Você está na página 1de 56

UPMC Universit

e Paris VI
Master (M2) in Fluid Mechanics
Internship at E.T.S.I.A., Universidad Polit

ecnica de Madrid
Electrical characterization of amorphous
silicon solar cells
Author:
Pablo Pe

nas
Supervisor:
Dr. Miguel Hermanns
27th August 2013
1
Technical abstract
This work begins with an introduction to the fundamentals of operation of an amorphous silicon
p-i-n cell in order to qualitatively explain its electrical behaviour. Other important features
such as carrier recombination and the shunt-leakage current are discussed.
A procedure to electrically characterize the current-voltage (I V ) behaviour of amorphous
silicon p-i-n cells is then presented. The electrical behaviour of the cell is described analytic-
ally by an equivalent circuit model, alongside its limitations and underlying assumptions. The
characterization process essentially consists in experimentally determining the unknown circuit
elements. A Variable Illumination Method is employed, where the experimental data is strictly
limited to various I V curves under specic illumination conditions. The simplicity of the
experimental set up, consisting purely of an appropriate light source and an I V measuring
station, is a main advantage of this particular characterization method.
A single-junction cell is rst characterized. All parameters except the eective -product and
built-in voltage could be extracted from aI V curve in the dark. The eective -product, an
indicator of the rate of carrier recombination and state of degradation of the cell, has been shown
to depend on the illumination spectrum and intensity. The I V curves predicted by the model
were observed to perform poorly under large reverse voltage biasing. This has been attributed
to a non-Ohmic voltage dependence of the shunt leakage current in this region. Therefore, an
extended model that takes into account shunt leakage current non-linearities is also presented.
A double-junction a-Si:H/a-Si:H cell is next characterized. The tandem cell was exposed to a
particular IR and UV bias light in order to excite each subcell dierently. It was not possible to
achieve single subcell excitation. Nevertheless, theoretical relations have been derived for such
scenario. Instead, it was concluded that UV light induced a considerably greater photocurrent
in one particular subcell than in the other, while IR light had the opposite eect. The charac-
terization process was then adapted to match these particular circumstances. The end result
was a set of analytical expressions relating graphically-obtained variables (open-circuit voltage
and resistance, short-circuit current and resistance) to the circuit parameters. These expressions
were seen to be supported by experiment. The characterization proved useful to determine the
state of degradation of one subcell relative to the other.
Further work is required to model the J V curves for the double-junction cell. This may be
done by rst achieving single subcell excitation, therefore making use of the theoretical relations
derived for this scenario.
2
3
Acknowledgments
The contribution of the project supervisor, Dr. Miguel Hermanns, of the Universidad Politecnica
de Madrid (UPM), merits special acknowledgment for his excellent guidance, reviews and sug-
gestions that have made this work possible.
The author wishes to thank the members of Instituto de Energa Solar (IES) of the UPM,
especially Yedileth Contreras for her great help with the apparatus set-up and experimental
measurements; and Ignacio Rey-Stolle for kindly providing the electrical characterization facil-
ities.
The author gratefully acknowledges the vital contribution of Alonso Pardo in the design and
assembly of the circuit and LED boards.
Finally, the author would like to express his gratitude towards Javier Izard from the company
Soliker, for providing the test cell and for sharing valuable insight on the electrical behaviour of
amorphous silicon solar cells.
4
Contents
1 Introduction 6
2 Electrical behaviour of a-Si:H solar cells 7
2.1 Fundamentals of operation of a p-i-n junction . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Diusion and voltage biasing . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.2 Carrier recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Single-cell electrical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 General equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Limitations of the model and underlying assumptions . . . . . . . . . . . 14
2.3 Extended model for reverse voltage biasing . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Shunt leakage current and shunt-busting . . . . . . . . . . . . . . . . . . . 16
3 Test cell and experimental technique 17
3.1 Test cell description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Apparatus description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4 Electrical characterization of a single-junction cell 19
4.1 Measurements in dark conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.1 Reverse voltage biasing model . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Measurements under illumination . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Model performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3.1 Cell response to changes in illumination intensity and spectrum . . . . . . 31
4.4 Sources of error and uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5 Electrical characterization of a double-junction cell 37
5.1 Model and general equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Experimental procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.3 Measurements in dark conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.4 Theoretical derivations for single subcell excitation . . . . . . . . . . . . . . . . . 41
5.4.1 SC current OC voltage relation . . . . . . . . . . . . . . . . . . . . . . . . 42
5.4.2 SC and OC resistance property . . . . . . . . . . . . . . . . . . . . . . . . 43
5.5 Measurements under illumination . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.5.1 Baseline assumptions at SC and OC operating conditions . . . . . . . . . 46
5.5.2 SC resistance-current relation . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.5.3 The V-R-J relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.6 Summary of ndings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6 Conclusions 54
6.1 Further work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1 Introduction
Thin-lm hydrogenated amorphous silicon (a-Si:H) p-i-n solar cells are extensively used in a wide
range of applications. They are commonly used in power sources for electronic devices such as
calculators and watches, batteries, photosensors and building integrated photovoltaics such as
semi-transparent building facades or window glazing. Amorphous silicon solar cells constitute
one of the most promising options for low-cost, large scale applications in photovoltaics.
The main advantage of thin-lm a-Si:H cells over crystalline silicon cells is their thinness ( 300
nm). This does not only imply a-Si:H cells having lower material costs, but it also makes them
more aesthetically attractive and exible in their applications. Furthermore, manufacturing
costs are potentially low since there is a greater cost reduction potential than in c-Si cells due
to the signicantly lower amount of silicon material required in its manufacture.
On the other hand, they suer from limited eciency ( 5 10%) partly due to light-induced
degradation (Staebler-Wronski eect) that manifests in the form of relatively high carrier re-
combination losses.
A popular solution to improve the cell conversion eciency is to stack cells of dierent band gaps
to form a multi-junction or tandem cell. It is of uttermost importance to be able to properly
characterize single-junction and tandem cells to allow for future improved designs.
This leads to the overall purpose of this work: to provide an electrical characterization method
for a-Si:H cells. Particularly, a double-junction a-Si:H/a-Si:H cell shall be characterized, treating
it rst as a single-junction cell and then as a double-junction cell.
The cell may be described analytically by an equivalent circuit. Equivalent circuits are a con-
venient way of characterizing and modelling cells. They provide insight on the physical processes
that take place within the cell and evaluation of the circuit parameters gives a clear picture of
the cells properties.
The characterization process essentially consists in experimentally determining the unknown
parameters of the equivalent circuit via a Variable Illumination Method. The experimental data
is strictly limited to various I V curves under specic illumination conditions. No knowledge
of the exact irradiance power, spectral response or QE of the individual subcells is required.
While for the single-junction cell characterization white light is best suited for the task, for
the tandem cell characterization, bias UV and IR illumination shall be employed to excite the
subcells dierently.
The main advantage of this characterization method over other methods is the simplicity of the
experimental set-up, consisting purely of an appropriate light source and an I V measuring
station.
This report is structured as follows. Section 2 covers the fundamentals of operation of p-i-n
junctions found in a-Si:H cells in order to understand the shape of their characteristic I V
curves. Next, the analytical, equivalent circuit model for a single-junction cell is introduced
alongside its limitations and underlying assumptions. The physical characteristics of the test
tandem cell, together with the measuring apparatus and light sources, are briey described in
Section 3. Section 4 is devoted to the characterization using the single-junction cell model. The
model performance is evaluated, and the common sources of error are exposed. In Section 5, the
equivalent circuit model shall be extended to describe a double-junction cell. The characteriz-
ation method for such then follows. Lastly, the concluding remarks followed by possible future
work are presented in Section 6.
6
2 Electrical behaviour of a-Si:H solar cells
This section qualitatively describes the physics of p-i-n junctions and other aspects such as car-
rier recombination in order to gain valuable insight on the electrical behaviour of hydrogenated
amorphous silicon cells. An equivalent circuit able to model such behaviour is then presented.
2.1 Fundamentals of operation of a p-i-n junction
The photodiode inside an a-Si:H based cell has a p-i-n structure. The rst layer consists in a thin
(usually 10 40 nm), p-type doped layer. Since it is negatively charged, holes are the majority
carriers. The second layer is refered to as the intrinsic (i ) layer. Typically, it has a thickness
d = 200 600 nm. The third layer is a thin, n-type doped layer (of similar thickness to the p-
layer). Since it is positively charged, electrons are the majority carriers. In thermal equilibrium,
electrons are donated from the n-layer to the p-layer. This generates an approximately uniform
built-in electric eld E
b
acting in the direction shown in Figure 1. A built-in voltage, V
b
, is also
generated across the p-i-n junction. The following expression may be used to relate the two:
|E
b
|
V
b
d
(1)
Figure 1: Schematic of a typical a-Si:H p-i-n cell. The TCO (transparent conducting oxide) and
AZO (aluminium zinc oxide) layers act as the front and back electrical contacts respectively.
Most of the photovoltaic generation of the electron-hole pairs takes part in the undoped i -
layer. Electron-hole pairs are created through photon absorption as depicted in Figure 2. A
photon with the right amount of energy may transfer such energy to an electron in the valence
band of the semiconductor material (a-Si:H), where it is tightly bound in a covalent bond
between neighbouring atoms. The electron becomes excited and jumps over to the higher
energy conduction band. An empty covalent bond is formed, referred to as a hole. This process
has been graphically represented in Figure 3. Electrons in the conduction band (and similarly
holes in the valence band) are free to move and therefore contribute to the electric current. Free
holes and electrons are referred to as carriers.
7
Figure 2: Schematic of the generation of electron-hole pairs in a p-i-n cell. Note that the
TCO forms the window layer, while the AZO layer acts as a reector to maximise the capture
probabiliy of the cell.
Figure 3: Band diagram representation of carrier photo-generation. The inicident photon is
represented by the curly orange arrow, whose energy is transferred to an electron in the semi-
conductor valence band. The excited electron jumps over to the conduction band, and a hole is
thus created in the valence band.
The electric eld drives the photo-generated free electrons and holes from the i -layer to the
n-type and p-type layers respectively. This is referred to as drift. Diusion (due to gradients in
carrier concentrations) drives carriers in the opposite senses. This has been illustrated in Figure
4. Drift and diusion forces are always in competition with each other.
The following current sign criterion shall be adopted. For the top p-i-n junction schematic in
Figure 4, the current driven by drift is larger than the current driven by diusion. Hence the net
output current (density) J shall be regarded as positive. For the bottom schematic, diusion
dominates over drift, generating a current in the opposite sense and consequently regarded as
negative. The dominant driving force is determined by the the external voltage V applied over
the cell, together with the illumination conditions. V directly aects the electric eld |E| as
follows:
|E|
V
b
V
d
(2)
8
Figure 4: Schematics of a p-i-n junction. The length of the arrows symbolises the magnitude
of the carrier uxes due to drift or diusion. Top: Case when drift dominates over diusion
(strong |E|). Operating point (A) in Figure 5 belongs to this regime. The net output current
is taken to be positive: J > 0. Bottom: Case when diusion dominates over drift (weak |E|).
Operating point (C) in Figure 5 belongs to this regime. The net output current is taken to be
negative: J < 0.
Consider a cell operating under intermediate illuminations. Its characteristic J V curve is
sketched in Figure 5. At operating point (A), the cell is operating in short-circuit conditions
(SC) since there is no external voltage applied (V = 0). At this particular point, E = E
b
. The
electric eld is strong, hence the current due to drift (J
drift
) is much bigger than the current due
to diusion (J
di
). As a result, the net current J is largely positive. In short, point (A) belongs
to the operating regime where J
drift
J
di
, portrayed in the top p-i-n schematic in Figure 4.
When V < 0 (reverse voltage biasing), the cell operates deeper in the J
drift
J
di
regime. This
is because |E| becomes even stronger, and so diusion eects in the i -layer become even more
negligible. The drift current is relatively insensitive to the applied electric eld [5] since it is
limited by the number of minority carriers rather than by the electric eld not being strong
enough. This is observed in Figure 5, where J is shown to increase minimally as V becomes
more negative.
9
Figure 5: Sketch of typical J V curves for an a-Si:H cell under illumination (solid line) and in
the dark (dashed line), labelled with the dierent main regimes of operation. MPP stands for
the Maximum Power Point of the cell.
The diusion current, on the other hand, is very sensitive to |E|; the mechanism relating the
two will be explained next in Section 2.1.1. When V > 0 (forward voltage biasing), |E| weakens
and the diusion current increases, eventually becoming comparable to the drift current as V
approaches the open-circuit voltage V
OC
. This happens often past the maximum power operating
point (MPP), where drift still dominates. Point (B) corresponds to open-circuit conditions (OC),
where V = V
OC
and J = 0. Therefore, this point belongs to the J
drift
= J
di
regime. If V is
increased further, |E| weakens even more and J
di
overcomes J
drift
, resulting in J < 0. Point (C)
belongs to this regime where J
drift
J
di
, portrayed in the bottom p-i-n schematic in Figure 4.
A cell operating in the dark behaves in the same way. However, at a given V , the diusion
current is always stronger than for illuminated conditions. This is because there are no electron-
hole pairs being photo-generated in the i -layer, hence the minority carrier diusion gradients
are steeper. As a result J
drift
dominates over J
di
only when V < 0. Note that at point (B),
V = 0 and J = 0, meaning the cell is essentially in thermal equilibrium.
2.1.1 Diusion and voltage biasing
Electrons diusing from the n-side to the p-side have to overcome an electrostatic potential
barrier of energy E
B
, where:
E
B
= E
C,p
E
C,n
= q(V
b
V ) (3)
E
C,p
is the conduction band energy at the p-layer, E
C,n
is the conduction band energy at the n-
layer, q is the carrier elementary charge. Clearly, E
B
decreases (increases) with forward (reverse)
voltage biasing. Note that the energy of the barrier is proportional to the electric eld |E| and to
the dierence in the electrostatic potential (
n

p
) across the p-i-n junction. This is pictured
in Figure 6.
10
Figure 6: (Electron) Energy band diagrams and their corresponding p-i-n junction operating
conditions. E
C
and E
V
denote the conduction and valence energy band respectively. Left: Case
where cell operates at short-circuit conditions (V = 0). The net current density will be J = 0 in
dark conditions, else J > 0 under illumination. Right: Case where cell operates under forward
voltage biasing (V > 0). Note that J > 0 if V < V
OC
, else J < 0 if V > V
OC
.
With no applied external voltage, the Fermi Level (E
F
) of the electrons at n and p is the same.
When V > 0, the energy input from the power source raises overall the energy of the electrons
at n, and the Fermi Level at n is shifted to a higher energy closer to E
C,p
. This means that
the probability of an electron in the conduction band at n having energy E > E
C,p
is greater.
Likewise, an electron photo-generated in the i -layer is more likely to have sucient energy to
overcome the now smaller energy barrier. Forward voltage biasing similarly reduces the energy
barrier in the valence band, (which of course faces the opposite direction), thus encouraging
hole diusion. The end result is the increase in J
di
with forward voltage biasing. Note that
the opposite happens with reverse voltage biasing, the energy barrier becomes larger, hence J
di
decreases.
2.1.2 Carrier recombination
Hydrogenated amorphous silicon contains an amphoteric dangling bond defect that can be neut-
ral, positively charged, or negatively charged. The lattice structure of a-Si:H is sketched and
compared to that of c-Si in Figure 7. Dangling bonds are empty covalent bonds, depicted as
dashed lines in Figure 7b.
Dangling bonds act as the main recombination centres for carriers. Recombination refers to
electrons in the conduction band losing energy and dropping down to the valence band, where
they are again bound in a covalent bond. High recombination rates result in small output
11
currents and low cell eciencies. It is obvious from Figure 7a that c-Si based solar cells are
absent of dangling bonds due to the organized, crystalline atomic arrangement of the semicon-
ductor material. This explains why c-Si cells oer small recombination current losses and higher
eciencies.
(a) c-Si structure (b) a-Si:H structure
Figure 7: Lattice structure of c-Si and a-Si:H. Dangling bonds are represented by dashed lines.
In a-Si:H cells, most of the bulk recombination in the i -layer occurs due to neutral dangling
bonds. However, near the p-i and n-i interfaces the dangling bonds may be charged. As the
illumination is reduced to low levels, the neutral zone in the i -layer shrinks. In the regions
near the p and n doped zones, dangling bonds are hence charged, locally weakening the electric
eld and increasing recombination [2]. This is known as interface recombination. In the doped
layers, the dopant atoms introduce many dangling bonds and do not contribute free electrons.
The recombination rate in these layers is so high that photons absorbed in doped layers do not
contribute to the power generated by solar cells [8]. The mere function of the doped layers is
to induce the built-in electric eld. a-Si:H cells suer from light induced degradation, known as
the Staebler-Wronski eect. This causes the cell eciency to signicantly decrease (by 15-30%)
during the rst few hundred hours of operation. It will be seen that the state of degradation
(hence the magnitude of recombination current losses) may be quantied in terms of the eective
mobility-lifetime () product of the photocarriers.
2.2 Single-cell electrical model
The electrical behaviour of an a-Si:H cell may be accurately described by the equivalent electrical
circuit model proposed in [1]. In this section, the equations of the model are rst presented and
the range of validity and underlying assumptions of the model are then discussed.
2.2.1 General equations
The equivalent circuit for a non-ideal a-Si:H cell is shown in Figure 8. Here, J
L
represents the
loss-free illumination current (or photocurrent) density generated by the cell. J
R
is the current
density lost due to carrier recombination in the i -layer. J
D
represents the ideal diode current
density and J
P
is the shunt leakage current density lost across the parallel or shunt resistance
R
p
. R
s
represents the series resistance of the non-ideal cell.
J is the net output current density of the cell, while V corresponds to the voltage across the cell
terminals. It is important to note that V is taken as positive (corresponding to forward voltage
biasing) when the n-terminal is at a higher external voltage potential than the p-terminal.
12
Likewise, J is dened to be positive when the current leaves the p-terminal, i.e. when drift
dominates over diussion. This means that this model takes the short-cicuit current (J
SC
) of
a cell under illumination to be positive. This is consistent with the voltage and current sign
criteria established back in Section 2.1. Note that the useful power P generated by the cell is
simply given by P = JV [W/m
2
].
Figure 8: Equivalent electrical circuit for a single-junction a-Si:H cell.
This circuit may be described mathematically by applying Kirchhos rst law, which states
that the sum of currents around any node in an electrical circuit must be zero:
J = J
L
J
R
J
D
J
P
(4)
The current term J
D
may be replaced by a more detailed expression describing the diode beha-
viour. Similarly, J
R
may be described through a recombination model proposed in [1], and J
P
may be expressed in terms of R
s
and R
p
. Equation (4) may therefore be fully written as:
J = J
L
J
L
d
2
()
e
[V
b
(V +JR
s
)]
J
0
_
e
V +JR
s
nV
Te
1
_

V +JR
s
R
p
(5)
The symbols J
0
, n, ()
e
correspond to the diode saturation current density, diode ideality
factor and eective carrier mobility and lifetime product respectively. It is also recalled that
d and V
b
are the i -layer thickness and built-in voltage. V
Te
is the thermal voltage, dened as
follows:
V
Te
=
k
B
T
q
(6)
where k
B
is Boltzmanns constant, T the absolute temperature and q the elementary charge.
Equation (5) will be referred to as the characteristic equation of the cell, which essentially
describes the relationship between cell current density J and cell voltage V . In the cell char-
acterization process, it is often more convenient to work in terms of the non-dimensional ideal
voltage of the cell, V

, and non-dimensional current density J

, rather than in terms of V and


J. V

and J

are dened as:


V

=
V +JR
s
V
Te
(7)
J

=
JR
s
V
Te
(8)
Note that V

represents the voltage drop across R


p
or the diode element, normalised by V
Te
.
J

represents the voltage drop across R


s
, normalised by V
Te
. J
0
and J
L
shall be likewise non-
dimensionalised as follows:
J

0
=
J
0
R
s
V
Te
(9)
J

L
=
J
L
R
s
V
Te
(10)
13
In terms of V

and J

, the characteristic equation of the cell (5) may be rewritten as follows:


J

= J

L
_
1
d
2
()
e
V
b
[1 V

(V
Te
/V
b
)]
_
J

0
_
e
V

/n
1
_

R
s
R
p
V

(11)
This will be referred to as the non-dimensionalised characteristic equation of the cell.
At this point, it is worth stating the well used term R denoting the cells characteristic resistance
in the V J plane:
R =
V
J
(12)
R is usually evaluated at short-circuit and open-circuit conditions, giving the short-circuit res-
istance R
SC
and open-circuit resistance R
OC
of the cell respectively. Useful information can be
extracted from these quantitities. It is therefore pertinent to introduce the term R

, dened as
the cells dimensionless resistance in the V

plane:
R

=
V

(13)
R

is obtained by dierentiating (11) with respect to J

. This gives:
R

=
_
J

L
d
2
()
e
V
b
[1 V

(V
Te
/V
b
)]
2
V
Te
V
b
+J

0
1
n
e
V

/n
+
R
s
R
p
_
1
(14)
Unfortunately, R

still depends on J

L
, a quantity which is a-priori unknown and dicult to
determine experimentally. It is therefore logical to eliminate J

L
from (14) through direct sub-
stitution. This is done rst by solving for J

L
in (11):
J

L
=
J

+J

0
_
e
V

/n
1
_
+
R
s
R
p
V

1
d
2
()
e
V
b
[1V

(V
Te
/V
b
)]
(15)
Substituting (15) into (14), an useful expression for R

is obtained:
R

=
_
_
J

+J

0
_
e
V

/n
1
_
+
R
s
R
p
V

1
d
2
()
e
V
b
[1V

(V
Te
/V
b
)]
d
2
()
e
V
b
[1 V

(V
Te
/V
b
)]
2
V
Te
V
b
+
J

0
n
e
V

/n
+
R
s
R
p
_
_
1
(16)
Note that in the absence of illumination, (J
L
= 0) the non-dimensional characteristic equation
and R

reduce to:
J

= J

0
_
e
V

/n
1
_
+
R
s
R
p
V

(17)
R

=
_
J

0
n
e
V

/n
+
R
s
R
p
_
1
(18)
2.2.2 Limitations of the model and underlying assumptions
This model diers from the typical equivalent circuit of P-N junction solar cells in the inclusion
of a current loss term J
R
, that strongly increases with forward voltage. This term takes into
account the recombination losses in the i -layer of the cell previously discussed in Section 2.1. It
is a function of the eective product (combines the lifetime and mobility of free electrons
and holes), that determines the state of degradation. This recombination function is taken from
the neutral dangling-bond recombination model described in [4].
The electrical model, or rather, the recombination current term has been developed under the
following assumptions:
14
constant, strong electric eld |E| within the i -layer,
diusion eects, which are supposed much weaker than drift eects, are neglected,
bulk recombination in the i -layer is determined by the neutral dangling bonds.
The rst two assumptions are applicable to illuminated p-i-n cells under small positive or neg-
ative voltage biasing, with thin i -layers and relatively low defect densities therein. For large
forward voltages the model loses its validity since diusion can no longer be neglected, as it
has already been seen. Looking back at Figure 5, the recombination model is most valid for
the J
drift
J
di
regime. Its accuracy near point (B), i.e. near OC conditions, will still be
acceptable even though the theoretical assumptions are no longer fullled.
The third assumption is only valid for suciently strong illuminations such that J
SC
1 A/m
2
[2]. It is recalled that at lower illumination levels than this, most of the recombination occurs
in the p-i and n-i interfaces. This is because at the end regions in the i -layer near the p and
n doped zones, dangling bonds are charged, locally weakening the electric eld (which may
no longer be approximated as uniform), thus increasing recombination. As the illumination
intensity is increased, charge defects are neutralized, the eld becomes more uniform and inter-
face recombination decreases. It has been found that the eective -product may vary under
signicant changes in irradiance, usually increasing as the illumination level increases [2].
Finally, a very important assumption is that the model takes the shunt-leakage current of the
cell to be Ohmic, i.e. always linearly dependent on the external voltage. This is often untrue
when the cell is operating under large reverse voltage biasing, where the model is thus no longer
valid. This shall be explained and discussed in the next section below.
2.3 Extended model for reverse voltage biasing
A cell will normally operate near its maximum power point (MPP) under signicant forward
voltage biasing. However, in a photovoltaic module, heavy shading of a particular cell may
force that cell to operate at large negative voltages. At large reverse voltage biasing, the model
described by (5) will show signicant deviations from the experimental J V curves. The reason
is that the model does not take into account the full physics of the reverse bias characteristic
of the cell. Two important eects are the diode avalanche breakdown at large negative voltages
and the non-linearity of the shunt leakage current.
The excess variable dark leakage current observed at low voltage biasing is commonly referred
to in the literature as shunt leakage current (I
SH
). In the equivalent circuit model given by
Equations (4) and (5), the shunt leakage current is assumed to be Ohmic (linearly dependent on
voltage). The term J
P
has been used to denote the strictly Ohmic shunt leakage current density.
Consequently, it has been represented as the current across a parallel or shunt resistance (R
p
),
as pictured in Figure 8. This model provides a satisfactory t for forward voltage biasing.
However, at high enough reverse voltage biases, the leakage current shows a non-linear voltage
dependence, where I
SH
|V |

, with 2 3 [7]. The non-linearity of I


SH
at large negative
voltages is clearly observed in Figure 18, where the measured current almost purely consists of
the shunt leakage current.
An extension term has been developed [11] in the form of an additional current term J
B
which
describes the diode avalanche breakdown and the shape of the reverse bias characteristic of the
cell. When the voltage reaches the breakdown voltage, V
br
, the cell will allow large reverse
currents to ow through it. The additional current term J
B
may be modelled as follows:
J
B
=
V +JR
s
R
p
_
1
V +JR
s
V
br
_

(19)
Here and are positive correction coecients, obtained experimentally. The full model hence
15
becomes:
J = J
L
J
R
J
D
(J
P
+J
B
) (20)
The term J
P
+J
B
= J
SH
is the eective shunt leakage current density. This model is especially
suited for V
br
< V < 0, and its accuracy has been validated, for example, in [3]. For V 0 the
new current J
B
may be neglected without any loss of accuracy.
There are some other recent models that explain this non-linear behaviour, such as the space-
charge-limited current model proposed in [7].
2.3.1 Shunt leakage current and shunt-busting
In a-Si:H p-i-n cells, the physical origin of the shunt conduction paths, along which the shunt
leakage current ows, has been attributed to lateral drift currents arising from dierently sized
electric contacts [9] or to local non-uniformities.
The p-type and n-type layers are very thin ( 10 nm). Therefore, doping inhomogeneities,
surface roughness or metal/contact material diusion can create possible shunt paths. According
to [7], the most likely way is through an aluminium incursion from the AZO contact layer to
the n-type layer. During the deposition of the AZO layer, Al can diuse into the a-Si:H cell to
form an Al lament that destroys part of the n-i junction. This is sketched in Figure 9. The
result is a localized p-i-metal structure along which the shunt leakeage current ows.
Figure 9: Schematic of a typical a-Si:H p-i-n cell layout with a shunt structure due to (alumin-
imum) contact diusion into the a-Si:H layer. The TCO layer forms the front electric contact,
while the AZO layer forms the back contact. The dashed red lines represent the paths of the
shunt leakage current (J
SH
). It ows in parallel to the ideal exponential diode current (J
D
),
whose paths are represented by the solid blue lines. At high forward voltage biases, J
D
J
SH
,
while at small forward voltages or negative voltages J
SH
J
D
.
The metal diusion hypothesis is reinforced by the shunt-busting phenomenon observed in a-
Si:H cells. Shunt-busting involves applying a reverse voltage bias to the cell for a certain period
of time (without exceeding its breakdown voltage), which causes the shunt leakage current to
decrease to a lower value. In the equivalent circuit model, this corresponds to a drastic increase
in the value of R
p
. It is likely that reverse voltage biasing forces the Al laments out of the
a-Si:H layer through oxidation or evaporation, eliminating thus the parasitic shunt path and
improving the cells electric yield.
16
3 Test cell and experimental technique
3.1 Test cell description
The test cell on which the electrical characterization was performed consists of a thin-lm,
double-junction a:Si:H cell. It has a total area of approximately 40 mm 40 mm. It is shown
in Figure 10.
Figure 10: Photograph of the test a-Si:H tandem cell (front view).
A diagram detailing its physical, layered structure is represented in Figure 11. It contains two
a-Si:H subcells of p-i-n structure for improved eciency. The top subcell is designed to absorb
high energy photons, while the bottom subcell mainly absorbs lower energy photons. This means
that the i -layer of the bottom subcell must have a lower band gap. Note that the i -layer of the
top subcell is somewhat thinner than that of the bottom subcell.
Figure 11: Diagram of the test cell portraying its layer arrangement. The approximate layer
thicknesses are indicated. Adapted from [10].
The tandem cell only has two electrical contacts. The TCO layer is a transparent (anti-reective)
coating that acts as the front contact, while the AZO back reector (chemically composed
17
of ZnO:Al) acts as the back contact. The front (high-emmisivity) glass and back glass are
supporting layers that essentially add structural rigidness and protect the cell.
The tandem cell has an an active cell area A (where photo-generation takes place) of dimensions
15 mm 30 mm. Therefore A = 4.5 10
4
m
2
. To avoid confusion, it should be claried that
although the measuring apparatus obviously reads the total current (I), the current density (J)
is used instead in the characterization process. It is simply given by:
J =
I
A
[A/m
2
] (21)
Consequently, this means that the cells series and parallel resistances will be automatically
normalised by the cell area, with units [m
2
].
3.2 Apparatus description
Measuring Station
All measurements were taken at the Instituto de Energa Solar (IES). The I V curves were
obtained through a 4-point measuring station, equipped with a variable resistor and linked to
a computer via the Agilent VEE Pro software for data acquisition and processing. It enabled
to perform automatic I or V sweeps between specied minimum and maximum values, for a
specied number of measuring points.
The test cell was always placed on top of a thermoelectric support (Peltier cooler), designed to
keep the test cell at the desired temperature.
Solar Simulator
The solar simulator essentially consists of a 1000 W Xenon lamp of adjustable height. It provided
the illumination used in the characterization of the single-junction cell described in Section 4.
LED board
For the characterization of the tandem cell described in Section 5, the illumination was provided
by a board containing 36 LEDs (in a 6 6 arrangement). Half of these were focused infrared
(IR) LEDs ( = 830 nm, = 5 mm, 70 mW/sr) while the other half were focused ultraviolet
(UV) LEDs ( = 405 nm, = 3 mm, 10 mW/sr). A photograph displaying the IR and UV
LED arrangement is shown in Figure 12.
Figure 12: Photograph of the LED board used as the IR and UV light source.
The LED board was connected to a circuit board. The LED intensities could be varied by
manually tuning trimming-type potentiometers. The UV LED branches were independendent
of the IR LED branches. This allowed to turn on just the IR LEDs or UV LEDs on their own
(partial operation), or all LEDs simultaneously. Note that the intensities of both UV and IR
LEDs were controlled individually via two separate potentiomenters.
18
4 Electrical characterization of a single-junction cell
The aim of this section is to provide an experimental method that can be used to systematically
determine the unknown elements of the equivalent circuit for a single-junction cell previously
shown in Figure 8, described mathematically by (11). The methods were applied to the test
tandem cell, but in this case it was treated as a single cell in all regards.
The circuit parameters R
p
, R
s
, J
0
, and n may be determined from a J V curve in the dark.
The remaining parameters, ()
e
and V
b
, may be readily evaluated from a set of J V curves
under suciently strong illumination. A variable illumination method (VIM) is employed, which
essentially consists in obtaining a set of J V curves under dierent illumination intensities.
It is important to note that all these parameters are temperature dependent. In this work, the
thermal variation of these parameters is not dealt with. All measurements were taken at 298 K,
which results in a constant thermal voltage, V
Te
= 25.7 mV.
4.1 Measurements in dark conditions
The non-dimensional characteristic equation for the cell operating in the dark, previouosly given
in Equation (17), is recalled:
J

= J

0
_
e
V

/n
1
_
+
R
s
R
p
V

(22)
When V

1, given that R
p
J

0
/R
s
1, the J

0
-dependent term in Equation (22) is negligible
with respect to the last term. Physically, this means that the leakage current is much larger
than the diode current. The characteristic equation simplies to:
J

=
R
s
R
p
V

(23)
Therefore, an experimental (J

) V

plot in the V

1 region will be a straight line of slope


m = R
s
/R
p
. This slope can be directly obtained through a linear regression t in the V

1
regime.
When V

1, the exponential term in (22) is dominant. Physically, this means that the
exponential diode current overwhelms the shunt leakage current. It is found that:
J

= J

0
e
V

/n
ln (J) =
1
n
V

+ ln
_
J

0
_
(24)
Therefore, a plot of ln (J

) V

can be approximated as a straight line of slope m = 1/n and


y-intercept c = ln (J

0
).
To represent the experimental curves in terms of V

and J

, the series resistance R


s
must be
known. R
s
can be found from the exponential region in the J V curve. This region is
described by (24) and the condition V

1 applies. Note that V

can be approximated as
V/V
Te
for voltages up to the start of the exponential region (since JR
s
V ). Therefore at the
exponential region the condition V/V
Te
1 must also apply.
An educated guess for R
s
must rst be taken. It is known that for the correct value of R
s
, the
ln(J

) V

curve (at the V

1 region) will be a straight line. If, on the other hand, the


guessed value is far from the correct one, the ln(J

) V

curve will show a strong curvature.


A linear regression t is then performed for V/V
Te
1 and the correlation coecient between
the linear t and the experimental curve is recorded. Then, R
s
is systematically varied and its
corresponding correlation coecient recorded. For the correct R
s
, the correlation coecient will
be a maximum.
19
It is important to note that the results depend on the minimum value of V/V
Te
that is considered
to be much greater than 1, which marks the lower bound of the linear t. The V/V
Te
lower limit
must be chosen carefully. At its optimal value of R
s
the greatest possible correlation coecient
must be attained and at the same time the largest possible portion of the exponential region of
the dark J V curve should be captured, all while ensuring that V/V
Te
1 is satised.
Experimental results
Figure 13 shows the experimentally obtained J V dark characteristic curve of the cell. It is
clearly linear up to V/V
Te
20. It is then followed by an exponential region that blows up
after V/V
Te
50. It can be seen that after V/V
Te
50 the exponential term in (22) must be
dominant, meaning that this portion of the curve lies in the V

1 region indeed.
Figure 13: J V experimental curve in dark conditions. Left: J V curve in the dark for small
forward voltages. A linear region up to V/V
Te
20 is observed. Right: J V curve in the dark
for large forward voltages. This portion of curve belongs to the exponential-dominated regime.
The correct value for R
s
was found to be 0.0032 m
2
. This result shall be veried shortly.
Using this value, experimental plots of J V

were constructed at the V

1 region (Figure 14)


and at the V

1 region (Figure 15). R


p
, J
0
and n could then be easily obtained. It may be
observed in Figure 15 that for the chosen value of R
s
, the ln (J

) V

curve indeed becomes


a straight line. The red solid line marks the linear regression t performed in the top graph in
order to model Equation (24). The t is again displayed in the bottom graph of Figure 15 to
verify that it lies and matches well the exponential region of the dark J

curve.
20
Figure 14: J

plot with a linear t in the V

1 region in order to get R


p
from its slope.
Figure 15: J

plots and linear t for the V

1 region at the chosen value of R


s
. Lower
limit of t: V/V
Te
52. Top: Semilog plot on which linear t is performed, enabling J
0
and
n to be found. Bottom: Plot in a log-free scale to verify that the exponential part of curve is
exclusively captured by the t.
The V/V
Te
lower limit of the t was set to 52. Figure 16 shows the correlation coecient
between the experimental points and the t for a range of values of R
s
. Clearly, the correlation
is maximised at R
s
= 0.0032 m
2
.
21
Figure 16: Correlation coecient of the linear t performed in the V

1 region, as a function
of R
s
. Lower limit of t: V/V
Te
52.
Lastly, the choice for the V/V
Te
lower limit being set to 52 is justied in Figure 17. It shows
that the maximum correlation is greatest when the linear t is set to start at V/V
Te
50 60.
For V/V
Te
> 65, the correlation coecient falls since the number of useful experimental points
start to run out. One can also observe the strong dependence of R
s
with the V/V
Te
lower limit.
In this case, V/V
Te
= 52 was seen as a proper lower limit according to the criteria previously
stated.
Figure 17: Dependance of the maximum correlation coecient and corresponding R
s
on the
V/V
Te
lower limit. Top: Maximum correlation coecient as a function of the chosen V/V
Te
lower
limit. Bottom: Correct R
s
corresponding to the maximum correlation coecient according to
the chosen V/V
Te
lower limit.
22
4.1.1 Reverse voltage biasing model
Should the extended model with the breakdown current term be required, a J

experimental
plot in the dark for the V

< 0 region may be used to estimate the coecients and presented


in (19). In the absence of illumination, at the reverse voltage bias region (V
br
< V < 0), the
J

0
-dependent terms in (22) are negligible, which suggests that the output current is purely made
up of the eective leakage shunt current: J = J
P
+J
B
. The extended characteristic equation
in terms of V

becomes:
J

=
R
s
R
p
_
1 +
_
1 V

V
Te
V
br
_

_
V

(25)
The graph in Figure 18 shows that the bare model described by (22) will always underestimate
J at suciently large negative voltages since it does not take into account the non-linearity of
the shunt leakage current nor the diode avalanche breakdown at high negative voltages. Typical
breakdown voltages for a-Si:H cells lie between -6 V and -8 V [6]. Hence, for this double-junction
cell in question a rough estimate for its V
br
is 16 < V
br
< 12 V, which is quite large.
The parameter may be determined considering the curve in the 100 < V

< 0 region, where


V

(V
Te
/V
br
) 1 and (25) may be approximated as:
J

=
R
s
R
p
[1 +]V

(26)
Therefore, a (J) V plot in such region may be approximated as a straight line of slope
m = (R
s
/R
p
)[1 + ]. The parameter was then determined by systematically varying it until
a good model-experiment matching was obtained.
Figure 18: Comparison between the experimental and modelled J

curves for reverse voltage


biasing. The extended model is given by (25) with V
br
= 16 V, = 4.2 and = 0.5.
23
4.2 Measurements under illumination
As it will be shown, the remaining unknown parameters, ()
e
and V
b
, are linked to the short-
circuit dimensionless resistance R

SC
of illuminated J V curves, dened as:
R

SC
=
V

SC
(27)
R

SC
may be obtained from the gradient evaluated at the short-circuit operating point (SC) in
an experimental J

plot. It is recalled that when the cell is operating at SC, J = J


SC
and
V = 0. This results in V

SC
= J

SC
= J
SC
R
s
/V
Te
. In order to determine ()
e
and V
b
, the cell
must be exposed to intensities of illumination (irradiances) that induce such current densities
J
SC
that the next two conditions are satised:
J

0
e
V

SC
/n

R
s
R
p
(28a)
J

0
e
V

SC
/n
J

SC
= V

SC
(28b)
An expression for R

SC
may be obtained through direct evaluation of Equation (16) at SC, noting
that the J

0
-dependent terms in the denominator are negligible in magnitude in comparison to
the rest of the terms if (28a) and (28b) are satised. The result is:
R

SC
=
V

SC
=
_
_
J

SC
+
R
s
R
p
V

SC
1
d
2
()
e
V
b[1V

SC
(V
Te
/V
b
)]
d
2
()
e
V
b
_
1 V

SC
(V
Te
/V
b
)

2
V
Te
V
b
+
R
s
R
p
_
_
1
(29)
It is algebraically convenient to introduce the following dimensionless parameter and variable
:
=
d
2
()
e
V
b
(30)
= (V

SC
) = 1 V

SC
(V
Te
/V
b
) (31)
Here , where 0 < < 1, is a measure of the subcells recombination current magnitude relative
to the illumination current J
L
. Typically, decreases with illumination strength. incorporates
the eect of irradiance (through V

SC
) on the recombination current. It should be noted that
is always constant (provided ()
e
remains invariant). In these new terms, the expression for
R

SC
reads:
R

SC
=
__
1 +
R
s
R
p
_
(1 )
( )
+
R
s
R
p
_
1
(32)
and , which contain the remaining unknown circuit parameters, may be grouped together
and solved for in the equation above:
(1 )
( )
= F
exp
=
[R

SC
]
1

R
s
R
p
1 +
R
s
R
p

1
R

SC
(33)
F
exp
implies that the grouped parameters form a function F (dependent on R

SC
) which may be
evaluated directly from experiment. Note that the approximation in (33) is only valid if:
R
s
R
p
[R

SC
]
1
1 (34)
24
which may not always be the case. If condition (34) is not satised, the full expression for F
exp
should be used, rather than just [R
SC
]
1
.
A method to determine ()
e
and V
b
is proposed. More precisely, this method may be employed
to obtain a representative value of ()
e
, mainly valid for irradiances of the same order of
magnitude as the experimental curves used in its nding. This is because ()
e
depends on
the intensity of the illumination and it cannot be assumed to be invariant across illuminations
of dierent orders of magnitude. V
b
must not necessarily be known beforehand.
The procedure involves systematically varying V
b
over a sensible range of values. For each
value of V
b
, the corresponding mean ()
e
is then computed from the experimental data, thus
obtaining a set of ()
e
V
b
pairs. The optimal ()
e
V
b
pair must be the one that gives the
best matching of the model against the experimental J V curves. If V
b
is too high or too low,
the modelled curves will not adjust themselves well to the experimental J V curves especially
near the open-circuit region. In particular V
OC
will be overestimated (or underestimated) if V
b
is too high (or too low).
Assuming a value for V
b
, ()
e
can be computed at a particular intensity of illumination by
substituting (30) into (33):
()
e
=
d
2
V
b
_

2
F
exp
1+F
exp
_ (35)
Note that and of course F
exp
are determined experimentally.
A simplied expression from which ()
e
can be quickly estimated can be attained. Recom-
bination losses must be assumed to be small ( 1) and the irradiance is taken to be weak
enough so that is close to 1 (i.e. V

SC
(V
b
/V
Te
) 1). Furthermore, the condition in (34) is
assumed to hold. In this case Equation (33) simplies to:
(1 ) =
1
R

SC
(36)
After substituting (30) and (31) into (36) and rearranging, the simplied, approximate expres-
sion for ()
e
introduced in [2] is obtained:
()
e, approx
=
d
2
V
b
(1 )R

SC
=
_
d
V
b
_
2
J
SC
R
s
R

SC

=
_
d
V
b
_
2
J
SC
R
SC
(37)
Experimental results
The parameter ()
e
was estimated by employing both the full approach given by (35) and
the simplied approach found and used in the literature, given by (37). This was done in order
to assess the accuracy of the simplied approach with respect to the full approach. The full
approach later proved to yield better and consistent estimates for the ()
e
V
b
pairs.
The cell was exposed to six dierent illuminations of white light produced by the solar simulator
(briey described in Section 3.2) ranging approximately from 0.3 to 1.8 suns. The experimental
J

curves are plotted in Figure 19, from which V

SC
and R

SC
were obtained. The induced
J
SC
lies between 20 and 120 A/m
2
in magnitude. After non-dimensionalisation, J

SC
was seen
to comply with conditions (28a) and (28b). R

SC
also complied with condition (34).
25
Figure 19: Experimental J

curves for dierent illuminations (approx. 0.3 1.8 suns). For


each curve, the short-circuit operating point (SC) is shown, along with the tted tangent with
gradient m (where m = 1/R

SC
).
Typically, the built-in voltage of a simple a-Si:H cell lies between 0.8 and 1 V. For a tandem cell,
the equivalent V
b
will approximately be the sum of each subcells V
b
. The minimum possible
value of V
b
must always exceed V
OC
attained at strong irradiances, in this case V
b
> 1.75 V.
Hence, for this particular tandem cell the range of V
b
considered was from 1.75 V to 2.1 V.
Figure 20 shows the variation of R

SC
and ()
e
with J

SC
(which is fairly proportional to
the irradiance) for V
b
= 1.85 V . The decaying trend in R

SC
with J

SC
is expected, and it
may be inferred from (29). On the other hand, ()
e
can be seen to remain constant at
this narrow range of irradiances considered. This is because the full approach is more precise
as it will be next seen. The dierence in ()
e
calculated by both approaches widens with
irradiance since (0 < < 1) moves away from 1 with increasing V

SC
(or irradiance). For both
approaches, an average value of ()
e
was computed by taking the mean ()
e
over the 6
dierent illuminations at a given V
b
. The average ()
e
vs V
b
is plotted in Figure 21. Note
that ()
e
1/V
b
2
according to the simplied expression given in (37).
Next, the validity of using the average ()
e
V
b
pairs to characterize the cell in this particular
experimental illumination range was evaluated. The validity of the pairs obtained by both
approaches were examined, i.e. by the full approach given in Equation (33) and the simplied
approach in (36).
In order to do this, the following functions were dened:
F
th
=
(1 )
( )
(38a)
F
th, approx
= (1 ) (38b)
F
exp
=
1
R

SC
(38c)
F
th
represents the grouped expression of and used in the full approach while F
th, approx
represents that of the simplied approach.
26
Figure 20: Variation of R

SC
and ()
e
with J

SC
(irradiance) for V
b
= 1.85 V, computed from
the J

curves in Figure 19. Top: Experimentally determined variation of R

SC
with J

SC
.
Bottom: Computed ()
e
vs J

SC
using both approaches.
Figure 21: Variation of the average ()
e
with V
b
obtained by the full and the simplied
approaches.
Using any pair of values for the average ()
e
and V
b
from Figure 21, the parameter and
variable (V

SC
) were computed, the latter spanning a sensible V

SC
range. The theoretical
F
th
J

SC
and F
th,approx
V

SC
curves were constructed and compared against the experimental
F
exp
V

SC
points. They are shown in Figure 22.
It is thus corroborated that both theoretical functions F are essentially invariant regardless of
the ()
e
V
b
pair used. It is clearly observed that F
th
correlates well with the experimental
function F
exp
for the whole range of illuminations considered. As a result, all illuminations
within this range will give a similar ()
e
value by this approach. This was seen on Figure
20. It may be concluded that the average ()
e
value is indeed representative over the entire
experimental illumination range.
However, there is a noticeable worse correlation between F
th, approx
and F
exp
. The average
()
e
value for the simplied approach is observed to be only accurate for the lower half of the
27
illumination range. Note that as the illumination strength is lowered, the dierence bewteen
simplied and full approaches decreases, and F V

SC
becomes a straight line. It is expected
that the validity of using the simplifed approach increases.
Figure 22: F
th
, F
th,approx
as functions of V

SC
, compared against experimental F
th,exp
V

SC
points. The variable thickness in the full model curve accounts for the fact the curve is composed
of several overlapping F
th
V

SC
subcurves, by taking several ()
e
V
b
pairs from Figure 21
over the whole V
b
range.
The nal step involved picking the optimal ()
e
V
b
pair from Figure 21 through direct
modelling of the experimental J V curves. For this particular illumination range, it was found
that the pair comprised by V
b
= 1.85 V and ()
e
= 4.6 10
12
m
2
/V gave the best matching
between the modelled and experimental J V curves.
There is a good chance that even when modelling single-junction cells, the value of V
b
that gives
the best t is not the same as the actual true value of the aSi:H cell. As previously mentioned
in Section 2.2.2, V
OC
will very likely be underestimed by the model when the true value of V
b
is used. In this case, it must be kept in mind that the cell is actually a double-junction tandem
cell. Here, V
b
loses part of its physical meaning since it is in fact the equivalent built-in voltage
of the combination of both subcells, and therefore must be treated as a tuning parameter whose
value should be adjusted to minimise the mismatch of the modelled and experimental J V
curves.
4.3 Model performance
The cell parameter values are summarised in Table 1.
With all elements the equivalent circuit now known, the J V curves modelled by (5) were
constructed. Since a closed-form exact solution of equation (5) or (11) is not available, a Newton-
Raphson iterative method was employed. Note that the photocurrent J
L
is required. It was
computed by directly specifying the desired J
SC
. Rearranging Equation (5) evaluated at SC
gives:
J
L
=
J
SC
_
1 +
R
s
R
p
_
+J
0
_
e
J
SC
R
s
nV
Te
1
_
1
d
2
()
e
(V
b
J
SC
R
s
)
(39)
28
Parameter Value Units
Starter values
V
Te
25.7 10
3
V
d 600 nm
Dark measurements
R
s
3.20 10
3
m
2
R
p
47.6 m
2
J
0
8.6 10
7
A/m
2
n 3.78
Illumination measurements
()
e
4.6 10
12
m
2
/V
V
b
1.85 V
Table 1: Parameter values for the single-cell equivalent circuit model.
The performance of the model with these parameters was assessed for forward and reverse voltage
biasing.
Forward voltage biasing
Figure 23 veries the good agreement between experimental and modelled J V curves with the
chosen parameters. However, V
OC
is slightly overestimated. The V
OC
predicted by the model
was found to be largely dependent on the ()
e
V
b
pair chosen. This particular pair gave good
agreement for a wide range of positive V at the expense of overestimating V
OC
. The fact that
the tandem cell is approximated as a single-junction cell is expected to undermine the model
accuracy. Furthermore, the underlying assumptions of the recombination model are no longer
valid near the OC region as previously stated in Section 2.2.2. Therefore the model is expected
to perform worse in this region.
Figure 23: Experimental and modelled illuminated J V curves for forward voltage biasing.
29
Ideally, the cell will mostly operate at voltages near the maximum power point (MPP). Again,
the model accurately describes the P V behaviour of the cell in this region. This is plotted in
Figure 24.
Figure 24: Experimental and modelled illuminated P V curves for forward voltage biasing.
Reverse voltage biasing
In the V < 0 regime, Figure 25 shows the bare model ts well the experimental curves up to
V = 3 V, quite a large value. For larger negative voltages, the model described by (5) will
always underestimate J since it does not take into account the non-linearity of the shunt leakage
current nor the avalanche diode breakdown at high negative voltages. The same model, but now
extended to include the J
B
term described in Section 2.3, faithfully solves this problem.
Extreme o-design operation is not an issue when dealing with a single cell. However, for
photovoltaic modules composed by strings of many cells in series, it is not unlikely that a
certain cell will be subject to some degree of shading, resulting in current mismatch. For heavy
shading, if each cell was tested individually, the output current of the shaded cell would be
signicantly lower than that of the rest of the unshaded cells. Note that the output current J
of a set of cells connected in series is dictated by the smallest J that corresponds to that of the
shaded cell. The operating point of the rest of the unshaded cells will forcefully move so that
every cell now produces the same J as that of the shaded cell.
As an example, suppose all unshaded cells are initially operating at the MPP. The reduction
in J imposed by the shaded cell will force the unshaded cells to operate at a higher V (hence
at a lower J) for a particular J V curve. As it can be seen from Figures 23 and 24 each
unshaded cell now produces less useful power, less J and generates extra (positive) voltage.
This extra voltage dierence must be cancelled by the shaded cell, assuming the whole module
is operating under a xed external voltage. In this case, the shaded cell operating point will
be at a large negative voltage. There is thus a danger of cell damage due to a hot spot, if the
power dissipation through the cell is high enough, and of avalanche diode breakdown of the cell
30
if the voltage across it is negative enough. The more cells there are in series, the worse it gets
for the shaded cell. This is corroborated in [6], where it is concluded that the power loss due to
shading decreases with increasing number of cells connected in series, but the risk of cell damage
increases.
Figure 25: Experimental and modelled J V curves for reverse voltage biasing at three distinct
illumination levels. The coecients for the extended model are = 4.2 and = 0.5.
4.3.1 Cell response to changes in illumination intensity and spectrum
The validity of the procedure for obtaining ()
e
and V
b
was conrmed by calculating these
parameters from a new set of weaker illuminations of dierent spectrum. The induced J
SC
was approximately one order of magnitude less than before (2 < J
SC
< 10 A/m
2
). This was
achieved by placing lters on top of the cell. The precise spectrum of the light that reached the
cell was not known. In the electrical model used, all parameters are assumed to be independent
of the illumination intensity or spectrum except ()
e
. Even on tandem cells, changing the
illumination spectrum has little eect on the shape of the J V curves provided all subcells
remain excited at a similar level. This means that all photocurrents (J
L
) generated by the
subcells are of similar orders of magnitude. A change in the illumination spectrum may cause
the photocurrent magnitudes of both subcells to shift by dierent amounts, and so will the
recombination currents. The eect on the J V curve will be ecienctly captured by re-
evaluating the value of ()
e
under this particular illumination.
The illumination spectrum under which both subcells must be excited may be known from the
spectral response of the tandem cell. The spectral response (SR) has units [A/W] and it is
dened by:
SR() =
J
SC
()
P
irr
()
=
J
SC
()
E()()
(40)
J
SC
() is the short-circuit current density generated by the cell under monochromatic light of
31
wavelength , P
irr
() is the illumination intensity, E() is the energy of a photon and () is
the photon ux per unit area and time.
The spectral response for the whole cell is plotted in Figure 26. This SR() portrays the
combined response of both subcells. When SR() > 0 , both subcells are being simultaneously
excited by the incident illumination, hence an output current J
SC
is generated. Note that the
spectral response of a tandem cell is always limited by one of its subcells.
Figure 26: Spectral Response of the tandem cell.
For the test tandem cell, the spectral response of the tandem cell hints that the incident illumin-
ation spectrum determined by the lters must have displayed at least one wavelength between
300 and 700 nm since both subcells were excited. The latter was veried experimentally since
J
SC
was observed to vary proportionally to the irradiation.
For V
b
= 1.85 V, the average ()
e
(obtained by the full approach) was 3.46 10
12
m
2
/V. Note
that it is around 25% lower than for the stronger set of illuminations (previously presented in
Section 4.2). This agrees with the physical explanation that at lower illuminations, the p-i and
i -n interface recombination increases due to the increased number of charged dangling bonds
and the fact that charge-assisted capture is much more likely than capture by neutral dangling
bonds [4].
Using this value and the rest of values from Table 1, the J V curves at both forward and
reverse voltage biasing in this illumination regime were modelled. They are plotted in Figures
27 and 28 respectively. The deviation of the experimental curves from the bare model at large
negative voltages is perhaps even more noticeable as the illumination is weakened. Once again,
V
OC
is overestimated by the model for the same reasons.
32
Figure 27: Experimental and modelled illuminated J V curves for the new set of weaker
irradiances with an altered spectral distribution.
Figure 28: Experimental and modelled illuminated J V curves for reverse voltage biasing for
the new set of weaker irradiances with an altered spectral distribution. The extended model
includes the J
B
term dened in Section 2.3.
33
4.4 Sources of error and uncertainties
All the elements in the electrical model are temperature-dependent. For consistent results, it is
therefore very important that all measurements are taken with a strict eye on the temperature,
with a recommended maximum allowance of 1

C oset. As an example, Figure 29 portrays the


strong eect of temperature on the dark J V curve of the cell. The computed parameters from
the dark J V curves at 25

C and 28

C are compared in Table 2:


Parameter T = 25

C T = 28

C Units
R
s
0.0032 0.0024 m
2
R
p
47.6 55.0 m
2
J
0
8.6 10
7
2.3 10
7
A/m
2
n 3.80 3.88
Table 2: Parameters computed from a dark J V curve measured at dierent temperatures.
Figure 29: Measured dark J V curves at dierent temperatures.
It is often dicult to maintain the cell at the desired temperature due to its tendency to heat
up when it is subject to suciently strong illuminations. To avoid this, the cell was placed atop
a Peltier cooler designed to maintain the cells temperature at a specied value, thus minimising
any sources of error coming from temperature-related eects.
Another main issue in calculating the parameters from the dark J V curve is that it relies
on a single curve. Discrepancies between dierent J V curves at the same temperature were
seen to occur. Errors in the measuring apparatus, which was quite temperamental, were not
uncommon, whether they were oscillations in the measurements or severe osets in the readings.
Therefore care must be taken in choosing a dark J V curve that is indeed representative at
the desired temperature by taking several curves. Alternatively, taking the mean of each of the
parameter values computed over several J V dark curves is perhaps a better option.
In order to obtain the desired I V curves, the measuring apparatus performed an automatic
voltage sweep between specied minimum and maximum V values, taking an inputted number
of measuring points in between. In order to obtain the smoothest curves possible, the apparatus
took an inputted number of I-readings (usually set as 3060) at each measuring point, and the
mean I was recorded as the nal value. Even so, small-scale oscillations were often unavoidable.
The reason for this is that the current measuring device is accurate to approximately 1 A. This
translates in J

being accurate to 2 10
4
. Most oscillations observed had an amplitude of this
same order of magnitude. Oscillations 50 times larger than this were nevertheless registered, as
34
clearly seen in Figure 30.
Figure 30: Close-up of the SC region for J

curve (full curve is plotted in Figure 19). SC


denotes the (smoothed) short-circuit operating point. The tangent to gradient line is essentially
a linear regression t performed between V
min
= 0.25 V and V
max
= 0.25 V.
The oscillations add to the uncertainty of accurately determining the gradients from a J

plot, e.g. at the SC point of an illuminated curve in order to evaluate ()


e
, or at the V

1
region of a dark curve in order to nd R
p
. The oscillations meant that the gradients were
forcefully determined through a linear regression t limited by arbitrary minimum and maximum
voltage values in the region of interest. Figure 30 shows the linear t constructed between
V
min
= 0.25 V and V
max
= 0.25 V used to determine R

SC
. R

SC
displays a strong dependence
on the minimum and maximum voltage values chosen as shown in Figure 31. The t is initially
centered around the SC point, until V
min
hits the rst experimental point at V = 0.25 V, after
which only V
max
is increased. This plot shows that the computed value of R

SC
hits a plateau
at V
max
V
min
= 0.3 0.7 V, which can be assumed to be around the true value.
Therefore care must be taken to ensure that the t indeed covers a sensible V range at the
region of interest.
Figure 31: Estimated R

SC
for the J

curve in Figure 30 through a linear regression t


limited by V
max
and V
min
, plotted as a function of the V length of the t.
35
4.5 Conclusions
An experimental method to evaluate the parameters of the electrical model has been presented.
The performance of the electrical model for a single-junction cell has then been assessed. The
model has been seen to accurately represent J V curves for dierent illumination regimes,
provided ()
e
(which has been shown to be spectrum and intensity dependent) is reevaluated
accordingly. Changes in the spectrum and intensity of the illumination will be captured by
()
e
. This is also true for tandem cells, provided all subcells are excited in a similar way.
For the double-junction cell this is not the case when for example, the bottom subcell is being
excited (by pure IR light) while the top subcell may be operating the dark. J
SC
will be very
small, while V
OC
will be signicantly greater than 0 given that the excited subcell is generating
a positive voltage dierence in OC conditions. Modelling this particular J V curve is beyond
the scope of the single-junction cell model.
The next logical step is to introduce a tandem cell equivalent electrical circuit, where each sub-
cell is now represented separately. A procedure to experimentally determine the unknown circuit
elements for each subcell is therefore presented in the next section. Note that the procedure
assumes the subcells cannot be accessed by electrical contacts separately. The potential advant-
ages of the tandem-cell model with respect to the single-cell model are that the former will be
able to properly capture the response of a tandem cell to (extreme) changes in the spectrum,
and be able to model the J V curves of each subcell individually.
36
5 Electrical characterization of a double-junction cell
A multi-junction or tandem cell may be treated as a set of N subcells connected in series.
Each subcell may be modelled independently by the single cell model presented in the previous
sections. In this case the characterization process shall be limited to the case of a double-junction
cell (N = 2).
5.1 Model and general equations
The electrical behaviour of an a-Si:H tandem cell may be modelled using the equivalent circuit
drawn in Figure 32. This equivalent circuit specically describes a double-junction cell. The
equivalent circuit is composed of two subcircuits connected in series. Each subcircuit has a
distinct set of elements and represents a single subcell. Note that each subcircuit is identical to
the single-junction cell equivalent circuit.
Figure 32: Equivalent circuit for a double-junction cell. It is treated as two subcells connected
in series. Each subcell is represented by a subcircuit comprised by a distinct set of elements.
Making use of the same notation for the various currents as for the single-junction cell model,
this circuit may be described mathematically by again applying Kirchos rst law. For the
general case of N subcells:
J = J
L,i
J
R,i
J
D,i
J
P,i
i = 1, 2 . . . N (41a)
V =
N

i=1
V
i
(41b)
where subscript i refers to the i
th
subcell.
Expanding (41a) as before gives the characteristic J V
i
equation for each subcell:
J = J
L,i
J
L,i
d
2
i
()
e,i
[V
b,i
(V
i
+JR
s,i
)]
J
0,i
_
e
V
i
+JR
s,i
n
i
V
Te
1
_

V
i
+JR
s,i
R
p,i
(42)
It is again more convenient to work in terms of the dimensionless quantities V

i
and J

, dened
37
as:
V

i
=
V +JR
s,i
V
Te
(43)
J

=
JR
s
V
Te
(44)
where R
s
denotes the combined series resistance of the whole cell:
R
s
=
N

i=1
R
s,i
(45)
Likewise, the subcell diode saturation current J
0,i
and subcell photocurrent J
L,i
may be non-
dimensionalised as follows:
J

0,i
=
J
0,i
R
s
V
Te
(46)
J

L,i
=
J
L,i
R
s
V
Te
(47)
It is algebraically convenient to introduce the non-dimensional parameter
i
and variable
i
:

i
=
d
2
i
()
e,i
V
b,i
(48)

i
= 1 V

i
(V
Te
/V
b,i
) (49)
Here
i
, where 0 <
i
< 1, is a measure of the subcells recombination current magnitude
relative to the illumination current J
L,i
. Typically,
i
decreases with illumination strength.
i
models the eect of the external voltage and irradiance on the the recombination current. It
decreases (increases) with forward (reverse) voltage biasing.
Using these terms, the non-dimensionalised characteristic equations of the tandem cell may be
written as:
J

= J

L,i
_
1

i

i
_
J

0,i
_
e
V

i
/n
i
1
_

R
s
R
p,i
V

i
i = 1, 2 . . . N (50a)
V

=
N

i=1
V

i
(50b)
V

, not unlike J

, depends on the combined resistance R


s
. From the denitions of V

i
in (43),
the combined resistance R
s
in (45) and Equation (50b), it follows that:
V

=
V +JR
s
V
Te
(51)
The dimensionless resistance R

of the tandem cell is dened by:


R

=
V

=
N

i=1
V

i
J

=
N

i=1
R

i
(52)
where R

i
is the non-dimensional resistance of the i
th
subcell, given by:
R

i
=
V

i
J

=
__
J

+J

0,i
_
e
V

i
/n
i
1
_
+
R
s
R
p,i
V

i
_

i
(V
Te
/V
b,i
)

i
(
i

i
)
+
J

0,i
n
i
e
V

i
/n
i
+
R
s
R
p,i
_
1
(53)
38
5.2 Experimental procedure
It has been shown in Section 3.1 that the tandem cell is composed to two subcells. The top
subcell (i = 1) is designed to absorb mostly high energy photons (short wavelength spectrum),
while the bottom subcell (i = 2) is optimised for the absorption of lower energy photons (long
wavelength spectrum).
As for the single-junction cell case, the aim of the following sections is to determine all unknown
circuit elements. In order to achieve this, the characterization process will once again rely on a
Variable Illumination Method. The experimental data available is strictly limited to just a set of
experimental J V curves obtained for dierent illumination conditions. From each curve, four
key quantities may be graphically obtained. These four quantities, vital in the characterization
process, are: J

SC
, R

SC
, V

OC
and R

OC
.
A J V curve under no illumination was initially recorded. It is essential to determine R
s
and
thus enabling the whole non-dimensionalisation process. The tandem cell was then exposed to
IR light at 830 nm (in otherwise dark conditions) in order to induce a much greater photocurrent
in subcell 2 than in subcell 1. The intensity of the IR light was varied and several J V curves
at dierent irradiances were recorded. Likewise, it was next exposed to UV light at 405 nm of
varying intensity in order to induce the opposite eect. Lastly, the cell was exposed to IR light
at a xed intensity in conjunction with variable intensity UV light.
The light source in all cases was the LED board already described in Section 3.2.
5.3 Measurements in dark conditions
The combined series resistance R
s
= R
s,1
+R
s,2
may be determined by following the exact same
procedure previously presented for the single-junction cell analysis. It is impossible to determine
R
s,1
or R
s,2
independently. This is not a problem since both V

and J

depend on R
s
and not
just on R
s,i
. Furthermore, it will be seen that the plots under dark conditions used for the
identication of parameters are identical to those used for the single-junction cell case.
The characteristic equations for a double-junction cell operating under no illumination are given
by:
+J

= J

0,i
_
e
V

i
/n
i
1
_

R
s
R
p,i
V

i
i = 1, 2 (54a)
V

= V

1
+V

2
(54b)
When V

i
1, given that R
p,i
J

0,i
/R
s
1, the J

0,i
-dependent term in (54a) is negligible with
respect to the last term, leaving:

R
p,i
J

R
s
= V

i
i = 1, 2 (55)
Substituting (55) into (54b) leads to:
J

=
R
s
R
p,1
+R
p,2
V

(56)
Therefore, an experimental (J

) V

plot in the V

1 region will be a straight line of slope


m = R
s
/(R
p,1
+R
p,2
). This slope can be directly obtained through a linear regression t in the
V

1 regime.
When V

i
1, the exponential term in (54a) is dominant, leaving:
J

= J

0,i
e
V

i
/n
i
i = 1, 2 (57)
39
Solving (57) for V

i
and substituting it into (54b) leads to:
ln (J

) =
1
n
1
+n
2
V

+
1
n
1
+n
2
ln K
d
(58)
where:
K
d
=
_
J

0,1
_
n
1
_
J

0,2
_
n
2
(59)
A plot of ln (J

) V

in the V

1 region can be approximated as a straight line of slope


m = [1/(n
1
+n
2
)] and y-intercept c = [1/(n
1
+n
2
)] ln (K
d
).
Experimental results
The dark J V curve used for the current tandem cell characaterization process was taken a
few months later than that used in the single-junction cell analysis. This time, value for R
s
was found to be 0.00254 m
2
, which is, as expected, of the same order of magnitude to that
obtained for the single-junction cell analysis. Values for n
1
+ n
2
and K
d
could then be easily
obtained from a linear t in the V

1 region as shown in Figure 33.


Figure 33: J

plots and linear t for the V

1 region at the chosen value of R


s
. Lower
limit of t, V/V
Te
50. Top: Semilog plot on which the linear t is performed, enabling n
1
+n
2
and K
d
to be found. Bottom: Plot in a log-free scale to verify that the exponential part of curve
is exclusively captured by the t.
On the other hand, R
p,1
+R
p,2
, which are the equivalent of R
p
in the single-junction cell analysis,
came out to be much larger. There is a recorded evolution for R
p
in time. Its value was observed
to increase from 5 m
2
[12] to 50 (by the time of the single-junction cell analysis), then
500 and nally 1000 (tandem cell analysis). Within this period, several reverse voltage bias
measurements were taken. Therefore, this drastic increase in R
p
, (analogous to a big reduction
of the shunt leakage current) is attributed to shunt-busting, described back in Section 2.3.1.
40
Figure 34: J

plot with a linear t in the V

i
1 region in order to get R
p,1
+R
p,2
from its
slope.
5.4 Theoretical derivations for single subcell excitation
Before embarking upon the characterization of the cell using measurements under bias IR and UV
illumination, it is worth detailing the expected cell behaviour for the case in which the incident
illumination only induces a photocurrent in a single subcell, and the remaining unexcited subcell
is forced to operate in dark conditions.
The purpose of this section is to derive useful relations and properties that may be used as
indicators for the case of single subcell excitation. As it will seen later in Section 5.5, none
of these properties were satised experimentally. This led to the conclusion that single subcell
excitation was not attained, hence an alternative characterization route was then developed.
The case where only subcell 2 is excited while subcell 1 remains in the dark (J
L,1
= 0, J
L,2
> 0)
will be considered in the following derivations.
Typical operating curves for both subcells and the resultant curve for the whole cell are sketched
in Figure 35. It clearly portrays the case where subcell 2 is excited, while subcell 1 operates in
the dark. Note that when the cell is operating at short-circuit conditions (SC), subcell 1 and
subcell 2 operate at 1, SC and 2, SC respectively.
The rst baseline assumption in this case is that at the open-voltage (OC) operating condition,
V

1,OC
= 0 always, hence V

OC
= V

2,OC
. Secondly, it is evident from Figure 35 that the proximity
of points 2, SC and 2, OC = OC implies that V

2,SC
V

2,OC
. The smaller J

SC
is, the closer to
each other these two points will be.
A third important baseline assumption is that the shunt leakage current must be Ohmic not only
when the subcell operates at forward voltage biases but also at reverse voltage biases (at least
in the range V

i,OC
< V

i
< 0). This implies that in dark conditions, the subcell will display a
linear J

characteristic curve within this negative voltage range.


41
Figure 35: J

sketch of the expected operating curves of each subcell when exposed to


illumination that excites only subcell 2. Subcell 1 is constrained to operate in the dark.
5.4.1 SC current OC voltage relation
Bearing the rst baseline assumption in mind, the characteristic equation for subcell 2 at OC
reads:
0 = J

L,2
_
1

2

OC
_
J

0,2
_
e
V

OC
/n
2
1
_

R
s
R
p,2
V

OC
(60)
where
OC
= 1 V

OC
(V
Te
/V
b,2
).
At the SC operating condition, the charactersitic equations for subcells 1 and 2 become:
J

SC
= J

0,1

R
s
R
p,1
V

1,SC
(61a)
J

SC
= J

L,2
_
1

2

2,SC
_
J

0,2
e
V

2,SC
/n
2

R
s
R
p,2
V

2,SC
(61b)
Note that at SC, subcell 1 is operating in reverse voltage biasing (V

1,SC
< 0). Here (61a) treats
the J

1
characteristic curve of the subcell in this quadrant (V

1
< 0, J

> 0) to be linear
(corresponding to an Ohmic shunt leakage current). It is modelled as a straight line of slope
m
rev
= R
s
/R
p,1
, and y-intercept c = J

0,1
. Furthermore, the magnitude of c is very small, and
may only be taken into consideration if J

SC
is of the same order of magnitude as J

0,i
. Otherwise
c may be treated to be 0 if J

SC
J

0,i
.
The second baseline assumption implies that
2,SC

OC
. It is thus valid to assume that:
1

2

2,SC
= 1

2

OC
(62)
After premultilying (61a) and (61b) by R
p,i
/R
s
, adding them together and then substituting in
(60) to eliminate J

L,2
, one gets:
(R
p,1
+R
p,2
)J

SC
R
s
=
R
p,2
J

0,2
R
s
_
e
V

OC
/n
2
e
V

2,SC
/n
2
_
+V

OC
(V

1,SC
+V

2,SC
) +
R
p,1
J

0,1
R
s
(63)
42
The term (V

1,SC
+V

2,SC
) = V

SC
= J

SC
may be neglected since it is much smaller than the rest
of the terms if R
p,i
/R
s
1. Using Equation (61a), an expression for V

2,SC
may be formulated:
V

2,SC
= V

SC
V

1,SC
= J

SC
+
R
p,1
(J

SC
J

0,1
)
R
s

R
p,1
(J

SC
J

0,1
)
R
s
(64)
Introducing this expression for V

2,SC
into (63) nally gives:
(R
p,1
+R
p,2
)J

SC
R
s
=
R
p,2
J

0,2
R
s
_
e
V

OC
/n
2
exp
_
R
p,1
(J

SC
J

0,1
)
n
2
R
s
__
+V

OC
+
R
p,1
J

0,1
R
s
(65)
This is a non-linear equation relating J

SC
and V

OC
. It may be further simplied depending on
the recorded magnitude of J
SC
with respect to J
0,i
, which is typically of the order 10
6
A/m
2
.
For suciently small irradiances, J

SC
may be comparable in magnitude to J

0,i
. It may also be
the case that both are always comparable in magnitude, regardless of the irradiance used. This
means that the slope m
rev
is small, indicating that the cell is well-built and has small parasitic
resistances. This implies large values for R
p,i
and a small R
s
. In any case, when J

SC
J

0,i
is
satised, the 2, SC and OC points will practically merge and the approximation V

2,SC
= V

OC
is valid. Equation (65) then simplies to a linear form:
(R
p,1
+R
p,2
)J

SC
R
s
= V

OC
+
R
p,1
J

0,1
R
s
(66)
It is thus expected that the magnitude for V

OC
is such that V

OC
V
1,SC
R
p,1
J

0,1
/R
s
is
complied. This means that if J

0,i
R
s
/R
p,i
then V

OC
1. This implies very small illumination
intensities. It should be borne in mind that it is often dicult to obtain suciently accurate
J V measurements in the extremely small 0 < V < V
OC
region under very small irradiances.
If on the other hand, J

SC
J

0,1
, the approximation V

2,SC
= V

OC
is no longer valid, but all the
J

0,1
-dependent terms in (65) may then be neglected. It keeps a nonlinear form, and may only
be solved through iteration:
(R
p,1
+R
p,2
)J

SC
R
s
=
R
p,2
J

0,2
R
s
_
e
V

OC
/n
2
exp
_
R
p,1
J

SC
n
2
R
s
__
+V

OC
(67)
5.4.2 SC and OC resistance property
The case where subcell 2 is excited while subcell 1 is in the dark is still being considered. This
property states that both R

SC
and R

OC
should tend to R
p,1
/R
s
as the irradiance exciting subcell
2 is increased.
Evaluating rst R

1,SC
, it is noted that V

1,SC
< 0 hence the J

0,1
-dependent terms in the denition
for R

1,SC
in (53) may be neglected provided J

0,1
R
s
/R
p,1
. This results in:
R

1,SC
=
R
p,1
R
s
(68)
R

2,SC
shall be evaluated next for the case where the cell is exposed to strong enough irradiances
that induce V

OC
V

2,SC
1. The recombination-free exponential term in (53) is dominant.
Hence:
R

2,SC

n
2
J

0,2
e
V

2,SC
/n
2
(69)
43
With increasing irradiance and V

2,SC
, R

2,SC
becomes more negligible in magnitude with respect
to R

1,SC
and as a result R

SC
R
p,1
/R
s
.
It must be emphasized that this applies only if the J

1
curve in the dark (of subcell 1) at
reverse voltage biases (for V

OC
< V

1,SC
< 0) is indeed described by a straight line of slope
m
rev
= 1/R

1,SC
= R
s
/R
p,1
. This occurs exclusively for Ohmic shunt leakage currents. This
may not always be the case, since the shunt leakage current of a particular subcell may possess
a strong voltage power law dependence. This results in an exponential-shaped working curve
like the one shown in Figure 18 (where the non-linear behaviour is noticeable at V

< 100). It
may be the case that the exponential behaviour is instead observed at smaller negative voltages
where V
1,SC
may lie. R

SC
will thus still tend to R

1,SC
, but it will no longer be equal to R
p,1
/R
s
since the Ohmic model equations are not valid.
For irradiances strong enough so that V

OC
1, the OC resistance may be written as:
R

OC
= R

1,OC
+R

2,OC
=
R
p,1
R
s
+
n
2
J

0,2
e
V

OC
/n
2

R
p,1
R
s
= R

SC
(70)
It is therefore concluded that as the irradiance increases, the magnitude of R

2,OC
with respect
to R

1,OC
eventually becomes negligible, so R

OC
should also tend to R
p,1
/R
s
.
5.5 Measurements under illumination
A selection of J

curves for isolated IR illumination and isolated UV illumination are plotted


in Figure 36. These curves provide enough evidence which supports the fact that under each
of these two illumination conditions, both subcells are simultaneously excited i.e. J
L,1
and J
L,2
are non-zero, and that one subcell is signicantly more excited than the other.
Figure 36: A selection of J

curves for isolated IR and isolated UV illuminations of dierent


intensities.
First of all, at the maximum irradiance for either pure UV or IR illumination, it is seen that
J

SC
0.02. This value is exceedingly high if say subcell 1 is taken to be operating in the dark,
44
while only subcell 2 is producing a photocurrent. A J

SC
so large cannot be attained for this
tandem cell, provided:
a) R
p,1
and R
p,2
are of the same order of magnitude, and
b) subcell 1 displays a linear J

1
characteristic for reverse voltage biasing in the region
comprised by V

OC
< V

1
< 0 (i.e. the shunt leakage current is Ohmic).
Experimental dark plots veried that the tandem cell displayed a linear J

characteristic
in the 100 < V

< 0 region, so it was assumed that these two conditions were satised.
Under these conditions, an estimate of the maximum expected value for J

SC
can be obtained.
As it has been shown in Section 5.4.1, when subcell 1 is operating under reverse voltage biasing
in dark conditions, its J

1
characteristic equation may be approximated as a straight line
through the origin, with slope m
rev
= R
s
/R
p,1
. The resultant J

SC
of the tandem cell is thus
determined by J

SC
= m
rev
V

1,SC
, where V

1,SC
is the value of V

1
when the tandem cell operates
in SC conditions, and it is negative.
It follows that V

1,SC
V

2,SC
V

2,OC
= V

OC
. At the maximum irradiance curves in Figure
36, V

OC
50. Giving R
p,1
a lower-bound value of 100 m
2
, the maximum J

SC
that may be
attained under these circumstances is estimated to be:
J

SC,max
= m
rev
V

i,SC

R
s
R
p,i
V

OC

0.0025
100
50 1.25 10
3
(71)
This upper-bound value for J

SC
is well under one order of magnitude less than the experimental
value at the strongest irradiances. In theory, as the irradiance is increased, J

SC
should tend to
this upper bound value. This was clearly not observed. Note that no restriction has yet been
bestowed upon the magnitude for the built-in voltage of subcell 2, nor on V

2,OC
. According
to the manufacturer, the built-in voltages of both subcells may be taken to be approximately
equal: V
b,1
V
b,2
. This is supported by the fact that the doping and thickness of the p and n
type layers are essentialy the same for both subcells.
It is therefore expected that V
b,i
0.9 1.0 V. For simplicity, V
b,i
shall be set to 0.925 V, which
is half of the value for V
b
found in the single-junction cell analysis. This sets the following upper
bound value for V

i,OC
:
V

i,OC,max
= V
b,i
/V
Te
35 (72)
The values for V

OC
clearly exceed this value at both IR and UV illumination conditions. Again,
this suggests that both subcells must be excited, implying that V

1,OC
V

2,OC
0. Another
argument to support this claim is that the R

SC
and R

OC
property described in Section 5.4.2
was not met. This property states that as the irradiance on subcell 2 is increased, (while subcell
1 remains in the dark), R

SC
and R

OC
should decay and tend to R
p,1
/R
s
. Not only they did
not tend to this value, but also for all irradiances the values for R

OC
were well below this limit.
Once again, this indicates that both subcells are photo-excited.
At this point, it is concluded that for IR and UV illumination, both subcells are excited. Fur-
thermore, it is evident that the IR and UV illumination excite the subcells in dierent ways,
given the noticeably dierent shapes of the J

between the two illumination regimes.


Figure 37 shows the J

curves under IR illumination only, UV only and the superposition


of both illuminations (IR+UV). All three curves correspond to the maximum irradiances that
could be produced by the LED board.
45
Figure 37: J

curves for IR only, UV only and IR+UV at maximum power.


Under IR+UV illumination conditions, the two subcells must be excited. V

OC
remains practic-
ally at the same value as for the isolated IR and UV illumination. This again proves that the
subcells must be excited by the isolated IR and UV illuminations hence producing a V

i,OC
1
and of the same order of magnitude.
On the other hand, the superposition of IR+UV light causes J

SC
to greatly increase from 0.02
to 0.07. This increase should be justied by one of the three following hypothetical scenarios
regarding the cell behaviour under IR and UV light exposure:
1) UV excites both subcells at a similar level (J
L,1[UV]
J
L,2[UV]
), while IR excites subcell 2
considerably more than subcell 1 (J
L,2[IR]
> J
L,1[IR]
).
2) IR excites both subcells at a similar level (J
L,1[IR]
J
L,2[IR]
), while UV excites subcell 1
considerably more than subcell 2 (J
L,1[UV]
> J
L,2[UV]
).
3) UV light excites subcell 1 considerably more than subcell 2 (J
L,1[UV]
> J
L,2[UV]
), while
IR light has the opposite eect: it excites subcell 2 considerably more than subcell 1
(J
L,2[IR]
> J
L,1[IR]
).
The J
SC
of a tandem cell is often closest to that of the limiting subcell. Its value is in fact
dictated by the illumination current of the limiting subcell. For scenario 1), subcell 1 is always
the limiting subcell regardless of the irradiance conditions. The corresponding J

SC
for IR+UV
illumination may be roughly estimated:
J

SC,[IR+UV]
J

L,1[IR+UV]
J

L,1[IR]
+J

L,1[UV]
J

SC,[IR]
+J

SC,[UV]
= 0.02 + 0.02 = 0.04 (73)
The same result applies for scenario 2) in which subcell 2 is always the limiting subcell. This
value for J

SC
is well below 0.07, meaning scenarios 1) and 2) are implausible. Therefore, the
only possible scenario that justies this is scenario 3).
5.5.1 Baseline assumptions at SC and OC operating conditions
It has been concluded that when exposing the cell purely to this particular IR or UV light (in
otherwise dark conditions), one subcell is considerably more excited than the other. The limiting
subcell, however, may not be treated to be operating in the dark. Useful relations may still be
formulated at the SC and OC points, assuming there is always one subcell producing a much
higher photocurrent.
Typical operating curves for both subcells and the resultant curve for the whole cell are sketched
in Figure 38 for the exposure to IR case. It clearly portrays the case where subcell 2 is much
46
more excited than subcell 1. Note that when the cell is operating at SC, subcell 1 and subcell
2 operate at 1, SC and 2, SC respectively.
Figure 38: J

sketch of the expected operating curves of each subcell when exposed to IR


illumination. Subcell 2 is visibly more excited than subcell 1.
At the SC operating condition, V = V
1,SC
+ V
2,SC
= 0 and J = J
SC
. In V

and J

, this
translates to:
V

SC
= V

1,SC
+V

2,SC
=
V
1,SC
+J
SC
R
s,1
V
Te
+
V
2,SC
+J
SC
R
s,2
V
Te
=
J
SC
R
s
V
Te
= J

SC
(74)
The experimental range of J
SC
does not exceed 1 A/m
2
, while V
i,SC
V
OC
lies within 0.1-1
V. This means that |V
i,SC
| J
SC
R
s,i
and consequently |V

i,SC
| V

SC
. This leads to the rst
relation:
V

2,SC
= V

SC
V

1,SC

= V

1,SC
(75)
It may be noted from the J

2
curve (corresponding to subcell 2) in Figure 38 that V

2
does
not vary much from the 2, SC to the 2, OC point due to the steep gradient. This provides a
second useful relation:
V

2,SC
V

2,OC
1 (76)
At OC, V

OC
= V

1,OC
+ V

2,OC
. As previously concluded, V

1,OC
and V

2,OC
must be of the same
order of magnitude, and much greater than unity. This constitutes the third relation:
V

1,OC
V

2,OC
1 (77)
At this point, it is useful to dene the experimental range of values in which V

1,SC
and V

2,SC
are conned for the set of IR (or UV) irradiances considered. From relation (77) above, it
is not unreasonable to approximate V

1,OC
or V

2,OC
as V

OC
/2. Consequently, V

1,SC
may be
approximated as:
V

1,SC
V

2,SC
V

2,OC

V

OC
2
(78)
For both IR and UV illuminations, Figure 36 reveals V

OC
to lie in the range 40 < V

OC
< 50.
Therefore, for the IR case, values for V

1,SC
and V

2,SC
must be roughly around 20 25. The
same range restriction applies to V

2,SC
and V

1,SC
under UV illumination.
47
This implies that strong enough IR irradiances able to induce J

SC
> 5 10
4
allow for a fourth
relation to be formulated:
J

SC

R
s
R
p,1
|V

1,SC
| (79)
Therefore, for these irradiances V

1,SC
(R
s
/R
p,1
) may be neglected when compared to J

SC
.
The gradient at 2, SC is considerably steeper than at 1, SC. The non-dimensional resistance
of the cell at SC may thus be approximated to be that of the least excited subcell. This may
be proven mathematically through an order of magnitude check for R

1,SC
and R

2,SC
through
careful consideration of Equation (53). Each resistance is of the same order of magnitude as
that of its respective dominant term, namely:
1
R

1,SC
J

SC

1
V
Te
10
2
10
2
10
2
= 10
6
(80a)
1
R

2,SC
J

0,2
e
V

2,SC
/n
2
10
7
10
4
= 10
3
(80b)
Note that relation (79) has been used.
i
, V
b
and n
i
are all O(1), and are left out from the
expressions above. This does not apply inside the exponential, where the estimate n
2
= 2 and
lower-bound value V

2,SC
= 20 are used. It is concluded that R

1,SC
R

2,SC
and so the fth
useful relation is:
R

SC
= R

1,SC
+R

2,SC
R

1,SC
(81)
Lastly, the registered J
SC
were at least of order 10
3
A/m
2
. The expected magnitude of J
0
, i is
of order 10
6
A/m
2
. This leads to the sixth relation:
J

SC
J

0,i
(82)
This allows to neglect J

0,i
when compared to J

SC
, provided the exponential term is not large.
When the cell is exposed to UV irradiation, the subcells exchange roles since subcell 1 is now
more excited than subcell 2. The same relations (75)-(81) apply, logically replacing subscript 1
by 2 and subscript 2 by 1.
5.5.2 SC resistance-current relation
A R

SC
J

SC
relation may be used to obtain R
p,i
and ()
e,i
. For the following derivation, the
IR illumination case shall be considered, where subcell 2 is much more excited than subcell 1.
R

SC
may be approximated as R

1,SC
as previously stated in (81). Given that V

1,SC
is negative,
all the J

0,1
terms in the denition for R

1,SC
in (53) can be neglected due to condition (82). The
expression for R

SC
reads:
R

SC
=
__
J

SC
+
R
s
R
p,1
V

1,SC
_

1
V
Te
/V
b,1

1,SC
(
1,SC

1
)
+
R
s
R
p,1
_
1
(83)
Considering only the IR irradiances which satisfy relation (79), V

1,SC
may be neglected in
Equation (83). Rearranging (83) then gives:
1
R

SC,[IR]
=

1
V
Te
/V
b,1

1,SC
(
1,SC

1
)
J

SC,[IR]
+
R
s
R
p,1
(84)
48
A 1/R

SC,[IR]
J

SC,[IR]
experimental plot was observed to give a straight line. This suggests
that
1,SC
(and
2,SC
) will not vary much in the narrow range of irradiances considered and
may be approximated to be constant. It has been shown that |V

1,SC
| is roughly conned within
20 < |V

1,SC
| < 25, which corresponds to variations in
1,SC
of less than 10%.
Therefore, taking the upper-bound value V

1,SC
= 25, and V
b,i
as 0.925 V, enables for
1,SC
to
be evaluated:

1,SC
= 1 V

1,SC
V
Te
V
b,1
1.7 (85)
Consequently, a 1/R

SC,[IR]
J

SC,[IR]
plot may be modelled as a straight line of slope m and
y-intercept c. Parameters R
p,1
and ()
e,1
are directly obtained through:
R
p,1
=
R
s
c
(86)
()
e,1
=
d
2
1
V
b,1

1
=
d
2
1
V
b,1
_

2
1,SC

1,SC
+
V
Te
mV
b,1
_
1
(87)
Similarly, under UV illumination, an analogous R

SC
J

SC
relation is obtained, essentially
interchanging subscripts 1 and 2 in Equation (84). It is written below for completeness:
1
R

SC,[UV]
=

2
V
Te
/V
b,2

2,SC
(
2,SC

2
)
J

SC,[UV]
+
R
s
R
p,2
(88)
where again
2,SC
may be evaluated as 1.7. R
p,2
and ()
e,2
are obtained from m and c in
analogous expressions.
Experimental results
Linear regression ts in order to extract m and c were performed on the 1/R

SC
J

SC
plots for
IR and UV illuminations (at suciently high J
SC
). These are shown in Figure 39.
Figure 39: 1/R

SC
J

SC
plots for IR and UV illumination. A linear regression t is performed
for each case. Note that the y-axis is is represented in logarithmic scale for a clearer view of the
y-intercept.
49
The computed values of R
p,1
and R
p,2
add up to match well the value obtained for R
p,1
+R
p,2
from the J

curve in the dark, which was evaluated to be 1170 m


2
.
1
is an order of
magnitude higher than
2
. This means that subcell 1 is much more degraded (i.e. it has heavier
recombination losses due to a higher dangling bond defect density) than subcell 2. This is
expected if the J

curves back in Figure 36 are compared. The shape of the IR illuminated


curves is dicated by that of subcell 1 (the limiting cell). Taking any of the IR curves, its J

is
observed to decay much more rapidly under positive V

than for the UV case, a clear symptom


of enhanced recombination.
5.5.3 The V-R-J relation
The V

OC
R

OC
J

SC
relation is a useful equation relating these three quantities (which may be
easily determined experimentally) to other circuit elements essential for the cell characterization.
The IR illumination case shall be considered again for its derivation. The equations for subcells
1 and 2 at short-circuit conditions (SC) read:
R
p,1
J

SC
R
s
=
R
p,1
J

L,1
R
s
_
1

1

1,SC
_
V

1,SC
(89a)
R
p,2
J

SC
R
s
=
R
p,2
J

L,2
R
s
_
1

2

2,SC
_

R
p,2
J

0,2
R
s
e
V

2,SC
/n
2
V

2,SC
(89b)
These two equations may be added up to give:
(R
p,1
+R
p,2
)J

SC
R
s
=
R
p,1
J

L,1
R
s
_
1

1

1,SC
_
+
R
p,2
J

L,2
R
s
_
1

2

2,SC
_

R
p,2
J

0,2
R
s
e
V

2,SC
/n
2
(90)
Note that the term V

1,SC
+ V

2,SC
= V

SC
= J

SC
is negligible when compared against the left
hand side of Equation (90) since R
p,i
/R
s
1. Consequently, it has been ommited. Focusing
next on the open-circuit (OC) operating point, the equations for subcells 1 and 2 at such point
are:
0 =
R
p,1
J

L,1
R
s
_
1

1

1,OC
_

R
p,1
J

0,1
R
s
e
V

1,OC
/n
1
V

1,OC
(91a)
0 =
R
p,2
J

L,2
R
s
_
1

2

2,OC
_

R
p,2
J

0,2
R
s
e
V

2,OC
/n
2
V

2,OC
(91b)
Small recombination currents may be assumed, which means that the eect of the external
voltage on the recombination current loss may be neglected:
1

i

i,SC
1

i

i,OC
(92)
This enables the removal of the unknown photocurrents J
L,i
in (90) through direct substitution
by employing equations (91a) and (91b), leaving
(R
p,1
+R
p,2
)J

SC
R
s
= V

OC
+
R
p,1
J

0,1
R
s
e
V

1,OC
/n
1
+
R
p,2
J

0,2
R
s
_
e
V

2,OC
/n
2
e
V

2,SC
/n
2
_
(93)
Next, an expression for R

OC
is required. Evaluating Equation (53) at OC, condition (77) implies
that the recombination-free exponential term is dominant. The expression for R

OC
simplies
to:
R

OC
= R

1,OC
+R

2,OC
=
n
1
J

0,1
e
V

1,OC
/n
1
+
n
2
J

0,2
e
V

2,OC
/n
2
(94)
50
To a good approximation, point 2, SC may be connected to point 2, OC in the V

plane by
a straight line of slope given by the gradient at 2, OC. This leads to the following expression
relating V

2,SC
and V

2,OC
:
V

2,SC
=
V

2
J

2,OC
J

SC
+V

2,OC
= R

2,OC
J

SC
+V

2,OC
= J

SC
n
2
J

0,2
e
V

2,OC
/n
2
+V

2,OC
(95)
Substituting this into (93) gives:
(R
p,1
+R
p,2
)J

SC
R
s
= V

OC
+
R
p,1
J

0,1
R
s
e
V

1,OC
/n
1
+
R
p,2
J

0,2
R
s
e
V

2,OC
/n
2
_
1 exp
_

SC
J

0,2
e
V

2,OC
/n
2
__
(96)
The double exponential may be simplied through a Taylor series expansion as the argument of
the outer exponential is small compared to unity:
(R
p,1
+R
p,2
)J

SC
R
s
= V

OC
+
R
p,1
J

0,1
R
s
e
V

1,OC
/n
1
+
R
p,2
J

0,2
R
s
e
V

2,OC
/n
2
_
J

SC
J

0,2
e
V

2,OC
/n
2
_
(97)
which leads to:
R
p,1
J

SC
R
s
= V

OC
+
R
p,1
J

0,1
R
s
e
V

1,OC
/n
1
(98)
Solving next for V

1,OC
gives:
V

1,OC
= n
1
ln
_
R
p,1
R
s
J

SC
V

OC
_
n
1
ln
_
R
p,1
J

0,1
R
s
_
= n
1
ln (
1
) n
1
ln
_
R
p,1
J

0,1
R
s
_
(99)
where:

i
=
R
p,i
R
s
J

SC
V

OC
(100)
Substituting equation (98) into (94) enables the elimination of V

1,OC
, therefore allowing to
rewrite R

OC
just in terms of V

2,OC
:
R

OC
=
R
p,1
n
1
R
s

1
1
+
n
2
J
0,2
e
V

2,OC
/n
2
(101)
Solving for V

2,OC
in this last expression gives
V

2,OC
= n
2
ln
_
R

OC

R
p,1
n
1
R
s

1
1
_
n
2
ln
_
J

0,2
n
2
_
(102)
Finally, adding together V

1,OC
in (99) and V

2,OC
in (102) gives V

OC
. The sought V-R-J relation
under IR illumination is attained:
V

OC,[IR]
= ln
_
_
R

OC,[IR]

R
p,1
n
1
R
s
1

1,[IR]
_
n
2
_

1,[IR]

n
1
_
ln
_
_
R
p,1
J

0,1
R
s
_
n
1
_
J

0,2
n
2
_
n
2
_
(103)
51
Similarly, for UV illumination, the V-R-J relation reads:
V

OC,[UV]
= ln
_
_
R

OC,[UV]

R
p,2
n
2
R
s
1

2,[UV]
_
n
1
_

2,[UV]

n
2
_
ln
_
_
R
p,2
J

0,2
R
s
_
n
2
_
J

0,1
n
1
_
n
1
_
(104)
The V-R-J relations (103) and (104) relate V

OC
, R

OC
and J

SC
and are otherwise dependent on
circuit parameters only. It should be emphasized that the V-R-J relation is only valid provided
one cell is considerably more excited than the other, and that all the baseline assumptions at
the OC and SC operating conditions are fully satised.
Experimental results
The V-R-J equation for IR illumination given in (103) was used to determine the diode ideality
factor n
1
and similarly, Equation (104) was used to nd n
2
.
The V-R-J equation is of the form V

OC
= ln (X) ln(K
b
), where K
b
is a constant. Considering
rst the V-R-J for IR illumination, the variable X depends on V

OC
, R

OC
and J

SC
, which are
known from experiment. Both X and K
b
depend on the circuit parameters R
p,1
, n
1
and n
2
.
R
p,1
is known to be 350 m
2
from the analysis detailed in Section 5.5.2. It is also known from
Section 5.3 that n
1
+n
2
= 3.73.
The variable X is computed for each experimental set of points. The known value for R
p,1
is
used, while a value for n
1
is initially guessed. Note that n
2
is simply given by: n
2
= 3.73 n
1
.
A V

OC
ln (X) plot is then constructed and a linear regression t is performed. A new value
for R
p,1
is recomputed from the y-intercept c = ln (K
b,[IR]
):
R
p,1,new
= R
s
_
K
b,[IR]
K
d
n
n
2
2
_ 1
n
1
(105)
where K
d
is dened in (59). It is likely that the recomputed resistance R
p,1,new
will dier
from the original one. Hence, n
1
is systematically varied. Each time X is recomputed and
R
p,1,new
is calculated, as shown in Figure 40. When R
p,1,new
equals the original value of 350
m
2
the corresponding n
1
may be taken to be the true ideality factor. Ideally, the slope of the
V

OC
ln (X) plot for this particular value of n
1
should be 1. In this case it was found to be a
bit higher.
The same procedure was performed for the UV illumination data, with the original R
p,2
accord-
ingly set to 760 m
2
. The dots in Figure 40 represent the nal values for n
i
when R
p,i,new
,
plotted in the horizontal axis, equals the original value. The sum of both ideality factors,
n
1
+ n
2
= 1.8 + 1.94 = 3.74, matches very well the value of 3.73 obtained from the analysis in
dark conditions.
5.6 Summary of ndings
The values of the circuit parameters, obtained experimentally from the various theoretical rela-
tions given in previous sections, are compiled in Table 3.
Subcell 1 is presumably more degraded than subcell 2 and suers from higher shunt leakage
current losses, judging from its smaller -product and lower shunt resistance (R
p,i
) value. Of all
parameters, only the dark saturation diode current of each subcell (J
0,i
) could not be evaluated
individually. This is consistent with the fact that single subcell excitation was not achieved. At
short-circuit conditions under say IR illumination, the magnitude of the diode current of the
non-limiting subcell (given by the J

0,2
-dependent term in Equation (90)) is much greater than
52
that of the limiting subcell (subcell 1). However, since the magnitude of the photocurrents are
not known, it was not possible to nd a useful expression containing J
0,2
in isolation.
Figure 40: Recomputed R
p,i,new
vs n
i
. The red dot for IR highlights n
1
= 1.8 when the
recomputed R
p,1,new
is 350 m
2
. The blue dot for UV highlights n
2
= 1.94 when R
p,2,new
equals
760 m
2
.
Parameter Tandem cell Subcell 1 Subcell 2 Units
Starter values
d
i
- 100 500 nm
V
b,i
- 0.925 0.925 V
V
Te
25.7 25.7 25.7 mV
Dark Measurements
R
s
0.00254 - - m
2
R
p,1
+R
p,2
1170 - - m
2
n
1
+n
2
3.73 - - m
2
J
n
1
0,1
J
n
2
0,2
8.25 10
24
- - A/m
2
Illumination Measurements
R
p,i
- 350 760 m
2
()
e,i
- 2.2 10
14
3.5 10
12
m
2
/V
n
i
- 1.8 1.94
Table 3: Parameter values for the tandem cell equivalent circuit model
53
6 Conclusions
An electrical characterization method for hydrogenated silicon p-i-n single-junction and double-
junction cells has been devised. In each case, the cell was modelled analytically through the
use of an equivalent circuit. The characterization process involved experimentally determining
the unknown circuit elements exclusively from J V curves in the dark and under specic
illumination conditions. The strength of this approach is the simplicity of the experimental
set-up, and that no knowledge of the exact irradiance power or QE of the cell is required at all.
For the single-junction cell characterization, white light was used. All parameters except the
eective -product and built-in voltage could be extracted from a J V curve in the dark. The
eective -product, an indicator of the rate of carrier recombination and state of degradation
of the cell, has been shown to depend on the illumination spectrum and intensity.
The limitations and assumptions of the model have been discussed. For the single-junction cell
case, the model was seen to perform well at small forward and negative voltage biasing. However,
the model-experiment matching at open circuit operating conditions is slightly worse, since the
expression for the recombination term is no longer valid. At large reverse voltage biases, the
model performed poorly. This has been attributed to a non-Ohmic voltage dependence of the
shunt leakage current in this region. Therefore, an extended model that takes into account shunt
leakage current non-linearities has also been presented.
For the double-junction cell characterization, the tandem cell was exposed to IR and UV bias
light in order to excite each subcell dierently. It was not possible to achieve single subcell
excitation. Nevertheless, theoretical relations have been derived for the case of single subcell
excitation. Instead, it was concluded that UV light induced a considerably greater photocurrent
in one subcell (subcell 1) than in the other (subcell 2), while IR light had the opposite eect,
exciting the latter (subcell 2) the most. The characterization process was then adapted to match
these particular circumstances.
6.1 Further work
This work leaves many matters that could be further investigated. The next logical step would be
to obtain the remaining parameter J
0,i
through experimentally achieving single-subcell excitation
or through other means. A method to compute the J V curves for the double-junction cell
model would then have to be developed.
The electrical characterization of modules composed by many cells in series is another issue
of particular interest. This could include the eects of current mismatch, shading, and diuse
irradiation on the overall module electrical response.
Lastly, a basic computational model based on the photocarrier equations restricted to a p-i-n
junction environment would prove useful to better understand the eect of e.g. dopant atom
concentrations and layer thicknesses on the cell eciency.
54
References
[1] Merten, J et al. Improved Equivalent Circuit and Analytical Model for Amorphous Silicon
Solar Cells and Modules. IEEE Transactions on Electron Devices, Vol. 45, No. 2, 1998.
[2] Asensi, J. M. et al. Analysis of the Role of Mobility-Lifetime Products in the Performance
of Amorphous Silicon p-i-n Solar Cells. Journal of Applied Physics, Vol. 85, No. 5, 1999.
[3] Quasching, V. and Hanitsch, R. Numerical Simulation of Current-Voltage Characteristics
of Photovoltaic Systems with Shaded Solar Cells. Solar Energy Vol. 56, No. 6, 1996.
[4] Hubin, J. and Shah, A. V. Eect of the recombination function on the collection in a p-i-n
solar cell. Philosophical Magazine B, Vol. 72, No. 6, 1995.
[5] Streetman, B.G. Solid State Electronic Devices. 4th edition, Prentice Hall Series, 1995, pp.
149-153.
[6] A. Johansson et al. Modelling Shading on Amorphous Silicon Single and Double Junction
Modules. World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003.
[7] Dongaonkar, S. et al W. Universality of non-Ohmic shunt leakage in thin-lm solar cells.
Journal of Applied Physics, Vol. 108, No. 12, 2010.
[8] Luque, A. Hegedus, G. Handbook of Photovoltaic Science and Engineering. John Wiley &
Sons Ltd., 2003, pp. 514-518.
[9] Martins, R. et al. Dependence of amorphous silicon solar cell performances on the lateral
drift current. Solar Energy Materials and Solar Cells, Vol. 45, No. 1, 1997.
[10] Trigo, J. F. et al. Progress in Optical Characterization of Amorphous Silicon Tandem Cells.
Photovoltaic Solar Energy Department (CIEMAT) and Grupo Unisolar (Soliker), 2011.
[11] Bishop, J. W. Computer Simulation of the Eects of Electrical Mismatches in Photovoltaic
Cell Interconnection Circuits Solar Cells, Vol. 25, 1988.
[12] Alves de Oliveira, A.S. Estudio del comportamiento espectral y electrico de celulas fotovol-
taicas en edicacion End of degree project, E.T.S.I.A (UPM), 2012, p. 168.
55

Você também pode gostar