Band tail states in microcrystalline silicon solar cells probed by photoluminescence and open circuit voltage

Full text

Page 1

Journal of Optoelectronics and Advanced Materials Vol. 7, No. 1, February 2005, p. 121 - 128


Invited Paper



BAND TAIL STATES IN MICROCRYSTALLINE SILICON SOLAR CELLS
PROBED BY PHOTOLUMINESCENCE AND OPEN CIRCUIT VOLTAGE




aInstitut für Photovoltaik, Forschungszentrum Jülich GmbH, 52425 Jülich, Germany
bCentral Laboratory for Solar Energy and New Energy Sources, Bulgarian Academy of
Sciences, 72 Tzarigradsko Chaussee, 1784 Sofia, Bulgaria



Thin film p-i-n microcrystalline silicon (µc-Si:H) solar cells were studied by
photoluminescence (PL) spectroscopy and open circuit voltage, Voc. By this means, the cause
of the commonly observed increase of Voc when the silane concentration in the gas phase for
the preparation of the intrinsic absorber layers is raised is addressed. The hot wire chemical
vapour deposition (HW-CVD) technique was used for the preparation of films and intrinsic
absorber layers of solar cells with efficiencies higher than 9%. By monitoring Voc and the PL
energy, EPL, as a function of temperature, information on the photogenerated carrier
distributions and the splitting of the quasi-Fermi energies was gained and the effect of states
in the gap was studied by comparison with results from a crystalline p-i-n diode. An increase
of EPL and Voc for increasing SC and decreasing temperature is attributed to the shift of the
excess carrier distributions to higher energies. It is proposed that this shift is limited by band
tail states. It is proposed that increasing SC leads to a reduction of the density of band tail
states, due to structural relaxation of the µc-Si:H network by the presence of hydrogen or
hydrogenated amorphous silicon.

(Received December 9, 2004; accepted January 26, 2005)

Keywords: Photoluminescence, Open circuit voltage, Solar cells, Microcrystalline silicon


1. Introduction

Microcrystalline silicon (µc-Si:H) thin films have attracted great interest because of their
potential applications in thin film electronics and thin film solar cells. High quality µc-Si:H can be
prepared at low substrate temperatures on a large variety of substrate materials such as glass,
stainless steel and even plastics. A significant advantage is the compatibility of the deposition
techniques with the quite mature amorphous silicon technology. The films are usually prepared from
silane highly diluted in hydrogen. By increasing the silane concentration (SC) in the hydrogen gas,
the microstructure of the material can be changed from highly crystalline to amorphous [1]. Previous
studies on µc-Si:H solar cells have shown an increased open circuit voltage (Voc) when the intrinsic
(i-) layers were prepared with increasing SC, irrespective of the deposition method, e.g. plasma
enhanced chemical vapour deposition (PE-CVD) [2-4] or hot wire chemical vapour deposition (HW-
CVD) [5]. In particular, µc-Si:H solar cells with i-layers deposited near the transition from
crystalline to amorphous growth exhibit the best performance [2,3,5]. Studies of the photo- and dark
conductivity on undoped material showed a continuous decrease with increasing SC [2]. It was also
found that the PL band, which is located below the band gap of crystalline silicon, shifts to higher
energies with increasing SC, i.e. with decreasing crystalline volume fraction [6-11]. This featureless
band is much broader than the luminescence spectra observed in high quality crystalline silicon, and
is about half the width of the photoluminescence band found in a-Si:H, e.g. [8,11,12]. The origin of
this PL band is not yet clear. Radiative recombination at defects in the crystalline phase, transitions

* Corresponding author: r.carius@fz-juelich.de
R. Cariusa*, T. Merdzhanovaa,b, S. Kleina, F. Fingera

Page 2

R. Carius, T. Merdzhanova, S. Klein, F. Finger


between states in the grain boundary region or transitions between localized tail states similar to the
case of amorphous silicon have been discussed [6-12]. However, there is evidence that the
recombination process involves localized states and the recombination processes show similarities
with those found in a-Si:H. Therefore, we assume that the PL originates from transitions between
localized band tail states similar to those in a-Si:H [7,13].

I
122
Isc
dark
V
Voc
illumination
(633nm)
Plmod
Isc
Voc
t
t
t

Fig. 1. Schematic representation of the current-voltage characteristics of a p-i-n solar cell in
the dark and under illumination (left); the time sequence of the modulation technique,
together with the expected signals to be measured (right).


Here, a systematic study of the influence of the SC (respectively the crystalline volume
fraction) on the PL properties and Voc of very high quality µc-Si:H solar cells is provided, in order to
obtain more information on the density and distribution of the tail states and its influence on Voc in
this solar cells. The emphasis is on the relationship between Voc, i.e. the splitting of the quasi-Fermi
energies and PL peak energy, i.e. the carrier distributions in the band tails.


2. Experimental details

Raman measurements in a back-scattering geometry were performed with the 488nm line of
an Ar-ion laser or the 647 nm line of a Krypton laser for excitation. Details of the set-up and the data
evaluation have been published elsewhere [1,5]. Photoluminescence (PL) spectra were measured
with a Fourier transform (FT) spectrometer (Bruker FS66v) and detected either with a LN2 cooled
Ge or an InAs detector. For the measurements, the samples were mounted in a continuous flow
helium cryostat, providing temperatures from 10 to 300 K. Measurements on solar cells were
performed using a modulation technique, in order to eliminate spurious PL signals from the glass,
the transparent conductive oxide (TCO) and the p-layer, and to detect the relevant carriers involved
in the photovoltaic process, as explained below. Here, the 632.8 nm line of a He-Ne laser was used
to provide homogeneous generation throughout the samples. The applied photon flux of about 1018
cm-2s-1 led to a 10 times higher short-circuit current density as compared to AM 1.5 illumination.
An external voltage was applied to the solar cells and switched from 0 V to Voc (20 Hz, rectangular
shape, 50% duty cycle) using a HP 8116A pulse generator.
Fig. 1(a) shows the I-V characteristics of a solar cell in the dark and under illumination, and
the two important parameters short circuit current, Isc, and open circuit voltage, Voc, are indicated.
The time sequence of the measurement, together with the expected signals to be measured is
schematically shown in Fig. 1(b). At Voc (i.e. I = 0) where full carrier recombination and no
extraction from the i-layer occurs, the maximum PL signal will be obtained. The minimum PL signal
will be obtained at Isc - where full carrier extraction and no recombination is expected in the ideal
case. With this technique, only those carriers relevant to the photovoltaic process are probed, and
spurious signals from the doped layers, glass substrate, etc. are suppressed. The change of the PL
signal upon changing the applied voltage from 0 V to Voc was detected by a lock-in technique, while
the FT-spectrometer was operated in step-scan mode. Isc was monitored with a digital oscilloscope
(Le Croy 9400). The contact area was typically 1 mm2. A Keithley 2400 source meter was used for
the Voc and Isc measurements, under steady state conditions. The investigated µc-Si:H solar cells
were prepared at a substrate temperature of Ts=185 °C in a p-i-n deposition sequence on textured
ZnO substrates with a ZnO/Ag back reflector. HW-CVD at a deposition pressure (Pd) of 3.5 - 5 Pa
was used to deposit the intrinsic layers. The microstructure of the i-layers with thickness between
0.85 and 1.3 µm was modified by varying the dilution of the silane process gas in hydrogen
(SC=[SiH4]/[SiH4+H2]), keeping the conditions for all other layers constant. Details of the

Page 3

Band tail states in microcrystalline silicon solar cells probed by photoluminescence…


123
deposition parameters are published elsewhere [5,14]. A commercially available p-i-n diode
(Hamamatsu S1190) of active diameter 1 mm was used as a reference.

620620
223344556677
SC (%) SC (%)
420 420
460 460
500 500
540 540
580580
Voc (mV)
Voc (mV)
HW-CVDHW-CVD
PE-CVDPE-CVD

Fig. 2. Open circuit voltage Voc at 300K as a function of the silane concentration in the gas
phase for two series of solar cells deposited by PE-CVD and HW-CVD.


3. Results

In Fig. 2, Voc is plotted as a function of the silane concentration used for the i-layer
preparation, for two series of solar cells prepared by HW-CVD and PE-CVD [2, 14]. The data points
for the highest SC values refer to those that still provide cells with good extraction properties. A still
higher SC leads to even higher Voc, but also to a degradation of the solar cell parameters (low Isc and
fill factor). With increasing SC, Voc increases almost linearly for the PE-CVD series, by about 100
mV. For the HW-CVD cells, the Voc is higher and the increase seems to be steeper at higher SC.
However, the overall change is very similar and is in line with observations from other series and
from other groups.
Fig. 3, the maximum of the photoluminescence energy (EPL) measured in the conventional
way at 10 K is plotted as a function of SC for two series of films (full symbols) prepared by PECVD
and HW-CVD at different substrate temperatures. The data suggest that EPL as a function of SC
exhibits a similar trend to that as a function of Voc, as shown in Fig. 1, although the magnitude of the
changes is different. A more detailed investigation of several sample series made by HW-CVD or
PE-CVD reveals a common trend, i.e. at low SC the increase of EPL is less pronounced (or even
constant) but at high SC, EPL changes more rapidly. In addition to the data for the films, results for
the intrinsic layer of the HW-CVD solar cells are shown. They exhibit the same trend as observed
for the films.
In Fig. 4, PL spectra obtained by the modulation technique on solar cells with different
silane concentrations (3 - 5.6%), measured at 150 K, are shown. For the lowest SC, the PL spectra
reveal a broad emission band with a maximum at about 0.86 eV (µc-Si-band) and a half-width of
about 0.14 eV, attributed to the microcrystalline phase.

1.1
12
silane concentration (%)

34567
0.9
1
PL energy (eV)
HW-CVDHW-CVD
PE-CVDPE-CVD

Fig. 3. Photoluminescence peak energy EPL measured at 10 K as a function of the silane
concentration for two series of films deposited by PE-CVD and HW-CVD and for the
intrinsic layers of the solar cells investigated in this study.

Page 4

R. Carius, T. Merdzhanova, S. Klein, F. Finger


124
0.7 0.8
photon energy (eV)
0.91 1.1
101
102
PL intensity (r.u.)
SC
5%
5.6%
5.2%
4%
3%
T=150K
Ex=1.96eV

Fig. 4. Modulated PL spectra of solar cells with i-layers prepared with different SC by
HW-CVD measured at 150K. Note the logarithmic intensity scale. The interference fringes
are tentatively corrected for the solar cells with 4% and 5% SC.


The luminescence energy is similar to that observed in a conventional photoluminescence
measurement on films at the same temperature. To demonstrate the shift of the µc-Si-band towards
higher energies with increasing SC (up to 0.91 eV), and the absence of any contribution at 1.35 eV
originating from an amorphous phase, a logarithmic intensity scale was chosen. Interference fringes
modify the shape of the PL spectra for solar cells with SC between 4% and 5.2%. The indicated
open circles and dashed line refer to tentatively corrected spectra. The PL intensity (given in relative
units) and the width of the PL band are very similar for all solar cells, and thus the quantum
efficiency of the PL does not change significantly.
In Fig. 5, EPL as a function of the silane concentration is shown for modulated spectra
measured at 125, 200 and 250 K. In addition, the data for the conventional PL measurements at
10 K shown in Fig. 3 are included for comparison. For the evaluation of EPL, the maximum of the
spectrum was taken. When this was not possible due to fringe patterns in the spectrum, the centre-of-
mass was chosen. This gave reliable values because of the symmetry of the spectra. All samples
exhibit a similar behaviour, i.e. with increasing SC an increase of EPL is observed. However, the
increase seems more pronounced for the PL measurement at 10 K. With increasing temperature, EPL
decreases for all samples, in agreement with results in the literature [10].

1.1
3456
silane concentration (%)
0.8
0.9
1
Epl (eV)
125 K
250 K
200 K
10 K
HW-CVD cells

Fig. 5. Photoluminescence peak energy EPL from modulation measurements on solar cells as
a function of the silane concentration measured at the indicated temperatures. Data for a
conventional PL measurement at 10 K are also shown.

Page 5

Band tail states in microcrystalline silicon solar cells probed by photoluminescence…


125
0100200300
T (K)
0.8
0.9
1.0
1.1
Epl (eV)
3%
5.6%
c-Si
SC


Fig. 6 Photoluminescence peak energy EPL as a function of temperature for the same devices
as in Fig. 5.


The dependence of EPL on the measurement temperature for the modulated spectra is shown
in Fig. 6 in more detail. Modulated spectra could not be obtained at all temperatures. At low
temperatures, the modulation effect decreased strongly for some of the samples. There is strong
evidence that this is related to the low mobility of charge carriers, which leads to poor extraction of
carriers [15]. At high temperatures, useful results could not be obtained because the low
luminescence efficiency causes a poor signal-to-noise ratio. In addition to the results obtained on the
microcrystalline solar cells, those obtained on the c-Si photodiode are included in the figure for
comparison. It is obvious that all microcrystalline samples exhibit a similar temperature dependence
in the investigated temperature range, but the results differ dramatically from the single crystalline
case. The c-Si diode has the highest Epl and exhibits only a weak change with temperature, whereas
EPL of the microcrystalline solar cell with SC=5.6 %, for example, shows a slightly lower energy at
low temperature and decreases strongly with increasing temperature. The increase of EPL with
increasing SC, as shown in Fig. 5, is obtained in the whole temperature range.
In Fig. 7, Voc is shown as a function of temperature for the same set of samples as in Fig. 6.
These measurements were done under the same conditions as for the modulated PL spectra, and the
data have been used as set-points for Voc (compare with Fig. 2). At 10 K, Voc of the c-Si diode is
about 1120 mV, and it decreases only weakly below 50 K. At higher temperatures, the decrease is
stronger and almost linear with temperature. Similar to the results of the PL measurements shown in
Fig. 6, the SC=5.6 % sample exhibits the highest value of all µc-Si:H solar cells, but the Voc of about
961 mV is significantly lower than that of the c-Si diode. At higher temperatures, the difference of
the Voc of c-Si and µc-Si:H becomes smaller, and at room temperature only a very small difference
remains. With decreasing SC, the Voc curves are shifted parallel towards lower values.
1200
0100200300
T (K)
600
700
800
900
1000
1100
Voc(mV)
3%
5.6%
c-Si
SC

Fig. 7. Open circuit voltage Voc as a function of temperature for the same devices as in Fig. 5.

Page 6

R. Carius, T. Merdzhanova, S. Klein, F. Finger


126
4. Discussion

As shown in Figs. 1 and 2, an increase in the silane concentration in the gas phase leads to
an increase of both the open circuit voltage and the photoluminescence energy. Raman and X-ray
diffraction measurements on the same samples reveal a decreasing crystalline volume fraction with
increasing SC, e.g. the 'Raman crystallinity', Xc
25% for SC=5.6%. The decrease of the crystallinity is usually accompanied by a broadening of the
X-ray diffraction peaks and a decreasing size of the crystalline grain clusters observed in TEM, see
e.g. [1] and references therein. This has led to the suggestion of a shift of the band gap of
µc-Si:H with decreasing crystallinity, due to quantum size effects, to explain e.g. the shift of EPL
[9,11]. We do not consider size effects here, because careful spectral response measurements (not
shown) do not reveal a shift of the band edge within experimental error (< 20 meV). Therefore, the
increase of almost 100 mV in Voc and EPL with increasing SC requires a different explanation.
A good correlation between the PL peak energy and microstructure, i.e. an increase of the
PL energy with decreasing crystalline volume fraction, was previously found for series of undoped
µc-Si:H thin films deposited at different substrate temperatures. It was tentatively interpreted as
being caused by changes in the density of states distribution due to changes in the microstructure [6].
In the following, we will discuss changes of the density of states distribution as a possible reason for
the alteration of both Voc and EPL.
As shown in Figs. 4 and 5, the PL spectrum is broad and featureless and located well below
the band gap of c-Si (1.16 eV at 10 K). In agreement with various other experiments that indicate
localized states close to the band edges, e.g. [16], the photoluminescence process is attributed to the
recombination of electrons and holes in localized states below the conduction band edge and the
valence band edge respectively [6]. The energy position and the shape of the PL spectrum are thus
determined by the distribution of carriers in these localized states and the transition probabilities for
radiative recombination. On the other hand, the distribution of the electrons and holes also
determines the Voc in the case of a quasi equilibrium condition, via the splitting of the quasi-Fermi
levels of the electrons and holes. Therefore, Voc and EPL are linked via the distribution of the excess
carriers. It is therefore expected that both values will exhibit similar trends, but the changes will
depend on the details of the density-of-states distribution (DOS). On the other hand, from the
investigation of both parameters, information on the DOS can be obtained. It should also be
mentioned here that the assumption of the quasi-equilibrium for the photogenerated carriers might
not be fulfilled for deep trap states at low temperatures. It is not clear how these conditions could be
described appropriately.
When compared to c-Si, the Voc of the µc-Si:H cells is lower at all temperatures but the
difference is largest at low temperatures and becomes smaller at high temperatures. At low
temperatures, the Fermi-Dirac distribution is sharp and the quasi-Fermi-level splitting is determined
by the integration over the DOS up to the quasi-Fermi-level. The smaller splitting of the quasi-
Fermi-level in the µc-Si:H cells is therefore strong evidence for a high DOS below the band gap in
these cells. These states shall be identified with the localized band tail states evidenced by other
techniques, e.g. [16, 17]. The much larger width of the PL band of the µc-Si:H films and solar cells
compared to their crystalline counterpart strongly supports this interpretation. Also the temperature
dependence of EPL is in line with this interpretation. The results can be discussed based on the well-
accepted model for radiative recombination between localized states in a-Si:H (see e.g. [13]), where
the temperature dependent shift of EPL is caused by a superposition of the shift of the optical gap and
the shift of the carrier distribution to lower energies. The latter is due to competing non-radiative
recombination processes.
The increase of EPL and Voc with increasing SC at a given temperature is interpreted by a
reduction of the density of band tail states. This results in a shift of the carrier distributions to higher
energy, and thus to a shift of the photoluminescence spectra to higher energies and to a larger
splitting of the quasi-Fermi levels, i.e. an increase of Voc, if the carrier concentration remains almost
constant. As shown in Fig. 4, the photoluminescence intensity does not change significantly, and
therefore the quantum efficiency remains fairly constant. Changes in the defect concentration can
therefore be ruled out, and the assumption of a constant carrier concentration seems appropriate.
Investigations of the short circuit current density jsc on the solar cells studied here reveal a decreases
RS [1], decreases from more than 65% for SC=3% to

Page 7

Band tail states in microcrystalline silicon solar cells probed by photoluminescence…


127
of jsc , i.e. a decreasing carrier collection, with decreasing temperature below about 200 K [15]. In
contrast, jsc of the c-Si solar cell remains almost constant down to 50 K whereas jsc of an a-Si:H solar
cell already starts to drop below 250 K. This indicates a decreasing carrier drift length in µc-Si:H,
probably due to a decreasing carrier mobility towards low temperature in agreement with results
from Hall measurements on µc-Si:H films which have been interpreted as evidence for tail states
[17]. Further support for the interpretation presented here stems from model calculations of PL
spectra for a given DOS and quasi-Fermi level splitting, i.e. Voc (see [18]), and from the results of
Voc as a function of the generation rate [15]. The results are consistent with the assumption that band
tail states act as traps for photogenerated carriers and limit the shift of the quasi-Fermi levels
towards higher energies.
Although there is only little known on the cause of the band tail states on a microscopic
scale, strain and interface states at the crystalline grains are possible sources for such states. We
propose that the reduction of the density of band tail states is due to structural relaxation of the
µc-Si:H network by the incorporation of hydrogen or amorphous hydrogenated silicon. This effect
would be similar to that obtained in polycrystalline-silicon thin film transistors, where a significant
reduction in the concentration of tail states has been observed after hydrogenation [19].


5. Conclusions

The change of EPL, and Voc, in µc-Si:H solar cells, as a function of the silane concentration
and temperature, can be interpreted assuming band tail states extending from the band edges into the
band gap of silicon. The decrease of EPL and Voc with increasing temperature is due to the change of
the carrier distributions in the band tails, and the accompanying shift of the quasi-Fermi energies to
lower energies. The increase of EPL and Voc for increasing SC is attributed to a reduced density of
band tail states. It is suggested that hydrogen or hydrogenated amorphous silicon surrounding the
crystalline phase leads to an improved lattice relaxation, and that thereby the density of band tail
states is reduced. The differences between HW-CVD and PE-CVD material and solar cells shall be
explored by a similar study.


Acknowledgements

We gratefully acknowledge technical support by M. Hülsbeck, J. Wolf and S. Michel, and
financial support by the Bundesministerium für Umwelt, Naturschutz und Reaktorsicherheit under
contract no. 0329814A.


References

[1] L. Houben, M. Luysberg, P. Hapke, R. Carius, F. Finger, H. Wagner, Phil. Mag. A
77, 1447 (1998).
[2] O. Vetterl, F. Finger, R. Carius, P. Hapke, L. Houben, O. Kluth, A. Lambertz, A. Mück,
B. Rech, H. Wagner, Solar Energy Materials and Solar Cells 62, 97 (2000).
[3] J. Meier, E. Vallat-Sauvain, S. Dubail, U. Kroll, J. Dubail, S. Golay, L. Feitknecht, P. Torres,
S. Fay, D. Fischer, A. Shah, Solar Energy Materials and Solar Cells 66, 73 (2000).
[4] O. Vetterl, R. Carius, L. Houben, C. Scholten, M. Luysberg, A. Lambertz, F. Finger,
H. Wagner, Mater. Res. Soc. Symp. Proc. 609, A15.2 (2000).
[5] S. Klein, F. Finger, R. Carius, B. Rech, L. Houben, M. Luysberg, M. Stutzmann, Mater. Res.
Soc. Symp. Proc. 715, A26.2 (2002).
[6] R. Carius, T. Merdzhanova, F. Finger, S. Klein, O. Vetterl, J. Materials Science: Materials in
Electronics 14, 625 (2003).
[7] R. Carius, T. Merdzhanova, F. Finger, Mater. Res. Soc. Symp. Proc. 762, A4.2 (2003).
[8] R. Carius, F. Finger, U. Backhausen, M. Luysberg, P. Hapke, H. Overhof, Mater. Res. Soc.
Symp. Proc. 467, 283 (1997).

Page 8

R. Carius, T. Merdzhanova, S. Klein, F. Finger


[9] D. Han, G. Yue, J. D. Lorentzen, J. Lin, H. Habuchi, Q. Wang, J. Appl. Phys.
87, 1882 (2000).
[10] G. Yue, J. D. Lorentzen, J. Lin, D. Han, Q. Wang, J. Appl. Phys. 88, 4904 (2000).
[11] A. Kaan Kalkan, S. J. Fonash, Shang-Cong Cheng, Appl. Phys. Lett. 77, 55 (2000).
[12] P. K. Bhat, G. Diprose, T. M. Searle, I. G. Austin, P. G. LeComber, W. E. Spear, Physica B
117 - 118, 917 (1983).
[13] R. A. Street, Adv. Phys. 30, 593 (1981).
[14] S. Klein, F. Finger, R. Carius, T. Dylla, B. Rech, M. Grimm, L. Houben, M. Stutzmann,
Thin Solid Films 430, 202 (2003).
[15] T. Merdzhanova, R. Carius, F. Finger, S. Klein, D. Dimova-Malinovska, Thin Solid Films
451 - 452, 285 (2004).
[16] S. Reynolds, V. Smirnov, C. Main, R. Carius, F. Finger, Mater. Res. Soc. Symp. Proc.
762, A3.4.1 (2003).
[17] R. Carius, J. Müller, F. Finger, N. Harder, P. Hapke in Thin Film Materials and Devices –
Developments in Science and Technology, eds. J.M. Marshall, N. Kirov, A. Vavrek,
J. M. Maud, (World Scientific Publishing Co.), p. 157 (1999).
[18] T. Merdzhanova, R. Carius, F. Finger, S. Klein, D. Dimova-Malinovska, J. Optoelectron. Adv.
Mater. 7(1), 485 (2005).
[19] J. R. Ayres, J. Appl. Phys. 74, 1787 (1993).
128