jap 2003 activation energy anna lys i stool
TRANSCRIPT
-
7/28/2019 Jap 2003 Activation Energy Anna Lys i Stool
1/4
Activation energy analysis as a tool for extraction and investigation of pnjunction leakage current components
A. Czerwinskia)
Institute of Electron Technology, Al. Lotnikow 32/46, 02-668 Warsaw, Poland
E. SimoenIMEC, Kapeldreef 75, B-3001 Leuven, Belgium
A. Poyai and C. ClaeysIMEC, Kapeldreef 75, B-3001 Leuven, Belgium and E. E. Department, KU Leuven,Kasteelpark Arenberg 10, B-3001 Leuven, Belgium
Received 3 March 2003; accepted 23 April 2003
The origin of p n junction reverse current is investigated by a method based on the analysis of the
leakage current activation energy. Its main advantages lie in the possibility to distinguish multiple
reverse-bias dependent leakage components and determine their mechanisms, which can be different
than the ShockleyRead Hall or field-enhanced generation mechanisms. This is illustrated for
state-of-the-art silicided shallow junctions, exhibiting a local Schottky effect, due to small-area
silicide penetrations. An estimate of the area of the Schottky or Shannon contacts follows from the
analysis. The method may be used for various semiconductor materials and leakage current origins.
2003 American Institute of Physics. DOI: 10.1063/1.1582553
I. INTRODUCTION
State-of-the-art complementary metaloxide
semiconductor CMOS ultralarge scale integration circuits
ULSI ICs contain a huge number of p n junctions. The
main ULSI IC parameters, like its leakage current, the
standby power limits and the dynamic random access
memory DRAM retention time, are determined by the p n
junction reverse leakage current (IR). As the integration
density continues to increase, the problem of leakage current
becomes more serious.1,2 Many techniques can be used to
analyze the current and the related lifetime,3 e.g., noncontactand nondestructive techniques, like microwave photoconduc-
tance decay and surface photovoltage. But these methods
investigate the sample volume over a distance of several mi-
crons, inappropriate in the case of small-size junctions in
ULSI ICs, where the leakage current originates mostly from
junction peripheries, i.e., edges and corners.4,5 The peripheral
current is measured on fully processed devices, and the
methods based on I V measurements are listed among the
best from a point of view of sensitivity.2,3 Here, an improved
method using I V measurements and analyses of the leak-
age current activation energy (Ea) will be shown, enabling
the separation of leakage current components with a differentphysical origin.
II. EXPERIMENT
Shallow np diodes compatible with submicron 0.18
m CMOS technology have been processed on Czochralski
150 mm p-type substrates. Shallow trench isolation STI
was applied. A retrograde p well was obtained by a deep
200 keV, 1.21013 cm2) and a shallow 55 keV, 1.5
1013 cm2) boron ion implantation, followed by a dopant
activation anneal 10 min, 850 C . The estimated projected
range (Rp) was 0.5 and 0.2 m below the silicon sur-
face, for the deep and shallow boron implantation, respec-
tively. The n region was made by an arsenic ion implanta-
tion 70 keV, 41015 cm2) and subsequent anneal 10 s,
1100 C . A junction depth around 0.1 m is expected. The
backend of the process consists of a cobalt silicide with tita-
nium capping layer, a tetraethylorthosilicate intermetal di-
electric layer and an Al Si Cu metallization. The maximum
silicidation temperature was 700 C. Reference junctions
were used, which were manufactured on Czochralski-grown
wafers with high initial oxygen concentration, internally get-
tered using a high-low-high temperature cycle and sur-
rounded by a local oxidation of silicon LOCOS isolation.
The internal gettering cycle consisted of a two-step pretreat-
ment, i.e., an oxygen outdiffusion at 1100 C for 6 h, and a
nucleation step at 750 C for 8 h under N2 atmosphere. Next,
diode processing was started by the field oxidation at 975 C
for 10 h, which served as the oxygen precipitation step. No p
well was fabricated in this case. The n region was obtained
by an arsenic ion implantation 70 keV, 31015 cm2) and
subsequent anneal 10 s, 1100 C30 min, 800 C . A stan-
dard Al metallization without a silicidation was applied.6
Junctions with different geometry were studied. The area
(A), the perimeter ( P), and the number of corners (NC) of
the measured test structures: large area SQ1 and medium
area SQ2 rectangles and meander ME1 junctions were:
A0.1, 0.001, 0.001 cm2, P1.3, 0.13, 8.04 cm, and NC4, 4, 320, respectively. The total leakage current IR is a
linear combination of the different components, given by:
IRAJAPJPNCJCIpar ,4 where JA (A/cm
2),
JP (A/cm), and JC A/corner are the planar, perimeter and
corner current densities and Ipar is a parasitic current.4 In thea Electronic mail: [email protected]
JOURNAL OF APPLIED PHYSICS VOLUME 94, NUMBER 2 15 JULY 2003
12180021-8979/2003/94(2)/1218/4/$20.00 2003 American Institute of Physics
Downloaded 12 Apr 2007 to 132.234.251.211. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
-
7/28/2019 Jap 2003 Activation Energy Anna Lys i Stool
2/4
first approximation, a combination of the currents measured
for a pair of neighboring structures allows to separate one of
the geometrical components. For example, the subtraction of
the SQ2 current from the SQ1 current leaves mainly the pla-
nar leakage current ( P/A ratio11.8), while the subtraction
of the SQ2 from the ME1 current yields the peripheral leak-
age current. The temperature-dependent I V static measure-
ments of diodes were done in the dark in the range from 15
to 120 C, with the temperature stability 0.1 C. High-
frequency 100 kHz and 1 MHz capacitancevoltage mea-
surements yielded a substrate doping of about 61014 cm3 for LOCOS and 2.51017 cm3 for STI di-
odes.
III. ACTIVATION ENERGY METHOD
A general relation between a physical variable and its
activation energy can be applied to the leakage current of a
p n junction, according to
IR T exp Ea /kT ] 1
with k as the Boltzmann constant, and T as the absolute
temperature. The slope of an Arrhenius plot, IR(T) vs 1/kT
yields the activation energy Ea .7
It is often assumed that EaDIFEg and EaGENEg/2,
where EaDIF and EaGEN are the activation energies for the
diffusion (IR ,DIF) and generation (IR ,GEN) currents, while Eg
is the silicon band gap, Eg(T)1.206 2.73104T eV .8
In fact, the dependence of measured leakage currents on tem-
perature is not purely exponential, as nonexponential prefac-
tors are present, e.g., due to the intrinsic concentration n iT3/2.8,9 The widely applied IR Ea relation from Eq. 1
implies that the thus derived Ea is different from the activa-
tion energy of the exponential term alone and is, therefore, to
be considered an effective value. In fact, ln(IR,GEN)Eg/2
(ETEi) ,10 when the impact of the electric field ( F) can
be neglected, where ET and Ei are the trap and the intrinsic
energy level, respectively. Only under the influence of the
electric field in the depletion region, EaGEN decreases, first
towards Eg/2, while for very high field conditions, values as
low as 0.150.25 eV are reported, as visible in Fig. 1.
In the literature, the simple Arrhenius plot representation
at a fixed reverse bias is generally used to extract the activa-
tion energy, even when the tunneling origin of the measured
leakage current has been acknowledged.11,12
Only in a fewcases the impact of the electric field in the junction on the
activation energy of the leakage current has been studied to
some detail, for standard p n diodes,13,14 or as a function of
the gate voltage for gated diodes,15 with fitting procedures
applied to determine the energy level of the dominant
generation-recombination centers,13,14 to derive the donor- or
acceptor-like nature of these centers,14,15 or to detect local
electric-field enhancement due to defects.14
In order to find out the origin of the junction leakage
current, techniques are proposed, which are based on a com-
bination of I V characteristics measured at nearby tempera-
tures T and TT, (T10 C in this work . First, instead
of a tedious application of the traditional Arrhenius represen-tation, one can use a pair of I V characteristics with the
effective activation energy characteristics Ea(VR) derived
directly from Eq. 1 , as follows:16
Ea 1/IR T dIR /d 1/kT . 2
Second, a pair of such Ea(VR) characteristics related to
two temperatures will be used to discriminate different
physical sources of leakage current. Each geometrical com-
ponent respectively, peripheral, or planar contains IR,DIF0 ,
the saturation diffusion current, independent of VR , and usu-
ally found by a linear extrapolation of IR at low biases to
zero depletion width.17
The leakage current obtained aftersubtraction of IR ,DIF0 from IR is usually and often wrongly
attributed to the IR ,GEN generation current. Instead, it will be
labeled IR,QG , and the related activation energy as EaQG ,
with QG for quasigeneration current, because IR ,GEN is
only a part of this remnant leakage current. Although the
analysis can be generalized to more than one additional leak-
age current source IR,ADD , for the sake of simplicity it is
assumed here that there are only two dominant bias-
dependent leakage-current components of different origin.
Although the physical mechanisms are generally unknown, it
is highly probable that one of them is the generation current,
thus they will be labeled IR,ADD and IR ,GEN . In other words,
IRIR ,DIF0IR ,QG , with IR ,QGIR,GENIR ,ADD .When IR,ADD is attributed to the nonconstant diffusion
current, dependent on VR and different from IR,DIF0 , then a
method with two characteristics Ea(VR ,T) and Ea(VR ,T
T) can be used for its determination.16 Such a noncon-
stant diffusion current may be found in junctions on IG,17 or
epitaxial substrates,17 in the peripheral leakage current of
LOCOS junctions,16 etc. But generally, the origin of IR ,ADD ,
characterized by an activation energy EaADD , is a priori un-
known. The presence of IR ,ADD not only increases the overall
leakage current of junctions i.e., deteriorates ICs perfor-
mance , but it also much disturbs the suppression of this
current. The determination of junctioncurrent sources is
FIG. 1. Activation energies for leakage currents of a silicided STI junction:
solid lines with small symbols related to EaQG(VR) of the quasigeneration
current IR,QG upperperipheral and lowerplanar determined from ex-
perimental I V characteristics after subtraction of the saturation current;
and full for peripheral components or empty for planar components sym-
bols related to EaADD(VR) and EaGEN(VR) energies of respectively, IR,ADDand IR,GEN currents. The latter currents are determined with the use of the
proposed method from IR,QG(VR) characteristics measured at various tem-
peratures, as shown in the inset for the peripheral current.
1219J. Appl. Phys., Vol. 94, No. 2, 15 July 2003 Czerwinski et al.
Downloaded 12 Apr 2007 to 132.234.251.211. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
-
7/28/2019 Jap 2003 Activation Energy Anna Lys i Stool
3/4
very difficult then, due to the mixed origin of measured val-
ues. No previous method enables an extraction of multiple
bias-dependent leakage current components of unknown ori-
gin. The proposed approach allows to extract them and find
out their origin using the knowledge of EaADD(VR) and
EaGEN(VR).
The proposed method considers two different leakage
current components, dependent on the reverse bias apart of
IR,DIF0) and characterized by unknown activation energies,defined by Eq. 1 . Using Eq. 2 , one can show that:
Ea,QG(IR ,GEN EaGENIR ,ADD EaADD)/(IR,GENIR,ADD).
Its derivative, with EaGEN and EaADD dependencies on T as-
sumed negligible,16,18 gives dEaQG/d(1/kT)IR,ADD GEN EaADD GENEaQG
2/ IR,QGIR ,ADD GEN , which finally
leads to
IR,ADD GENIR ,QG dEaQG/d 1/kT / EaADD GEN
EaQG2 dEaQG/d 1/kT 3
with dEaQG/d(1/kT) determined from two EaQG(T) charac-
teristics. Two solutions are possible, related to the IR ,ADD and
IR,GEN currents, as underlined by ADD GEN notation.Knowing one of them, the second one results from IR,QGIR ,GENIR,ADD . The true physical meaning of each solu-
tion needs to be revealed, based upon the extracted Ea values
and their dependence on bias.
When EaADD GEN is known, Eq. 3 alone yields
IR,ADD GEN (VR) for each voltage VR . For the general case of
unknown EaADD GEN , an input EaADD GEN initial guess is
assumed in Eq. 3 , and IR,ADD GEN (T1) and IR ,ADD GEN (T2)
currents for two temperatures T1 and T2 are calculated using
experimental IR ,QG results. From these two IR ,ADD GEN (T)
values, an output EaADD GEN (VR) is found with the use of
Eq. 2 . The iteration procedure, leading independently for
each VR from the initial guess to the solution, stops whenboth input and output EaADD GEN (VR) values coincide within
an assumed accuracy.
IV. RESULTS AND DISCUSSION
The method applied to the peripheral leakage current of
nonsilicided LOCOS Czochralski-Si junctions gives an al-
most constant EaADD(VR) close to 1.2 eV, which is a EaDIFvalue being often found from measurements for silicon at
room temperature.19 The attribution of IR,ADD(VR) to a bias-
dependent diffusion current is also confirmed by the indepen-
dent results of the high-temperature method.16 On the other
hand, a large diversity of IR ,ADD(VR) and EaADD(VR) valuesis found for silicided STI junctions. For the peripheral leak-
age current, a significant IR ,ADD component is revealed,
which is attributed to a Schottky current (IR,Sch), because of
EaADD values and their dependence on bias shown in Fig. 1.
For silicided p n junctions small regions of silicide penetra-
tion e.g., so called silicide spikes are often reported, which
results in an excessive leakage current,2024 typical for either
a Schottky or quasi-Schottky Shannon junction,20 depend-
ing on the silicide-penetration length and the junction depth.
From the IR ,Sch magnitudes, found at different tempera-
tures and at the same reverse bias e.g., VR0.4 or 0.8 V in
Fig. 2 , i.e., at constant electric field, the Schottky-contact
area (A S) is extracted from the extrapolation of the plot
ln(IR,Sch/T2A** )ln(AS)q B0(qF/4)
1/2 /kT to 1/kT
0, where A** is the effective Richardson constant, q is the
elementary charge, B0 is the asymptotic Schottky barrier
height at zero electric field and is the semiconductor per-
mittivity. For low biases contact areas in the range below 1
m2 are revealed Fig. 2 , which is in agreement with the
literature data.21 However, the approach used before was
based on the analysis of the forward current,21 which is in-
applicable here because of the smaller partial contribution
of the Schottky current. As visible in Fig. 2, the contact area
increases at higher VR e.g., two times at VR change from 0.4to 0.8 V , when the junction depletion-region extends to-
wards shorter shallower silicide penetrations, so also at the
change from forward to reverse bias. The slope of the
ln(IR,Sch/T2A** ) plot at small bias corresponds to the barrier
height at this bias, found to be slightly above 0.6 eV. When
silicide penetrations are in direct contact with the low-doped
p substrate or well region, a localized Schottky junction oc-
curs. The reported barrier heights for CoSi2 contacts to low-
doped p-Si range from 0.4221 to 0.48 eV.22 If the silicide is in
contact with the junction depletion region, but not with the
low-doped substrate, then higher values are reported, depen-
dent on depths of the junction and of the silicide penetra-
tions, e.g., 0.86 eV for CoSi2 /p-Si low doped .23 The effec-tive activation energy EaADD is increased in comparison with
the barrier height by the presence of the T2 prefactor in the
Schottky current equation,9 while, on the other hand, it is
lowered by the dependence on the electric field.
Leakage-current densities for Schottky or Shannon junc-
tions are much larger than in a standard p n junction,20 so
even a very small Schottky contact impacts the leakage cur-
rent, especially in the case of very good quality p n junc-
tions. Although the presence of various leakage mechanisms
have been often suggested, previous evaluations of the
Schottky leakage current concerned only cases when the
whole p n junction leakage current was related to silicide
FIG. 2. The IR,ADD/T2A** plot for a IR,ADD component of peripheral leak-
age current in a silicided STI junction, found at different temperatures and at
the same reverse bias ( VR0.4 or 0.8 V , and attributed to the Schottky-
contact current (IR,Sch). The Schottky-contact area at various biases is ex-
tracted as the crossing of the plot with the ordinate axis. The slope of the
plot for small bi as corresponds to the barrier height . A**
120 A cm2 K2 is applied. The inset presents the percentage contribu-
tion of this IR,Sch current to the IR,QG current at 45 C.
1220 J. Appl. Phys., Vol. 94, No. 2, 15 July 2003 Czerwinski et al.
Downloaded 12 Apr 2007 to 132.234.251.211. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp
-
7/28/2019 Jap 2003 Activation Energy Anna Lys i Stool
4/4
penetration.12,20,23,25 A clear presence of only a partial con-
tribution of IR ,Sch to the overall leakage is revealed for the
peripheral leakage Figs. 1 and 2 . When the temperature
decreases, the IR ,GEN contribution becomes relatively more
dominant. Besides the barrier height decrease with electric
field, and the revealed A S increase with VR , the barrier
height of a Shannon diode also changes when the junction
depletion region extends with VR .20
The inset in Fig. 2 shows comparable contributions ofthe Schottky and the generation currents for the case of pe-
ripheral leakage current shown in Fig. 1. They constitute also
most of the overall reverse current of silicided source/drain
junctions, due to the general dominance of the peripheral
component in small ULSI junctions, and the insignificance of
peripheral IR,DIF0 saturation current, as revealed. For higher
temperatures, in the range of 70 80 C, typical for DRAM
operation, the peripheral current shows the dominant contri-
bution of IR ,Sch and still a negligible IR,DIF0 . Oppositely,
IR,DIF0 dominates in the planar junctions in this temperature
range.
The vulnerability of junction peripheral regions in com-
parison with planar ones to the creation of silicide penetra-tion, are supposedly related to critical conditions e.g., an
increased mechanical stress of the silicide layer at junction
peripherals, close to the STI isolation. The silicide penetra-
tions may be also the cause of a stronger tunnelingcurrent
increase and activation energy decrease with increasing VR ,
because of the electric-field enhancement occurring in the
vicinity of defects.14 Finally, it is believed that the method
can be used for various semiconductor materials and leakage
current origins.
ACKNOWLEDGMENT
A.C. acknowledges partial financial support by the StateCommittee for Scientific Research, Poland, under Grant No.
7 T11B 084 20.
1 K. Kim, C.-G. Hwang, and J. G. Lee, IEEE Trans. Electron Devices 45,
598 1998 .2 H. Uchiyama, K. Matsumoto, T. Mchedlidze, N. Nisimura, and K. Yam-
abe, J. Electrochem. Soc. 146, 2322 1999 .3 J. E. Park, D. K. Schroder, S. E. Tan, B. D. Choi, M. Fletcher, A. Bucz-
kowski, and F. Kirscht, J. Electrochem. Soc. 148, G411 2001 .4 A. Czerwinski, E. Simoen, C. Claeys, K. Klima, D. Tomaszewski, J.
Gibki, and J. Katcki, J. Electrochem. Soc. 145, 2107 1998 .5 H.-D. Lee, S.-G. Lee, S.-H Lee, Y.-L. Lee, and J.-M. Hwang, Jpn. J. Appl.
Phys., Part 1 37, 1179 1998 .6 J. Vanhellemont, E. Simoen, A. Kaniava, M. Libezny, and C. Claeys, J.
Appl. Phys. 77, 5669 1995 .7 A. Poyai, E. Simoen E, C. Claeys, A. Czerwinski, and E. Gaubas, Appl.
Phys. Lett. 78, 1997 2001 .8 M. A. Green, J. Appl. Phys. 67, 2944 1990 .9 S. M. Sze, Physics of Semiconductor Devices Wiley Interscience, New
York, 1981 .10 D. K. Schroder, IEEE Trans. Electron Devices 29, 1336 1982 .11 T. Hamamoto, S. Sugiura, and S. Sawada, IEEE Trans. Electron Devices
45, 1300 1998 .12 H.-D. Lee, IEEE Trans. Electron Devices 47, 762 2000 .13 M. J. J. Theunissen and F. J. List, Solid-State Electron. 28, 417 1985 .14 A. Czerwinski, Appl. Phys. Lett. 75, 3971 1999 .15 M. Rosar, B. Leroy, and G. Schweeger, IEEE Trans. Electron Devices 47,
154 2000 .16 A. Czerwinski, E. Simoen, A. Poyai, and C. Claeys, J. Appl. Phys. 88,
6506 2000 .17 Y. Murakami and T. Shingyouji, J. Appl. Phys. 75, 3548 1994 .18 S. Zhu, X.-P. Qu, R. L. Van Meirhaeghe, C. Detavernier, G.-P. Ru, F.
Cardon, and B.-Z. Li, Solid-State Electron. 44, 2217 2000 .19 H. Aharoni, T. Ohmi, M. M. Oka, A. Nakada, and Y. Tamai, J. Appl. Phys.
81, 1270 1997 .20 E. C. Jones and N. W. Cheung, J. Vac. Sci. Technol. B 14, 236 1996 .21 Q. Wang, C. M. Osburn, and C. A. Canovai, IEEE Trans. Electron Devices
39, 2486 1992 .22 B. S. Chen and M. C. Chen, IEEE Trans. Electron Devices 43, 258
1996 .23 G. P. Schwartz and G. J. Gualtieri, J. Electrochem. Soc. 133, 1266 1986 .24 A. Czerwinski, J. Katcki, A. Poyai, E. Simoen, C. Claeys, J. Ratajczak,
and E. Gaubas, Mater. Sci. Semicond. Process. 4, 105 2001 .25 D. Z. Chi, W. D. Wang, S. J. Chua, and S. Ashok, J. Appl. Phys. 92, 7532
2002 .
1221J. Appl. Phys., Vol. 94, No. 2, 15 July 2003 Czerwinski et al.
Downloaded 12 Apr 2007 to 132.234.251.211. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp