University of Ljubljana

Faculty of Mathematics and Physics

Seminar -1st year, 2nd cycle

Solar Cells

Author: Katja Vozel

Advisor: izred. prof. dr. Denis Arčon

Ljubljana, May 2011

Abstract
This seminar is a short overview of current photovoltaics. Application of p-n junctions within solar cells is
discussed. We give the upper limit to the solar cell efficiency. Three generations of solar cells, crystalline
silicon solar cells, thin-film solar cells and photoelectrochemical cells, are described by means of their
operational principles and physics behind. We also compare efficiencies of current devices.

1
Contents
1 Introduction 2

2 Timeline of solar cells 3

3 Analysis of p-n junction 4

3.1 Current-voltage characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

4 Solar cell structure 6

5 Basic definitions 7

6 Efficiency: thermodynamic limit and detailed balance 8

7 Crystalline solar cells 9

8 Thin-film solar cells 10

8.1 Amorphous silicon solar cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

9 Novel solar cells 12

9.1 Operation principles of dye-sensitized solar cells . . . . . . . . . . . . . . . . . . . . . . . . . 12
9.2 Nanocrystaline junctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
9.3 Characteristics and practical aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

10 Comparison of PV devices 14

11 Conclusion 14

1 Introduction
Photovoltaic effect is emergence of electric voltage in a system exposed to solar radiation. With absorbtion
of photons, charge carriers are excited into conduction band. The mechanism of light induced electron
transition to a higher energy state is similar to that of photoelectric effect, where a photon carrying sufficient
amount of energy frees an electron from the surface of a metal. The photoelectric effect was explained by
Albert Einstein in 1905. Converting solar radiation into electrical energy is called photovoltaics (PV).
Devices exploiting PV effect are called solar cells, also photovoltaic cells or photovoltaic devices.
Global electricity consumption amounts to approximately 2·1014 kWh per year [1]. Could solar cells
satisfy the world’s hunger for energy and how much of the land would be needed? The Sun provides 1000
W/m2 of power density for a surface perpendicular to the Sun’s rays at sea level on a clear day. The actual
power at specific area varies with seasons and depends on the geographic position of the area. It must be
taken into account that the Sun shines only during the day and that the angle of Sun’s rays varies during
the day, if the PV devices do not rotate. We also estimate, that there is 70% of sunny days in a year.
With a little of calculation, we arrive at average power density of 120 W/m2 [2, 3, 4] coming from the Sun,
which we multiply by PV device efficiency of 10% to get the power density obtained from PV. Dividing the
consumption by the power density obtained from solar cells, we get the area of solar cells needed, which
is approximately 1012 m2 or 0.7% of Earth’s land [5]. By making a similar calculation only for Europe, a
continent with relatively high population and little sun, we find that we should cover approximately 8%
of Europe land to provide the whole Europe with electricity. That is quite a lot, but fortunately, Sahara
dessert, with area of 9.4·1012 m2 [6] is close enough to be exploited. After these estimates, we see that
solar energy could be one of the most promising energy sources, alternative to the currently dominating
fossil fuels. The problem is, however, very high cost of PV, particularly crystalline silicon solar cells, which
currently have the highest known efficiency. Although there are cheaper alternatives to silicon, they lack of
either efficiency or chemical stability, or can not be prepared by fast processing techniques [7]. In principle,
efficiency of solar cells is limited by the band gap of the material used, because the band gap has to match
a part of solar frequency spectrum. It is also conected with impurities in the material. Cheaper materials
often contain a lot of impurities, which can act as recombination centres for photogenerated charge carriers,

2
thus lowering the efficiency and stability of a cell. Chemical instability refers to light induced degradation
in amorphous silicon solar cells or to the limited number of redox cycles for electrolytes in novel solar cells.
PV is world’s fastest growing power generation technology. From 2008 until now, the global contribution
of PV to electricity production raised from 0.06% to 0.5% [1, 8]. The rapid, we could say exponential growth
of installed PV capacity from 2000 to 2010 was strongly influenced by national subsidies. Based on the
capacity installed and connected to the grid at the end of 2011, PV can provide roughly 2% of the electricity
demand in Europe, up from 1.15% at the end of 2010 [9]. In Slovenia, 0.44% of electricity demand was
provided by solar cells in 2011 [9]. Currently, around 80% of solar cells is made of polycrystaline or single
crystal silicon [10]. These are reffered to as first generation solar cells. Second generation solar cells comprise
of thin films of materials like amorphous silicon, cadmium telluride and copper indium (gallium) diselenide.
First and second generation solar cells use p-n monojunctions or heterojunctions to separate charge carriers.
Third generation solar cells include photoelectrochemical or dye-sensitized cells, organic cells and nano-
technology. Their working principles differ significantly from those of first and second generation cells. The
working mechanisms of novel cells require much less material than is needed for production of the first and
second generation solar cells. But these novel cells are not yet completely developed and their efficiencies
remain low, cases up to 10%, in comparison to first generation devices with typical efficiencies around 20%
and second generation devices with typical efficiencies between 10% and 19% [10, 11].
In this seminar we make a short glance at the history of PV in section 2. In section 3, we make a short
analysis of p-n junction in connection with its use in solar cells. We present the structure of a solar cell in
section 4. In the next section we explain basic physical principles of PV and we introduce some important
definitions considering PV. In section 6, the limiting efficiency of a single photovoltaic device is given. In
the next three sections, 7, 8 and 9, we discuss specifics of structure, physics and properties of three types
of solar cells: single crystal silicon solar cell, amorphous silicon solar cell and dye sensitized solar cell. We
also compare operation and efficiencies of devices in section 10.

2 Timeline of solar cells

French physicist Edmund Becquerel discovered photovoltaic effect in 1839 [12]. He noticed that by illu-
minating silver coated platinum electrode in electrolite, he could produce electric current. Several years
later, in 1876, William Adams and Richard Day observed light-induced photocurrent in selenium with
two heated platinum contacts. Probably the first large area solar cell containing selenium and gold was
made by Charles Fritts in 1894. It had efficiency of around 1%. In the following years the focus was on
coper-coper oxide thin film structures, which were Schottky barrier devices, where a semitransparent layer
of metal was deposited on top of a semiconductor. Photovoltaic effect of such devices is related to a barrier
to current flow at one of the metal-semiconductor interfaces. The teory of metal-semiconductor barrier was
developed by Schottky, Neville Mott and others [12]. The disadvantage of Schottky contacts is that the
required unimpeded exchange of the minority carriers is unavoidably coupled with a high level of surface
recombination at the metal contact [13].
In the middle of the 20th century, good quality silicon wafers were developed for applications in solid
state electronics. For the first time, considerable quantities of power were produced by photovoltaic cells
consisting of crystalline silicon. These were p-n junction devices, which had some advantages over Schottky
barriers, for example much better rectifying action, which means that the charge carriers move only in the
desired direction. Devices with p-n junction could reach better efficiencies. First silicon cell was made by
Capin, Fuller and Pearson in 1954 with 6% efficiency. The next few decades, efficiency was steadily rising,
but the cost of photovoltaic systems was too high to compete with other energy sources. Photovoltaics were
therefore used mainly as space sattelite energy supply, as the importance of small weight and reliability
outhweighted high costs.
At the same time, other materials were being researched. It was found out that cadmium sulphide,
gallium arsenide, indium phosphide and cadmium telluride p-n junctions should have equal or better
efficiencies than silicon p-n junction. Despite that, silicon remained the most popular material, due to
advances of silicon technology for microelectronics industry [12]. Oil industry crisis followed in 1970s, which
led to several fundations in research of renewable energy sources. During that time, deep understanding of
photovoltaics was gained. New strategies and materials were proposed to make photovoltaics cheaper, for
instance photoelectrochemical junctions, polycristalline silicon, amorphous silicon, organic conductors and
so on. None of the mentioned technologies have taken over the market of photovoltaics so far.

3
 Since 1970s, interests in renewable energy sources as alternative to fossil fuels have been growing. In the
late 1990s, production of solar cells expanded 15-25% per year, which also reduced their cost. Solar cells
have become competitive at remote locations, in navigation systems, telecomunications and as additional
power in grid connected loads at peak use [12].

3 Analysis of p-n junction

This section reveals the operational principle of most common solar cells and also serves as a basis for
further discussion.
As we know, an electric field spontaneously occurs at the interface between p- and n-type semiconduc-
tors, in so called space charge region (SCR). The built-in potential difference accross the p-n junction, Vbi
is [12, 14]  
kB T Nd Na
Vbi = ln , (1)
q n2i
where Nd and Na stand for density of donor and acceptor atoms. q is the absolute value of the charge of
an electron. The width of the space charge region was also calculated before [12, 14] and is given by
s  
2εs 1 1
wscr = | − wp | + wn = + Vbi , (2)
q Na Nd

where εs is the local permittivity of the semiconductor. We chose the origin of the longitudinal axis at
the interface between differently doped semiconductors. wp and wn are widths of charge regions on the
p- and n-side. According to equations 1 and 2, the built-in potential difference increases with increasing
doping, while the width of the SCR decreases. In solar cells both high potential difference and wide SCR
are needed for effective drift of charge carriers towards the right direction, so a compromise must be made
when choosing the doping level. The SCR is highly resistant part of p-n junction, while the bulk of the
junction has very low resistance for electric current. Any applied bias would have the highest drop in the
SCR.
A hole current flows from the n- to the p-side of the junction, known as hole generation current. It is
minority carrier current, because it results from generation of holes when electrons are thermally excited
from valence to conduction band on the n-side. A hole must be generated within a diffusion lenght from
SCR and when it reaches the SCR, it is immediately driven to the p-side by strong electric field. Similarly,
there flows an electron generation current from the p- to the n-side of the junction. Another hole current
flows from the p-side to the n-side of the junction, known as hole recombination current. It results from
the holes, that reach the SCR with sufficient energy to surmount the potential barrier of the SCR. The
number of such holes is proportional to e−q∆V /kB T and thus depends strongly on the potential step ∆V
across the junction. If a forward (positive) bias is applied, such that it raises the potential of the p-side
with respect to the n-side, the potential barrier is lowered and the recombination current is stronger. In
that case, the electric current flows with little resistance through the p-n junction, dominated by majority
carrier current. On the contrary, if a negative bias is applied, the potential barrier is increased and only
little current flows through the device. This is known as the rectifying action of a diode.
In equilibrium, when no external bias is applied, the net electric current is zero. When a forward bias
V is applied, the net charge current density, flowing in the direction of recombination currents, is given by
[14]
J = (Jhgen + Jegen )(eqV /kB T − 1), (3)
where Jhgen and Jegen are hole and electron generation current densities. The situation is considerably
altered, when a p-n junction within a photovoltaic material is illuminated. The generation currents, re-
sulting from photogenerated minority charge carriers, are increased. The so called photocurrent becomes
the dominant electric current, flowing in the opposite direction as recombination current. We take the
potential drop accross the junction to be the difference between the built-in bias Vbi and the applied bias
V , Vj = Vbi − V . We are trying to find solution for electric current density flowing through a photovoltaic
device under illumination and applied bias. In general, problem consists of solving a set of coupled dif-
ferential equations for the hole density, the electron density and the electrostatic potential, given specified
forms for the photogeneration, the recombination and the hole and electron currents [12]. It is a complex

4
problem, but analytic solutions can be found by using two approximations, which allow the hole and elec-
tron currents to be decoupled. The first one is so called depletion approximation, which assumes there
are no free charge carriers in the space charge region around the junction. The electric field vanishes at
fixed distances from the junction. Within the neutral regions away from the junction, majority carrier
densities have their equilibrium value and variations in minority carrier densities determine the current. In
neutral regions, there is no electric field and currents of minority carriers are only diffusive. The depletion
approximation allows the solutions in neutral p and n-regions to be decoupled.
In the second approximation we assume the recombination process to be linear in the minority carrier
density. It is sometimes called superposition approximation [12], because the effect of bias and illumination
are decoupled, so the solution to the current is the sum of both.
We give some of the important steps in solving the problem of electric current density. When a device
is illuminated with light with energy W , the rate of charge generation, g, is given by [12]
g(W, x) = (1 − R(W ))α(W )bs (W )e−α(W )x , (4)
where R is the reflectance of the material, bs (W ) is the incident photon flux, which will be explained in
section 5 and α(W ) is the absorption coefficient, which is given by the matrix element, calculated with use
of the Fermi golden rule [12]. Integrating over photon energies, we obtain the integrated charge generation
rate G. In the presence of both the electric field E and charge carrier density gradient, the carrier current
density can be written as the sum of a term proportional to the density gradient - diffusion current, and a
term proportional to the field - the drift current [12, 14]:
Jp (~r) = −qDp ∇p + qµp Ep and Jn (~r) = qDn ∇n + qµn En. (5)
Here, Jp denotes charge current density, as a consequence of motion of holes and Jn denotes charge current
density, as a consequence of motion of electrons. µp and µn are mobilities while Dp and Dn are the
corresponding diffusion constants of holes and electrons, respectively. If we combine equations 5 with the
steady state continuity equations, take the current generated by a spectrum containing a single wave lenght
and reduce the problem to one dimension, x, then the transport equations in the case of zero electric field
- in the neutral p- and n- regions - simplify to [12]:
d2 p p − p0 g(W, x) n2i
− + = 0, p0 = (6)
dx2 L2p Dp Nd
for holes in the n-region and
d2 n n − n0 g(W, x) n2i
− + = 0, n0 = (7)
dx2 L2n Dn Na
for electrons in the p-region. Lp and Ln stand for diffusion lenghts of holes and electrons, depending on
the charge carrier lifetime τ , p p
Lp = τp Dp , Ln = τn Dn . (8)
At the boundary with the SCR, the condition for density of holes is given by [12]
n2i qV /kB T
p − p0 = (e − 1), x = wn . (9)
Nd
The boundary condition at the outer surface is determined by the surface recombination rate Sp ,
dp
−Dp = Sp (p − p0 ). (10)
dx
It follows that the hole current in the n-region and the electron current in the p-region, which we obtain
in a similar way, can be written as [12]
dp dn
jp (W, x) = −qDp , jn (W, x) = qDn . (11)
dx dx
The current in the space charge region can be determined from carrier continuity. It can be written
as integral of difference between charge recombination rate U and charge generation rate G over the SCR:
[12] Z wn
Jscr = q (U − G)dx. (12)
−wp

5
This is equal for electrons and holes within the space charge region. The net current is given by the sum of
hole and electron currents at any point and is constant through the device in steady state. General solution
for p can be obtained analitically by solving equation 6. From 11 and after intergrating over energy, we
obtain general solution for Jp and similarly we obtain solution for Jn . We are not going to write down
the whole solutions, which can be found in [12]. There are two separate contributions to the net current
density. The first one is proportional to the incident spectral photon flux density and the second one is
proportional to eqV /kB T − 1 as a consequence of applied bias.
With solutions for current densities, we can calculate the dark current Jdark , the case where there is
no contribution from illumination, and the short-circuit current Jsc , the case where V = 0. The J is
illustrated in Figure 1 as a function of applied bias V . By adding short circuit current and dark current,
current-voltage characteristics of a solar cell, which will be discussed in section 5, is constructed:
J(V ) = Jsc − Jdark (V ) = Jsc − Jm,0 (eqV /mkB T − 1). (13)
Jm,0 is material-dependent constant and m is ideality factor that describes deviation from ideal diode
behavior with m = 1. The last term in expression 13 is indeed the same one obtained by general analysis
of p-n junctions [14]. The additional current density Jsc is a consequence of light illumination. The sign of
the currents is chosen so that photocurrent is positive. With increasing V the flux emmited from a solar
cell increases and the net current decreases. What is more, the latter expression provides the open-circuit
voltage - the voltage at which the current is zero,
mkB T Jsc
Voc = ln( + 1). (14)
q Jm,0
If the V is increased above the Voc , the cell begins to act like a light emmiting device. Voc must always be
less than Eg /q, where Eg is the band gap of the semiconductor used.
Maximum
power point
Current density
Jsc

Jm
Current density, J

Power
density

Vm Voc
Bias voltage, V

Figure 1: Current-voltage (black) and power-voltage (grey) characteristics of an ideal diode (ideal pho-
tovoltaic device). Vm and Jm denote voltage and current density at maximum power point. Reproduced
from [12].

3.1 Current-voltage characteristics

The voltage, which is a result of cell illumination when the sides of the cell are isolated, is called open-circuit
voltage Voc (denoted in Figure 1). The electric current, which emerges when the sides of the cell are short
circued, is called short circuit current Isc . When a load with a resistance RL is added in the circuit, voltage
V and current I developed by the cell under illumination are such that V = IRL . I(V) is determined
by current-voltage characteristics of the cell under given illumination [12]. The electric current is roughly
proportional to the illuminated area and that is why it will be useful to define short circuit current density
Jsc .

4 Solar cell structure

We present the structure of a c-Si photovoltaic device. Crystalline silicon solar cell is composed of two basic
layers, the emitter, which is n-doped, and the base, which is p-doped semiconductor silicon(see Figure 2).

6
The front contact has to be designed in a way that prevents shading of the front surface. Because the bulk
of the electron-hole pairs is generated near the surface, the position of p-n junction is also near the surface.
This prevents recombination of charge carriers before they reach the junction, where they are separated by
electric field.

Figure 2: Structure of crystalline silicon solar cell. Reproduced from [10].

5 Basic definitions
Module
A single solar cell area is typically about 100 cm2 , it generates a dc photovoltage of 0.5 -1 V and pho-
tocurrent of several amperes [12]. Because of low voltage and electric current, cells are connected into
modules, which produce standard voltage of 12 V. Modules are further connected into strings. Due to the
fluctuations in solar irradiation during the day, PV systems also include components for charge regulation
and storage. Parts of PV system are illustrated in Figure 3.

(a) (b) (c) +V

(d)

Power
PV generator conditioning Load

0V Storage
Module (Battery (dc)
or Grid (ac))
0
+12V

Figure 3: (a) Solar cell with surface contacts. (b) Cells conected in a module. (c) Modules are connected
in series into strings and in parallel into an array, to produce sufficient current and voltage. (d) Integration
of charge regulation and storage. Reproduced from [12].

Quantum efficiency
Quantum efficiency ηQE of a solar cell is the probability that an incident photon will deliver one electron
to the external circuit. The resulting short circuit photocurrent density depends upon quantum efficiency:
Z
Jsc = q bs (W )ηQE (W )dW . (15)

Here, q is unit charge, W denotes energy of photons and bs is the incident spectral photon flux density,
that is number of photons with energy between W and W + dW falling on a cell per unit area and per unit
time. The solar spectrum reaching surface of the Earth is presented in Figure 4.
Quantum efficiency is a property of a solar cell, which means it is related to absorption coefficient and effi-
ciency of charge separation but does not depend on incident spectrum. It is desirable to have a high ηQE at
energies where also solar photon flux density is high, that is somewhere between wavelenghts of 400 and 1300
nm.

7
By combining cells that operate best at differ-
ent wavelenghts, we can exploit larger part of so-
lar spectrum. These are so called tandem cells.
Although tandem originally referred to two cells,
it is now also used for arrangements with more
than two cells. Each cell in the arangement con-
verts a part of solar spectrum. The top cell
consists of a wide gap semiconductor. It con-
verts the short wavelength part of the spectrum
and transmits the other part to the cells below
[10].
Figure 4: Solar spectrum at the surface of the
Fill factor and efficiency
Earth. Reproduced from [15].
Current-voltage characteristics of a solar cells has a
point of maximum power, as seen in Figure 1. At the
point of maximum power density, we denote current density with Jm and voltage with Vm . The load, we
attach to the solar cell circuit, has the optimum resistance given by Vm /Jm . Jm and Vm should be as near
to Jsc and Voc as possible. We define the fill factor:
Jm Vm
FF = , (16)
Jsc Voc
which would be ideally one. For very good crystalline silicon solar cells, the fill factor is above 80% [10].
General efficiency η of a PV device is the ratio of maximum output power density to incident radiation
power density,
Jm Vm Jsc Voc F F
η= = . (17)
Ps Ps
For most cells, the efficiency is temperature dependent. With increasing temperature, the energy of charge
carriers is increased, leading to higher difussion current densities. The cell emits more strongly. According
to equation 14, with higher diffusion current density Jm,0 , the open-circuit voltage is lowered, which leads
to lower efficiency. Efficiencies are measured under Standard Test Conditions (STC), that is Air Mass 1.5
spectrum (Air Mass is lenght of the light path to the Earth surface relative to the path at highest position of
the sun), incident power density of 1000 W/m2 and temperature of 25◦ C. The efficiencies of solar modules
are always somewhat lower than the laboratory efficiencies of single solar cells.

Parasitic resistances
Every solar cell has two parasitic resistances, the series resistance Rs and the parallel or shunt resistance
Rsh . The series resistance is the resistance of the material and contacts of the cell, through which electric
current flows. Parallel resistance results from current leakage because of imperfect rectifying action. For
highly efficient cell Rs should be as low and Rsh as high as possible, otherwise the fill factor is considerably
lowered.

6 Efficiency: thermodynamic limit and detailed balance

Detailed balance is a way to determine maximum possible efficiency of a solar cell. It is based on balancing
different particle fluxes in the cell. The detailed treatment can be found for example in [16, 17]. A simple
form of detailed balance can be used to calculate the maximum efficiency of a two-band photoconverter,
which has only two electronic levels with a bandgap Eg between them. It is an approximation for a basic
c-Si or GaAs PV device containing p-n junction. In the calculation, the solar spectrum is fixed at Air
Mass 1.5 spectrum (see Figure 4). The short circuit current density Jsc , which is a result of excitation of
an electron to the upper band via photon absorption, as well as the dark current density Jdark , which is a
result of electron relaxation from upper to lower band, depend on the energy gap Eg . The power obtained
from the solar cell is a product of applied bias and net current density J = Jsc − Jdark , and is therefore
also Eg -dependent. From equation 17 we can anticipate that neither very small or very large band gap
would lead to maximum efficiency, because the Voc increases [13] and Jsc decreases with increasing band
gap (because less of the solar spectrum is absorbed). It can be shown, that the power of a cell has a

8
maximum at energy gap of 1.4 eV [12]. Dividing the maximum power by the power coming from the sun,
we obtain the maximum efficiency of 33% for a two-band photoconverter [12].
As can be seen from equation 15, the efficiency depends on the shape of solar spectrum bs (W ), which
varies slightly with the composition of atmosphere at different places on earth at different times. Neverthe-
less, the most important is the temperature of the light source - the Sun. If the Sun had lower temperature,
the optimum band gap and the maximum efficiency would both be lower. Clearly, if the temperature of the
Sun was equal to the temperature of the cell, there would be no net photoconversion. On the other hand,
if the temperature of the Sun would increase relative to the cell, so would the photoconversion efficiency.
The main limiting factor to efficiency of PV is that only part of the solar spectrum that matches the
band gap can be converted into electricity. No photons with energy less than Eg contribute to the output
power. Absorbed photons with energy W > Eg contribute only qV < Eg of electrical energy and the rest
of the energy is lost as heat. The efficiency of PV cells can be enhanced considerably by applying different
techniques, for example multijunction cells [10, 12], multiple band gaps [18], multiple absorption, light
concentration etc. Nevertheless, the efficiency is limited at least in the thermodynamic limit. Maximum
obtainable conversion efficiency of solar radiation into electricity can be roughly estimated by using Carnot
relation [10]. For the temperature of the Sun of 5760 K and the temperature of surroundings of 300 K, the
maximum efficiency would be 85% at cell temperature of 2470 K [10].

7 Crystalline solar cells

We move now to some specific examples of today’s solar cells, their properties and the physics behind.
Most of present solar cells are manufactured from semiconductor Silicon [19], which can conduct electricity
when electrons are excited from valence to conduction band via photon absorption. Silicon has a bandgap
of ∼ 1.12 eV which corresponds to wavelength within infrared, λ ∼ 1.11 µm [10]. Theoretical limit to
the efficiency for such device is 30% [12]. Crystalline silicon (c-Si) is an indirect gap semiconductor, which
means that both a photon and a phonon are needed to raise the electron into the conduction band. Because
of simultaneous involvement of both a photon and a phonon, indirect semiconductors have lower absorption
probability than direct semiconductors. Thus, in order to absorb long wave part of solar spectrum, silicon
layers should be thick, and consequently large amounts of material must be used. The minimal thickness
of a crystalline silicon solar cell is 100 µm [20].
Another crystalline photovoltaic material is gallium arsenide (GaAs). It has a band gap of 1.42 eV
at room temperature, almost the ideal one. It is a direct gap material which means it absorbs better
than silicon which reduces the mass of devices significantly. Solar cells made of GaAs are only a few µm
thick [20]. What is more, while silicon solar cell efficiency reduces with increasing temperature because of
more recombination events, connected with more phonons, and reduced band gap, GaAs is not as much
temperature sensitive. GaAs also has a better radiation hardness than silicon, so the extraterrestrial
radiation does not affect its performance so much over cell’s lifetime. Its cost is significantly higher than
that of silicon and it is primarily appropriate for space applications and light concentration systems.

9
8 Thin-film solar cells

Figure 5: Schematic structures and band diagrams of three types of thin film devices: a-Si (top left), CIGS
(top right) and CdTe (bottom). Reproduced from [10].

Because crystalline solar cells are expensive to produce, alternative photovoltaic materials are being re-
searched that can be grown more cheaply. Various methods are used to deposit semiconductor layers as
thin films directly onto useable substrates such as glass, metals, roofing sheets and plastics [19]. So far,
amorphous silicon (a-Si) has been most widely used, for example in small consumer devices such as calcu-
lators and in solar power plants. Apart from a-Si, polycristalline cadmium telluride (CdTe), polycristalline
copper indium disellenide (CuInSe2 and related compounds, often denoted by CIGS) and microcrystalline
thin film silicon (p-Si) have also been well developed. These materials all have higher efficiencies (CdTe
has efficiency above 16% [10] and the other two have efficiencies above 19% [10]) compared to a-Si (with
efficiency below 10% [10]). However, high efficiencies of devices made of CIGS and CdTe were achieved in
laboratories and are connected with complicated deposition techniques, not really adapted for the indus-
trial large-scale production. There are several technical problems making production of efficient modules
difficult [10]. Some elements used, for example In, Ga and Te are rather rare and hence long term use may
be questionable. Schematic structures and band diagrams of a-Si, CIGS and CdTe solar cells are presented
in Figure 5. a-Si solar cell includes a p-i-n junction, junction of p-doped whereas the other two include
various heterojunctions. The p-i-n junction is similar to p-n junction, the only difference is that there is a
region of intrinsic semiconductor between p-doped and n-doped regions.
Thin film cells are often multijunction devices - they include more than one junction between differently
doped semiconductors. The use of heterojunctions increases efficiency of the devices in comparison to
monojunctions, because thin film materials often cannot be doped equally well n-type and p-type. In figure
5 we see that CIGS solar cell includes heterojunctions between CIGS and CdS and between CdS and ZnO.
The highly doped CdS layer with a band gap of 2.4 eV serves as a window, which reflects the electrons
and therefore reduces losses due to surface recombination of charge carriers generated in CIGS by short
wavelenght light. The CdS also enables transport of electrons with minimum series resistance. However,
heterojunctions have some disadvantages that lower the efficiency of a cell. The main problem is that
defect states, encouraging the non-radiative recombination of charge carriers, appear at the heterojunction
because of differences between crystal structure or lattice constants of the two materials [12].
In comparison to single crystal materials, polycristalline and amorphous materials contain much more
defects which act as traps and recombination centres. Consequently, diffusion lenghts are shorter, doping
is more difficult, resistivity is increased and physical parameters like conductivity, minority carrier lifetime

10
and diffusion constant become carrier density dependent. The latter makes transport equations as well as
their solutions rather complicated. Suitable thin film materials should be good absorbers so the cell can be
thin, solving the problem of small diffusion length, therefore allowing less pure materials to be used. Here,
we discuss some physics and properties of an a-Si solar cell.

8.1 Amorphous silicon solar cell

The difference between single crystal and a-Si material is that
a-Si lacks of long range crystal order. The neighbourhood of
an atom inside a lattice is very similar to that of crystalline
silicon atom, with a few degrees variation in the bond angles.
The slightly altered local structure causes that momentum of
electrons in valence and conduction band is largely undeter-
mined, selection rules for photon absorption in crystalline sil-
icon are no longer valid, so that a-Si behaves like a direct gap
semiconductor with high absorption coefficient (at energy 2.5
eV, for example, absorption coefficient of c-Si is approximately
104 /cm, while absorption coefficient of a-Si is approximately
105 /cm). In Figure 6 (top), comparison between crystalline
and amorphous silicon density of states is presented. Band
tails and dangling bond states occur in a-Si. Dangling bond
states are due to defects in the lattice, where an Si atom is
bonded to only three of the neighbouring atoms, leaving the
fourth valence band orbital unused. These dangling bonds
may be negatively charged, positively charged, or neutral. The
dangling bond states cause energy levels deep in the band gap.
This kind of material can not be effectively doped, as the dan-
gling states (with density more than 1016 cm−3 [12]) capture
extra charge carriers. This so called background density of de-
fects (also background doping) can be reduced by saturating Figure 6: Top: Comparison between c-Si
the dangling bonds with atomic hydrogen. An atom of hy- and a-Si density of electron states. Bot-
drogen forms a bond with the unpaired electron in a neutral tom: Position of chemical potential in pres-
dangling defect, so that it can no longer trap a charge carrier. ence of defects.
The density of dangling defects is then reduced by a factor of
10. Still, the doping efficiency remains low in comparison to
crystalline silicon. The position of chemical potential is moved towards defect levels as shown in Figure 6
(bottom), which enhances the activation energy of mayority carriers. In p-type a-Si this activation energy
is around 0.4 eV [12]. These large activation energies limit the size of the built-in bias, increasing the
difference between the built-in bias and the band gap. Density of states as seen in Figure 6 leads to wide
distribution of time constants for any transient process as well as strong dependence on the occupation
of the states and hence carrier density. As we have mentioned before, mobility, lifetime and diffusion
lenght are density dependent and difficult to determine. We shall mention some changes to be made when
calculating transport. Firstly, the tail density of states g is often modelled as
Nt Wk −E c
g(W ) = e B T0 , (18)
kB T0
where Nt is volume density of tail states, Ec is the energy of the lower boundary of conduction band and
T0 describes the depth of the tail. And secondly, the dangling bond states can be modelled by two discrete
levels in the band gap representing unoccupied and singly occupied defect state for which rate equations are
constructed for capture and emission of carriers. Typical values of parameters in undoped hydrogenated
a-Si are: minority carrier lifetime of 10 to 20 µs, diffusion length of 0.1 µm and carrier mobility of 10−7
to 10−4 m2 V−1 s−1 . By doping we increase the density of defects and so diffusion lenghts are furtherly
reduced. Because of small diffusion lengths, p-i-n junction is used in a-Si devices rather than p-n juction.
We should mention an important limitation to the a-Si cell efficiency, namely the light induced degra-
dation. When illuminated, some of the Si-H bonds are broken, which increases the density of dangling
bonds and reduces efficiency. This process can be reversed by annealing at a few hundred degrees Celsius

11
and therefore a-Si may perform better in warmer enviroments. Typically, the efficiency is reduced by about
30% in the first 6 months of operation and the fill factor falls by about 10%.

9 Novel solar cells

Novel technologies include dye-sensitized solar cells, organic materials and nanotechnology [19]. These
new material as well as operational concepts tend to lower the production cost further. The possibility
of depositing thin layers of photovoltaic materials onto light weight flexible substrates and wide range of
colors makes them appropriate for placing onto buildings, clothing, mobile phones and so on. Apart from
reducing the cost, further development of third generation solar cells is focussing on improving efficiency,
stability and durability. Here, we shortly represent operation of dye-sensitized solar cells.

9.1 Operation principles of dye-sensitized solar cells

Dye sensitized solar cells or also Grätzel solar cells [21] are photoelectrochemical cells. A basic type of
electrochemical cell does not include a p-n or other junction between two solid inorganic semiconductors,
but is composed of semiconductor electrode, an electrolyte layer and counter electrode, covered with Pt,
as shown in Figure 7. The basis for both electrodes is the transparent TCO glass. TCO is short for
transparent conductive oxides. Photoelectrochemical cell converts light to electric power with no net
chemical change behind [22]. At the interface between semiconductor and electrolyte, a space charge layer
is formed, depending on type of both semiconductor and electrolyte. In equilibrium, the chemical potential
of electrons in the solid µn is equal to the redox potential of the electrolyte Eredox . Photons exceeding
the band gap energy produce electron hole pairs that are separated by electric field in the space charge
layer. The electrons diffuse to the end of electrode and are transported through the external circuit with
resistance. The holes are driven to the electrolyte interface, where they oxidize a molecule of redox pair.
The oxidized molecule is reduced again by an electron that re-enters the cell through the external circuit.
In a dye-sensitized solar cell, absorption of photons is enhanced by using an appropriate dye, absorbed
on the semiconductor surface. Light causes transition of an electron in the dye to exited state S∗ , as seen
in Figure 7. Such an excited electron can tunnel to the semiconductor quicker than relaxation occurs -
typical times of tunneling are of order 10−10 s and the lifetime of the excited state is of order 10−8 s [21].
After the electron is transferred to the semiconductor, it again diffuses to the end of the electrode and flows
through the external circuit back to the cell. The oxidized dye molecule is reduced again in redox reaction
with electrolyte. The electrolyte is reduced, again, by the electron that re-enters the cell at the counter
electrode. The whole sequence of reactions can be written down as follows, where the semiconductor used
is TiO2 and the electrolyte is I− −
3 /I . The latter reaction is catalyzed by Pt.

TiO2 |S + hν → TiO2 |S∗ N (19)

TiO2 |S∗ → TiO2 |S+ + e- (20)

TiO2 |2S+ + I− → TiO2 |2S + I−
3 (21)
I− −
3 + 2e → I
−
(22)
The potential difference between electrodes is given by the difference between chemical potential of
semiconductor and redox potential of electrolyte.
There are other processes within the dye-sensitized cell, that are undesired and decrease the efficiency
of the cell. These are: decay of excited state of the electron from the dye, recombination of the electron
with the dye and recombination of the electron with the electrolyte [21].

12
Figure 7: Left: Structure and operation of a dye-sensitized solar cell. The TCO glass at the sides is not
sketched. The incident light excites an electron in a dye molecule into a higher state. Such an excited
electron tunnels onto a TiO2 molecule. The electron then diffuses to the end of the electrode, enters the
external circuit and re-enters the cell through the opposite electrode. The oxidized dye molecule is reduced
by the electrolyte and the electrolyte is reduced by the re-entering electron. Reproduced from [23]. Right:
Surface of a mesoporous anatase film (scanned by SEM) prepared from a hydrothermally processed TiO2
colloid. Scale is 50 nm. Reproduced from [22].

9.2 Nanocrystaline junctions

It could be shown that only dye molecules directly attached to the semiconductor surface are able to
efficiently inject charge carriers into the semiconductor with a quantum yield of more than 90% [10]. In
order to absorb as much light as possible, minutely structured materials with an enourmous internal surface
area were developed. Mesoporous oxide films, made up of arrays of tiny nano - crystals are used instead of
flat semiconductor electrode. The pores are filled with semiconducting or conducting material. Through
such a junction of two interpenetrating networks charge carriers can move very easily. The structure is
typically 10 µm thick and has a surface area available for dye absorption over a thousand times larger than
that of a flat electrode [22]. Examples of oxides used are TiO2 , ZnO, SnO2 , Nb2 O5 , CdSe and other. An
example of mesoscopic material used for dye-sensitized solar cell electrode is presented in Figure 7 (right
hand side).

9.3 Characteristics and practical aspects

Today’s dye-sensitized solar cells reach the short-circuit photocurrent of more than 20 mA/cm2 [22], the
open-circuit voltage of 1 V and efficiency of more than 11% [24] in a laboratory under STC. These char-
acteristics make dye-sensitized solar cells competitive with other solar cells, together with low production
cost. The main advantage of dye-sensitized solar cells is that the electric current results from majority,
and not minority carrier transport. This means that no recombination events within the semiconductor
can happen. This allows use of less pure materials, simple processing and reduces cost. However, impure
materials can reduce the cell’s lifetime.
The problem is, however, that electrolyte used must sustain as much as 108 operating cycles, needed
for cell operation of 20 years [22], which is desired lifetime for a photovoltaic cell. Liquid electrolytes pre-
pared in combination with some appropriate solvents provide a system able to pass the standard stability
qualification tests for outdoor applications [22]. But, there is also a problem with leakage of liquid elec-
trolyte, evaporation and possibility of destruction because of low temperatures during winter. Therefore,
much effort has been made in order to replace the liquid electrolyte with either a gel electrolyte, solid-state
electrolyte, or p-conducting polymer material [10]. No efficient solid state electrolytes, alternative to liquid
electrolytes have been discovered so far [25].

13
10 Comparison of PV devices
Table 1 presents some efficiencies, illuminated areas, open-circuit voltages, short-circuit currents and fill
factors of different types of solar cells. Table 2 presents the same characteristics for some PV modules. The
existing modules have somewhat lower efficiencies than single solar cell devices. The highest cell efficiency
(over 30%) was achieved with multijunction devices, where one of the materials used is GaAs. This is not
surprising as we have already said that GaAs is a strong absorber with nearly ideal band gap.
In general, devices with higher short-circuit current have lower open-circuit voltage. It is because Jsc
and Voc depend on the width of the band gap - the Jsc increases and the Voc decreases with increasing band
gap, as we have mentioned earlier.

Classification Efficiency Area Voc J sc FF Test centre

(%) (cm2) (V) (mA/cm2) (%) (and date)

Silicon
Si (crystalline) 25.0 0.5 4.00 0.705 42.7 82.8 Sandia (3/99)
Si (multicrystalline) 20.4 0.5 1.002 0.664 38.0 80.9 NREL (5/04)
Si (thin film transfer) 16.7 0.4 4.017 0.645 33.0 78.2 FhG-ISE (7/01)
III-V cells
GaAs (crystalline) 26.1 0.8 0.998 1.038 29.7 84.7 FhG-ISE (12/07)
GaAs (thin film) 26.1 0.8 1.001 1.045 29.5 84.6 FhG-ISE (07/08)
GaAs (multicrystalline) 18.4 0.5 4.011 0.994 23.2 79.7 NREL (11/95)
InP (crystalline) 22.1 0.7 4.02 0.878 29.5 85.4 NREL (4/90)
Thin film chalcogenide
CIGS (cell) 19.4 0.6 0.994 0.716 33.7 80.3 NREL (1/08)
CdTe (cell) 16.7 0.5 1.032 0.845 26.1 75.5 NREL (9/01)
Amorphous/nanocrystalline Si
Si (amorphous) 9.5 0.3 1.070 0.859 17.5 63.0 NREL (4/03)
Si (nanocrystalline) 10.1 0.2 1.199 0.539 24.4 76.6 JQA (12/97)
Photochemical
Dye sensitised 10.4 0.3 1.004 0.729 22.0 65.2 AIST (8/05)
Organic
Organic polymer 5.15 0.3 1.021 0.876 9.39 62.5 NREL (12/06)
Multijunction devices
GaInP/GaAs/Ge 32.0 1.5 3.989 2.622 14.37 85.0 NREL (1/03)
GaInP/GaAs 30.3 4.0 2.488 14.22 85.6 JQA (4/96)
GaAs/CIS (thin film) 25.8 1.3 4.00 — — — NREL (11/89)

Table 1: Efficiencies and other characteristics of different solar cells. NREL = National renewable energy
laboratory, FhG-ISE = Fraunhofer Institut für Solare Energiesysteme; JQA = Japan Quality Assurance;
AIST = Japanese National Institute of Advanced Industrial Science and Technology. Stability of organic
and dye-sensitised solar cells was not investigated. Reproduced from [11].
Classification Efficiency Area Voc I sc FF Test centre
(%) (cm2) (V) (A) (%) (and date)
Si (crystalline) 22.9 0.6 778 5.60 3.97 80.3 Sandia (9/96)
Si (large crystalline) 20.3 0.6 16300 66.1 6.35 78.7 Sandia (8/07)
Si (multicrystalline) 15.5 0.4 1017 14.6 1.37 78.6 Sandia (10/94)
Si (thin-film polycrystalline) 8.2 0.2 661 25.0 0.320 68.0 Sandia (7/02)
CIGSS 13.5 0.7 3459 31.2 2.18 68.9 NREL (8/02)
CdTe 10.9 0.5 4874 26.21 3.24 62.3 NREL (4/00)
a-Si/a-SiGe/a-SiGe (tandem) 10.4 0.5 905 4.353 3.285 66.0 NREL (10/98)

Table 2: Efficiencies and other characteristics of different modules. Reproduced from [11].

11 Conclusion
Solar cells currently used contain p-n junctions. Theoretically, with no reflection of light, perfect absorption
and no losses inside the material or at the contacts, the efficiency of a solar cell containing a single junction
could reach 33%. The actual maximum efficiencies are around 25% with use of the best materials, which are
very expensive. Various alternatives to the expensive devices are being reasearched. The main goal for the
future is to further reduce the cost of PV modules, so the PV can be more widely used. Apart from reducing
amount of material used, the price is also reduced indirectly by increasing the efficiency of PV devices.
The efficiency is most effectively increased by using tandem cells in combination with light concentration
systems, where mirrors are used to focus sunlight on to the surface of PV modules. Efficiencies above 30%
are possible [10]. The limiting factor to the efficiency is heating of the system exposed to light irradiation
with power of several hundreds of suns.
In the past 10 years, the price of all known types of PV modules has fallen significantly. For example,
the price of electricity produced by c-Si modules has fallen from ∼0.40 e/kWh to ∼0.30 e/kWh and for

14
a-Si modules the price has fallen from ∼0.20 e/kWh to ∼0.15 e/kWh. The price of electricity produced
by CIGS modules was ∼0.17 e/kWh in 2003 and has fallen to ∼0.09 e/kWh until 2010 [26]. All of the
mentioned modules have a lifetime of at least 20 years [10, 27]. During this time the efficiency falls for
about 20% (see, for example [27]).
There are many installation possibilities for PV modules, among the best known being installation on
the roofs of the houses. In urban areas, the modules may be connected to the grid system, while at remote
locations, they are independent systems and have to be combined with a good charge storage system. There
are some ideas for the future, considering installation of PV, for example in large plain deserts, which are
extremely sunny locations. The electric energy produced in deserts would have to be transported to other
parts of the world, which could be done by using power lines, or transport of liquified hydrogen. Water is
split into hydrogen and oxygen by electrolysis and hydrogen can then serve as the energy carrier [10].

References
[1] http://en.wikipedia.org/wiki/Electricity generation (12. 5. 2012).
[2] http://www.indexmundi.com/world/electricity consumption.html (24. 5. 2012).
[3] http://hypertextbook.com/facts/1998/ManicaPiputbundit.shtml (24. 5. 2012).
[4] http://en.wikipedia.org/wiki/Insolation (24. 5. 2012).
[5] http://en.wikipedia.org/wiki/Earth (24. 5. 2012).
[6] http://en.wikipedia.org/wiki/Sahara (24. 5. 2012).
[7] F. C. Krebs, Fabrication and processing of polymer solar cells: A review of printing and coating
techniques. Solar Energy Materials & Solar Cells, 93, 394-412 (2009).
[8] http://www.ren21.net/Portals/97/documents/GSR/GSR2011 Master18.pdf (12. 5. 2012).
[9] EPIA - Global market outlook for photovoltaics until 2016, p. 61, (2012).
[10] A. Goetzberger, V.U. Hoffmann, Photovoltaic Solar Energy Generation. Springer-Verlag Berlin Hei-
delberg. ISBN 3-540-23676-7 (2005).
[11] M. A. Green, K. Emery, Y. Hishikawa and W. Warta, Solar Cell Efficiency Tables (Version 33). Prog.
Photovolt: Res. Appl., 17, 85–94 (2009).
[12] J. Nelson, The Physics of Solar Cells. Imperial College Press, London. ISBN 1-86094-340-3 (2003).
[13] P. Würfel, Physics of solar cells: from priciples to new concepts. Wiley-VCH verlag GmbH & Co,
KgaA, Weinheim. ISBN 3-527-40428-7 (2005).
[14] N. W. Ashcroft, N. D. Mermin, Solid State Physics. ISBN-13: 978-81-315-0052-1 (2011).
[15] http://continuingeducation.construction.com/article.php?L=68&C=488&P=6 (11. 5. 2012).
[16] W. Shockley, HJ Queisser, Detailed balance limit of efficiency of pn junction solar cells. J. Appl. Phys.,
32, p. 510 (1961).
[17] G. L. Araujo and A. Marti, Absolute limiting efficiencies for photovoltaic energy conversion. Sol.
Energy Mater. Solar Cells, 33, 213 (1994).
[18] http : //gregbonett.org/dalal.pdf (27.4.2012).
[19] http://www.med.govt.nz/sectors-industries/energy/pdf-docs-library/energy-data-and-
modelling/technical-papers/pv-in-new-zealand.pdf (8.4.2012).
[20] http://photovoltaics.sandia.gov/docs/PVFSCGallium Arsenide Solar Cells.htm (15. 5. 2012).
[21] A. Dergan, Študija prenosa naboja na površini titanovega dioksida z elektronsko paramagnetno reso-
nanco. Diplomsko delo, Ljubljana (2011).
[22] M. Grätzel, Photoelectrochemical cells. Nature, 414, 338-344, 15 November (2001).
[23] http://www.scielo.br/scielo.php?script=sci arttext&pid=S0103-50532003000600005 (8.4.2012).
[24] J. Bouclé and J. Ackermann, Solid-state dye-sensitized and bulk heterojunction solar cells using TiO2
and ZnO nanostructures: recent progress and new concepts at the borderline. Polym Int, 61, 355–373
(2012).
[25] Ionic liquids: Theory, Properties, New Approaches. Edited by Alexander Kokorin. InTech, Rijeka.
ISBN 978-953-307-349-1 (2011).
[26] http://www.solarthinfilms.com/active/en/home/photovoltaics/the pv facts/cost evolution of the
different pv modules.html (12. 5. 2012).
[27] http://www.nanosolar.com (21. 5. 2012).

15