Physics

Physics Research Publications

Purdue University Year 

Guard Ring Simulations for n-on-p

Silicon Particle Detectors
O. Koybasi G. Bolla

D. Bortoletto

This paper is posted at Purdue e-Pubs.

[Link] articles/1272
2978 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 57, NO. 5, OCTOBER 2010

Guard Ring Simulations for n-on-p Silicon Particle

Detectors
Ozhan Koybasi, Gino Bolla, and Daniela Bortoletto

Abstract—We propose a new guard ring geometry for n-on-p sil- n-type substrate with only a few on p-type silicon [1]. How-
icon particle detectors for high luminosity applications. The per- ever, in last few years, p-type silicon as sensor material has
formance of the guard ring structure is evaluated with simulations shown promising results for radiation hardness [2]–[5] and it
up to a radiation fluence of 1 1015 neq cm2 using an existing
three level trap model for p-type FZ silicon. The post-irradiation is now considered as a strong candidate among the scientific
performance improvement of guard rings with floating field plates community to reduce the detrimental effects due to the harsh
pointing towards the sensitive region is demonstrated. The break- radiation environment. One advantage of the n-on-p sensor
down behavior of the guard ring structure is studied as a function technology over the more conventional p-on-n one is that it
of oxide charge, field plate length, and oxide thickness. collects electrons which have slightly larger trapping time
Index Terms—Device simulations, field plates, guard rings, and higher mobility in silicon compared to holes. Also, it has
n-on-p silicon particle detectors, radiation hardness. been demonstrated that during annealing, the trapping times
increase for electrons and decrease for holes [6]–[8]. In the
I. INTRODUCTION p-on-n geometry, the junction moves to the back-side after
space charge sign inversion (SCSI) caused by radiation-in-
ILICON microstrip and pixel detectors used in the trackers
S of high luminosity collider experiments are exposed to
charged particles (which cause ionizing displacement damages)
duced negative space charges and, therefore, the high electric
field is on the side opposite to the readout electrodes. If the
detector is operated at partial depletion, the holes deposited in
with very high fluences due to their proximity to the interaction the depletion region spread over many readout electrodes due
point. The displacement damage induces defects in the silicon to the low electric field near the front surface, resulting in a
crystal, leading to higher depletion voltages. Therefore, the degraded resolution. On the other hand, the junction is always
tracking devices must be operated at high voltages to maintain located on the readout-side in n-on-p sensors due to non-SCSI,
full depletion or at least partial depletion of the substantial part so it is not necessary to fully deplete the sensor to create a
of the detector volume, which is required to get a reasonable high electric field underneath the readout electrodes. In the
signal to noise ratio from the detector. The heavily damaged n-on-n configuration, the collected carriers are also electrons
cutting edge of the sensor is a region of high generation and and it has already been shown with extensive studies that these
recombination and it may cause the leakage current to increase sensors are substantially more radiation hard than the p-on-n
by orders of magnitude if the depletion region of the sensor technology [5]. However, this approach is the most expensive
extends to this edge. Furthermore, the positive oxide charge one as it requires double-side processing, which makes the
which increases with radiation induces a conductive electron n-on-p technology extremely appealing. Further discussions on
channel at the silicon/oxide interface. Guard rings are needed these different planar detector technologies can be found in [9].
to shield the sensitive region from the silicon surface and dice Although numerous studies on guard ring designs for p-on-n
line leakage currents. The guard ring geometry should also and n-on-n silicon sensors can be found in the literature
establish a uniform potential drop along the silicon surface to [10]–[16], no standard guard ring design for the relatively
avoid large potential drops over very short distances that may new n-on-p technology has been established yet. Recently, a
lead to local breakdowns at voltages much lower than the bias preliminary study on the simulation of n-on-p guard rings for
voltage required for detector operation. the ATLAS upgrade has been reported [17]. In this paper, we
In earlier times, almost all the developments in silicon take a more thorough approach and investigate the impact of
detector technology for high energy physics were made using all design parameters with electrical simulations to develop
an optimum guard ring structure for high voltage operation of
Manuscript received December 15, 2009; revised April 21, 2010 and July n-on-p silicon detectors at high fluences for the upgrade of the
19, 2010; accepted July 26, 2010. Date of current version October 15, 2010. LHC trackers. The proposed designs and some test structures
This work was supported in part by the U.S. Department of Energy under Grant
DE-FG02-91ER40681 and in part by the National Science Foundation under
dedicated for systematic comparison of experimental behavior
Cooperative Agreement PHY 0612805 – UCLA Subaward 1000 G HD870. with simulations will be implemented along with the CMS pixel
O. Koybasi is with the Department of Physics and Birck Nanotechnology sensors at SINTEF, Norway. Although no experimental data
Center, Purdue University, West Lafayette, IN 47907 USA (e-mail: okoy-
basi@[Link]).
to validate the simulation results of the guard ring structures
G. Bolla and D. Bortoletto are with the Department of Physics, Purdue exist yet, simulations with the same tools and models were able
University, West Lafayette, IN 47907 USA (e-mail: [Link]@[Link]; to predict the experimentally observed breakdown behavior
bortolet@[Link]). of different devices, namely 3D CMS pixel detectors, with a
Color versions of one or more of the figures in this paper are available online
at [Link] satisfactory accuracy. For 4-electrode (4E) and 2-electrode (2E)
Digital Object Identifier 10.1109/TNS.2010.2063439 3D CMS pixel sensors that have actual breakdown voltages of
0018-9499/$26.00 © 2010 IEEE
KOYBASI et al.: GUARD RING SIMULATIONS FOR n-ON-p SILICON PARTICLE DETECTORS 2979

Fig. 1. Optimum guard ring structure for an oxide charge of 4 2 10 cm .

85 V and 110 V, respectively, the corresponding simulated

breakdown voltages are 80 V and 95 V [18].
The simulations have been performed with Synopsys
Sentaurus TCAD [19] which is a 1D/2D/3D finite element
semiconductor simulation package capable of simulating the
processing and electrical, thermal, and optical characteristics
of semiconductor devices made with various semiconductor
materials. Sentaurus Device takes a mesh made of discrete
elements as an input structure and solves the Poisson’s equation
along with carrier continuity equations at every grid point on
the mesh to calculate the electrical properties such as current
and capacitance and physical quantities such as carrier distri-
bution, carrier mobility and potential distribution inside the
semiconductor. In addition to electron, hole, acceptor, and
donor charges, the Poisson’s equation takes into consideration Fig. 2. Leakage current versus reverse bias voltage as a function of substrate
the charges associated with radiation-induced trap levels. The doping concentration for the guard ring structure in Fig. 1.
dynamic behavior of traps is described by Shockley Read Hall
recombination which depends on trap concentration, energy,
and carrier cross sections. Doping dependent mobility degrada- with a substrate thickness of 300 and substrate doping of
tion, mobility saturation at high fields and mobility degradation . Although the substrate doping concentration
at interfaces were taken into account with the mobility models and the doping profile of the guard rings do not play a role in
used. The physical models employed in the simulation include determining the optimum guard ring geometry as long as they
band gap narrowing at high doping concentration and high are within the range of typical values, the breakdown voltage for
temperatures as well. Sentaurus has several models that can be a given geometry is influenced by these parameters. Therefore,
activated to simulate the impact ionization which is the free it is worth investigating how the breakdown of the guard ring
carrier generation mechanism leading to avalanche breakdown. structure changes with substrate and guard ring doping before
The impact ionization model suggested by Okuto and Crowell discussing the impact of other parameters on the design.
[20] was used to evaluate the breakdown behavior of the guard Leakage current versus reverse bias voltage characteristics of
ring structure. Since the length of guard rings is quite large the guard ring structure in Fig. 1 with four different substrate
compared with the substrate thickness and the width of the doping concentrations is shown in Fig. 2. As one would expect,
sensor periphery, simulations over a 2D cross section orthog- the substrate with the highest resistivity yields the highest break-
onal to the length of guard rings can provide accurate enough down voltage (the reverse bias voltage at which the leakage cur-
information regarding the 3D guard ring behavior. The mesh rent starts to rise exponentially). However, although the break-
was refined near the silicon/oxide interface with a maximum down increases considerably (by 200 V) when the substrate
height of 0.15 and width of 0.8 to improve the accuracy doping is reduced from to , it
of the results although a convergence study has shown that the varies very slightly over the range of substrate doping concen-
breakdown voltage does not vary with mesh refinement near tration from to . Higher sub-
the silicon/oxide interface. strate resistivity is desired since it also leads to a lower deple-
tion voltage even though it requires the inactive portion of the
II. SIMULATION WITH NO RADIATION DAMAGE sensor to be larger as it will be discussed later. The increase of
Fig. 1 shows the 2D cross section of the seven p+ guard leakage current with lower substrate doping is attributed to the
ring structure used in the simulation, optimized for an oxide increasing depleted volume for an applied bias voltage, which is
charge of . The simulations were carried out also true after the full-depletion (across substrate thickness) at
2980 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 57, NO. 5, OCTOBER 2010

Fig. 3. Different p+ doping profile depths and corresponding breakdown Fig. 4. Different values of peak p+ doping concentration and corresponding
voltages. breakdown voltages.

all doping levels is attained due to the lateral depletion towards

the detector edge.
The effect of the guard ring doping profile has been investi-
gated as follows.
i) The peak guard ring doping concentration is kept at a
constant value and the depth of the profile is varied.
ii) The depth of the profile is fixed and the peak doping of
guard rings is changed.
Fig. 3 suggests that the doping profile should be as shallow as
possible. A breakdown voltage of 1080 V has been achieved
with a profile depth of while it drops to 850 V for
a profile depth of 3.5 . This is a remarkable observation as
one would expect a deeper implant, which has a smaller curva-
ture, to lead to a lower electric field and consequently a higher
breakdown voltage. Indeed, the maximum electric field in sil-
icon remains lower for structures with deeper implants up to
voltages close to the start of impact ionization, after which the
maximum electric field increases with a higher slope as a func-
tion of bias voltage. The deeper the implant, the lower the bias
voltage at which the electric field changes slope. In this case,
the guard ring design with a deeper implant reaches the elec-
tric field value that initiates the breakdown at lower voltages.
In [21], it has been demonstrated that the breakdown voltage
increases with strip implant depth for a p-on-n strip detector,
and the same trend has been reproduced with our simulation
tools and models. The only difference between the two situa-
tions is that the p+ implants are biased in the strip detector case
while they are floating in the guard ring case. Deeper biased
strips lead to a lower electric field due to smaller implant cur-
vature and this trend persists up to the realization of the break- Fig. 5. a) Potential drop and b) electric field along a cut line parallel to the
device surface at a depth of 100 nm from silicon/oxide interface for an oxide
down unlike floating guard rings in which the maximum electric 2
charge of 4 10 cm .
field changes slope and this change occurs at lower bias voltages
for deeper implants as discussed previously. In conclusion, the
breakdown voltage is enhanced with implant depth for biased sufficiently high to compensate the surface channel if the oxide
implants whereas the opposite statement holds true in the case charge reaches up to . This value of oxide charge
of floating implants. requires a peak p+ ring doping of at least to
According to Fig. 4, the breakdown voltage increases with de- prevent surface charge from inversion.
creasing peak p+ ring doping but the increase is not significant. In Fig. 1, the width of p+ guard rings and the spacing be-
The minimum doping of the p+ rings is limited by the maximum tween them have been set in such a way that equal amount of
value that oxide charge may reach. Although it seems to yield a potential drop occurs at each guard ring. This makes the corre-
higher breakdown, a peak doping of will not be sponding electric field peaks more or less the same, minimizing
KOYBASI et al.: GUARD RING SIMULATIONS FOR n-ON-p SILICON PARTICLE DETECTORS 2981

Fig. 6. Leakage current versus reverse bias voltage for different values of oxide
charge. The decrease in breakdown voltage at oxide charges lower than 4 2
10 cm is due to the non-uniform distribution of potential along the guard
2
rings while at oxide charge higher than 4 10 cm it is mainly due to more
abrupt potential drop at each guard ring.

the maximum electric field throughout the entire structure. In

optimizing the geometry, a fixed oxide charge of
was assumed. The electric field maxima are located at the lateral
p-/p+ boundaries facing the diode where the electron channel
is interrupted and the potential drops. Fig. 5 shows the poten-
tial drop and the corresponding electric field distribution along
the detector surface at a depth of 100 nm from the silicon/oxide
interface. With this potential distribution, the device is able to
sustain reverse biases exceeding 1000 V as shown in Fig. 6. For
lower or higher values of oxide charge, the guard ring perfor- Fig. 7. a) Potential drop and b) electric field along a cut line parallel to the de-
vice surface at a depth of 100 nm from silicon/oxide interface for oxide charges
mance degrades significantly. The breakdown voltage decreases
as the deviation of oxide charge from increases.
2 2
of 5 10 cm and 1 10 cm . The detector is reverse biased at 350
V which is just below the breakdown voltage.
At low oxide charges, since silicon surface conductivity is low,
the majority of the potential drops at the diode and innermost
guard rings. The potential drops more evenly at all guard rings was also confirmed with simulation. However, one n+ guard ring
with increasing surface conductivity. This goes in parallel with between the sensitive region and the innermost p+ guard ring is
an increasing slope of potential drop at each guard ring. After needed for biasing to the ground potential to make the field uni-
a very uniform field distribution is reached at an oxide charge form near the outermost pixel/strip.
of , further increase in oxide charge makes the When the sensor is reverse-biased, the depletion region
potential drop at each guard ring steeper due to increasing sur- spreads out laterally from the p-n junction toward the cutting
face conductivity, resulting in higher electric field peaks and edge of the sensor. If the space charge region reaches out to the
consequently lower breakdown voltage. Fig. 7 shows the po- edge which is a very effective generation center due to crystal
tential drop and the corresponding electric field for two extreme damage caused by cutting, charge is injected into the depleted
values of oxide charge, and , at zone from the scribe line, increasing the leakage current. In
a reverse bias voltage of 350 V which is just below the break- order to prevent the depletion region from reaching the edge,
down. Note that the uniformity of potential distribution does not the surface of the sensor periphery must be terminated by a
change significantly with further increase in oxide charge once wide p+ implant whose width is denoted by W in Fig. 1. The
it becomes even for a certain oxide charge. Therefore, if the minimum W required to prevent the depletion region from
guard ring structure needs to be optimized for higher values of making contact with the cutting edge at reverse biases of 1000
oxide charge, this cannot be accomplished by only modifying V and (bias voltage required for full depletion) has
the width of guard rings and the spacing between them but it been simulated for different values of substrate doping. The
also requires the number of guard rings to be increased. In this results are presented in Table I. For the guard ring structure,
case, the applied bias voltage will be shared between more guard the depletion voltages associated with each substrate doping
rings, resulting in less potential drop and a lower electric field are higher than those for an un-structured diode, which would
at each guard ring. be 7 V, 35 V, and 70 V for substrate doping of ,
Since the area under the oxide is already n+ due to the posi- , , respectively. This is because
tive oxide charge, adding n+ guard rings does not affect the po- of the lateral depletion of the guard ring structure as well as the
tential distribution and therefore the breakdown behavior. This traversal depletion.
2982 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 57, NO. 5, OCTOBER 2010

TABLE I TABLE II
W AS A FUNCTION OF SUBSTRATE DOPING AT BIASES OF 1000 V AND V THE UNIVERSITY OF PERUGIA TRAP MODEL

N , W, and V represent the substrate doping, minimum outermost p+

width, and depletion voltage, respectively.
of , and some older references such as [24] have
III. SIMULATIONS WITH RADIATION DAMAGE reported a value of .
A trap model has been proposed by the University of Pe-
The incident radiation creates defects in the silicon lattice, rugia [23] to describe radiation damage caused by proton irra-
forming extra energy levels in the band gap of silicon that act diation in p-type FZ silicon. According to the model, the ra-
as generation-recombination and charge trapping centers. The diation generates two acceptor levels positioned slightly above
primary effects of these bulk traps are the mid band gap and one donor level located far below the mid
i) an increase in leakage current proportional to the radia- band gap. The details of the trap model are presented in Table II.
tion fluence: Since the acceptor states are close to the mid band gap, they will
generate electron-hole pairs, increasing the leakage current. A
(1) small portion of the acceptor states will be occupied and there-
fore negatively charged, increasing the effective p-type doping
where V is the depleted volume, is the fluence, and while the unoccupied acceptor states will trap excess electrons
is the damage rate constant; in the conduction band. The function of the donor state, which
ii) an increase in depletion voltage due to increasing effec- is far below the mid band gap, is to trap excess holes from the
tive doping concentration : valence band.
This trap model accurately predicts the increase in leakage
(2) current and effective doping concentration due to irradiation
but not the trapping behavior. The carrier cross sections in this
where q is the electronic charge, d is the substrate thick- model have been modified at University of Glasgow to accu-
ness, is the permittivity of free space, and is the rately predict the trapping rates [25]. As long as the ratio of
dielectric constant of silicon; electron cross section to hole cross section is constant,
iii) a decrease in charge collection efficiency due to in- the depletion voltage changes very slightly with altering each
creasing carrier trapping. cross section. Therefore, the carrier cross sections have been
In addition to the displacement damage that introduces bulk- modified in such a way that they yield trapping rates that match
defects, the ionizing radiation by charged particles also gives the experimental data but at the same time the ratio
rise to an increase in positive oxide charge near the silicon/oxide is kept constant not to impact the effective doping concentra-
interface. A part of the ionizing radiation-generated carriers in tion which is predicted accurately by the University of Perugia
the oxide recombine immediately while the surviving electrons model. The parameters of the modified model are shown in
and holes drift in opposite directions under the effect of the Table III. This model yields a current-related damage rate of
oxide field. Since the electrons have a relatively high mobility in which is about 30% higher than the
, they are quickly swept out of the oxide. experimental value of as mea-
The holes, on the other hand, have a very low mobility in the sured at 293 K after an 80 minute annealing at 60 [26]. The
due to numerous shallow hole traps leakage current can change by more than 30% under different
that exist in the oxide. The holes move very slowly in direction annealing conditions. If the aim of the simulation is to model
of oxide field from one shallow trap to another and when they the effective doping concentration and trapping behavior rather
reach the silicon/oxide interface, where many deep hole traps than the amount of leakage current, the model seems to work
exist, they either recombine with electrons from silicon or get quite well. For guard ring simulations, only the change in effec-
trapped there permanently. The overall effect is an increase in tive doping concentration is relevant so either radiation damage
the oxide charge. This increase is not an everlasting process and model can be used.
the oxide charge saturates at densities up to a few The post irradiation performance of the guard ring structure
after a radiation dose of some kilograys due to the finite number has been evaluated with simulations using the modified trap
of existing traps. In [22], it has been observed that the oxide model in Table III. Fig. 8 and Fig. 9 show the breakdown be-
charge saturates at for p-type FZ and MCz havior of the guard ring structure after irradiation for different
silicon with surface orientation and at values of fluence and oxide charge, respectively. The breakdown
for n-type FZ silicon with surface orientation after a ra- behavior degrades significantly with irradiation. For an oxide
diation dose of 150 krad whereas [23] claims a saturation value charge of , the structure is able to survive only
KOYBASI et al.: GUARD RING SIMULATIONS FOR n-ON-p SILICON PARTICLE DETECTORS 2983

TABLE III potential drops at the inner guard rings since the depletion
THE MODIFIED VERSION OF THE UNIVERSITY OF PERUGIA TRAP MODEL region does not reach the outer ones.
The breakdown voltages in Fig. 8 and Fig. 9 are not high
enough for very high luminosity applications, so further im-
provements are needed.

A. Field Plates
The electric field after irradiation can be reduced by using
The modified parameters are shown in bold. field plates pointing towards the sensitive region as depicted in
Fig. 10. The field plates are at the potential of the silicon re-
gions they are in contact with. Therefore, in this configuration,
they are at lower potential than the underlying silicon as the po-
tential drops from the active region towards the edge. This forms
a negatively biased p-bulk MOS structure, which may result in
depletion of free carriers from the silicon surface or accumula-
tion of holes at the silicon/oxide interface, depending on the flat
band voltage and potential difference between the field plates
and the underlying silicon. Fig. 11 shows the electron density
and electric field distribution around the two innermost guard
rings of the optimized sensor periphery with field plates at a re-
verse bias of 900 V for a radiation fluence of
and an oxide charge of . The electron layer disap-
pears at the silicon/oxide interface regions underneath the por-
tion of the field plates on the left of the p+ rings due to the lower
Fig. 8. Leakage current versus reverse bias voltage at different fluences. The electrostatic potential of the field plates with respect to the sil-
2
oxide charge is 1 10 cm . icon surface. With the use of field plates, the potential drop starts
underneath the left edges of the field plates where secondary
electric field peaks in silicon are formed in addition to the ones
at the p-/p+ boundary facing the active region while the highest
field spots over all the structure are located in the oxide due to
the large potential difference across its thickness. The break-
down voltage improves with the creation of secondary electric
field peaks since the electric field will be distributed over more
spots with a lower peak value at each spot.
The n+ guard rings in Fig. 10 are used for the contact of the
field plates to the silicon. In principle, it would be possible to
make these contacts on the p+ guard rings if they were wide
enough, which would give the same potential distribution. How-
ever, the inner p+ rings are too narrow to put the metal con-
tacts on (according to SINTEF design rules) and increasing their
width will change the potential distribution in silicon. That is
why n+ rings, which have no effect on potential distribution,
Fig. 9. Leakage current versus reverse bias voltage for different values of oxide
2
charge after a radiation fluence of 1 10 n =cm . are used for metal contacts. Indeed, the three outermost p+ rings
are wide enough to be used for metal contact and an alternative
structure that gives identical performance would be one with the
up to reverse biases of 200 V at a fluence of outer three field plates in contact with the p+ rings instead the
while the breakdown voltage is as low as 50 V when the oxide n+ rings next to them. If the field plates extend outwards, the
charge rises to at the same fluence. The de- electric field is not reduced since the field plates will be at the
crease in breakdown voltage after irradiation is attributed to two same potential as the underlying silicon between the contact n+
reasons. ring and the p+ ring on its right, and a higher potential than the
i) Due to increasing surface conductivity as a consequence underlying silicon at the location of the p+ ring and beyond,
of increasing oxide charge, the potential drop at each p+ which causes further build of electron layer at the interface of
guard ring becomes steeper, resulting in higher electric and Si.
fields. The simulation of the breakdown behavior of the guard ring
ii) Depletion voltage increases due to the radiation-induced structure with field plates as a function of radiation fluence is
bulk traps. Therefore, for a given applied bias, the volume shown in Fig. 12. With the assumption that the oxide charge
of space charge region decreases with irradiation and all saturates at , a breakdown voltages of 900 V,
2984 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 57, NO. 5, OCTOBER 2010

Fig. 10. Guard ring structure featuring field plates extending towards the sensitive region. The width and spacing of p+ guard rings are identical to those in Fig. 1.

Fig. 12. Breakdown behavior of the guard ring structure with field plates as a
function of radiation fluence. The oxide thickness is 1 m and the field plate
length, L, which is defined in Fig. 10 is 6 m. The oxide charge is assumed to
2
saturate at 1 10 cm .

B. Oxide Thickness
The flat band voltage depends on the thickness of the oxide
between the field plates and silicon as well as the oxide
charge

(3)

where is the dielectric constant of . Therefore, the

Fig. 11. a) Electron density distribution (in cm ) and b) electric field dis- oxide thickness should play an important role in charge car-
tribution (in V/cm) around the two innermost guard rings. The oxide charge is
2
1 10 cm and oxide thickness is 1 m. The ratio of the horizontal scale to rier distribution and consequently in breakdown voltage. Simu-
vertical scale is 1/10. The white and brown lines represent the boundary of the lations with different oxide thicknesses were performed for var-
depletion region and the interface of p-type and n-type regions, respectively. ious oxide charge densities as shown in Fig. 13. For the sil-
icon regions underneath the field plates, only the product of
oxide charge and oxide thickness affects the carrier distribu-
1100 V, and 1400 V have been attained after fluences of tion. However, for the silicon regions not covered with field
, , and , respec- plates, the oxide thickness does not matter and the oxide charge
tively. Comparing with Fig. 9 which shows the simulation re- is the only parameter influencing the surface charge. For two
sults for the same fluences in the absence of field plates, it can be devices having the same product of oxide thickness and oxide
seen that the breakdown voltage has been enhanced significantly charge but different values of oxide charge and oxide thickness,
with the use of field plates. Furthermore, of the 300 the one with higher oxide charge density should have a lower
substrate thickness is depleted before the breakdown at flu- breakdown voltage because the potential drop at the locations
ence of while the detector reaches full deple- where electron channel is interrupted will be sharper. This ex-
tion well below the breakdown for fluences of plains why, for example, the breakdown voltage corresponding
and as indicated by the simulation. to an oxide thickness of 1.5 and oxide charge density of
KOYBASI et al.: GUARD RING SIMULATIONS FOR n-ON-p SILICON PARTICLE DETECTORS 2985

Fig. 14. Breakdown voltage versus field plate length as a function of oxide
Fig. 13. Breakdown behavior as a function of oxide charge for different oxide
2
thicknesses at a fluence of 1 10 n =cm .
2
charge at a fluence of 1 10 n =cm . The oxide thickness is 1 m.

the electron channel underneath the field plate is not compen-

is higher compared to the one corresponding sated due to the large shift in flat band voltage to the more neg-
to an oxide thickness of 1.0 and oxide charge density of ative values with increasing oxide charge, so essentially no sec-
although they have the same product. ondary electric field peak is formed underneath the field plate
As it can be seen in Fig. 13, the simulations indicate an optimum edge. The relatively small changes in breakdown voltage as a
oxide thickness of 1.5 and 1 for oxide charge densi- function of field plate length for these oxide charges are a con-
ties of and , respectively, re- sequence of the changes in the uniformity of electric field peaks
garding breakdown voltage. For equal to at the p-/p+ boundaries facing the sensitive region. The break-
the optimum thickness is 0.5 and for down voltages for the highest two values of oxide charge, which
the optimum is 0.2 . Roughly speaking, it seems that, for are below 250 V, can be enhanced by reducing the oxide thick-
a given , as the deviation of the product of and ness, and consequently, the shift in flat band voltage.
from increases, the breakdown voltage
decreases. IV. CONCLUSIONS
Device simulations have shown that optimized guard rings
C. Field Plate Length without field plates can withstand reverse bias voltages above
Obviously, the length of field plates influences the carrier dis- 1000 V without any radiation damage for an oxide charge
tribution in silicon. Therefore, the optimum field plate length of . A breakdown voltage above 900 V can
that yields the highest breakdown must be determined. The ef- be achieved with the proposed structure featuring field plates
fect of the field plate length is investigated as a function of oxide pointing toward the active region after a radiation fluence of
charge for an oxide thickness of 1 . Characterizing the field , assuming the oxide charge saturates at a
plate length by the distance L between the lateral boundary of density of . However, for oxide charge values of
the underlying p+ ring facing the p-n junction and the left end and above, the field plate solution to improve
of the field plate as depicted in Fig. 10, the breakdown voltage the breakdown after exposure to high ionizing radiation doses
versus field plate length for different oxide charges is shown in fails for an oxide thickness of 1 and the breakdown voltage
Fig. 14. The dependence of breakdown voltage on field plate decreases to below 250 V regardless of the field plate length.
length is non-monotonic. For oxide charges of For the field plate solution to become effective for these values
and , the breakdown voltage initially increases of oxide charge, the oxide thickness should be diminished
with field plate length as the maximum electric field spot at the to reduce the flat band voltage shift to such a value that the
p-/p+ boundary facing the active region decreases while the sec- potential difference between field plate and underlying silicon
ondary electric field peak underneath the left edge of the field can compensate the electron channel at the silicon/oxide in-
plate increases, and therefore, the electric field distribution over terface. However, this will be at the expense of decreasing the
the two spots of the guard ring becomes more even. This trend breakdown voltage for lower oxide charges. Therefore, the road
continues up to field plate lengths of 4 and 5 for oxide map for further optimization will be identified after measuring
charges of and , respectively, the saturated oxide charge density on the fabricated devices.
with corresponding breakdown voltages of 980 V and 680 Our future work will also address the optimization of guard
V. The decrease of the breakdown voltage with further increase rings (within the limits of design rules) for fluences above
of the field plate length is a result of the changing uniformity of to see the upper fluence limit at which these
the electric field distribution over all guard rings as the amount devices can be operated. Due to the lack of experimental data
of potential drop at a certain guard ring affects the potential dis- to support most of the simulation observations, test structures
tribution over the ones on its right. For oxide charge values of that feature different doping profiles, guard ring dimensions
and , it has been observed that and spacing, field plate lengths, and oxide thicknesses will be
2986 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 57, NO. 5, OCTOBER 2010

fabricated at SINTEF to confirm if the experimental behavior [12] M. Da Rold, A. Paccagnella, A. Da Re, G. Verzellesi, N. Bacchetta,
follows the same trend as the predictions of the simulation. In and R. Wheadon et al., “Radiation effects on breakdown characteristics
of multiguarded devices,” IEEE Trans. Nucl. Sci., vol. 44, no. 3, pp.
case of any offset between the simulation and experimental 721–727, Jun. 1997.
data, the parameters of the models will be tuned to improve the [13] B. S. Avset and L. Evensen, “The effect of metal field plates on multi-
convergence of the theory to the measurement results. guard structures with floating p+ guard rings,” Nucl. Instrum. Meth. A,
vol. 377, pp. 397–403, 1996.
[14] M. Da Rold, N. Bacchetta, D. Bisello, A. Paccagnella, G. F. D. Betta,
ACKNOWLEDGMENT and G. Verzellesi et al., “Study of breakdown effects in silicon multi-
guard structures,” IEEE Trans. Nucl. Sci., vol. 46, no. 4, pp. 1215–1223,
The authors would like to thank T.-E. Hansen for helpful Aug. 1999.
discussions. [15] M. Boscardin, L. Bosisio, A. Candelori, G. F. D. Betta, S. Dittongo,
and P. Gregori et al., “An improved termination structure for silicon
radiation detectors with all p-type multiguard and cut-line implants,”
REFERENCES IEEE Trans. Nucl. Sci., vol. 50, no. 4, pp. 1001–1007, 2003.
[1] G. Bolla, D. Bortoletto, C. P. Grim, R. L. Lander, Z. Li, and S. Willard, [16] C. Y. Chien, H. S. Cho, G. W. Liang, X. B. Xie, W. Huang, and Z. Li,
“First results on radiation damage studies using n+/p/p+ diodes fabri- “New designs of CMS silicon pixel detectors for radiation hardness,”
cated with multi-guard ring structures,” Nucl. Instrum. Meth. A, vol. Nucl. Instrum. Meth. A, vol. 455, pp. 564–575, 2000.
435, pp. 178–186, 1999. [17] M. Benoit, A. Lounis, and N. Dinu, “Simulation of radiation damage
[2] G. Casse, P. P. Allport, P. R. Turner, S. Marti i Garcia, and M. Lozano, effects on planar pixel guard ring structure for ATLAS inner detector
“Performances of miniature microstrip detectors made on oxygen en- upgrade,” IEEE Trans. Nucl. Sci., vol. 56, no. 6, pp. 3236–3243, Dec.
riched p-type substrates after very high proton irradiation,” Nucl. In- 2009.
strum. Meth. A, vol. 535, pp. 362–365, 2004. [18] O. Koybasi, D. Bortoletto, T. E. Hansen, A. Kok, T. A. Hansen, and N.
[3] G. Casse, P. Allport, and A. Watson, “Effects of accelerated annealing Lietaer et al., “Design, simulation, fabrication, and preliminary tests of
on p-type silicon micro-strip detectors after very high doses of proton 3D CMS pixel detectors for the Super-LHC,” IEEE Trans. Nucl. Sci.,
irradiation,” Nucl. Instrum. Meth. A, vol. 568, pp. 46–50, 2006. Aug. 2010, to be published.
[4] G. Casse, P. Allport, and A. Greenall, “Response to minimum ionising [19] “Synopsys TCAD Manuals” Synopsys Inc., 2007 [Online]. Available:
particles of p-type substrate silicon microstrip detectors irradiated with [Link]
neutrons to LHC upgrade doses,” Nucl. Instrum. Meth. A, vol. 581, pp. [20] Y. Okuto and C. R. Crowell, “Threshold energy effect on avalanche
318–321, 2007. breakdown voltage in semiconductor junction,” Solid-State Elec-
[5] A. Affolder, P. Allport, and G. Casse, “Studies of charge collection ef- tronics, vol. 18, no. 2, pp. 161–168, 1975.
ficiencies of planar silicon detectors after doses up to 10 n cm [21] A. K. Srivastava, A. Bhardwaj, K. Ranjan, Namrata, S. Chatterji, and
and the effect of varying diode configurations and substrate types,” R. K. Shivpuri, “Two-dimensional breakdown voltage analysis and op-
Nucl. Instrum. Meth. A, vol. 604, pp. 250–253, 2009. timal design of a microstrip detector passivated by a dielectric,” Semi-
[6] G. Kramberger, M. Batic, V. Cindro, I. Mandic, M. Mikuz, and M. cond. Sci. Technol., vol. 17, no. 5, pp. 427–434, May 2002.
Zavrtanik, “Annealing studies of effective trapping times in silicon de- [22] H. F. W. Sadrozinski, C. Betancourt, R. Heffern, I. Henderson, J.
tectors,” Nucl. Instrum. Meth. A, vol. 571, pp. 608–611, 2007. Pixley, and A. Polyakov et al., “Total dose dependence of oxide
[7] G. Kramberger, V. Cindro, I. Mandic, M. Mikuz, and M. Zavrtanik, charge, interstrip capacitance and breakdown behavior of sLHC
“Influence of trapping on silicon microstrip detector design and perfor- prototype silicon strip detectors and test structures of the SMART
mance,” IEEE Trans. Nucl. Sci., vol. 49, no. 4, pp. 1717–1723, 2002. collaboration,” Nucl. Instrum. Meth. A, vol. 579, pp. 769–774, 2007.
[8] J. Weber and R. Klingenberg, “Free charge carriers trapping properties [23] M. Petasecca, F. Moscatelli, D. Passeri, and G. U. Pignatel, “Numerical
in neutron-irradiated DOFZ silicon pad detectors,” IEEE Trans. Nucl. simulation of radiation damage effects in p-type and n-type FZ silicon
Sci., vol. 54, no. 6, pp. 2701–2705, 2007. detectors,” IEEE Trans. Nucl. Sci., vol. 53, no. 5, pp. 2971–2976, 2006.
[9] L. Rossi, P. Fischer, T. Rohe, and N. Wermes, Pixel Detectors: From [24] D. D. Pitzl, N. Cartiglia, B. Hubbard, J. Leslie, K. O’Shaugnessy, and
Fundamentals to Applications. Germany: Springer, 2006. W. Rowe et al., “Study of radiation effects on AC coupled silicon strip
[10] L. Evensen, A. Hanneborg, B. S. Avset, and M. Nese, “Guard ring detectors,” Nucl. Phys. Proc. Suppl., vol. 23B, pp. 340–346, 1991.
design for high voltage operation of silicon detectors,” Nucl. Instrum. [25] D. Pennicard, C. Fleta, R. Bates, V. O’Shea, C. Parkes, G. Pellegrini,
Meth. A, vol. 337, pp. 44–52, 1993. and N. Tartoni, “Simulations of radiation-damaged 3D detectors for the
[11] G. A. Beck, A. A. Carter, T. W. Pritchard, J. R. Carter, D. J. Munday, Super-LHC,” Nucl. Instrum. Meth. A, vol. 592, pp. 16–25, 2008.
and D. Robinson et al., “Radiation-tolerant breakdown protection of [26] M. Moll, “Radiation damage in silicon particle detectors—microscopic
silicon detectors using multiple floating guard rings,” Nucl. Instrum. defects and macroscopic properties,” Ph.D. dissertation, Hamburg,
Meth. A, vol. 396, pp. 214–227, 1997. 1999.