See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228885845

Third-generation photovoltaics

Article in Proceedings of SPIE - The International Society for Optical Engineering · November 2007
DOI: 10.1016/S1369-7021(07)70278-X

CITATIONS READS

480 6,891

1 author:

Gavin Conibeer
UNSW Sydney
329 PUBLICATIONS 9,236 CITATIONS

SEE PROFILE

All content following this page was uploaded by Gavin Conibeer on 25 March 2014.

The user has requested enhancement of the downloaded file.

 Third-generation
photovoltaics
Third-generation approaches to photovoltaics (PVs) aim to achieve high-
efficiency devices but still use thin-film, second-generation deposition
methods. The concept is to do this with only a small increase in
areal costs and hence reduce the cost per Watt peak1 (this metric
is the most widely used in the PV industry). Also, in common with
Si-based, second-generation, thin-film technologies, these will use
materials that are both nontoxic and not limited in abundance. Thus,
these third-generation technologies will be compatible with large-scale
implementation of PVs. The approach differs from first-generation
fabrication of high-quality, low-defect, single-crystal PV devices that have
high efficiencies approaching the limiting efficiencies for single-bandgap
devices but use energy- and time-intensive techniques.

Gavin Conibeer
ARC Photovoltaics Centre of Excellence, School of Photovoltaic and Renewable Energy Engineering, University of New South Wales , Sydney
NSW 2052, Australia
E-mail: [email protected]

Third-generation approaches to PVs aim to decrease costs to devices that limits efficiencies to either 31% or 41%, depending
well below the $1/W level of second-generation PVs to $0.50/W, on concentration ratio (Fig. 1). This requires multiple energy
potentially to $0.20/W or better, by significantly increasing threshold devices. There are several approaches to achieve such
efficiencies but maintaining the economic and environmental cost multiple energy threshold devices1,2.
advantages of thin-film deposition techniques (Fig. 1 shows the The two most important power-loss mechanisms in single-
three PV generations)1. Increasing efficiency strongly leverages bandgap cells are the inability to absorb photons with energy less
lower costs because the smaller area required for a given power than the bandgap (1 in Fig. 2) and thermalization of photon energies
also reduces balance-of-systems costs, such that efficiency exceeding the bandgap (2 in Fig. 2). These two mechanisms alone
values well above 30% could dramatically decrease these costs amount to the loss of about half of the incident solar energy in solar
per Watt. To achieve such efficiency improvements, devices aim cell conversion to electricity. Multiple threshold approaches can in
to circumvent the Shockley-Queisser limit for single-bandgap principle utilize this otherwise lost energy. Such approaches do not in

42 NOVEMBER 2007 | VOLUME 10 | NUMBER 11 ISSN:1369 7021 © Elsevier Ltd 2007

 Third-generation photovoltaics REVIEW

possible, or limiting efficiency of a given device. This was first applied

to PVs by Shockley and Queisser3, and is used by many authors to
model solar cells. According to this analysis, the number of electrons
flowing from an ideal solar cell through an external circuit is equal
to the difference between the number of photons absorbed over the
energy range El to Eh and the number of photons the device emits over
the same energy range. The formulation is not detailed here but uses
the Planck equation for the spectrum of light received from the sun
and a modified Planck equation for the light emitted by the cell, the
latter exponentially enhanced by the forward bias of the cell, just as for
a light-emitting diode1,4–6. The approach is an ideal limiting efficiency
one that assumes zero contact resistance and junction losses, infinite
mobility, and 100% luminescent efficiency.
The approach predicts ideal current density versus voltage, or J–V,
characteristics. This way of describing the generation process is more
fundamental and is equivalent to the more usual ideal diode equation
Fig. 1 Efficiency and cost projections for first- (I), second- (II), and third-
generation (III) PV technologies (wafer-based, thin films, and advanced thin
description, but is easier to understand as it is just the balance of
films, respectively). particles into and out of the cell1. It gives limiting efficiencies of
31.0% and 40.8% for unconcentrated and maximally concentrated
light, respectively, for an optimized bandgap of 1.3 eV and 1.1 eV,
respectively (see section on tandem cells below), as first calculated by
Shockley and Queisser3. For Si and GaAs bandgaps of 1.12 eV and
1.45 eV, respectively, the limiting efficiencies are both ~29% for
unconcentrated light. Thus, the record efficiencies in each case
of 24.7%7 and 25.1%8 at one-sun illumination indicate devices
approaching the radiative limit.
Another advantage of the particle balance approach is that it can
compare different types of solar cell. The high and low energy limits
(Eh and El) can be modified to allow for the particular absorption range
of the specific energy threshold. The detail of this can determine
whether the device neatly splits the spectrum into separate segments,
Fig. 2 Loss processes in a standard solar cell: (1) nonabsorption of below-
bandgap photons; (2) lattice thermalization loss; (3) and (4) junction and as in a tandem solar cell, or has overlapping energy thresholds
contact voltage losses; (5) recombination loss (radiative recombination is competing for photons, as in an intermediate energy level device. The
unavoidable). detail of the way the terms are added describes whether the device
fact disprove the validity of the Schockley-Queisser limit, rather they has the same current through each element – series connection – and
avoid it by the exploitation of more than one energy level – in some is sensitive to the particular spectrum (as in a tandem cell), or whether
form – for which the limit does not apply. The limit that does apply the device has different currents through the elements – parallel
is the thermodynamic one, shown in Fig. 1, of 67% or 86.8% (again connection – as in some of the other devices described below. (Such
depending on concentration). parallel connections would also require appropriate voltage matching.
Three families of approaches have been proposed for applying This is approached differently in the various devices described.)
multiple energy levels2: (a) increasing the number of energy levels;
(b) multiple carrier pair generation per high energy photon or single Multiple energy level approaches
carrier pair generation with multiple low energy photons; and The concept of using multiple energy levels to absorb different
(c) capturing carriers before thermalization. Of these, tandem cells, an sections of the solar spectrum can be applied in many different
implementation of strategy (a), are the only ones that have, as yet, device structures. The ideal limiting efficiencies for these are often
been realized with efficiencies exceeding the Shockley-Queisser limit. identical for a given number of energy levels. Hence their differences
In order to decide which of the many possible approaches are are manifest in the degree to which each overcomes nonidealities.
equivalent and which offer genuine fundamental advantages, the This includes any inability of a particular cell design to select photon
approach of detailed particle balance is used to calculate the maximum absorption at its optimum energy level in the cell, the presence of

NOVEMBER 2007 | VOLUME 10 | NUMBER 11 43

REVIEW Third-generation photovoltaics

parasitic processes (usually associated with defects), and the ease of series design more sensitive to spectral variations. Furthermore, they
manufacture and the abundance of appropriate materials. become increasingly spectrally sensitive as the number of bandgaps
increases. For space-based cells this is not a great problem because of
Tandem or multicolor cells the constant spectrum, but for cells designed for terrestrial use, it is
The tandem or multicolor cell is conceptually the easiest configuration significant because of the variability of the terrestrial solar spectrum.
to understand. It belongs to strategy (a) of increasing the number This is particularly the case at the beginning and end of the day
of energy levels. Solar cells consisting of p-n junctions in different when the spectrum is significantly red-shifted by the thickness of the
semiconductor materials of increasing bandgap are placed on top of atmosphere. Nonetheless, the much greater ease of fabrication of
each other, such that the highest bandgap intercepts the sunlight first in-series devices makes them the design of choice for most current
(Fig. 3). devices.
The elegance of the approach – first suggested by Jackson9 The particle balance limiting efficiency depends on the number of
in 1955 – is that both spectrum splitting and photon selectivity subcells in the device. For 1, 2, 3, 4, and ∞ subcells, the efficiency η is
are automatically achieved by the stacking arrangement. This 31.0%, 42.5%, 48.6%, 52.5%, and 68.2% for unconcentrated light, and
complementarity makes the approach difficult, if not impossible, to 40.8%, 55.5%, 63.2%, 67.9%, and 86.8% for maximally concentrated
beat for an ideal device. To achieve the highest efficiency from the light. Hence, the efficiency increases with the number of subcells in
overall tandem device, the power from each cell in the stack must be both cases, but the efficiency gain decreases with each subsequent
optimized. This is done by choosing appropriate bandgaps, thicknesses, subcell1,4,10.
junction depths, and doping characteristics, such that the incident
solar spectrum is split between the cells most effectively. The next III-V tandems
requirement is to extract electrical power from the device most The highest quality, and hence highest efficiency, tandem devices
effectively. Two configurations are used: either a ‘mechanically stacked’ are made using single-crystal III-V materials. These are grown
cell, in which each cell in the stack is treated as a separate device with monolithically by epitaxial processes such as metal organic vapor phase
two terminals for each; or an ‘in-series’ cell with each cell in the stack epitaxy (MOVPE). Epitaxial techniques are very expensive but give
connected in series, such that the overall cell has just two terminals on very high quality crystalline material. Epitaxial growth requires control
the front and back of the whole stack. For a fixed solar spectrum and of the lattice parameter at a constant value; and bandgap control is
an optimal design, these two configurations give the same efficiency. also required for a tandem cell. It is the flexibility of the III-V group
But for a real, variable spectrum, the mechanically stacked design gives of compounds that lends it to the growth of such cells, usually lattice
greater flexibility because of the ability to optimize the I-V curve of matched on a Ge substrate. This requires most devices to be based
each cell externally and then connect them in an external circuit. The on the AlAs/GaAs system, which has a lattice parameter close enough
reduced flexibility of just optimizing the I-V curve for the whole stack, to that of Ge at 5.66 Å to avoid dislocations. Nonetheless, the twin
because the same current must flow through each cell, makes the in- requirements of lattice parameter and bandgap control mean that

Fig. 3 A simplified schematic of a three-bandgap tandem solar cell. The bandgap of each cell decreases from the front to the back, giving both spectrum splitting
and photon selectivity.

44 NOVEMBER 2007 | VOLUME 10 | NUMBER 11

 Third-generation photovoltaics REVIEW

ternary (or even quaternary) compounds are required for a three cell and hence has a higher defect density and lower efficiency, but which
stack (e.g. GaInP/GaAs/GaInAs cells11,12). can be produced by much cheaper, low-energy intensity deposition
Other researchers are investigating the InN/GaN/AlN system. These methods and uses elements and compounds that are not scarce or
three compounds have a nearly constant lattice parameter at ~3.2 Å toxic. Such devices do not need concentration to reduce the cost per
and a bandgap range easily covering that of interest13,14. Another Watt. This thin-film approach thus tackles the twin requirements of
approach, which is proving to be very successful, is to avoid the third-generation devices, namely low cost per Watt and the use of
requirement for strict lattice matching by growing a partially strained nontoxic and abundant materials.
stack of cells approximately lattice matched to Ge. Such ‘metamorphic’
cells do have higher defect concentrations, but this is offset by the a-Si tandems
increased flexibility in bandgap design – the current world record Amorphous silicon (a-Si)* cells are used for single-junction cells, but
efficiency for any PV device is held by just such a GaInP/GaInAs/Ge tend to give efficiencies of only about 4–5% because of high defect
metamorphic device at 40.7% under 240 suns15, the first device to concentrations associated with the lack of crystallinity17. These
exceed 40%. These efficiency values are significant fractions of the efficiencies can be boosted in tandem cells with a-Si as a top cell with
limiting efficiencies, indicating very high radiative efficiencies. one or two lower cells of an alloy with Ge (a-Si:Ge), which lowers
The next stage is to move onto four- or even five-bandgap cells. the bandgap. These cells are in-series devices that are grown by thin-
These not only have potentially higher efficiencies but also have film processes such as chemical vapor deposition (CVD) or other
higher voltage and lower current than three-bandgap cells. This means vacuum deposition techniques. The lack of a need for crystallization
that series resistance losses are lower, an important consideration for and the vapor-phase deposition mean that much less energy is required
concentrator cells16. for the process, and the use of raw materials tends to be low for the
thin layers deposited – a few hundred nanometers. Such two- or three-
Concentrator systems bandgap stacks can give efficiencies as high as 13% in the laboratory18,
The expense of the growth techniques and of the compounds used but it is difficult to transfer these to production where efficiencies are
means that such devices are usually designed for use in optical around 10%. They also have the same problem of spectral sensitivity
concentrator systems operating at a few hundred suns. This means shared by all in-series tandems.
that only a small area of the very efficient but also very expensive
cell material is required at the optical focus of a relatively cheap Si nanostructure tandems
concentrator. Potentially, this can bring the cost per Watt of electricity Another approach is to retain both the advantages of crystalline
generated down to low levels16. Concentration also gives the higher material and thin-film deposition but to avoid the high costs of
limiting efficiencies mentioned above because the sun effectively fills epitaxial III-Vs by use of thin-film crystalline Si, which is crystallized by
all of the sky as far as the cell is concerned. This in turn means that a post-growth solid phase crystallization anneal19. Such single-junction
photons emitted by the cell must be emitted towards the sun. Hence, cells are now in production at efficiencies just under 10%20. To
the solid angle over which the cell must accept light is the same boost the efficiencies of these cells in a tandem and retain the other
as that over which it emits photons, giving the least possible loss. advantages of third-generation approaches, research is underway on
Furthermore, tandems are well suited to concentrator systems because engineering wider bandgaps for Si-based materials using quantum
as the number of cells in the stack increases, the voltage-to-current confinement in nanostructures.
ratio increases, thus decreasing resistive losses in the high current This bandgap engineering can be done using either quantum wells
densities of concentrator cells15. However, concentrators require direct (QWs) or quantum dots (QDs) of Si sandwiched between layers of a
sunlight and hence do not work with an overcast sky, unlike flat-plate dielectric based on Si compounds such as SiO2, Si3N4, or SiC21. For
cell modules. They also make the overall system more complex and sufficiently close spacing of QWs or QDs, a true miniband is formed
less modular. Hence, they tend to be more appropriate for large-scale creating an effectively larger bandgap. For QDs of 2 nm (QWs of
systems located in inland areas where cloud cover is low. Nonetheless, 1 nm), an effective bandgap of 1.7 eV results – ideal for a tandem cell
another important advantage for tandem devices is that concentrator element on top of Si. These layers are grown by thin-film sputtering or
systems work best if optimized for operation in the middle of the day CVD processes followed by a high-temperature anneal to crystallize the
when sunlight is strongest. Over this period, the spectrum is least Si QWs/QDs. The matrix remains amorphous, thus avoiding some of
variable and hence spectral sensitivity is less significant. the problems of lattice mismatch. Much work remains on passivation
of defects, formation of junctions, and connection to a Si cell, but the
Thin-film tandems
An alternative approach to reducing the cost per Watt is to use *Or, strictly, the alloy of a-Si with hydrogen. The hydrogen bonds to the dangling
bonds in the amorphous structure and significantly passivates defects. At 1.8–1.9 eV,
material that is not of as high a quality as epitaxial III-V materials the bandgap of the alloy is substantially greater than that of Si at 1.1 eV.

NOVEMBER 2007 | VOLUME 10 | NUMBER 11 45

REVIEW Third-generation photovoltaics

Fig. 4 Si nanostructure/Si tandem cell: the nanostructure cell consists of Si QWs or QDs in an amorphous dielectric matrix connected by a defect tunnel junction to a
thin-film Si cell.

approach has great potential, although again, as an in-series tandem, it that would be very difficult to arrange in practice. Hence, without
suffers from the problem of spectral sensitivity. The conceptual design this rather artificial modification, there is a problem with photon
of a complete device is shown in Fig. 4. selection – although the device nonetheless collects photons that
would otherwise not be absorbed. However, note that the current now
Intermediate-level cells: impurity PV and only has to be equal across the two lower energy levels while the main
intermediate band solar cells current across the bandgap is independent. This reduces the spectral
The approach with these devices is to introduce one or more energy sensitivity and compensates to some extent for the reduced photon
levels within the bandgap such that they absorb photons in parallel selectivity.
with the normal operation of a single-bandgap cell. As such, it is also These additional sub-bandgap absorbers can either exist as discrete
an implementation of strategy (a). This semi-parallel operation offers energy levels in an impurity PV (IPV) cell, or as a continuous band
the potential to be much less spectrally sensitive but to still give high of levels nonetheless isolated from the conduction and valence
efficiencies. bands – the intermediate band solar cell (IBSC) shown in Fig. 522,23.
Such a device has the same limiting efficiency as a three-level Both devices can absorb two below-bandgap photons to create
tandem – 63% under maximum concentration, 48% under one sun one electron-hole pair at the bandgap energy, but the IBSC has the
– because it has the same number of energy thresholds. However, advantage that the delocalization of carriers in its continuous band
this calculation does not take into account spectral sensitivity and means that these photons do not necessarily have to be absorbed
assumes ideal properties such as ideal photon selection, a potential by the same electron. This gives a much longer lifetime to the
problem with intermediate level devices. Strictly to ensure complete intermediate level, allowing much more time for absorption of the
photon selectivity, it is necessary to modify the absorption and second photon. To maximize this advantage, the intermediate band
emission ranges of the device such that the photon energy ranges do should be half-filled with electrons – i.e. it should have a Fermi level
not overlap. This implies limited width conduction and valence bands at half the band energy, as illustrated in Fig. 5 – such that absorption

Fig. 5 The intermediate band solar cell. Below-bandgap photons are absorbed by the two transitions to and from the intermediate level contributing to
photocurrent, in parallel to normal operation of the cell.

46 NOVEMBER 2007 | VOLUME 10 | NUMBER 11

 Third-generation photovoltaics REVIEW

of an electron from the valence band or emission of an electron to the great deal of experimental evidence showing production of up to seven
conduction band are equally likely. electron-hole pairs for the absorption of a high energy photon – i.e. a
The IPV cell is made by incorporation of deep-level defects in a quantum efficiency (QE) of seven – with high luminescence efficiencies.
cell – the optimum is at one third of the bandgap energy. An example The ratio of the photon energy to the bandgap energy (Ehν/Eg) must
that has been suggested is B in SiC24 and another suggested and tried be ≥ QE. This phenomenon was first seen in PbSe QDs but has now
experimentally is In in Si25. Such defects also increase the probability been seen in quite a wide range of QDs fabricated from II-VIs and
of radiative and also nonradiative recombination, as both are enhanced other materials, including Si, which is significant for possible future
for narrower energy gaps. This can be offset somewhat by arranging large-scale implementation. However, as yet the phenomenon has only
for defect incorporation only to occur deep in the cell away from the been observed with absorption spectroscopy measurements.
junction. This allows normal absorption of short wavelength light and The formulation for the limiting efficiency of such a device is
separation of minority carriers with little enhanced recombination. But given elsewhere32. The efficiency is calculated in a similar way to
it also allows absorption of long wavelength light below the bandgap a single-bandgap cell but with the current enhanced by an energy-
deep in the cell. In practice, however, these conflicting requirements dependent quantum yield term, QY(E). The value of this depends on
make the cell too much of a compromise and no advantage has yet the particular model used to simulate the data. It is always equal to
been shown25,26. one, up to a threshold energy above the effective bandgap energy, E0.
Formation of an intermediate band for an IBSC has been suggested QY then increases to two, indicating the production of two electron-
in some III-V, II-VI, and chalcopyrite systems, usually alloyed with hole pairs above this energy. This threshold energy varies with material
a transition metal27,28. One specific example is Cr-doped ZnS29. and is usually equal to ~3E0, but is sometimes as low as 2E0, yielding
They have also been attempted experimentally using the confined efficiency limits of 36% and 42%, respectively, at one sun33.
energy levels of a GaInAs/GaAs QD superlattice30. These devices have A device based on this approach requires a means of allowing the
demonstrated several of the indicators of true IBSC operation, although multiple electron-hole pairs to be separated, transported, and collected
they have not yet achieved an efficiency advantage. Nonetheless this in a bulk structure. This is the subject of ongoing research32.
seems likely in the near future, particularly if they are operated under
concentration. Modulation of the spectrum: up/down
conversion
Multiple carrier excitation One of the practical problems with both the intermediate-level and
Carriers generated from high-energy photons (at least twice the multiple-carrier generation designs is that they require good optical
bandgap energy) absorbed in a semiconductor can undergo impact properties (close to the radiative limit), as well as good electrical
ionization events resulting in two or more carriers close to the bandgap properties, to reduce nonradiative recombination and allow carrier
energy. This approach is an implementation of strategy (b). But impact transport – compounding the difficulty of making a good device.
ionization has a vanishingly small probability in bulk material. Recently, Separation of these functions into separate elements of a combined
it has been discovered that this process can be much more efficient in device would thus make a device much easier to optimize.
QDs31,32 (Fig. 6). The exact mechanisms involved are not yet entirely All the devices discussed above use the incident solar spectrum,
clear, but they are related to the reduced requirement for conservation albeit concentrated in some cases. The principle limitation on PV
of crystal momentum in the small spatial volume of a QD. There is a efficiencies arises from the polychromatic nature of this spectrum.

Fig. 6 Multiple-carrier generation in QDs: a high-energy photon is absorbed at a high energy level in the QD, which then decays into two or more electron-hole pairs
at the first confined energy level. Energy is conserved but momentum conservation is relaxed.

NOVEMBER 2007 | VOLUME 10 | NUMBER 11 47

REVIEW Third-generation photovoltaics

Thus, potentially, modification of the incident spectrum could allow a The limiting efficiency for a UC is the same as for the IBSC. The
single-junction standard PV cell to operate at a higher efficiency. The difference between the two is that the UC is electrically separate from
problem is then to create a device that either absorbs a photon of at the cell, although still optically coupled. The limiting efficiency is given
least twice the bandgap energy and emits two photons incident on the as 48% under one sun36, the same as the IBSC. This is higher than
cell (a down-converter or DC), or absorbs at least two below-bandgap that for a DC because the solar spectrum has a long high-energy tail.
photons and emits one above-bandgap photon (an up-converter or Hence, for any given cell bandgap, there will be more photons to be
UC). Schematics of these two devices are shown in Fig. 7, both of up-converted at half the bandgap than down-converted at twice the
which are implementations of strategy (b). bandgap. The bandgap of a Si cell is close to the optimum bandgap for
The DC is placed in front of a standard cell and can boost current a DC but is rather lower than that for a UC at about 2 eV.
by converting ultraviolet (UV) photons to a larger number of visible Experimental progress has been made using UCs in rare earth
photons. However, the DC does require that more visible photons are elements, specifically Er. Er is doped in a matrix that provides a specific
emitted than high-energy photons absorbed, i.e. its QE must be greater separation of the atoms, absorption of 1500 nm photons to the first
than unity†. High-DC QEs (although still less than one) are exhibited Er energy level of two adjacent atoms can then boost one electron
by some luminescent phosphors and by porous Si or Si nanocrystals34. to a higher energy level, resulting in emission of a 980 nm photon,
It may also be that the multiple-exciton generation in QDs discussed above the bandgap of Si. A very small but measurable increase in
above can be adapted for use as a DC. However, as exceeding unity current has been detected using a UC based on this principle37. This
is a difficult, although not impossible, requirement, research is mostly has a reasonable quantum efficiency of about 4%, but the very narrow
focused on UCs. Put underneath a fairly standard single-junction cell‡, bandwidth for absorption makes the device impractical, with little
a UC can boost current using below-bandgap photons that are not prospect for improvement based on the rare earths, although the
normally absorbed. As the UC does not interrupt the incidence of general approach is very promising and other UC systems may well fare
photons on the front surface, even a very low efficiency UC gives a better.
small current boost and hence an efficiency increase.
A DC material absorbs short wavelength photons over a narrow Hot carrier cells
energy range, as shown in Fig. 7. Carriers generated in this material The final option for increasing efficiencies – strategy (c) – is to allow
decay radiatively via a midgap level to give two photons at or just absorption of a wide range of photon energies but then to collect the
above the bandgap of the cell. Specific differences in the assumptions photogenerated carriers before they have a chance to thermalize. A
compared with those for the IBSC (mainly that the lower level must be hot-carrier solar cell is just such a device that offers the possibility of
half the upper level) give the limit for a DC of 36.7% under one sun35. very high efficiencies (the limiting efficiency is 65% for unconcentrated
illumination) but with a structure that could be conceptually simple
†The threshold QE in order to achieve a net gain in current is actually greater than compared with other very high efficiency PV devices – such as
unity, as some emitted light will go back to the sun. Nonetheless, this fraction can be multijunction monolithic tandem cells. For this reason, the approach
kept to below 10% by careful optical matching of refractive indexes.
lends itself to thin-film deposition techniques with their attendant
‡This requires a bifacial cell in which grid contacts on the front and back allow the
low material and energy usage costs and the ability to use abundant,
cell to be operated with light incident on either face. In fact, bifacial Si cells are quite
routinely made and have the added advantage that below-bandgap light is transmitted. nontoxic elements.

(a) (b)

Fig. 7 (a) Down-converter in front of a single-junction cell. (b) Up-converter behind a single-junction cell.

48 NOVEMBER 2007 | VOLUME 10 | NUMBER 11

 Third-generation photovoltaics REVIEW

electrons in an antenna tuned to the wavelength of light – hence the

devices have to be of the order of a few hundred nanometers. The
broadband incoherent nature of the solar spectrum also requires a wide
range of antenna sizes to match all the wavelengths and the need to
arrange two directions of polarization. In addition to these practical
problems, it has also been shown that the approach can only achieve
48% even under ideal conditions45.
Thermal approaches include TPV46, in which a narrow bandgap cell
is illuminated by black- or gray-body radiation from a hot source but at
a lower temperature than the sun. Efficiencies can be boosted by use of
a selective emitter that only allows light just above its bandgap to be
incident on the cell, the rest being reflected back to reheat the primary
emitter. Thermophotonics1,47 is a variation in which the thermal source
Fig. 8 Band diagram of the hot carrier cell that requires slowed carrier cooling in heats a luminescent diode similar to the TPV cell, this then illuminates
the absorption and collection of carriers through selective energy contacts. the cell with a spectrum strongly peaked just above their common
bandgap. These approaches would normally use waste heat from an
The concept underlying hot carrier solar cells is to slow the rate of industrial process or similar and hence not be PV, but they can be
photoexcited carrier cooling, which is caused by phonon interaction coupled to an emitter heated by solar thermal energy. In practice, the
in the lattice, to allow time for carriers to be collected while they large number of different elements, with their multiplying inefficiencies,
are still at elevated energies (‘hot’). This allows higher voltages to be and the need to thermally insulate some elements from others make
achieved by the cell1,38,39. It thus tackles the major PV loss mechanism the approaches very difficult to optimize.
of thermalization of carriers (2 in Fig. 2). In addition to an absorber In Fig. 2, one of the loss mechanisms is from radiative
material that slows the rate of carrier relaxation, a hot carrier cell must recombination (loss 5). In most devices this is assumed to be a
allow extraction of carriers from the device through contacts that minimum loss that cannot be reduced for a cell at the radiative
accept only a very narrow range of energies (selective energy contacts), limit, i.e. no nonradiative recombination. This is necessary as a
as shown in Fig. 8. reciprocal device that can absorb solar wavelengths must also be
Initial experimental progress has been made on selective energy able to emit those same wavelengths48. However, it is possible that a
contacts using double-barrier resonant tunneling structures, with a nonreciprocal device could re-use some of this emitted radiation and
single layer of Si QDs providing the resonant level40. The problem boost efficiencies beyond the radiative limit. Such a device is known
of slowing carrier cooling is very difficult. It has been observed at as a multiport circulator, where incoming light incident on port 1 is
very high illumination intensities via a phonon bottleneck effect in emitted at port 2 and incident light on port 2 is emitted at port 31,49.
which carrier energy decay mechanisms are restricted. Compounds In principle, such a circulator can be used to redirect light emitted
with large mass difference between their anions and cations have a from a solar cell onto a second cell, light from this cell can then go
gap in their allowed phonon modes that can slow down these decay through a second circulator to a third cell, etc. It has been shown that
mechanisms and enhance the bottleneck effect41. Examples are GaN such an approach can boost efficiencies to 93%, the Landsberg limit,
and InN, with some experimental evidence for slowed cooling in the for an infinite number of circulators illuminating tandem cells with an
latter42. Theoretical work on replicating this effect by modifying the infinite number of bandgaps1,5. The obvious practical difficulties of this
phononic band structures of QD nanostructure superlattices43 will soon are offset to some extent by the fact that most of the efficiency gain
be attempted experimentally. Nonetheless, the hot carrier cell, while is obtained with the first circulator and, as we have seen, a tandem
promising, is still a long way from demonstration. cell has most of its increase in efficiency for the first few layers. The
nonreciprocal rotation of the polarization of light by a magnetic field
Other approaches can be and is used to fabricate such esoteric-sounding circulator
A few other approaches have been suggested for increasing PV devices for microwave and laser optics50. However, the complexity
efficiency, including quantum antennas, thermophotonics or of the components and the very small efficiency gains make such an
thermophotovoltaics (TPVs), and circulators. These are not part of the approach applied to PVs only appropriate for theoretical considerations.
main discussion of multiple energy level third-generation PVs, but brief
descriptions are given below. Conclusions and future directions
The idea of a quantum antenna is to use the wave nature of The combined methodology of using multiple energy thresholds
light rather than its particle nature44. In-coming light waves oscillate and low cost processes with abundant nontoxic materials in third-

NOVEMBER 2007 | VOLUME 10 | NUMBER 11 49

REVIEW Third-generation photovoltaics

generation devices offers significant leverage in the pursuit of implementation of such techniques could dramatically decrease
significantly lower cost per Watt PVs. Many of these devices use cost per Watt with spectral robustness as they are compatible with
the unique flexibility of QW and QD nanostructures to optimize conceptually relatively simple thin-film devices.
absorption, carrier generation, and separation. The technique of The even more esoteric approaches of circulators, quantum
detailed balance for calculating limiting efficiencies illustrates the antennas, and thermophotonics/TPVs are probably impractical even
similarity in several third-generation concepts, despite their different if they can be shown to work theoretically. Combinations of two or
approaches, as it depends on the number of energy levels involved. more approaches are also a possibility, e.g. the combination of both
In practice, however, the actual efficiencies and ease of optimization a UC and a DC on the same cell6, or the use of a UC with a tandem
depend on the different physical approaches. Efficiency, spectral cell. Alternatively, either an intermediate-level or a multiple-carrier
robustness, and cost/ease of manufacture are important for a generation device could be used as a DC rather than for producing
robust technology that can supply very significant increases in PV excess carriers directly. The use of detailed balance analysis
implementation. No device matches all these goals yet. helps to show the equivalencies and complementarities between
Tandem PV devices are the best developed so far and further these approaches. Also, no doubt, other approaches not yet conceived
improvement, whether by increased concentrator system efficiency will be possible as understanding of the topic continues to improve.
or by reduced cost and increased efficiency of thin-film designs, can
leverage much lower overall costs per Watt. However, these devices
Acknowledgments
tend to suffer from poor spectral robustness. Work on intermediate
The author acknowledges the support of The ARC Photovoltaics Centre of
level devices and up/down conversion is at a much earlier stage,
Excellence, UNSW, during the collection of information for, and the writing of,
but promises increased efficiencies and greater spectral robustness, this article. The Centre of Excellence is supported by the Australian Research
potentially with thin-film-type materials. The more advanced concepts Council (ARC) and the Global Climate and Energy Project (GCEP), the latter
administered by Stanford University. The author would also like to thank his
of multiple-carrier generation and hot carrier cells are further away colleagues in the Centre of Excellence for their work on and input to Third
still and still have serious theoretical questions to answer. Nonetheless, Generation PV concepts.

REFERENCES 25. Keevers, M. J., and Green, M. A., Sol. Energy Mater. Sol. Cells (1996) 41–42, 195
1. Green, M. A., Third Generation Photovoltaics: Ultra-High Efficiency at Low Cost, 26. Würfel, P., Sol. Energy Mater. Sol. Cells (1993) 29, 403
Springer-Verlag, Berlin, (2003) 27. Palacios, P., et al., Phys. Rev. B (2006) 73, 085206
2. Nelson, J., The Physics of Solar Cells, Imperial College Press, London, (2003) 28. Yu, K. M., et al., Appl. Phys. Lett. (2006) 88, 092110
3. Shockley, W., and Queisser, H. J., J. Appl. Phys. (1961) 32, 510 29. Tablero, C., Phys. Rev. B (2006) 74, 195203
4. Würfel, P., J. Phys. C (1982) 15, 3967 30. Marti, A., et al., Phys. Rev. Lett. (2006) 97, 247701
5. Brown, A., Thesis: Ultimate Efficiency Limits of Multiple Energy Threshold 31. Schaller, R. D., and Klimov, V. I., Phys. Rev. Lett. (2004) 92, 186601
Photovoltaic Devices, University of New South Wales, Sydney, Australia (2003)
32. Hanna, M. C., and Nozik, A. J., J. Appl. Phys. (2006) 100, 074510
6. Conibeer, G. J., et al., 19th Euro. PVSEC, Paris, France, (2004), 274
33. Klimov, V., Appl. Phys. Lett. (2006) 89, 123118
7. Zhao J., et al., Prog. Photovoltaics (1999) 7, 471
34. Švrcek, V., et al., Thin Solid Films (2004) 451, 384
8. Green, M. A., et al., Prog. Photovoltaics (2007) 15, 425
35. Trupke, T., et al., J. Appl. Phys. (2002) 92, 1668
9. Jackson, E. D., Proc. Conf. on the Use of Solar Energy, Tuscon, Arizona, (1955),
122 36. Trupke, T., et al., J. Appl. Phys. (2002) 92, 4117

10. Brown, A. S., and Green, M. A., Prog. Photovoltaics (2002) 10, 299 37. Shalav, A., et al., Appl. Phys. Lett. (2005) 86, 013505

11. Takamoto, T., et al., Prog. Photovoltaics (2005) 13, 495 38. Würfel, P., Sol. Energy Mater. Sol. Cells (1997) 46, 43

12. Wanlass, M., et al., 4th World Conf., PV Energy Conversion, Hawaii, (2006), 729 39. Ross, R. T., and Nozik, A. J., J. Appl. Phys. (1982) 53, 3813

13. Trybus, E., et al., J. Cryst. Growth (2006) 288, 218 40. Jiang, C.-W., et al., 21st Euro PVSEC, Dresden, Germany, (2006), 168

14. Jani, O., et al., 4th World Conf., PV Energy Conversion, Hawaii, (2006), 20 41. Conibeer, G., and Green, M. A., 19th Euro PVSEC, Paris, France, (2004), 270

15. King, R. R., et al., Appl. Phys. Lett. (2007) 90, 183516 42. Chen, F., and Cartwright, A. N., Appl. Phys. Lett. (2003) 83, 4984

16. King, R. R., et al., 21st Euro PVSEC, Dresden, Germany, (2006), 124 43. Conibeer, G, et al., 21st Euro PPVSEC, Dresden, Germany, (2006), 90

17. Meier, J., et al., Thin Solid Films (2004) 451, 518 44. Bailey, R. L., J. Eng. Power (1972), 73

18. Yang, J., et al., 26th IEEE PVSC, Anaheim, CA, (1997), 563 45. Corkish, R., et al., 3rd World Conf. on PV Energy Conversion (Osaka, 2003), 2682

19. Aberle, A. G., J. Cryst. Growth (2006) 287, 386 46. Couts, T., In: Clean Electricity from Phtovoltaics. Archer, M. D., and Hill, R., (eds.),
Imperial College Press, London, (2001)
20. Basore, P. A., 4th World Conf., PV Energy Conversion, Hawaii, (2006), 2089
47. Catchpole, K. R., et al., 3rd World Conf., PV Energy Conversion, Osaka, Japan,
21. Conibeer, G., et al., Thin Solid Films (2005) 511–512, 654 (2003), 270
22. Luque, A., and Martí, A., Phys. Rev. Lett. (1997) 78, 5014 48. Kirchoff, G., Ann. Phys. (1860) 19, 275
23. Luque, A., et al., Physica B (2006) 382, 320. 49. Ries, H., Appl. Phys. B (1983) 32, 153
24. Richards, B. S., et al., 3rd World Conf., PV Energy Conversion, Osaka, Japan, 50. Fujii, Y., J. Lightwave Technol. (1991) 9, 1238
(2003), 2738

50 NOVEMBER 2007 | VOLUME 10 | NUMBER 11

View publication stats