![Schematics of the defect distribution within a given volume. (a) Distribution of point defects after ion implantation. (b) Distribution of DLs (blue circles) after high T annealing. (c) Areal density of DLs [ corresponds to a plan view of (b) ] . (d) Dislocation line density.](https://www.researchgate.net/profile/Jan-Kruegener/publication/269688468/figure/fig1/AS:295279757676548@1447411588617/Schematics-of-the-defect-distribution-within-a-given-volume-a-Distribution-of-point_Q320.jpg)
Content uploaded by Jan Krügener
Author content
All content in this area was uploaded by Jan Krügener on Dec 17, 2014
Content may be subject to copyright.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
IEEE JOURNAL OF PHOTOVOLTAICS 1
Electrical and Structural Analysis of Crystal Defects
After High-Temperature Rapid Thermal Annealing
of Highly Boron Ion-Implanted Emitters
Jan Kr¨
ugener, Robby Peibst, F. Alexander Wolf, Eberhard Bugiel, Tobias Ohrdes, Fabian Kiefer, Claus Sch¨
ollhorn,
Andreas Grohe, Rolf Brendel, and H. J¨
org Osten
Abstract—Ion implantation of boron is a promising technique
for the preparation of p-type emitters in n-type cells. We use rapid
thermal annealing with temperatures up to 1250 °C and annealing
durations between 6 s and 20 min to anneal the implant-induced
crystal defects. Experimental J0eis compared with simulated and
measured defect densities. Perfect dislocation loops are identified
to be the dominating defect species after rapid thermal annealing
(RTA) above 1000 °C. Even for emitters with J0evalues around
40 fA/cm2, defects are present within the valleys of the textured sur-
faces after annealing. On textured Al2O3-passivated boron emit-
ters, we measure J0eof 38 fA/cm2for a sheet resistance around
80 Ω/after very short annealing processes (1 min at 1200 °C).
Index Terms—Boron, crystal defects, ion implantation, photo-
voltaic, rapid thermal annealing (RTA), silicon.
I. INTRODUCTION
COMMONLY, p-type Czochralski-grown (Cz) substrates
are used for monocrystalline silicon solar cells in produc-
tion. Unfortunately, p-type Cz wafers suffer from degradation
problems caused by the presence of boron–oxygen complexes
[1].
For high-efficient solar cells, one has to avoid the aforemen-
tioned problems. Several approaches to solve these problems
are possible, e.g., an improvement of the p-type material (lower
oxygen concentration, change to different group III dopant) or
a change from p-type to n-type substrates. n-type substrates re-
Manuscript received July 11, 2014; revised September 3, 2014 and October
14, 2014; accepted October 22, 2014. This work was supported by the German
Federal Ministry for Economic Affairs and Energy under Contract 0325480C.
J. Kr¨
ugener and E. Bugiel are with the Institute of Electronic Materials and
Devices, Leibniz University Hannover, D-30167 Hannover, Germany (e-mail:
kruegener@mbe.uni-hannover.de; bugiel@mbe.uni-hannover.de).
R. Peibst, T. Ohrdes, and F. Kiefer are with the Institute for Solar Energy
Research Hamelin, D-31860 Emmerthal, Germany (e-mail: peibst@isfh.de;
ohrdes@isfh.de; kiefer@isfh.de).
F. A. Wolf is with BOSCH Corporate Research, D-70839 Gerlingen-
Schillerh¨
ohe, Germany (e-mail: f.alex.wolf@gmx.de).
C. Sch¨
ollhorn and A. Grohe are with Bosch Solar Energy, D-99310
Arnstadt, Germany (e-mail: claus.schoellhorn@bosch.com; andreas.grohe@
directphotonics.com).
R. Brendel is with the Institute for Solar Energy Research, Hamelin and
with the Laboratory of Nano and Quantum Engineering, Leibniz University
Hannover, D-30167 Hannover, Germany (e-mail: brendel@isfh.de).
H. J. Osten is with the Institute of Electronic Materials and Devices, Germany
as well as with the Laboratory of Nano and Quantum Engineering both from the
Leibniz University Hannover, D-30167 Hannover, Germany (osten@mbe.uni-
hannover.de)
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JPHOTOV.2014.2365468
quire suitable doping techniques for the preparation of p-type
emitters with low saturation current densities. Due to the com-
mercial availability of tools for ion implantation of solar cells,
ion implantation has recently attracted a lot of attention [2]–
[4]. Compared with classical emitter formation by diffusion,
ion implantation offers the possibility of single-sided and even
laterally patterned doping (e.g., for selective emitters) without
sacrificial layers (e.g., phosphorus silicate glass), which could
lead to a simplification of the overall process flow. Furthermore,
problems related with the p-type emitter formation by boron dif-
fusion (e.g., lifetime degradation by boron-rich layer formation
[5]) can be avoided by using ion implantation.
However, ion implantation creates a large amount of crystal
defects. These defects may act as recombination centers them-
selves or may mediate recombination by capturing metal impu-
rities [6], [7]. In order to cure the implant-induced defects and,
furthermore, to activate the implanted ions, a high-temperature
Ttreatment is needed after implantation. In microelectronics,
rapid thermal annealing (RTA) at Taround 1000 °Cforsome
seconds is used for the activation of implanted dopants [8]. In
microelectronic industry, it is desirable to use short annealing
procedures to avoid broadening of doping profiles (i.e., to keep
the dopant diffusion length short). However, such a thermal bud-
get is not sufficient to dissolve all defects; thus, residual defects
remain within the implanted region [9].
Since broadening of doping profiles is not as critical in pho-
tovoltaics, a higher thermal budget can be applied for dissolving
the residual defects. Here, two ways exist to increase the thermal
budget: 1) increasing the temperature or/and 2) increasing the
annealing duration. So far, experimental studies were performed
in quartz furnaces, implying maximum Tof 1050 °C and long
annealing times up to 1 h (see [10]). Higher T, which can be
achieved by using RTA, could lead to a faster dissolution of crys-
tal defects and, thus, enables shorter process duration. The high
process flexibility of RTA allows covering a wide range of ther-
mal budgets and is, therefore, well-suited to evaluate the needed
thermal budget for the dissolution of implant-induced defects.
Here, we present the investigations of the influence of RTA
of boron ion-implanted silicon on the electrical and structural
emitter properties. For symmetrical double-sided textured sam-
ples, the emitter saturation current densities (J0e) are mea-
sured using the method of Kane and Swanson [11]. The ob-
tained results are compared with numerical device simulations
with respect to doping profiles obtained from process simula-
tions (defect and dopant distribution after ion implantation and
2156-3381 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2IEEE JOURNAL OF PHOTOVOLTAICS
subsequent annealing). We compare the experimental J0ewith
calculated defect densities from process simulations and find
correlations. The simulated defect densities are calculated us-
ing a new model, which is adapted to the high temperatures
we used here [12]. The occurrence of structural defects and
their distribution is experimentally investigated using trans-
mission electron microscopy (TEM). Perfect dislocation loops
(DLs) are identified to be the dominating defect species after
RTA above 1000 °C. Even for emitters with J0evalues around
40 fA/cm2, defects are present within the valleys of the textured
surfaces after annealing.
II. EXPERIMENT AND MODELING
We prepare symmetrical p+-n-p+samples by double-sided
ion implantation of elemental boron (2 ×1015 cm−2at 10 keV)
through a 20 nm-thick SiO2screening oxide. Prior to the im-
plantation, the 6 Ω·cm n-type Cz substrates (156 ×156 mm2
pseudosquare, 180 μm thick as-cut) are saw damage-etched, tex-
tured, and standard RCA [13] cleaned followed by a dry thermal
oxidation. During ion implantation (VIISta HCS from Varian),
the substrates are tilted by 6°toward 100. This angle is
caused by the used implant setup. Subsequent defect annealing
is conducted in a standard RTA tool (SHS-2900 from Mattson)
under inert ambient (dry N2) for temperatures between 900 and
1250 °C and annealing durations tbetween 6 s and 20 min
(plateau time). Reference samples are annealed in conventional
quartz furnace for 20 min at 1050 °C, since these parameters
are known to result in good electrical properties [14].
The samples are passivated on both sides by 15 nm Al2O3
deposited by atomic layer deposition plus a subsequent forming
gas anneal at 350 °C for 10 min. Effective lifetimes, emitter
saturation current densities, and sheet resistances (Rsheet)are
measured by quasi-steady-state photo conductance method.
With SENTAURUS process [15], we simulate the resulting
dopant and defect distribution after ion implantation and sub-
sequent annealing. The textured surface of the wafers has a
significant impact on the occurrence of implant-induced de-
fects after annealing [16]. Unfortunately, no theoretical models
exist so far, which allow us to take this effect into account
in process simulations. Therefore, the comparison of the ex-
perimentally observed and simulated defect densities is purely
qualitative. A very basic approach to deal with this problem is
to simulate the defect distribution for the different local implant
geometries, and to calculate the overall defect distribution by
using the area-weighted arithmetic average. Here, we use areal
fractions of 6/7 for the pyramid sidewalls and 1/7 for the valleys
in between the pyramids. These fractions are characteristic for
the texture used here and are extracted from the results in [16].
Thus, we perform simulations for two different implant angles:
1) 54.7°for the sides of the pyramids and 2) 0°for the valleys
in between the pyramids. In order to reflect the influence of the
surface texture, we adjust the ion dose on the pyramid sides by
a factor of 1/3, which is the ratio of planar to textured surface
area. For the valleys, the ion dose does not need to be adjusted,
since the ion beam hits the surface perpendicular [16].
Theoretical J0evalues are calculated, assuming an Al2O3sur-
face passivation. For the corresponding surface recombination
Fig. 1. Schematics of the defect distribution within a given volume. (a) Dis-
tribution of point defects after ion implantation. (b) Distribution of DLs (blue
circles) after high Tannealing. (c) Areal density of DLs [corresponds to a plan
view of (b)]. (d) Dislocation line density.
velocity, we use a parameterization, which is based on the results
of Black et al. [17]. Experimental doping profiles are measured
by electrochemical capacitance–voltage method (ECV).
After high temperature annealing, DLs will be the dominant
defect species [18]. For the formation and evolution of the DLs,
we use a new model [12], which adapts the existing model of
Zographos et al. [19] to the high temperatures used here. From
our simulations, we extract DL densities and relate them with
measured J0evalues. In addition, experimental defect densities
are measured with TEM (G2 F20 TMP from Tecnai). Plan view
and cross section samples are prepared by ion milling (691 PIPS
from Gatan).
From simulated doping profiles, we calculate theoretical J0e
values with SENTAURUS device [20], intentionally omitting
any influence of a residual defect distribution. A scaling factor
is needed to reflect the impact of the surface texture, since the
simulations are performed assuming planar surfaces. For similar
surface passivation, Richter et al. reported a 2.1 higher J0eon
textured compared with planar surfaces [21]. Together with the
quasi-defect-free J0evalues (assuming an Al2O3surface passi-
vation), we can determine the influence of the defect quantity
on J0e.
III. RESULTS AND DISCUSSION
A. Simulated Defect Densities
Ion implantation results in a large amount of point defects like
vacancies and interstitials [see Fig. 1(a)][8]. These point defects
agglomerate during annealing to {311}defects, which subse-
quently transform into DLs [22]. For the temperatures used here,
DLs are expected to dominate the defect population [18]. Dis-
location line densities ¯ρLare calculated from the areal density
of DLs NL. The latter is a parameter well known in microelec-
tronic experiments [23] and is defined as the integrated loop
concentration DL[see Fig. 1(b)]across the sample thickness
NL=dx ·DL(1)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
KR ¨
UGENER et al.: ELECTRICAL AND STRUCTURAL ANALYSIS OF CRYSTAL DEFECTS AFTER HIGH-TEMPERATURE RTA 3
Fig. 2. Simulated areal densities of DLs NL(a) for 0°and 2 ×1015 cm−2at 10 keV, (b) for 54.7°and 1.2 ×1015 cm−2at 10 keV, and (c) the resulting weighted
density of DLs (see text for details). Simulated dislocation line densities ¯ρL(d) for 0°and 2 ×1015 cm−2at 10 keV, (e) for 54.7°and 1.2 ×1015 cm−2at 10 keV,
and (f) the resulting weighted density of DLs (see text for details). Lines of constant NLand ¯ρLare also shown as guides to the eye.
NLcorresponds to a plan view of the defect distribution as il-
lustrated in Fig. 1(c). For nonamorphizing B implants, DLs only
appear in a localized defect band [see Fig. 1(d)], situated in the
vicinity of the projected range [24], [25]. The mean dislocation
line density in the defect band ¯ρLcan, therefore, be obtained by
using NL, the mean loop radius rL, and the defect band width
w[see Fig. 1(c)]
ρL=2πrLNL/w. (2)
Fig. 2 summarizes the simulated values for NL[(a)–(c)] and
¯ρL[(d)–(f)] for the investigated range of annealing tempera-
tures and annealing times. Fig. 2 also contains lines of constant
¯ρLand NL, respectively. The defect density shows an expo-
nential dependence on the annealing temperature at constant
annealing time. The simulations show higher defect densities
for perpendicular implantation [see Fig. 2(a) and (d)]compared
with implantation on a tilted surface [see Fig. 2(b) and (e)].The
difference at a given Tand tis about one order of magnitude for
NLand half an order for ¯ρL. In the following, we will focus on
the results of the weighted defect densities [see Fig 2(c) and (f)]
since these are assumed to reflect the impact of surface texture
on the overall defect densities the best.
For Tabove 1050 °C, the data point for the longest anneal-
ing talways shows NLbelow 1 ×107cm−2. For such a low
density, J0eis assumed not to be limited by Shockley–Read–
Hall (SRH) recombination mediated by implant-induced defects
anymore. Rather other recombination mechanisms like Auger
recombination and SRH surface recombination are expected
to dominate the minority carrier lifetime. This assumption is
supported by recently reported ion-implanted boron emitters
with as low saturation current densities as 55 ±8fA/cm
2
(78 Ω/) after annealing for 20 min at 1050 °C [14]. ¯ρLas a
function of the thermal budget shows a very similar behavior, al-
though its quantity has a weaker dependence on Tcompared with
NL(see Fig. 2). For the longest annealing time for every tem-
perature above 1050 °C, ¯ρLis always well below 5 ×107cm−2.
B. Structural Characterization by Transmission
Electron Microscopy
For a qualitative verification of our simulation results, we
perform TEM investigations after ion implantation and subse-
quent RTA. In Fig. 3(a), selection of representative TEM images
after annealing is shown. For an identification of the present de-
fect species, dark field TEM under weak-beam conditions is
commonly used [26]. Compared with bright field images, more
distinct details of single defects are visible. That way, we can
identify faulted and perfect DLs as the prominent defect species
to be present for the used implant and annealing conditions. For
example, single faulted and perfect DLs are marked by the ar-
rows in Fig. 3(a) and (b), respectively. After 20 min of annealing
at 900 °C, many defects of various types are visible in the corre-
sponding TEM images [Fig. 3(a)-plan view and Fig. 3(b)-cross
section]. Mainly faulted and perfect DLs and {311}defects
are observed. This corresponds to other experimental results
on planar surfaces with similar annealing conditions [22]. For
all annealing conditions, defects are mostly located within the
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4IEEE JOURNAL OF PHOTOVOLTAICS
Fig. 3. Dark field weak-beam TEM images of textured Si(100) surfaces after
ion implantation of 2 ×1015 boron ions/cm2and subsequent annealing. The
different thermal budgets are given in the upper right corners of each image.
All images instead of (b) (cross section) are plan view images. The arrow I in
(a) marks a single faulted and the arrow II in (b) a single perfect DL. The dashed
line in (d) highlights the base area of a single-texture pyramid.
valleys of the texture. This inhomogeneous defect distribution
is caused by geometrical effects as discussed in [16]. Since the
textured surface consists of randomly distributed (in size and
height) pyramids, several local implant angles exist. The vari-
ation in the implantation angle comes in hand with different
local ion doses and projected ranges and, thus, a different de-
fect distribution. The valleys in between the texture pyramids
receive the highest ion dose due to perpendicular implantation
and, therefore, show the highest defect density. Although the
tips of the pyramids receive the same ion dose as the valleys, the
surface proximity is assumed to lead to faster defect dissolution
(e.g., discussion in [16]).
Cross-sectional images between two texture pyramids [see
Fig. 3(b)]show a similar defect distribution: Many defects are
within the valleys and less on the sides of the pyramids. The
contrast pattern in the Si on the right-hand side of Fig. 3(b)
is caused by thickness variations of the sample [25]. For an
increase in the thermal budget, especially for an increase in T,a
decrease in the observed defect density is found.
Within the investigated area, some single-texture pyramids
of various sizes are visible, e.g., as highlighted with the dashed
lines in Fig. 3(d). Furthermore, areas with different contrast
are observed: The valleys in between the pyramids seem to be
brighter than the tips of the pyramids. This effect is caused by
the varying thickness across the sample surface.
A quantitative analysis of the defect densities is difficult due
to the influence of the local surface morphology on the defect
occurrence as mentioned above [16]. Thus, a random texture on
Si(1 0 0) with small pyramids of lower height should exhibit
Fig. 4. Comparison of area weighted (see text) simulated (full symbols) and
experimentally determined (boxes) areal densities of DLs (sum of perfect and
faulted). The box values are determined from four corresponding TEM images,
respectively.
higher defect densities than a surface with larger and higher
pyramids caused by a higher circumference/area ratio of the
first. Another source of inaccuracy is caused by the TEM mea-
surement itself. The analysis is restricted to very narrow and
thin areas of the TEM sample, where smaller pyramids are
dominant. Therefore, the defect densities measured by TEM are
local rather than global values reflecting the defect densities for
the whole sample. The analysis of larger areas, including big-
ger and smaller pyramids, which allows the extraction of more
reliable data is not possible due to preparation (smaller pyra-
mids may vanish during ion milling) and methodical limitations
(thickness of higher pyramids is too large for TEM analysis)
caused by the textured surface structure.
Fig. 4 shows a comparison of the simulated and the experi-
mentally observed areal density of DLs for representative ther-
mal budgets. The simulated defect densities are calculated as
described in the experimental section by using the arithmetic
average of the defect densities on the pyramid sides (6/7 frac-
tion) and of the valleys in between the pyramids (1/7 fraction).
After annealing for 20 min at 900 °C, the simulation predicts
a slightly lower defect density compared with the measured
one, but compares well to what is reported for planar surfaces
[22]. For 1050 °C (5 min) and 1100 °C (1 min), simulation and
measurement are in very good agreement. Increasing the tem-
perature above 1100 °C results in an offset between simulation
and experiment by about one order of magnitude, whereas the
simulation predicts a higher defect density. This difference is
not understood so far. Aside from the offset between experi-
ment and simulation, the qualitative trends are the same. For
a quantitative verification of the used model for high anneal-
ing temperatures (T>1050 °C), further experiments on planar
surfaces are required.
C. Measured Saturation Current Densities and Comparison
With Defect Densities
Fig. 5 summarizes the measured J0evalues after RTA and
passivation as a function of the thermal budget [see Fig. 5(a)]
and the resulting sheet resistance Rsheet [see Fig. 5(b)],re-
spectively. For low T(1050 and 1100 °C), a clear correlation
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
KR ¨
UGENER et al.: ELECTRICAL AND STRUCTURAL ANALYSIS OF CRYSTAL DEFECTS AFTER HIGH-TEMPERATURE RTA 5
Fig. 5. Measured J0evalues. (a) As a function of the thermal budget and
(b) as a function of the resulting sheet resistance. The inset in (a) denotes the
highlighted line scans.
between the thermal budget and J0eis observed. Increasing the
annealing duration for a given Tleads to lower J0eas well as
increasing Tfor a given duration. For Tabove 1100 °C, no such
direct correlation exists [see inset in Fig. 5(a)]. Such a saturation
behavior can be expected for annealing conditions, which imply
the dissolution of implant-induced defects to a quantity where
other recombination paths, e.g., Auger recombination or surface
recombination, become dominant.
J0eas a function of the sheet resistance follows a similar trend
see Fig. 5(b)].ForTup to 1100 °C, J0eand Rsheet decrease for
increasing annealing duration. The decrease in Rsheet for longer
annealing durations originates from a deeper doping profile with
a lower peak doping concentration, which implies higher carrier
mobility. This behavior of Rsheet is observed even for Tabove
1100 °C.
Within the Trange discussed above, another effect should be
mentioned: The dissolution of recombination active interstitial
boron-clusters (BIC) [27], which are often observed after high-
dose boron implantation. BICs dissolve at around 900 °C with a
time constant of a few minutes (see [28]). Increasing the Tleads
to even shorter decay constants. Furthermore, our simulations
show that for Tof 1050 °C and above, almost all boron atoms
are electrically active [see Fig. 6]. Only for Tbelow 1050 °C
and annealing durations in the range of a few 10 min, larger
amounts of boron atoms bound in BICs are found [see Fig. 6].
Fig. 6. Simulated total and active B profiles after implantation and annealing.
With respect to the annealing times used here and for Tof
1050 °C and above, an influence of BIC on J0eis, therefore,
most unlikely.
This assumption is supported by recent studies in which ex-
perimental effective charge carrier lifetimes from Pawlak et al.
[28] are compared with simulated densities of DLs, disloca-
tion line densities, and densities of BICs [29]. Pawlak et al.
found a strong decrease of the effective lifetime for doses of
1×1015 cm−2and above [28]. The authors suppose that the in-
creasing number of BICs, which they assume to be the dominant
recombination centers, is responsible for the lifetime reduction
at higher boron doses [28]. On the other hand, the correspond-
ing simulations of Wolf [29] show that the strong decrease of
the effective lifetime as observed by Pawlak et al. [28] is most
likely caused by insufficiently cured implant-induced defects.
Nevertheless, a large variety of possible BIC configurations
have been reported [30], [31], and only little is known about their
electrical properties [28]. In addition, measurements, which al-
low determining the possible existence of BICs, like the com-
bination of secondary ion mass spectroscopy and ECV are very
difficult to use on textured surfaces. Another possible method
for the detection of BICs is TEM. For example, Bonifelli et al.
have shown for epitaxially grown silicon layers the occurrence
of larger BICs, whereas smaller BICs could not be detected by
TEM [31]. Therefore, it is not possible to exclude aprioriany
kind of impact on the results we present here, especially after
annealing below 1050 °C.
We measure J0evalues as low as 38 ±2fA/cm
2for a sheet
resistance of 81 Ω/after very short annealing processes (e.g.,
1 min at 1200 °C). These excellent values are similar to those
found in recent investigations of diffusion-based boron emitters
[21]. In Fig. 5(b), additional furnace annealed reference samples
are shown which underline the assumption that above a certain
thermal budget (below a certain defect density), J0eseems not
to be limited by implant-induced defects.
Next, we compare simulated ¯ρLand measured J0evalues by
taking the minority carrier lifetime τdinto account. The lifetime
as a function of ¯ρLcan be described by the following equation
(neglecting a weakly varying logarithmic term) [32]:
τd=1
2πρLDn/p
.(3)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6IEEE JOURNAL OF PHOTOVOLTAICS
Fig. 7. Comparison between the annealing Tdependence of the square of
experimental J0e(open symbols) and simulated (filled symbols) NL(a) and
¯ρL(b). The lines (dashed and continuous) are guide to the eye. (c) J0eas a
function of the weighted ¯ρL; the dashed line is a linear fit for all data points
displayed.
Here, Dn/p is the minority carrier diffusivity. In classical
diode theory, J0efor p-type emitters is
J0e=qDn/pn2
i
NA√Dnτd
(4)
with qdenoting the elemental charge, niis the intrinsic carrier
concentration, and NAis the emitter doping concentration, re-
spectively. From (3) and (4), a linear dependence of J0eon ¯ρL
is expected, whereas both values can directly be compared.
In Fig. 7, the simulated values of NL[see Fig. 7(a)]and ¯ρL
[see Fig. 7(b)]are compared with the measured J2
0evalues as a
function of the annealing temperature. In the low Trange, J2
0e
shows a strong dependence on T. It decreases by around three
orders of magnitude with increasing Tfrom 900 to 1050 °C. The
same qualitative trend is visible for ¯ρL, although the decrease is
only two orders of magnitude. This behavior can be explained
as follows: for low annealing temperatures, many small faulted
loops are present, which transform into large perfect loops for
Taround 1000 °C [33]. This transition should have significant
effects on the dislocation-induced strain fields in the emitter
region, and it is most likely the reason for the steep decrease
of NLwhen increasing Tfrom 900 to 1050 °C. Since strain
fields are known to increase the minority carrier recombination
in this region [6], [7], the steep decrease of NLexplains the
high efficiency of increasing Tin the regime between 900 and
1050 °C. In addition to this transformation of DLs, a signifi-
cant dissolution of BICs takes place for Tbelow 1050 °C(see
discussion above), which also could have an impact on the re-
sulting J0e. Nevertheless, our results indicate that the remaining
implant-induced defects should be the dominating factor on J0e.
For the Trange above 1000 °C, perfect loops dominate the
loop population, and Ostwald ripening occurs at a reduced speed
[23]. Thus, ¯ρLwhich reflects the summed circumference of all
DLs per volume [see Fig. 1(d) and (2)] should, therefore, be
a sufficient measure of the recombination activity. Annealing
above 1000 °C results in much smaller ¯ρLand J2
0ecompared
with the low Trange. For Thigher than 1050 °C, the simulation
predicts an exponential decrease of ¯ρLwith constant exponent
with respect to the studied T.J2
0eshows a decrease with respect
to the annealing duration of the same order of magnitude as the
decrease of ¯ρL. Only for the highest Tof each of the split groups
5 min, 1 min, and 30 s, the decrease of J2
0ehas saturated.
Fig. 7(c) shows the correlation of J2
0eand the corresponding
weighted ¯ρL. For low defect densities, the graph shows a cloud-
like distribution and not the expected clear linear relationship.
This behavior is not surprising, since almost all of the used
annealing conditions result in very low defect densities [see
Fig. 5(a)]. In that case, J0eis only determined by the varying
dopant profiles resulting from the different annealing processes.
For higher defect densities, a linear correlation between J2
0e
and ¯ρLcan be observed [see linear regression in Fig. 7(c)].
Unfortunately, only two data points are available in the highly
defective range, whereas firm conclusions cannot be drawn.
Furthermore, the intersection of the linear regression with the
J2
0eaxis corresponds a J0eof 35 fA/cm2, which is very close
to results of diffusion-based (free of implant-induced defects)
boron emitters with a similar sheet resistance [see [21] and the
furnace annealed references in Fig. 5(b)].
A further quantification of the impact of implant-induced de-
fects on J0eis done by calculating theoretical J0evalues from
simulated doping profiles. These J0evalues solely correspond to
the recombination activity of the doping profile, and the recom-
bination at the passivated surface, and neglect any effect of the
ion implantation or the resulting defects. Fig. 8(a) exemplarily
shows the simulated doping profile after annealing for 20 min at
1050 °C, together with a corresponding measured ECV profile
and a diffusion-based profile from Richter et al. (45 fA/cm2at
88 Ω/) [21]. The agreement of the simulated and measured
doping profiles supports the validity of the used models.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
KR ¨
UGENER et al.: ELECTRICAL AND STRUCTURAL ANALYSIS OF CRYSTAL DEFECTS AFTER HIGH-TEMPERATURE RTA 7
Fig. 8. (a) Doping profiles from process simulation and EC–Vafter ion im-
plantation and subsequent annealing for 20 min at 1050 °C. Additionally, a
diffusion-based profile from [21] is shown. (b) Measured (open symbols) and
calculated saturation current densities (full symbols) as a function of the thermal
budget. The dotted line denotes J0eas reported for diffused emitter profile E2
in (a) [21]. The lines (dashed and continuous) are a guide to the eye.
Fig. 8(b) summarizes the calculated and measured J0evalues
for the investigated parameter space, assuming an Al2O3surface
passivation for the simulations. Again, we use the diffusion-
based doping profile E2 [diamonds in Fig. 8(a)]from Richter
et al. [21] as a reference [dotted line in Fig. 8(b)]. The simulation
predicts J0eto be in the range of 28–48 fA/cm2[see Fig. 8(b)]
and is, therefore, comparable with the diffusion-based reference.
The experimentally measured J0eis in the range from 38 ±2
to 1400 ±25 fA/cm2. For the shortest annealing duration at a
given temperature, the highest differences between experiment
and simulation are observed, up to 1376 fA/cm2for 20 min
annealing at 900 °C. Aside from implant-induced defects, this
difference is most likely caused by boron bound in BICs [see
Fig. 6]. For temperatures of 1050 °C and above, the difference of
the simulated and measured results is in the range of ±7fA/cm2,
except for the lowest annealing temperature at a given anneal-
ing duration. The latter seems to be limited by implant-induced
defects. This is supported by the corresponding simulated dis-
location line densities [see Fig. 7(c)]; e.g., after annealing for
1 min at 1100 °C(J0e=91.5±5fA/cm), a high dislocation
line density of 2.2 ×108cm−2is predicted.
Altogether, the presented results allow estimating the impact
of the residual damage after ion implantation of boron and sub-
sequent annealing on the resulting saturation current densities
[linear regression in Fig. 7(c)]. Nevertheless, more work with
respect to this matter has to be done in order to obtain a more
quantitative and predictive model, e.g., for device simulations.
IV. CONCLUSION
We have performed investigations of the influence of high
temperature RTA on the resulting emitter saturation current
densities of ion-implanted B emitters. Experimental J0eval-
ues are compared with dislocation line densities obtained from
SENTAURUS process simulations using an extended model of
DL formation for temperatures up to 1250 °C. For ¯ρLbelow
2×108cm−2,J0eseems to no longer be dominated by ¯ρL.
On textured surfaces, J0evalue of 38 ±2fA/cm
2are obtained
on 81 ±1Ω/Al2O3-passivated boron emitters after very short
RTA processes (1 min at 1200 °C). TEM analysis shows the oc-
currence of mainly DLs located in the valleys of the texture,
even for electrically well-performing emitters. Furthermore, we
find a qualitative agreement between the model of the evolution
of DLs at temperatures above 1000 °CbyWolfet al., as well as
the experimentally observed defect densities in TEM.
REFERENCES
[1] K. Bothe, R. Sinton, and J. Schmidt, “Fundamental boron-oxygen-related
carrier lifetime limit in mono- and multicrystalline silicon,” Prog. Photo-
voltaics, Res. Appl., vol. 13, pp. 287–296, 2005.
[2] W. S. Ho, Y.-H. Huang, W.-W. Hsu, Y.-Y. Chen, Y.-Y. Chen, and
C. W. Liu, “Ion implanted boron emitter n-silicon solar cells with wet
oxide passivation,” in Proc. IEEE 37th Photovoltaic Spec. Conf., 2011,
pp. 001058–001060.
[3] T. S. B¨
oscke, D. Kania, C. Sch¨
ollhorn, D. Stichtenoth, A. Helbig,
P. Sadler, M. Braun, M. Dupke, M. Weis, A. Grohe, J. Lossen, and H.-J.
Krokoszinski, “Fully ion implanted and coactivated industrial n-type cells
with 20.5% efficiency,” IEEE J. Photovoltaics, vol. 4, no. 1, pp. 48–51,
Jan. 2014.
[4] Y.-W. Ok, A. D. Upadhyaya, Y. Tao, F. Zimbardi, K. Ryu, M.-H. Kang, and
A. Rohatgi, “Ion-implanted and screen-printed large area 20% efficient n-
type front junction Si solar cells,” Sol. Energy Mater. Sol. Cells, vol. 123,
pp. 92–96, 2014.
[5] M.A, Kessler, T. Ohrdes, B. Wolpensinger, and N.-P. Harder, “Charge
carrier lifetime degradation in Cz silicon through the formation of a boron-
rich layer during BBr3 diffusion processes,” Semicond. Sci. Technol.,
vol. 25, pp. 055001-1–055001-9, 2009.
[6] M. Seibt, R. Khalil, V. Kveder, and W. Schr¨
oter, “Electronic states at
dislocations and metal silicide precipitates in crystalline silicon and their
role in solar cell materials,” Appl. Phys. A, vol. 96, pp. 235–253, 2009.
[7] V. Kveder, M. Kittler, and W. Schr¨
oter, “Recombination activity of con-
taminated dislocations in silicon: A model describing electron-beam-
induced current contrast behavior,” Phys. Rev. B, vol. 63, pp. 115208-
1–115208-11, 2001.
[8] H. Ryssel and I. Ruge, Ion Implantation. Chichester, U.K.: Wiley, 1986.
[9] S. N. Hong, G. A. Ruggles, J. J. Wortman, and M. C. ¨
Ozt¨
urk, “Mate-
rial and electrical properties of ultra-shallow p+-n junctions formed by
low-energy ion implantation and rapid thermal annealing,” IEEE Trans.
Electron Devices, vol. 38, no. 3, 476–486, Mar. 1991.
[10] A. Lanterne, S. Gall, Y. Veschetti, R. Cabal, M. Coig, F. Mil´
esi, and
A. Tauzin, “High efficiency fully implanted and co-annealed bifacial n-
type solar cells,” Energy Procedia, vol. 38, pp. 283–288, 2013.
[11] D. E. Kane and R. M. Swanson, “Measurement of the emitter saturation
current by a contactless photoconductivity decay method,” in Proc. IEEE
18th Photovoltaic Spec. Conf., 1985, pp. 578–583.
[12] F. A. Wolf, A. Martinez-Limia, D. Stichtenoth, and P. Pichler, “Modeling
of the annealing of dislocation loops in implanted c-Si solar cells,” IEEE
J. Photovoltaics, vol. 4, no. 3, 851–858, May 2014.
[13] W. Kern and D. A. Puotinen, “Cleaning solutions based on hydrogen
peroxide for use in silicon semiconductor technology,” RCA Rev., vol. 31,
pp. 187–206, 1970.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8IEEE JOURNAL OF PHOTOVOLTAICS
[14] F. Kiefer, R. Peibst, T. Ohrdes, T. Dullweber, J. Kr¨
ugener, H. J. Osten,
C. Sch¨
ollhorn, A. Grohe, and R. Brendel, “Study of the influence of the
boron emitter profile on VOC and JSC losses in fully ion implanted n-type
PERT solar cells,” Physica Status Solidi, 2014, to be published.
[15] (2012). Sentaurus Process User Guide, Ver. G-2012.06 (Synopsys, Inc.,
Mountain View, CA, USA).
[16] J. Kr¨
ugener, E. Bugiel, R. Peibst, F. Kiefer, T. Ohrdes, R. Brendel, and
H. J. Osten, “Structural analysis of textured silicon surfaces after ion
implantation under tilted angle,” Semicond. Sci. Technol., vol. 29, no. 9,
pp. 095004-1–095004-7, 2014.
[17] L. E. Black, T. Allen, K. R. McIntosh, and A. Cuevas, “Effect of boron
surface concentration on recombination at the p-Si-Al2O3 interface,” J.
Appl. Phys., vol. 115, pp. 093707–1–093707-9, 2014.
[18] A. Florakis, T. Janssens, E. Rosseel, B. Douhard, J. Delmotte, E. Cor-
nagliotti, J. Poortmans, and W. Vanderhorst, “Simulation of the anneal
of ion implanted boron emitters and the impact on the saturation current
density,” Energy Procedia, vol. 27, pp. 240–246, 2012.
[19] N. Zographos, C. Zecher, and I. Avci, “Efficient TCAD model for the
evolution of interstitial clusters, {311}defects, and dislocation loops in
silicon,” in Proc. Mater. Res. Soc., vol. 994, pp. 0994–F10–01, 2007.
[20] (2012). Sentaurus DeviceUser Guide, Version G-2012.06 (Synopsys, Inc.,
Mountain View, CA, USA).
[21] A. Richter, J. Benick, and M. Hermle, “Boron emitter passivation with
Al2O3and Al2O3/SiNxstacks using ALD Al2O3,” IEEE J. Photo-
voltaics, vol. 3, no. 1, pp. 236–245, Jan. 2013.
[22] I. Avci, “Loop nucleation and stress effects in ion-implanted silicon,”
Ph.D. dissertation, Univ. Florida, Gainesville, FL, USA, 2002.
[23] C. Bonafos, D. Mathiot, and A. Claverie, “Ostwald ripening of end-of-
range defects in silicon,” J. Appl. Phys., vol. 83, pp. 3008–3017, 1998.
[24] K. S. Jones, S. Prussin, and E. R. Weber, “A systematic analysis of defects
in ion-implanted silicon,” Appl. Phys. A, vol. 45, pp. 1–34, 1988.
[25] M. Milosavljevic, G. Shao, M. A. Lourenco, R. M. Gwilliam, and K. P.
Homewood, “Engineering of boron-induced dislocation loops for efficient
room-temperature silicon light-emitting diodes,” J. Appl. Phys., vol. 97,
pp. 073512-1–073512-7, 2005.
[26] D. B. Williams and C. B. Carter, Transmission Electron Microscopy—
Par t 3: Im agi n g. New York, NY, USA: Springer, 2009.
[27] S. Mirabella, E. Bruno, F. Priolo, D. De Salvador, E. Napolitani, A. V.
Drigo, and A. Carnera, “Dissolution kinetics of boron-interstitial clusters
in silicon,” Appl. Phys. Lett., vol. 83, pp. 680–682, 2003.
[28] B. J. Pawlak, T. Janssens, S. Singh, I. Kuzma-Filipek, J. Robbelein,
N. E. Posthuma, J. Poortmans, F. Cristiano, and E. M. Bazizi, “Studies of
implanted boron emitters for solar cell applications,” Prog. Photovoltaics,
Res. Appl., vol. 20, pp. 106–110, 2012.
[29] F. A. Wolf, Section 6.3.1, “Modeling of annealing processes for
ion-implanted single-crystalline silicon solar cells,” Ph.D. dissertation,
Univ. Erlangen, Erlangen, Germany. [Online]. Available: http://nbn-
resolving.de/urn:nbn:de:bvb:29-opus4-49068, p. 121, 2013.
[30] O. Cojocaru-Mir´
edin, E. Cadel, F. Vurpillot, D. Mangelinck, and
D. Blavette, “Three-dimensional atomic-scale imaging of boron clusters
in implanted silicon,” Scripta Materialia, vol. 60, pp. 285–288, 2009.
[31] S. Boninelli, S. Mirabella, E. Bruno, F. Priolo, F. Cristiano, A. Claverie,
D. De Salvador, G. Bisognin, and E. Napolitani, “Evolution of boron-
interstitial clusters in crystalline Si studied by transmission electron mi-
croscopy,” Appl. Phys. Lett., vol. 91, pp. 031905–1–031905-3, 2007.
[32] C. Donolato, “Modeling the effect of dislocations on the minority carrier
diffusion length of a semiconductor,” J. Appl. Phys., vol. 84, pp. 2656–
2664, 1998.
[33] G. Z. Pan, K. N. Tu, and A. Prussin, “Size-distribution and annealing
behavior of end-of-range dislocation loops in silicon-implanted silicon,”
J. Appl. Phys., vol. 81, pp. 78–84, 1997.
Authors’ photographs and biographies not available at the time of publication.
