Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
art ic l e i nf o a b s t r a c t
Article history: The most straightforward way to map the photo-induced short circuit current density (Jsc) in solar cells is
Received 19 March 2015 light beam-induced current (LBIC) mapping. Recently several methods for Jsc imaging based on camera-
Received in revised form based photoluminescence and illuminated lock-in thermography imaging were proposed. This letter
30 June 2015
reports an alternative method for Jsc imaging, which is solely based on the evaluation of dark lock-in
Accepted 17 July 2015
thermography images. This method is particularly advantageous to improve the accuracy of dark lock-in
Available online 6 August 2015
thermography based local efficiency analysis of solar cells.
Keywords: & 2015 Elsevier B.V. All rights reserved.
Quantitative evaluation
Photocurrent imaging
Lock-in thermography
2D device simulation
Local analysis
The local short circuit current density Jsc,i (i¼ position index) is Jsc imaging method based on illuminated lock-in thermography
one of the basic local parameters for the efficiency of solar cells. (ILIT) was proposed [12]. This method is based on measuring the
Together with the local dark saturation current density J01,i and its bias-dependent dissipated heat due to the thermalization of the
ideality factor (a measure for injection-dependent recombination) local photocurrent in the depletion region under weak reverse
it governs the local open circuit voltage potential Voc,i in the cell. In bias, well before avalanche multiplication takes place. Eventual
most of previous imaging methods for efficiency-related solar cell ohmic shunts may be corrected by using an additional dark lock-in
parameters [1–5] Jsc,i is assumed to be distributed homogeneously thermography (DLIT) image taken at the cell bias of the ILIT
across the cell. However, in particular in multicrystalline silicon
measurement. Manipulation of the thermalization power density
solar cells, not only J01 but also Jsc is distributed inhomogeneously
across the pn-junction was used already earlier for imaging local
and even may limit the local conversion efficiency [6]. The classical
avalanche multiplication factors [13]. This ILIT method should
way to create an image of Jsc is light beam-induced current (LBIC)
work well, it is robust to series resistance variations and is easily
mapping [7]. However, this is a sequential and therefore time-
applicable to arbitrary excitation spectra, since the excitation
consuming method, and if AM 1.5 data should be obtained, it is
experimentally expensive, since it needs an LBIC system with needs not to be pulsed [14]. As for other light-induced methods
different wavelengths [8]. Based on an alternative one-diode such as PL imaging and LBIC, the accuracy of these results is
model of illuminated solar cells [9], a photoluminescence (PL) directly influenced by the homogeneity of the illumination
based method for imaging Jsc was proposed [10]. However, very intensity, which may, however, be corrected if known.
recent simulations have revealed that the accuracy of PL-based J01 In this letter another method for imaging Jsc is proposed, which
imaging in local low-lifetime regions, on which this Jsc imaging is based solely on the evaluation of DLIT images. The advantage of
method is based on, is insufficient [11]. The main reason is that this method, compared to PL-based Jsc imaging, is that it should be
this method strongly depends on the model of independent more accurate. Its advantage to the existing ILIT method is that the
diodes, each of them being connected with the terminals by an new method appears as a by-product improving the accuracy of
individual series resistor, which does not hold for a solar cell. DLIT-based local efficiency analysis [5,15] without needing any
Hence, this PL-based method is obviously not accurate. Recently a new measurements. Since this is a dark measurement, is also not
influenced by the inhomogeneity of an illumination system. In
n
Corresponding author. particular, it can be applied afterwards to measurements per-
E-mail address: breiten@mpi-halle.mpg.de (O. Breitenstein). formed in the past.
http://dx.doi.org/10.1016/j.solmat.2015.07.027
0927-0248/& 2015 Elsevier B.V. All rights reserved.
O. Breitenstein et al. / Solar Energy Materials & Solar Cells 143 (2015) 406–410 407
Fig. 2. PC1D simulation of (a) nsc(z) of standard cell, and (b) nsc(z) of PERC cell for two values of z and different illumination conditions (940 nm: red, 859 nm olive, 780 nm
blue, AM 1.5 black). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 3. PC1D results (symbols) and fit (lines) of Jrec,sc over J01 for the standard cell (a) and the PERC cell (b). The bulk lifetimes are given.
lifetime condition by evaluating Jsc and Voc. The very weak for monochromatic illumination are crossing each other close to
e
dependence of J01 on the illumination wavelength is due to the Jrec,sc ¼0 and J01 = J01, which is the emitter part of J01. This indicates
different rear surface excess carrier concentrations for different that, for this near-IR illumination, the recombination under short
effective excitation depths, leading to different J01rear contribu- circuit condition is indeed governed by bulk and backside
tions. We see that, as expected, for low values of J01 (high τb), Jrec,sc recombination, hence under this condition recombination in the
depends linearly on J01 and for larger J01 (low τb) sub-linearly. The emitter is negligible. This does not hold for AM 1.5G illumination,
results for AM 1.5G are lying closest to 850 nm. This means that, at which also contains short-wavelength contributions.
least in the cases considered here, monochromatic illumination at For a given solar cell to be investigated and a certain illumi-
850 nm approximates AM 1.5G illumination quite well with nation condition, as a rule the value of Jgen is not exactly known.
respect to Jsc imaging. Only for 940 nm, which corresponds to the
The PC1D-based values in Table 1 cannot generally be used, since
highest penetration depth, the values of Jrec,sc are significantly
they also depend on the illumination intensity, the cell thickness,
larger, as are the values of n(z) in Fig. 1.
and on optical properties like the reflectivity, the texturization,
The simplest way to describe a linear and then slightly bent
and the backside reflectance. However, if for the investigated solar
curve is a parable. Therefore the empiric method proposed here
consists in approximating this dependence by a square function cell the average value oJsc 4 is known, e.g. from flasher mea-
with three parameters A, B, and C surements, the method proposed here can be applied never-
theless, and the number of parameters can even be reduced from
2
Jrec,sc = A J01 − B J01 −C (2) three to two. From Eqs. (1) and (2) follows:
The parameter A defines the slope of the curves in Fig. 2, 2
Jsc, i = Jgen − A J01, i + BJ01, i + C (3)
parameter B defines the degree of non-linearity towards high bulk
lifetimes, and parameter C defines the average absolute value of Jsc. The average of (3) over the whole area is
The lines in Fig. 3 are the best fit results after Eq. (2), the para-
2
meters for all illumination conditions are given in Table 1. We see A J01, i − B J01, i
that Eq. (2) approximates the dependencies quite well, only for
⟨Jsc ⟩ = Jgen + C − ∑
i
N (4)
very low lifetimes (1 and 3 ms) and 940 nm illumination there are
minor departures. Also these simulations show that an illumina- (N ¼number of pixels). This leads to the final result, which only
tion wavelength of about 850 nm acts closest to AM 1.5G. All fits contains the parameters A and B
O. Breitenstein et al. / Solar Energy Materials & Solar Cells 143 (2015) 406–410 409
2
A J01, i − B J01, by Eq. (5) after fitting the parameters A and B for optimum
i 2
Jsc, i = ⟨Jsc ⟩ + ∑ − A J01, i + B J01, i agreement between (c) and (f). For comparison, Fig. 4(e) shows
i
N (5)
ILIT-Jsc data of this cell measured after [12]. Fig. 4(d) shows a plot
It seems now that the parameter C has no physical meaning of DLIT-based Jsc versus blurred LBIC-based Jsc data in the region
anymore. In reality it governs the average value of Jsc. Only if this framed in Fig. 4(c). The data in this region are not disturbed by any
average value is known, C may be replaced leading to Eq. (5). busbar or edge effects. This scatter plot was found to be the most
The method is tested on two industrial 156 156 mm2 sized accurate tool for adjusting the parameters A and B. Note that the
multicrystalline solar cells. The one, from which images are shown Local I–V method replaces the data in the busbar regions by
here, showed a flasher Jsc of 34.9 mA/cm2 and at the ILIT investi- averaged data outside of this region. Therefore in (f) also the
gation a mean Jsc of 32.8 mA/cm2, all other Jsc images are referred busbars contain Jsc data, in contrast to LBIC- and also to ILIT-Jsc,
to the latter value. For comparison a Jsc image of this cell was also where these regions are shadowed. This difference has been
measured by the ILIT-based method [12]. The DLIT measurements regarded here by setting the average value of DLIT-Jsc slightly
were performed at 10 Hz partly by using the latest version of the higher than that of the other Jsc images (34 instead of
“PV-LIT” system of the company InfraTec GmbH Dresden (Ger- 32.8 mA/cm2).
many) [18] and partly by using a custom-designed LIT system by We see in Fig. 4 that the general agreement between the
the company IRCAM GmbH [19]. The ILIT measurements were blurred LBIC-Jsc, the ILIT-based Jsc, and the DLIT-based Jsc is good.
performed at 40 Hz by using the commercially available system As expected from Eqs. (1) and (2), where the linear contribution of
“IR-MultitoolTM” by the company “Aescusoft GmbH” [20]. LBIC parameter A is dominant, the Jsc images look qualitatively like the
measurements were made monochromatically at 780, 840, and inverted J01 image. Fig. 4(c) shows that for most of the data the
940 nm wavelength, and a map for AM 1.5G was derived from quantitative correlation between DLIT- and LBIC-Jsc is also rea-
integrating the spectrum in conjunction with several monochro- sonable. Only in regions with very high J01 (lowest Jsc) DLIT-Jsc
matic LBIC maps [8]. The experimentally fitted parameters A and B slightly overestimates the recombination activity of the defects.
for these illumination conditions can be found in Table 1. This is probably caused for this wavelength by the too simple
Fig. 4 shows results of monochromatic illumination at 940 nm, quadratic approximation made in Eq. (2). The spatial resolution of
since here the LBIC contrast was largest. Fig. 4(a) shows the J01 the ILIT-based J01 image in Fig. 4(e) is better than that of the DLIT-
image of this cell, which was obtained by evaluating 2 DLIT images based image (f) since for (e) a higher lock-in frequency of 40 Hz
measured at flock-in ¼10 Hz and V ¼ þ0.55 and þ0.6 V by “Local was used than the 10 Hz used for (f). The fit leads for 940 nm
I–V” [5], assuming a homogeneous effective series resistance of illumination to parameters A¼ 3.5 109 and B ¼2.5 1020 cm2/A,
Rs ¼ 0.6 Ω cm2. Fig. 4(b) shows an LBIC image taken at 940 nm which are also given in Table 1. These fitting parameters, as well as
wavelength, which is scaled to an average of 32.8 mA/cm2. Fig. 4 the ones experimentally obtained for the other illumination con-
(c) shows the same LBIC data artificially blurred by the same point ditions, significantly deviate from the simulated values. The
spread function, which is also responsible for the thermal blurring parameter A appears generally somewhat smaller than expected
of the DLIT data. Fig. 4(f) shows the DLIT-based Jsc image obtained and parameter B is clearly larger. Since these data were obtained
Fig. 4. (a) DLIT-based J01 (0–8 pA/cm2), (b) LBIC at 940 nm (20–40 mA/cm2), (c) artificially blurred LBIC at 940 nm (20–40 mA/cm2), (d) DLIT-based Jsc versus blurred LBIC-
based Jsc in the region framed in (c), (e) ILIT-based Jsc (20–40 mA/cm2), (f) DLIT-based Jsc (20–40 mA/cm2).
410 O. Breitenstein et al. / Solar Energy Materials & Solar Cells 143 (2015) 406–410
on two different cells (AM 1.5 þ940 nm vs. 780 þ830 nm; the first Acknowledgment
cell containing significantly stronger J01 maxima, see Fig. 3), it
cannot be excluded that at least parameter B, which is used to The authors are grateful to D. Hinken and K. Bothe (ISFH,
describe the nonlinearity of the Jsc(J01) dependence, also depends Hameln, Germany) and to M. Padilla and E. Schäffer (Fraunhofer
on the used cell. Maybe in future a physically more justified ISE) for performing the LBIC mappings, to J. Greulich and H. Höffler
approximation than (2) can be found to describe this dependence. (Fraunhofer ISE) for valuable discussions and providing the special
We believe that these deviations between the simulations and the P doping profile describing the PERC cell, to I. Geisemeyer and S.
experiments are due to the fact that the simulations were made Wasmer (Fraunhofer ISE) for LIT measurements, and to InfraTec
for a homogeneous bulk lifetime, but in reality the lifetime in the GmbH Dresden (Germany) [18] for cooperation and providing the
poor quality regions is governed by local defects like grain PV-LIT system used for part of this investigation.
boundaries, as can be clearly seen in the LBIC image Fig. 4(b). It is
well known that the influence of such local defects has to be
averaged differently to that of homogeneous SRH defects for References
describing their influence on luminescence and on J01 [21]. For
describing luminescence imaging, and also for Jsc imaging, the [1] T. Trupke, E. Pink, R.A. Bardos, M.D. Abbott, Spatially resolved series resistance
of silicon solar cells obtained from luminescence imaging, Appl. Phys. Lett. 90
lifetime averaging has to occur linearly. Hence, in worst case, low (2007) 093506.
lifetime regions do not contribute to the average. Therefore the [2] M. Glatthaar, J. Haunschild, M. Kasemann, J. Giesecke, Warta, S. Rein, Spatially
relative influence of low lifetime regions on the averaged lumi- resolved determination of dark saturation current and series resistance of
silicon solar cells, Phys. Status Solidi (RRL) 4 (2010) 13–15.
nescence or Jsc signal cannot be larger than the areal fraction of [3] M. Glatthaar, J. Haunschild, R. Zeidler, M. Demant, J. Greulich, B. Michl,
these regions. J01, on the other hand, may assume very high values W. Warta, S. Rein, R. Preu, Evaluating luminescence based voltage images of
in low lifetime regions. Therefore the influence of small local low- silicon solar cells, J. Appl. Phys. 108 (2010) 014501.
[4] C. Shen, H. Kampwerth, M. Green, Spatially resolved photoluminescence
lifetime defects is much stronger on J01 than on the average life- imaging of essential silicon solar cell parameters, in: Proceeding of the 38th
time and on Jsc. This is the reason why, for predicting solar cell IEEE Photovoltaic Specialists Conference, Austin, USA, 2012, pp. 1855–1859.
efficiencies, in particular Voc, 1/ τeff (which is for small bulk life- [5] O. Breitenstein, Nondestructive local analysis of current-voltage characteristics
times proportional to J01) has to be averaged instead of τeff [21].
of solar cells by lock-in thermography, Sol. Energy Mater. Sol. Cells 95 (2011)
2933–2936.
The strongly increased J01 in positions of recombination-active [6] B. Michl, M. Rüdiger, J.A. Giesecke, M. Hermle, W. Warta, M.C. Schubert, Effi-
grain boundaries has been proven experimentally in [22] and ciency limiting bulk recombination in multicrystalline silicon solar cells, Sol.
Energy Mater. Sol. Cells 98 (2012) 441–447.
recently also by high-resolution PL imaging in [23]. The para- [7] J. Marek, Light beam-induced current characterization of grain boundaries,
meters A and B in Table 1 describe a correlation between J01 and J. Appl. Phys. 55 (1984) 318–326.
Jsc. If, for homogeneous lifetime distributions (as assumed in PC1D) [8] M. Padilla, B. Michl, B. Thaidigsmann, W. Warta, M.C. Schubert, Short-circuit
current density mapping for solar cells, Sol. Energy Mater. Sol. Cells 120 (2014)
and for strongly inhomogeneous distributions (as for multi- 282–288.
crystalline cells) different averaging rules for J01 and Jsc have to be [9] O. Breitenstein, An alternative one-diode model for illuminated solar cells,
applied, this leads to different values for A and B in both cases, IEEE J. Photovolt. 4 (2014) 899–905.
[10] O. Breitenstein, H. Höffler, J. Haunschild, Photoluminescence image evaluation
which explains the differences in Table 1. If the method described of solar cells based on implied voltage distribution, Sol. Energy Mater. Sol.
here is applied to multicrystalline silicon cells, experimentally Cells 128 (2014) 296–299.
obtained values of A and B should be used. [11] O. Breitenstein, J. Bauer, D. Hinken, K. Bothe, The reliability of thermography-
and luminescence-based saturation current density imaging, Sol. Energy
It might be asked for the relevance of an imaging method, Mater. Solar Cells 137 (2015) 50–60.
which needs two fitting parameters (A and B) to be applied. Note, [12] F. Fertig, J. Greulich, S. Rein, Spatially resolved determination of the short-
however, that these two parameters have to be determined for a circuit current density of silicon solar cells via lock-in thermography, Appl.
Phys. Lett. 104 (2014) 201111.
certain cell type and illumination condition only once by com- [13] O. Breitenstein, J. Bauer, J.-M. Wagner, A. Lotnyk, Imaging physical parameters
paring the results of this method to that of another established Jsc of pre-breakdown sites by lock-in thermography techniques, Progr. Photovolt.
imaging or mapping method like LBIC or ILIT-based Jsc imaging Res. Appl. 16 (2008) 676–685.
[14] F. Fertig, J. Greulich, St Rein, Short-circuit current density imaging of crystal-
[12]. Once these parameters are known, the method can be line silicon solar cells via lock-in thermography: robustness and simplifica-
applied to all other cells of this type without further calibration. tions, J. Appl. Phys. 116 (2014) 184501.
The results in Table 1 show that the parameters A and B strongly [15] O. Breitenstein, Local efficiency analysis of solar cells based on lock-in ther-
mography, Sol. Energy Mater. Sol. Cells 107 (2012) 381–389.
depend on the illumination wavelength, but less on the cell type. [16] PC1D, Version 5.9, http://www.engineering.unsw.edu.au/energy-engineering/
This shows that the method should also work with cells having pc1d-software-for-modelling-a-solar-cell.
slightly different parameters. In particular, this method overcomes [17] K. Bothe, D. Hinken, Quantitative luminescence characterization of crystalline
silicon solar cells, in: G.P. Willeke, E.R. Weber (Eds.), Academic Press (Bur-
the assumption of a homogeneous Jsc previously made in the DLIT- lington), Elsevier, 2013, p. 299.
based “Local I–V” method [5,15], for which the software is avail- [18] See http://www.infratec.eu/.
able [24], without needing any additional measurements. In the [19] See http://www.ircam.de/.
[20] See http://www.aescusoft.com/.
latest version of this software the automatic calculation of Jsc(x,y) [21] R.A. Sinton, Predicting multi-crystalline solar cell efficiency from lifetime
by the method proposed here is already included as an option. One measured during cell fabrication, in: Proceeding of the 3rd World Conference
limitation of our method is that, by describing the whole cell by on Photovoltaic Energy Conversion, Osaka, Japan, 2003, pp. 1028–1031.
[22] S. Rißland, O. Breitenstein, High resolution saturation current density imaging
DLIT-based J01 and the two parameters A and B, it assumes at grain boundaries by lock-in thermography, Sol. Energy Mater. Sol. Cells 104
homogeneous values of the IR emissivity. Therefore it is advisable (2012) 121–124.
to make the DLIT measurements by applying local emissivity [23] O. Breitenstein, J. Bauer, D. Hinken, K. Bothe, Towards an improved Laplacian-
based photoluminescence image evaluation method, Sol. Energy Mater. Sol.
correction. In conclusion, the DLIT-based Jsc imaging method Cells (2015) http://dx.doi.org/10.1016/j.solmat.2015.06.020, in press.
described here may become a valuable contribution to make DLIT- [24] See http://www.max-planck-innovation.de/en/.
based local efficiency analysis more realistic.