterstripcapacitancesareestimatedby meansof anAC analysis.
1 Introduction.
TheGLASTtrackingsystemusessiliconmicrostripdetectorsto tracktheelectron-positron
pairsresultingfrom gamma-rayconversionin thin leadfoils [1,2]. It consistsof a square�����arrayof nearlyidenticaltowermodules.Eachtowerhasascintillatorvetocounteron
thetop,followedby a ��� -layersilicon-striptrackerand,finally, acesiumiodidecalorime-
ter. Eachtrackinglayerusestwo planesof single-sidedsilicon strip detectorswith strip
orientedat � � degreeswith respectto eachother. Thedetectors,cut from � -inch wafers,
have a size of about � � �"!$#&% and are wire bondedinto ”ladders” ' �(!)# long. Four
of these”ladders”composea detectorplane.Whenoperatinga system,with sucha high
channelcount,in space,two limitationsappear: theavailability of powerfor theelectron-
icsandthedifficulty of dissipatingtheresultingheat.In orderto satisfytheseconstraints,
both the numberof channelsandthe readoutelectronicspower consumptionshouldbe
minimised.
Thebaselinedetectoris realisedon * -typewaferhaving a resistivity of �,+.-0/1!$# ,
anda thicknessof� ����23# . The readoutelectronicsis AC coupledto the 465 implants,
whicharebiasedby meansof integratedpolysiliconresistors.Thisdesignforeseesastrip
pitchof 7�� 892:# , andanimplantwidth of ; �92:# . Themetaloverhangstheoxide� 23# be-
yondtheimplantin orderto reducetherisk of breakdown. Theresultingnumberof strips
perdetectoris��� 8 , thus,usinga � � -channelreadoutelectronics,< chipsarenecessaryto
collectall theinformation.Recently, theconstraintson thepower consumptionhasbeen
revised,forcingareductionof thenumberof chips.It resultsin areductionof thenumber
of stripsperdetector.
In thispaper, theelectricalperformancesof thedetectorareevaluated,varyingboth
the strip pitch andwidth. First, a DC analysisis doneto estimatethe electric field in
thecritical region. Then,anAC analysisis appliedin orderto predictthe interstripand
backplanecapacitances.
2 Simulation technique.
Thesimulationis carriedout with ATLAS Device SimulationSoftwareproducedby Sil-
vaco[3]. A 2D approachis adoptedsince,for thesimulatedvalues,thepartialderivative
alongthe strip length is negligible. The lateralboundaryconditionsarea crucial point
whenselectingthe lengthof thestructureto besimulated.Mainly, therecanbetwo ap-
proaches.Oneis to build a very long cell in orderto keepthe lateralboundariesdistant
from the region of interest.Thedrawbackis a high numberof meshpointsand,conse-
quently, a long computationaltime. Theotherapproachis to choosea lateralboundary
2
1/2 implant widthfield plate length
pitch
depth
n-bulk
metal metalcathodes
p+ implant
n+ implant
p+ implant
anode
oxide
Figure1: Basiccell usedfor thesimulations.
whereonly theverticalcomponentof theelectricfield is present.As a result,we obtain
a very narrow cell having a lengthequalto thestrip pitch (fig. 1). This is optimalfor the
electricfield analysisbecausea high numberof pointscanbeusedto refinethemeshat
the edgesof the implantswithout exceedingthe computationaltime. Nevertheless,this
structureis not suitablefor the interstrip capacitanceevaluation,becausethe coupling
betweennon-adjacentstripsis neglected.As a matterof fact, this calculationrequiresa
lowernumberof meshpointswith respectto theDC analysis,thuswecanjoin morecells
without increasingthecomputationalcost.
TheDC solutionis obtainedsolvingbothPoissonandcarriercontinuityequations.
Thisanalysisallowsfor theevaluationof thebreakdown risk. It is known thatbreakdown
occurswhenfreecarriersareacceleratedup to a point wherethey have sufficient energy
to generatemorefree carrierswhencolliding with the atomsof the crystal. In orderto
acquiresufficient energy two conditionsmustbe met: the electricfield is enoughhigh
(morethanabout ' � ���>=@?"A�!)# ), andthe meanfreepathof the carriersis long enough
to allow accelerationto a sufficienthigh velocity. Only whenthegenerationrateof these
free carriersis sufficiently high this processleadsto avalanchebreakdown. In orderto
take into accounttheseparameters,animpactionizationmodelis includedin thesimula-
tion.
Beforestartingthe simulations,two parametersmustbe defined:the oxide charge den-
sity andthe 465 implantprofile. The former, for thesake of simplicity, is consideredall
localisedat the B3C@DEB�CGF % interface,while agaussianshapeis chosenfor thelatter.
It is worth noting that, in thesimulations,the 4 5 implanthasbeendirectly connectedto
3
themetal.Thisdoesnot introduceanappreciableerrorsincetheflat bandvoltage,related
to thethin AC couplingoxidelayer, is verysmall.
The AC analysisis performedjoining two elementarycells. It allows for the re-
Cback
1 2 3
bulk
oxide Cac
C"int
C’int
1 2 3
bulk
oxide Cac
Backplane capacitance Interstrip capacitance
Figure2: Structureadoptedfor theAC analysis.
constructionof thestripcapacitance,oneof theparametersdeterminingthenoiselevel of
the readoutelectronics.It is givenby the sumof two capacitances:one( HJILKNMPO ) derives
form thecouplingbetweenthe implantandthebackplanelayer, while theother( HJQSRUT ) is
dueto thecouplingbetweentheconsideredstrip andall theothers.We do not take into
accounta third tiny contributiongivenby theAC couplingcapacitor, sincethesimulation
is performedona structurein which themetalis connecteddirectly to theimplant.
Theprocedureadoptedfor thesimulationis thefollowing: first, thedevice is broughtat
a biasvoltagerising thepotentialof theanode.Then,in orderto evaluateHJILKNMPO , a small
signal(with a frequency of �UVXWZY ) is appliedto theanode,andthecapacitancebetween
thecentralstrip (strip n. 7 ) andtheanodeitself is checked. As far as HJQ[RUT is concerned,
a small AC signalis appliedto the cathoden. ' andthe capacitancesH]\Q[R�T and H]\ \QSRUT are
measured.In this way thecouplingto moredistantneighboursis neglected,andthetotal
capacitanceis givenby:
HJT�^_T�`aHJILKbMPO�cdHJQ[RUT9`eHJILKNMfO"cd7 �hg H \Q[R�T chH \ \Q[R�T_i (1)
Thecalculationof thebackplanecapacitanceasa functionof thebiasvoltagepermitsthe
reconstructionof thefull depletionvoltageby meansof the �UA�H]%ILKNMfO curve.
Thetablebelow shows thepitch/widthvaluesusedin thesimulations.
4
width pitch width/pitch23# 23#50 175 0.28650 208 0.24050 235 0.21350 265 0.18960 175 0.34360 208 0.28960 235 0.25560 265 0.226
Table1: Pitchandwidth valuesusedin thesimulations.
3 DC analysis.
First, a simulationis donein orderto estimatethe impactof the metaloverhangon the
electricfield intensityin thevicinity of thejunction. Fig. 3 shows themapof theelectric
203j
204j
205j
206j
207j
208j
209j
210j
211
-2
-1
0k1
2
3l4
Microns
Mic
rons
metal
n-type bulk
oxide
p+ implant
E Field (V/cm)
03.74e+047.48e+041.12e+051.5e+051.87e+052.24e+052.62e+052.99e+053.37e+053.74e+05
Figure3: Map of the electricfield strengthfor the 7 '�;(A ;�� structure(the biasvoltageis7�� �(? ).
field intensityfor the 7�'�;�A ;�� (pitch/width)structurewith a metaloverhangof� 23# , and
anoxidecharge mn^_o of � � �p�rq$qtsuA�!$#&% . In thesilicon, thehighestelectricfield is located
at the implant edgeand in the proximity of the metalborder. Without field-plates,the
5
lines of the electric field would be all accumulatedat the edgeof the implant. Fig. 4
shows theprofile of theelectricfield strength,takenalonga line crossingtheedgeof the
465 implant, for thesamestructure:gwv i with, and
gwx i without field-plate.Theusefulness
0.2 0.4 0.6 0.8 1 1.2
0
50000
100000
150000
200000
250000
300000
350000
Line crossing the junction edge (micron)
Elec
tric fie
ld (V
/cm)
(a)
(b)
Figure4: Electricfield profile alonga cut-linecrossingthe implantedgefor the 7�' ;(A ;��structure:
gwv i with, andgfx i withoutfield-plate.
of the field-plateis clearly visible: the peakheight is reducedby two times. Further-
more,thedevice without field-platepresentsa peakhigherthanthecritical value,hence
potentiallydangerous.Fig. 5, showing the electroncurrentdensity, demonstratesthat
thereis generationof free carriersin that region, but not high enoughto determinean
avalanchebreakdown. Accordingto thesimulations,it occursat a biasvoltageof about
��� �(? . Plainly, the calculationdoesnot take into accountthe presenceof local defects
whichcouldgenerateenoughfreecarriersto reachthebreakdown ata lowervoltagewith
respectto the ideal case.Thus,the useof the metaloverhangis advisablein a detector
with suchcharacteristics.
Now, let usanalysetheinfluenceof thepitch/widthsizeontheelectricfieldstrength.
Fig. 6 and7 display the maximumvalueof the electricfield asa function of the bias
voltagefor thestructureslisted in Table1. As expected,it dependsbothon the implant
pitchandwidth. Anyway, thepeakintensityis below thecritical valuefor all thestructures
up to abiasvoltageof '�� �"? .
6
208y
209y
210 211y
-2
-1
0
1
2y
Micronsz
Mic
rons
e- Current Magnitude (A/cm2)
0{1.73e-103.47e-10|5.2e-10}6.94e-10~8.67e-10�1.04e-091.21e-091.39e-091.56e-091.73e-09
Figure5: Theelectroncurrentdensityplot evidencesa carriergenerationat theedgeoftheimplant.
1.0e+05
1.5e+05
2.0e+05
2.5e+05
3.0e+05
100 150 200 250 300
Ele
ctr
ic f
ield
(V
/cm
)
�
Bias Voltage (V)
implant width = 50µm
full depletion pitch = 175µmpitch = 208µmpitch = 235µmpitch = 265µm
1.0e+05
1.5e+05
2.0e+05
2.5e+05
3.0e+05
100 150 200 250 300
Ele
ctr
ic f
ield
(V
/cm
)
�
Bias Voltage (V)
implant width = 60µm
full depletionpitch = 175µmpitch = 208µmpitch = 235µmpitch = 265µm
Figure6: Magnitudeof theelectricfieldpeakasa functionof theappliedvoltage,for structureshaving an implant ;���23#wide.
Figure7: Magnitudeof theelectricfieldpeakasafunctionof theappliedvoltage,for structureshaving an implant ����23#wide.
7
1.2e+05
1.4e+05
1.6e+05
1.8e+05
2.0e+05
2.2e+05
2.4e+05
0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34
Ele
ctric
fiel
d (V
)
w/p
Bias voltage = 200V
implant width = 50µmimplant width = 60µm
Figure8: Electricfield peakasa functionof thewidth/pitchratio.
Also theoxidechargedensityplaysanimportantrole: thehighertheconcentration
thestrongertheelectricfield. As anexample,fig. 9 showstheeffectof varyingtheoxide
1.0e+05
1.5e+05
2.0e+05
2.5e+05
3.0e+05
3.5e+05
100 150 200 250 300
Ele
ctric
fiel
d (V
/cm
)
�
Bias Voltage (V)
235/50 structure
Qox = 1*1011 q/cm2
Qox = 2*1011 q/cm2
Figure9: Electricfield peakasa functionof thebiasvoltagefor differentoxidecharges.
charge densityfrom � � ���(q$q to 7 � ���(q$q�suA !)#&% for the 7�'�;�A ;�� structure. The higher
concentrationbrings the electricfield over the critical valuewhenwe apply more than
7�; �(? .
8
4 AC analysis.
In orderto verify the quality of the simulations,a comparisonbetweenthe calculations
andexperimentaldatais performed. The dataare taken from [4], wherea carefulex-
perimentalstudyregardingthecapacitancesis done.Oneof thetestedstructuresis very
similar to our detector, having a thicknessof� ����23# anda bulk resistivity of �"+�- . The
siliconis ��������� -type,leadingto ahigheroxidechargeconcentrationwith respectto the
��������� crystaltypeforeseenfor GLAST. Threestructureswith a constantwidth/pitch
ratio, equalto �(���p; , aremeasured.Figures10, and11 show a comparisonbetweenex-
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 50 100 150 200
Cback (
pF
/cm
)
pitch (µm)
Backplane capacitance
Bias Voltage = 500VSimulation
Experimental data
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
0 50 100 150 200
Cin
t (p
F/c
m)
�
pitch (µm)
Interstrip capacitance
Bias Voltage = 500VQox=1*1011 q/cm2
Qox=2*1011 q/cm2
Qox=3*1011 q/cm2
Experimental data
Figure10: Backplanecapacitanceof thetestdetector.
Figure11: Interstripcapacitanceof thetestdetector.
perimentaldataandsimulationresultsfor bothinterstripandbackplanecapacitance.The
agreementis foundto bequitegood,thusvalidatingthesimulationprocedure.It is worth
noting that the valueof the interstripcapacitancedependson the electronaccumulation
layer underthe B3C�DXB3C�F % surface,that in its turn dependson many parameter. In par-
ticular, suchparametersare: the fixed oxide charge density, the bias voltage,and the
environmentalhumidity. It is quite difficult to estimatenumericallythe influenceof the
lastparameter. In aqualitativeway, negativecharges(polarisedwatermolecules)arecol-
lectedon theuncoveredoxidesurface,partially compensatingthepositive oxidecharge
[5]. In our casetheeffective valueof m�^_o&`�7 � ���rq$qts�A�!$#&% is found to fit in a proper
way theexperimentaldata.
Let’snow simulatethecapacitancesof thestructureslistedin Table1.
9
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
180 200 220 240 260
Cba
ck (p
F/cm
)
pitch (µm)
GLAST detector - Backplane capacitance
Bias voltage = 250V
implant width=50µmimplant width=60µm
Figure12: Backplanecapacitanceasa functionof thepitch.
0
1
2
3
4
5
6
0 50 100 150 200 250 300
1/C
back
2 (cm� 2
/pF2 )
Ubias (V)
175/50208/50235/50265/50
Figure13: �UA H�%ILKbMPO asafunctionof thebiasvoltagefor structureshaving animplant ;���23#wide.
10
Fig. 12 shows the backplanecapacitancefor a biasvoltageof 7 ;��(? . It doesnot
changesubstantiallyvarying the width from ; � to ����23# , while, asexpected,increases
with thepitch. HJILKNMfO givesinformationconcerningthedepletionof thedetector. Fig. 13 is
a plot of ��A�H�%ILKbMPO asa functionof theappliedvoltagefor thestructureshaving animplant
;���23# wide. For eachcurve, the changeof the slopeindicatesthat the full depletion
voltageof thedetectoris reached.Increasingthepitch from ��< ; to 7�� ;92:# it risesfrom
��' � to �p<��(? . Thesevaluesdependslightly ontheoxidechargeconcentration;in thiscase
weusedmn^_o�`�� � ���(q$qGsuA�!$#&% .Theinterstripcapacitanceis plottedin figure14. As onecansee,thevaluedepends
consistentlyonboththeimplantwidth andpitch.
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
180 200 220 240 260
Cin
t (pF
/cm
)
�
pitch (µm)
GLAST detector - Interstrip capacitance
265/50
235/50
208/50
175/50
Bias voltage = 250V
implant width = 50µmimplant width = 60µm
Figure14: Interstripcapacitanceasa functionof thepitch usinga biasvoltageof 7�;��(?andanoxidecharge mn^�o�`�� � �p� q$q suA !)# % .
The total capacitanceasa function of the pitch is presentedin fig. 15. As it is
evidencedexperimentallyin [4], this valuedependsonly on the ratio betweenthe two
quantities,on thecontraryto theelectricfield behaviour (fig. 8). Thecapacitancegrows
linearlywith the ��A�4 ratio,asdemonstratedin fig. 16.
11
1
1.1
1.2
1.3
1.4
1.5
1.6
180 200 220 240 260
Cto
t (pF
/cm
)
�
pitch (µm)
GLAST detector - Total capacitance
implant width = 50µmimplant width = 60µm
Figure15: Total capacitanceasa functionof thepitch.
1
1.1
1.2
1.3
1.4
1.5
1.6
0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34
Cto
t (pF
/cm
)
�
w/p
GLAST detector - Total capacitance
f(w/p) = 0.89 + 1.3*(w/p)
implant width = 50µmimplant width = 60µm
Figure16: Total capacitanceasa functionof thewidth/pitchratio.
12
5 Conclusions.
The resultsof the capacitancestudy indicatethat the best-performingstructure,among
thoseanalysed,hasa pitch of 7���;�23# and an implant ;���23# wide. The DC analysis
performedfor this structuregivesa maximumfield underthecritical value,beingabout
7(��< � ��� = ?�A�!$# at a biasvoltageof '�����? . Still, having sucha high valueof theelectric
field, weconsiderapitchof 7 ��;�23# asa limit for animplantwidth of ;���23# . For a larger
pitch it is necessaryto increasealsotheimplantwidth.
6 Acknowledgements
The authorwould like to thank A. Vacchi for the continuousencouragementandsup-
port. Many thanksalso to V. Bonvicini, L. Bosisio,andA. Rashevsky for the fruitful
discussions.
References
[1] P. Michelsonet al., “GLAST LAT”, Responseto AO 99-OSS-03,StanfordUniv.,
(1999).
[2] E.D. Bloom et al., “Proposalfor GLAST”, SLAC-R-522(1998).
[3] SILVACO International,ATLAS User’s manual(1998).
[4] CMS Collaboration,Addendumto theCMS tracker TDR, CERN/LHCC2000-016
(2000).
[5] A. Bischoff et al., Nucl. Instr. andMeth.A326, 27-37(1993).
13