Sei sulla pagina 1di 10

619

IEEE T R A " r m N S ON NUCLEAR SCIENCE, VOL. 39, NO. 4,1992

Large Area Cylindrical Silicon Drift Detector

W. Chen, H. Kraner, Z. Li, P. Rehak


Brookhaven National Laboratory, Upton NY 11973

E. Gatti, A. Longoni, M. Sampietro


Politecnico di Milano, 32 Piaeza Leonard0 da Vinci, 20133 Milano, Italy*

P. Holl, J. Kemmer
TU Miinchen, 8046 Garching and
KETEK GmbH, Hauptstrafle 41d, 8048 Haimhausen, Germany

U. Faschingbauer, B. Schmitt, A. Worner, J. P. Wurm


Max-Planck-Institut for Nuclear Physics, D-6900 Heidelberg 1, Germany

Abstract I. INTRODUCTION
An advanced silicon drift detector, a large area cylin-
The semiconductor drift detector [I] uses the trans-
drical drift detector, was designed, produced, tested and
port of electrons in a direction parallel to the large de-
installed in the NA45 experiment. The active area of the
tector surface. The direction of the transport is imposed
detector is practically the total area of a 3 inch diame-
by means of rectifying junctions at different potentials on
ter wafer. Signal electrons created in the silicon detector
both sides of the wafer. The shape of the junctions as
by fast charged particles drift radially outside toward an
well as the applied potential on them can vary to achieve
array of 360 anodes located on the periphery of the de-
different drift geometries. Up to now only linear geome-
tector. The drift time measures the radial coordinate
tries and cylindrical geometries with the drift of electrons
of the particle's intersection; the charge sharing between
toward the center were realized [2-31.
anodes measures the azimuthal coordinate.
An ideal detector for the experiment NA45 at the
The detector provides unambiguous pairs of r,d co- CERN SPS [4] should measure x and y coordinates
ordinates for events with multiplicities up to several of several hundred charged particles in an unambiguous
hundred. Its use in the experiment aims at a position way. The required resolution is about 20 pm in both co-
resolution of 20 pm (rms) in each direction giving about ordinates within a circular area with a radius of 3 cm.
2 loe two-dimensional elements. A cylindrical silicon drift detector can fulfill those re-
There is a small hole in the center of the detector quirements of the experiment. In reality the cylindrical
to allow the passage of the noninteracting particle beam. geometry of the detector is very close to the ideal geome-
The longest drift distance is about 3 cm. The nomi- try for most of fixed target experiments with unpolarieed
nal value of the drift field is 500 V/cm resulting in a beams.
maximum drift time of 4 ps. Fig. 1 shows the detector and its supporting assem-
bly. The active part is practically the entire 3 inch silicon
wafer with a small hole in the center. Signal electrons
produced by fast particles drift inside the silicon radi-
This manuscript has been authorized under contract
number DE-AC02-76CH00016 with the U.S. Department ally towards the outside edge of the detector. Close to
of Energy. Accordingly, the U. S. Government retains the outside edge there is an array of 360 anodes a t the
same radius each collecting electrons from 1 deg seg-
a non-exclusive, royalty-free licence to publish or repro-
ments. Anodes are bonded to traces on a ceramic board
duce the published form of this contribution, or allow
to which the silicon wafer is glued. The traces on the
others to do so, for U.S. Government purposes.
ceramic board contact traces on an FR-4 (G10) mother
(*) This research is also supported by the Italian INFN, board leading to the inputs of charge sensitive pream-
MURST and CNR. plifiers. The contacts are accomplished by a layer of
elastomer interconnectors [5]. The mechanical forces for
all connectors are provided by 12 screws. The ceramic

0018-9499/92$03.00 0 1992 IEEE


The fourth section shows some results obtained from
laboratory tests and from the run of the detector at the
NA45 experiment at CERN. Results from the experiment
are very preliminary and more beam results will be
published later. A list of conclusions closes the paper.

11. TRANSPORT OF ELECTRONS

A. Theoretical considerations
During the drift time the electrons diffiise in the drift
direction (r) and in a direction perpendicular to the drift
(4). The diffusion in the drift direction is not influenced
by the drift field. However, in the 4 direction the effect
of the diffusion is combined with the defocusing caused
by the radial divergence of the drift field.
The general problem of the electron transport in a
radial field can be simplified in our case of a constant
radial drift velocity v. We treat the motion of electrons
in a coordinate system having the origin a t the center
of gravity of the entire electron cloud. (Moving away
Figure 1: Photograph of the n-side of the cylindrical from the center with a constant radial velocity v . ) Let
detector within its assembly. The detector is the silicon us define direction X = ~4 perpendicular to the radius
wafer of 3 inch diameter with a small hole in the center direction at the center of the electron cloud. The mean
to allow the passage of the noninteracting beam. Anodes radial velocity of an electron a t X has a component of
are connected to the outside preamplifiers. A mother velocity v x = v $ in the X direction. The net flux of
board for the preamplifiers fills the outside radius. (160 electrons in the X direction thus can be written as:
") X an
j x = nu- - D -
T ax
board holding the detector can be thus quickly discon- where n(t,X ) is the electron density and D is the diffusion
nected from the FR-4 mother board and heated to as constant of electrons in silicon. Combined with the
high a temperature as 3 O O O C for a possible annealing of continuity equation we obtain the diffusion equation for
the radiation damage. electrons in X direction
The preamplifiers are of a hybrid type [6] with 3
channels in each unit. The photograph shows the sockets
into which the preamplifier units are plugged. The
outside diameter of the assembly is 160 mm.
Each preamplifier drives about 5 m long 50S2 cable where TO is the initial radial coordinate of the electron
with a gaussian shaper [6] a t the receiving end. Outputs cloud.
of the shapers are sampled at 50 MHz by fast 6 bit The solution of the above differential equation is
nonlinear flash ADCs [7]. a Gaussian distribution of n(t,X) with respect to the
The present paper treats mainly the design aspects coordinate x
of the detector and is organized as follows. The second n ( t , X ) = -e
1 -q
2~
&U
section studies the transport of electrons in a cylindrical
geometry. The consequences of the analysis of the where the time dependence enters through U
transport for the pitch of anodes and for the number
of read out channels is shown. The third section treats
the flow of the leakage current generated at the Si -
Si02 interface into a sink anode. The diversion of the
leakage current away from the signal anodes improves The second term is the square of the initial spread of
the performance of' the detector. Moreover, the increase the electron cloud multiplied by the square of the
of the current generation on the Si - Si02 interface due radii ratio. This is the simple geometrical projection of
to anticipated radiation damage effects less the detector. the initial sigma to the final radius. The first term is
The detector is thus expected to be radiation harder than due to the diffusion of electrons during the drift. The
one would first estimate. well known expression for the diffusion in a linear field
62 1

0 I I

D - between modes
o - 10% of anode pitch

X - 40% of anode pitch

9
0
~~
200.0 I
400.0 I
600.0 I
800.0 1000.0
I
I I I
30.0 40U.O 600.0 800.0 1000.0
Anode pitch Iuml Anode pitch [uml

Figure 2: Optimal resolution of an ideal cylindrical Figure 3: Optimal resolution in a linear geometry under
detector in the X coordinate obtainable by the charge the same signal and the same drift path as at previous
division as a function of the anode pitch. 20000 electrons Fig.
were produced at radius of 6 mm and drifted towards an
anode ring a t a radius of 32 mm.
receiving the electron cloud is used in an optimal way
to determine precisely the anode coordinate X. However,
is multiplied with a factor, which is the square root of
in the cylindrical case, the position X determined at the
the geometrical factor for the original spread. Physically
anode ring by charge division is projected to the initial
it means that the diffusion during the drift is multiplied
radius. The resolution obtained a t the anode ring is
by an average geometrical factor smaller than the factor
thus improved by the ratio of the initial and the anode
corresponding to a full drift length. If we project the
radii. No such projection is done in the case of a linear
measured diffusion at the radius of anodes back to the
geometry and the anode resolution is the resolution of
origin of the ionization, the width of diffusion is smaller
the measurement. We can see that for an anode pitch
than the diffusion width in a linear geometry.
of 500 p m (selected in the design) and for the origin
B. Position Resolution of ionization TO equal to 6 mm the cylindrical case has
The difference between the the diffusion during the about three times better resolution than the linear case.
drift in cylindrical and linear geometry is larger when the The difference between the resolution in cylindrical and
electrons were created closer to the center, that is, for TO linear geometry decreases with an increase of PO.
smaller. Fig. 2 and Fig. 3 show the optimum resolution There are several resolution curves on Fig. 2 and
along the x coordinate as a function of the anode pitch Fig. 3. The lowest curve (the best resolution) corresponds
in a cylindrical and linear geometry respectively. Signal to a case where the cloud of signal electrons arrives in
electrons were produced 6 mm from the center and between two neighbour anodes. The charge division
drifted toward anode ring located a t radius of 32 mm. gives the position of the center of the cloud with a
Due to the defocusing effects during the diffusion in the very good accuracy even when the anode pitch is larger
cylindrical geometry the electrons arrive at the anode then the spread of the cloud. The highest curve (the
ring with a Y I = 400pm to be compared with a spread worst precision) corresponds to the arrival of the signal
in a linear case of 170 p m only. The position along electrons right a t the center of an anode. There is
the anode direction is determined by a charge division aot enough signal on the neighbour anodes if the anode
method. That means that the full signal from every anode pitch is larger than the spread of the electrons. The
622

Figure 4: Closer look on the n-side of the cylindrical Figure 5: Photography of the part of the n-side of the
detector and its assembly. The magnification is about 2.6 detector under a magnification of 8 times. Anode pitch
times. is 500 p m .
curves between these two extremes have the center of the
electrons shifted a t 10% of the anode pitch. The r.m.s of the read out channel is proportional to the anode size.
the electron cloud is 400 p m in the Fig. 2 and we start The total noise considered for the position determination
to see the dependence of the resolution on the coordinate depends only on the spread of electrons and there is no
relative to the anode position for the anode pitch above deterioration of the resolution with a finer anode pitch.
500 p m . C. Implementation
The other interesting detail shown in Fig. 2 is the
worsening of resolution with a smaller anode pitch (region Fig. 4 shows the coexistence of the linear and the
between an anode pitch of 200 p m to 500 p m ).The cylindrical geometry on the detector. Due to the limi-
effect is due to the contribution of the series noise tations in the production of masks for the detector, the
of the preamplifiers to the final resolution. In this cylindrical detector is in reality a polygon with 120 sides.
detector the preamplifiers are external as opposed to ones The detector is linear within 3" sectors and then there
integrated on the silicon wafer. The total preamplifier is a 177' angle between two neighbor sectors. The elec-
input capacitance is dominated by the capacitance of tric field is applied on 241 concentric rings (or polygons
the first transistor and the capacitance of the connection with 120 sides) just resolved in Fig. 4. The signal elec-
and is independent of the anode size and pitch. The trons are transported in the middle of the wafer, that is,
amplifier noise of an individual anode channel is thus also about 125 p m from the surface. The electric field i n the
independent of the anode pitch. For smaller anode pitch middle of the wafer does not have a sharp angle and the
the signal is read through more channels adding noise field under a 177" angle between two sectors is rounded.
from more preamplifiers into the position determination. At small radii where the linear dimension of a 3 O sec-
The worsening of the resolution can be avoided by tor is less than 200 p m the drift field inside the wafer is
integrating the first transistor onto the wafer of the practically cylindrical.
detector [8]. When the capacitance of the first transistor There are other features visible on Fig. 4. The
is made equal to the anode capacitance the noise of black spots almost randomly scattered across the silicon
U1

Figure 7: Negative potential in a radial cross section


(called “Y”) of the detector. Equipotentials imposed at
both surfaces of the detector are the rectifying junctions.
Figure 8: The same part of the detector n-side under a
Surface between rectifying junctions is covered by S O z .
magnification about 25 times.
Potential a t this surface depends on the global design of
the detector.

detector are integrated resistors of the voltage divider. 111. SURFACE PHENOMENA
There are 240 resistors implanted on each side of the
wafer. An effort was made to place the resistors on the The second important feature of the design of the
wafer in such a way that the power dissipated in the cylindrical drift detector is the collection of the current
resistors is distributed uniformly across the whole wafer generated on the Si - Si02 interface on the guard (sink)
area. The drift field is constant and therefore the voltage anode. The design of drift detector was made in such a
difference between the neighbour rings is constant. The way that the electrons generated on the interface are not
area covered by individual rings depends on the radius of collected at the detector anode. The designed detector
the ring. To dissipate the same power per unit area the collects all electrons generated at the Si - Si02 interface
power dissipated per resistor should be proportional to on a guard anode. The total leakage current collected
the radius of the ring. As a compromise, the dissipation on the detector anodes is the sum of the bulk generation
is constant for 12 neighbouring rings and then changes current and junction diffusion currents only.
to some other value for the next 12 rings. In total, there Fig. 5 and Fig. 6 shows the n-side of the detector
are 20 different groups of resistors on each side of the under larger magnifications 8 and 25 times respectively.
detector. The right values of the drift voltages must be Individual rings are well visible. The white rings are
applied to the rings at the boundary of each group. The the rectifying p+n junctions. The white appearance of
bond wires visible a t the left bottom of Fig. 4 bring the the junction is due to the aluminum which covers the
drift voltages from an outside divider to every 12-th ring. junctions. The remaining silicon areas are covered by
The bonds at the left hand side of Fig. 4 are connec- thermally grown SiOz.
tions between the anodes and the gold plated traces on Fig. 7 shows the negative potential in a radial cross
a ceramic board. Holes in a FR-4 board above the ce- section of the detector. Equipotentials imposed at both
ramics are exit holes for the forced radial air flow which surfaces of the detector are the rectifying junctions. The
cools the detector. picture shows very clearly the realization of the main
Figure 8: Detail of negative potential in a cross secti.on
of the detector close to the surface. The potentid is

valley in the potential energy for electrons (negative po- bubbles on figures where the negative potential is plotted)
tential). In this valley the signal electrons are transported are immediately absorbed by surrounding p + n junctions.
from the point of creation by the ionization of fast parti- Electrons, however, (one can visualize the motion of
cle to the anode. Now let us concentrate our attention, electrons as heavy balls) fall to the minimum of the
however, to secondary valleys located right below sur- potential energy right at the interface. To follow these
face covered by thermally grown SiOz. The existence electrons we have to know the shape of the interface in
of the secondary valleys is a consequence of fixed posi- the coordinate not shown in Fig. 7.
, tive charges in the Si02 close to the Si - Si02 interface. In a case of cylindrical symmetry or in a fully sym-
The presence of these charges bends the energy band in metrical geometry of polygon of 120 sides, the potential
such a way that electrons are held close to the interface is identical at each cross section independent of the az-
(simple electrostatic attraction between mobile electrons imuthal angle 4. Electrons falling onto the interface have
within the silicon and fixed positive charges in the oxide no way to escape. They start to accumulate at the re-
close to the interface). gions of secondary valleys forming an accumulation layer
Potential shown in Fig. 7 was calculated assuming at the interface. Due to the negative electric charge of
that there are no electrons a t the Si - Si02 interface electrons the positive oxide charges are partially compen-
to compensate positive charge of the SiOz. This is sated and the band bending at the interface decreases.
one extreme condition which is hard to realize and is After a while an equilibrium situation is reached where
not very desirable. The surface of silicon under areas there is the maximum density of electron in the accumu-
covered by Si02 is depleted, and a relatively large current lation layer at the interface. The boundary between the
is generated on the Si - Si02 interface [9]. Holes secondary valley and the main valley has disappeared at
generated at the interface (one can visualize holes as one point along the secondary valley. Mobile electrons in
625

Figure 10: Microphotography of the part of the n-side Figure 11: Microphotography of the “river” region
of the detcctor under a magnification of 170 times. Two under a rnagniflcatisn of 600 times. Flow of electrons ia
different aluminum layers are visible. The non sintered towards left bottom of the picture. Entrance opening of
aluminum covers Si02 and bridges above the “river” to the river is always smaller to prevent excess drainage of
connect the sintered aluminum (rougher surface) covering electrons from the Si - SiOa interface.
rectifying junctions. At the bottom left there is a
structure which allows “rivers” current to reach the sink
anode at the outside radius while avoiding signal anodes. following ring is interrupted as well. These interruptions
form “rivers” where the surface current flows.
Several “rivers” are visible as radial lines in Fig. 6
the accumulation layer form almost an equipotential sur- and Fig. 6 . These “rivers” have to carry away all current
face and an arrival of one electron anywhere along the generated a t the Si - Si02 interface. At a small radius of
interface leads to the emission of one electron from the the detector, the surface is small and there is only little
secondary valley into the main valley which is collected current which was generated a t a smaller radius to be
later a t the signal anode of the detector. carried away. At larger radii there is larger quantity of the
To prevent the inclusion of electrons generated a t generated current and also current generated upstream
Si - Si02 interface from the secondary valley in the must flow in “rivers”. The number of “rivers” increases
signal anode of the detector a part of the interface from 2 at the center to 40 at the largest radius of the
charge must be drained away from the interface before detector.
the maximum electron density is reached. One way The draining of surface generated current should be
of draining is by breaking the cylindrical symmetry by done at a correct rate. Not enough drainage would let
incorporating of small openings into the rectifying rings. the surface leakage current reach the detector anode.
Small openings let the electrons accumulated in a region Too much drainage, however, would deplete larger part
between two rectifying rings fall into the next inter-ring of the interface surface and increase the amount of the
region. The negative potential in the next region is surface leakage current. Fig. 8 shows the detail of the
similar to the previous one but the potential difference of negative potential a t the interface where there are still
one ring (6 V). Thus electrons generated a t the interface some electrons left at the interface. The potential barrier
move from one inter-ring region into the next region. The to prevent electrons generated at Si - Si02 interface is
626

a
i
e
E
z
a
0
0
0
O
X
0

I A L L ANODES
1 ~ x y ~ x ~ ~ x ~ ~ ~ x - x - x . - x - x - - ~

Slope=7.2hA/cmZ
t I I I I I I T I I I ‘

13 25 37 49 61 73 85 97 1 0 9 121 133145157
I I RING # I I

5 10 15 20 25
s (cm2)

Figure 12: Microphotography of one resistor of the volt-


age divider under a magnification of 600 times. Resistors Figure 13: Dependence of the leakage current flowing
are made during the same production step as the rectify- into all signal anodes and the leakage current flowing to
ing junctions. Resistance of the implant is about 8kn/U. the sink anode on the active sise of the B11 detector.
Changing the width of the connection between two rings
different values of resistance can be obtained.
are needed to put the entire ring a t the same potential.
still present. Conductivity due to the electrons left on the The second aluminum layer was non sintered, thereby
eliminating the possibility of “spiking” at the oxide cut
interface is just right to carry generated electrons from
corners.
location of origin to “rivers”. This direction corresponds
to a vertical flow between rectifying junctions as shown Microphotography of Fig. 11 shows details of ring
interruption to form a “river”. Surface electrons flow
in Fig. 9. Electrons generated a t the upper (bottom)
half of the picture are transported downwards (upwards) toward the left bottom of the picture. The entrance
towards the “river”. The electric field along the interface opening of the “river” is the part of the river to avoid the
is small enough that it does not introduce any azimuthal excessive drainage of the surface electrons. Quality of the
component of the drift field for signal electrons moving alignment and of the photolithography can be seen from
this picture and from the microphotography shown in
iu the middle of the main valley in the radial direction.
Fig. 10 shows a “river” terminating at the sink Fig. 12. Registration of alignment to 5 2 pm is required
in this design.
anode outside signal anodes. The rectifying electrode
circumvallating signal anodes is interrupted between two
anodes to let the surface current to pass signal anodes IV.RESULTS
and flow into the sink anode. The second layer of non
sintered aluminum covering partially Si02 is visible as A. Tests in Laboratory
a smooth metallization layer. This layer defines the Several cylindrical drift detectors were produced at
potential a t the top of Si02 surface and also makes BNL and TU Munich. Fig. 13 shows the functioning of
bridges across “rivers”. The bridges connect sintered the designed collection mechanism for the surface current
aluminum a t the top of rectifying junctions of the same qualitatively. The leakage current flowing to the all signal
ring after being interrupted by “rivers”. Connections anodes as well as the leakage current flowing to the sink
627

r-
5
I9 863
Envia 21703
812. Ring ,169 OL -7OOV r g “229 oL -5OOV 04-04-1991 Me07 177.;
RHS 118.2
Sun anode Is O.‘L282E+04 nR. Rvrg. grd 0.3154E+04 nR 3M) UOFLW O.DOOOE+W

280
320

240
1
F
OVRW O.OOaOE+W

200

160

120

80

LO

0
320 3 0
CHANNELS FIRED

Figure 15: Frequency of hits as a function of the anode


number or the azimuthal coordinate. 3O sectors are
well visible. The entries into histogram were selected
Figrire 14: Leakage current for each of the 360 anodes
according to the information from a silicon pad detector
of a cylindrical detector B12. The ring # 169 was biased
right in front of the cylindrical silicon drift detector.
a t -700 V. The upper curve (left linear scale) shows the
Only,events with no hit in an aeimuthal sector in the
current flowing into the sink anode.
pad detector were entered. A hole in the frequency of
hits at arround the anode number 230 corresponds to the
location of the selected pad.
anode are plotted as function of the active size of the
detector. The measurement was done with the detector
B11. This detector has no defect between rings number
13 and number 169 a t the inner radius. The slope of Three figures presented here (Fig. 16, Fig. 16 and Fig. 17)
the leakage current to all signal anodes is 7.2 nA/cm2. are preliminary.
The slope of the leakage current to the sink anode is
180 nA/cm2. The improvement (decrease) of the leakage
current is a factor of 25. We have other indications that V. CONCLUSIONS
the surface leakage current was drained too hard by about
a factor of 3 which gives a more realistic improvement An advanced silicon drift detector, a large area
factor of 8. We would like to stress that the diversion of cylindrical drift detector, was designed, produced, tested
the surface generated leakage current into the sink anode and installed in the NA45 experiment. The detector
made the detector less sensitive to potential radiation provides unambiguous pairs of r,+ coordinates for events
damage of the Si - Si02 interface. with multiplicities up to several hundred. The position
Fig. 14 shows the leakage current of individual an- resolution of the detector of 20 p m (rms) in each direction
odes in B12 detector. 90 % of the detector surface was provides about 2 10’ two-dimensional elements.
active and most of the anodes have the leakage current Although somewhat complicated, it has been pos-
below 1 nA under operating field and bias. sible to produce nearly fully operational devices on the
whole area of 3 inch diameter wafers. The surface cur-
B. Beam P e r f o r m a n c e rent is efficiently collected by a sink anode as designed.
In summer of 1991 the B12 cylindrical detector was The observed anode current is primarily bulk generated
installed into the NA45 Experiment a t the CERN SPS. current at an average level of 7nA/cm2.
628

160 -
k
I ID

70 -
w -

50 -

40 -

DRllTlYE Dl5lR ALL

Figure 17: Frequency of hits as a function of drift time.


The entries into the histogram were selected according
to the pad detector. This time only events with no hits
between two radii were plotted.

DRl.r[TIUE DISTR ALL 6. Preamplifiers and shapers were developed at BNL


and produced by ASTER TECHNOLOGY Inc., P.O.
Box 729, Ramsenburg, NY 11960.
Figure 16: Frequency of hits as a function of drift time.
No selection of entries into the histogram. The increase 7. ADCs were produced by Dr. B. Struck, 2000 Tang-
of number of events per bin with the increase of the drift stedt/Hamburg, Germany.
time is due to the physics of the particle production in 8. P. Rehak et al., Nucl. Instr. and Meth. m,168
the experiment. (1990).
9. A.S. Grove, Physics and Technology of Semicondmc-
tor Devices, J, Wiley and Sons, Inc., New York, 1967,
REFERENCES Page 298.

1. E. Gatti and P. Rehak, Nucl. Instr. and Meth. 225,


608 (1984).
2. P. Rehak et al., Nucl. Instr. and Meth. 248, 367
(1986).
3. P. Rehak et al., IEEE Transactions on Nucl. Science
36, 203 (1989).
-
4. U. Faschingbauer et al., Proposal to the SPSC CERN,
SYSC 88-25, SPSC/P 237.
5. GD type low resistance connectors from Shin-Etsu
Polymer America, Inc., 34135 7-th Street, Union
City, CA 94587.

Potrebbero piacerti anche