Energy Policy 28 (2000) 989}998

Photovoltaics: technology overview

M.A. Green*
Centre for Photovoltaic Engineering, University of New South Wales, Sydney, NSW 2052, Australia
Received 24 May 2000

Abstract

Solar electricity produced by photovoltaic solar cells is one of the most promising options yet identi"ed for sustainably providing
the world's future energy requirements. Although the technology has, in the past, been based on the same silicon wafers as used in
microelectronics, a transition is in progress to a second generation of a potentially much lower-cost thin-"lm technology. Cost
reductions from both increased manufacturing volume and such improved technology are expected to continue to drive down cell
prices over the coming two decades to a level where the cells can provide competitively priced electricity on a large scale. The
subsidised, urban residential rooftop application of photovoltaics is expected to provide the major application of the coming
decade and to provide the market growth needed to reduce prices. Large centralised solar photovoltaic power stations able to provide
low-cost electricity on a large scale would become increasingly attractive approaching 2020.  2000 Elsevier Science Ltd. All rights
reserved.

1. Introduction combining the Nobel prize winning work of several of the

most important scientists of that century. The German
Photovoltaics involves the direct conversion of sun- scientist, Max Planck, began the century engrossed in the
light into electricity in thin layers of material known as problem of trying to explain the nature of light emitted
semiconductors with properties intermediate between those by hot bodies, such as the sun. He had to make assump-
of metals and insulators. Silicon, the material of microelec- tions about energy being restricted to discrete levels to
tronics and the information age, is the most common match theory and observations. This stimulated Albert
semiconductor. In the latter half of the 20th century, silicon Einstein, in his `miraculous yeara of 1905 (Stachel, 1998),
photovoltaic solar cells started to be used mainly to gener- to postulate that light was made of small `particlesa, later
ate small amounts of electricity in remote areas where there called photons, each with a tiny amount of energy that
was no conventional source of electricity. In the 21st cen- depends on the photon's colour. Blue photons have
tury, photovoltaics will grow to maturity. Almost everyone about twice the energy of red photons. Infrared photons,
will be aware of photovoltaics since photovoltaic solar cells invisible to the eye have even less energy. Ultraviolet
will be on the roof of their home or that of their neighbours photons, the cause of sunburn and skin cancer, are also
* be they in one of the growing megacities across the globe invisible but carry even more energy than the blue ones,
or in a remote rural village (Green, 2000). accounting for the damage they can do.
This paper reviews the current state of solar cell tech- Einstein's radical suggestion led to the formulation
nology, outlines developments expected over the coming and development of quantum mechanics, culminating in
few decades and considers implications for energy policy. 1926 in Edwin SchroK dinger's wave equation. Wilson sol-
ved this equation for material in solid form in 1930. This
allowed him to explain the di!erence between metals,
2. Brief history good conductors of electricity and insulators; also the
properties of semiconductors with their intermediate
Solar cells have their origins from some of the most electrical properties. Electrons, the carriers of electrical
important scienti"c developments of the 20th century, charge, are free to move around in metals, allowing
electrical currents to #ow readily. In insulators, electrons
* Tel.: #61-2-9385-4018; fax: #61-2-9662-4240. are locked into the bonds holding the atoms of the
E-mail address: m.green@unsw.edu.au (M.A. Green). insulator together. They need a jolt of energy to free them

0301-4215/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 0 1 - 4 2 1 5 ( 0 0 ) 0 0 0 8 6 - 0
990 M.A. Green / Energy Policy 28 (2000) 989}998

from these bonds, so they can become mobile. The same

applies to semiconductors, except a smaller jolt is needed
* even the red photons in sunlight have enough energy
to free an electron in the archetypical semiconductor,
silicon.
Russel Ohl discovered the "rst silicon solar cell by
accident in 1940 (Riordan and Hoddeson, 1997). He was
surprised to measure a large electrical voltage from what
he thought was a pure rod of silicon when he shone
a #ashlight on it. Closer investigation showed that small
concentrations of impurities were giving portions of the
silicon properties dubbed `negativea (n-type). These Fig. 1. Incoming sunlight is converted to an electrical current #ow in
properties are now known to be due to a surplus of a load such as a lamp connected between the cell contacts.
mobile electrons with their negative charge. Other re-
gions had `positivea (p-type) properties, now known to
be due to a de"ciency of electrons, causing an e!ect incoming sunlight is thereby converted into electrical
similar to a surplus of positive charge (something close to energy consumed by this load.
a physical demonstration of the mathematical adage that The cell operates as a `quantum devicea, exchanging
two negatives make a positive). photons for electrons. Ideally, each photon of su$cient
William Shockley worked out the theory of the devices energy striking the cell causes one electron to #ow
formed from junctions between `positivea and `negativea through the load. In practice, this ideal is seldom reached.
regions (p}n junctions) in 1949 and soon used this theory Some of the incoming photons are re#ected from the cell
to design the "rst practical transistors. The semiconduc- or get absorbed by the metal contacts (where they give up
tor revolution of the 1950s followed, which also resulted their energy as heat). Some of the electrons excited by the
in the "rst e$cient solar cells in 1954. This caused enor- photons relax back to their bound state before reaching
mous excitement and attracted front-page headlines at the cell contacts and thereby the load.
the time (Riordan and Hoddeson, 1997). The electrical power consumed by the load is the
The "rst commercial use of the new solar cells was on product of the electrical current supplied by the cell and
spacecraft, beginning in 1958. This was the major com- the voltage across it. Each cell can supply current at
mercial application until the early 1970s, when oil em- a voltage from 0 V to a maximum in the 0.5}1.0 V range,
bargoes of that period stimulated a re-examination of the depending on the particular semiconductor used for the
cells' potential closer to home. From small beginnings, cell.
a terrestrial solar cell industry took root at this time and
has grown rapidly, particularly over recent years, to
US$1 billion per year in sales, by the end of the 20th 4. Cell technology
century (Perlin, 1999). Increasing international resolve to
reduce carbon dioxide emissions as a "rst step to reigning in 4.1. Silicon wafers
the `Greenhouse E!ecta, combined with decreasing cell
costs, sees the industry poised to make increasing impact The technology used to make most of the solar cells,
over the "rst two decades of the new millennium. fabricated so far, borrows heavily from the microelec-
tronics industry (Green, 1982).
The silicon source material is extracted from quartz,
3. Operating principles although sand would also be a suitable material. The
silicon is then re"ned to very high purity and melted.
Fig. 1 is a schematic of a solar cell under illumination. From the melt, a large cylindrical single crystal is drawn
Light entering the cell through the gaps between the top (Fig. 2), usually of 10}15 cm diameter and 1 m or more in
contact metal gives up its energy by temporarily releasing length, weighting several tens of kilograms. The crystal,
electrons from the covalent bonds holding the semicon- or `ingota, is then sliced into circular wafers, less than
ductor together; at least this is what happens for those half a millimetre thick, like slicing bread from a loaf.
photons with su$cient energy. The p}n junction within Sometimes this cylindrical ingot is `squared-o!a before
the cell ensures that the now mobile charge carriers of the slicing so the wafers have a `quasi-squarea shape that
same polarity all move o! in the same direction. If an allows processed cells to be stacked more closely side-
electrical load, such as the lamp shown in Fig. 1, is by-side.
connected between the top and rear contacts to the cell, Most of this technology is identical to that used in the
electrons will complete the circuit through this load, much larger microelectronics industry, bene"ting from
constituting an electrical current in it. Energy in the the corresponding economies of scale. Since good cells
 M.A. Green / Energy Policy 28 (2000) 989}998 991

Fig. 2. Growth of a cylindrical silicon crystal.

Fig. 4. Structure of a typical commercial cell with textured surface and

screen-printed contacts.

on the surface, with the silicon structure that remains

determined by crystal directions that etch very slowly.
The square-based pyramids apparent in Fig. 4 that are
formed by this process are similar in shape, if not in size,
to the great pyramids of Egypt. These pyramids are very
Fig. 3. (a) Directional solidi"cation of a large block of multicrystalline
silicon from a melt; (b) Sawing of large block into smaller ingots prior to e!ective in reducing re#ection from the cell surface.
slicing into multicrystalline wafers. (Light re#ected from the side of a pyramid will be re#ec-
ted downwards, getting a second chance to get coupled
in).
The all-important p}n junction is then formed. The
can be made from material of lower quality than that impurity required to give p-type properties (usually
used in microelectronics, additional economies are ob- boron) is introduced during crystal growth, so it is al-
tained by using `o!-speci"cationa silicon and `o!-speci- ready in the wafer. The n-type impurity (usually phos-
"cationa silicon wafers from this industry. phorus) is now allowed to seep into the wafer surface by
As the photovoltaics industry matures, it will increas- heating the wafer in the presence of a phosphorus source.
ingly use technology optimised for its own requirements. This gives a thin skin of phosphorus-doped material
An example is the increasing use of `multicrystallinea around the entire wafer. The skin along the wafer edge is
silicon wafers. The starting ingot is formed simply by removed (that along the rear is rendered inactive during
solidifying the molten silicon slowly in its container (Fig. the rear contacting step).
3(a)). This ingot can be massive, weighing several hun- Next, the top and bottom contacts are applied using
dreds of kilograms. It is sawn into pieces of a more metal particles (usually silver) suspended in a paste with
manageable size (Fig. 3(b)) and then sliced into wafers. other additives. This paste is `screen-printeda onto the
Techniques for growing silicon in the form of ribbons cell surface in the desired pattern using a simple process
from the melt have also been developed. These have the similar to that used to print patterns onto T-shirts. After
advantage that no slicing is required (Green, 1982). printing, the paste is dried and heated at high temper-
ature, leaving the metal particles agglomerated together.
4.2. From wafers to cells A very thin layer of insulating material is sometimes
added to the top cell surface as an antire#ection coating,
Some manufacturers make their own wafers while similar to the coating used on high-quality camera lenses.
others buy them from wafer suppliers. In either case, the Such coatings are always used for `multicrystallinea wa-
"rst step in processing a wafer into a cell is to etch the fers, since the pyramidal-texturing approach is not e!ec-
wafer surface with chemicals to remove damage from the tive for such wafers and this alternative approach is
slicing step. essential to control re#ection.
The surface of crystalline wafers is then etched again All the equipment required for this process is available
using a chemical that etches at di!erent rates in di!erent `o! the shelfa from the microelectronics industry (the
directions through the silicon crystal. This leaves features `hybrida or `thick "lma industry sector). This, combined
992 M.A. Green / Energy Policy 28 (2000) 989}998

with the small number of processing steps, has made this

sequence almost universally used by solar cell manufac-
turers. However, the penalty paid for its simplicity and
convenience is lower cell performance than is inherently
available from the staring wafers. The screen-printed
silver pastes are also quite expensive.
After fabrication, the cells are soldered together and
packaged under a sheet of glass into a weatherproof
package known as a module. Generally, 36 cells are
packaged into a module, since this is the required number
to generate enough voltage to allow charging of a lead-
acid battery.

4.3. Improved technology

Fig. 6. Buried contact solar cell.
The previous `screen-printeda sequence was developed
in the early 1970s and produces cells with performance
typical of this era. As shown in Fig. 5, there has been over
50% relative improvement in laboratory silicon cell per-
formance since that time. Only 20% of the cost of produ-
cing a standard module is due to cell processing costs.
Both these and the remaining 80% of the costs are able to
be reduced per unit power output by increasing cell
e$ciency (since these costs are related to cell and module
area). This means that cell processing costs per unit area
can double if this results in 20% improvement in cell
performance (or triple for 40% improvement). Given that
the screen-printing sequence is simple but still reasonably
expensive, due to the costs of the required metal pastes,
such a trade-o! is feasible.
The most successful commercialisation to data of an
improved, higher performance sequence has involved the Fig. 7. Heterojunction with intrinsic thin layer (HIT) cell.
buried contact cell of Fig. 6. This cell o!ers 20}30%
improvement in output with virtually no increase in cell
processing costs (Bruton et al., 1997). The overall relative
cost advantage of the technology is therefore close to its comes from better quality surface regions that allow
relative e$ciency advantage. The improved performance much better response to blue light, absorbed close to the
surface, and much lower electrical resistance and optical
losses due to the improved top contacting scheme.
A more recently commercialised approach is the Het-
erojunction with intrinsic thin layer (HIT) cell of Fig. 7.
This combines crystalline silicon technology with that of
amorphous silicon, discussed below. The HIT cell would
be expected to give some of the improvements of the
buried contact sequence in the areas mentioned, al-
though not to the same extent. The rear processing of the
cell is improved compared to the buried contact sequence
and the cell responds to light from both directions, a fea-
ture that can be used to advantage in some applications.
For the processing of multicrystalline wafers, the use of
silicon nitride as an antire#ection coating has advantages
known for some time. These arise from the presence of
hydrogen in this layer, arising from its presence in one of
the source gases (SiH ) used in the deposition process.

Fig. 5. Evolution of silicon laboratory cell e$ciency. The dark squares The hydrogen di!uses into the silicon and is e!ective in
show improvements demonstrated by the author's group. reducing detrimental activities at the boundaries between
 M.A. Green / Energy Policy 28 (2000) 989}998 993

the individual grains in multicrystalline material. The use

of nitride is expected to be more widely adopted in the
future for such multicrystalline material, in particular.

5. Thin-5lm solar cells

5.1. Thin-xlm advantages

The potential for on-going cost reductions is the key

reason for con"dence in a signi"cant role for photovol-
taics in the future.
Rather than the wafer-based technology of the Fig. 8. Thin-"lm approach.
previous section, the future belongs to `thin-"lma
technology. In this approach (Fig. 8), thin layers of from perfection means that the perfect ordering over
semiconductor material are deposited onto a supporting large distances is no longer possible. Amorphous mater-
substrate, or superstrate, such as a large sheet of glass. ial has much lower electronic quality, as a consequence,
Typically, less than a micron thickness of semiconductor and originally was not thought suitable for solar cells.
material is required, 100}1000 times less than the thick- However, producing amorphous silicon by decomposing
ness of a silicon wafer. the gas, silane (SiH ), at low temperature, changed this

Reduced material use with associated reduced costs is opinion. It was found that hydrogen from the source was
a key advantage. Another is that the unit of production, incorporated into the cell in large quantities (about 10%
instead of being a relatively small silicon wafer, becomes by volume), improving the material quality. Hydro-
much larger, for example, as large as a conveniently genated amorphous silicon cells very quickly found use
handled sheet of glass might be. This reduces manufac- in small consumer products such as solar calculators and
turing costs. digital watches, their main use so far.
Silicon is one of the few semiconductors inexpensive The problem with outdoor use is that some of the
enough to be used to make solar cells from self-support- bene"cial e!ect of hydrogen becomes undone under
ing wafers. However, in thin-"lm form, due to the bright sunshine and the cell performance degrades. Ini-
reduced material requirements, virtually any semicon- tially, there was hope that some simple material-related
ductor can be used. Since semiconductors can be formed solution could be found. When this did not happen, the
not only by elemental atoms such as silicon, but also only alternative was to design around it. Cells had to be
from compounds and alloys involving multiple elements, developed that could work well with material of de-
there is essentially an in"nite number of semiconductors graded quality, rather than of the starting quality.
from which to choose (Berger, 1997).
At present, solar cells made from "ve di!erent thin-"lm 5.2.2. Amorphous cell design
technologies are either available commercially, or close Since the amorphous silicon quality is much poorer
to being so. Over the coming decade, one of these is than crystalline silicon, a di!erent cell design approach is
expected to establish its superiority and attract invest- required. The most active part of a p}n junction solar cell
ment in major manufacturing facilities that will sustain is right at the junction between the p- and n-type region
the downward pressure on cell prices. As each of these of the cell (Fig. 1). This is due to the presence of an electric
thin-"lm technologies has its own strengths and weak- "eld at this junction.
nesses, the likely outcome is not clear at present (Zweibel With amorphous silicon cell design, the aim is to
and Green, 2000; Kazmerski, 1997; Partain, 1995). stretch out the extent of this junction region as far as
possible so almost all the cell is junction. This is done by
5.2. Amorphous silicon alloy cells having the p- and n-type doped regions very thin, with an
undoped region between them. The strength of the elec-
5.2.1. Properties tric "eld established in this undoped region is nearly
Given its success in wafer form, silicon is an obvious constant and depends on this region's width. The poorer
choice for development as a thin-"lm cell. Early attempts the quality of amorphous silicon, the stronger the "eld
to make thin-"lm polycrystalline silicon cells did not needs to be for the device to work well and hence the
meet with much success. However, starting from the thinner the device needs to be.
mid-1970s, very rapid progress was made with silicon in For degraded material, it turns out that the cell needs
`amorphousa form. In amorphous silicon, the atoms are to be thinner than the thickness required to absorb all the
connected to neighbours in much the same way as in the useable incident sunlight. The way around this is to stack
crystalline material but accumulation of small deviations several cells on top of one another so that light not
994 M.A. Green / Energy Policy 28 (2000) 989}998

absorbed by an upper one passes through to an underly- mean that modules would have to be carefully disposed
ing cell. of or, preferably, recycled after their useful life was "nish-
As discussed in Section 6, this works best if the mater- ed. However, there may be some problems in gaining
ial in the underlying cells is varied so that each responds market acceptance in what is likely to be mainly
progressively better to the redder light that is transmitted a `greena market over coming years.
to it. By alloying silicon with germanium, a material There are also only limited known resources of tellu-
chemically similar to silicon but much scarcer, this is rium (Zweibel and Green, 2000). If all identi"ed reserves
readily achieved. were converted into cells of the present designs overnight,
The best commercial amorphous silicon cells presently they could generate 10% of the world's present electricity
use three cells stacked on top of one another, with pro- use (a steadily decreasing percentage inde"nitely, if re-
gressively more germanium in the bottom two (Zweibel and cycled at end of life).
Green, 2000; Kazmerski, 1997; Pantain, 1995). Each cell is An even more promising technology at the moment, in
very thin, only 100}200 nm thick. This ensures reasonable the author's opinion, is one based on the ternary com-
stability (only about 15% degradation in output when pound, copper indium diselenide (CuInSe ). As if three

exposed to bright sunlight). However, the stabilised e$cien- elements were not enough, this compound is often
cy is quite poor, only 6}7% for the best commercial mod- alloyed with copper gallium diselenide (CuGaSe ) and

ules, according to manufacturers' data sheets. copper indium disulphide (CuInS ), giving material with

This low e$ciency, even with the sophisticated cell up to "ve elements involved (Kazmerski, 1997). The n-
design involved, is expected to make it di$cult for this type layer in these devices consists of a layer of cadmium
technology to be competitive in the long term. However, sulphide, as in the previous cadmium telluride cells. An
the low temperatures involved in making these cells alternative for this layer is being sought, to eliminate the
mean that they can be deposited onto low-temperature toxic cadmium.
substrates such as plastics. This makes them especially Small area laboratory cells have demonstrated e$cien-
suitable for consumer products. cy close to 19%, despite the "ne-grained polycrystalline
material used (Zweibel and Green, 2000; Kazmerski,
5.3. Thin-xlm, polycrystalline compound semiconductors 1997; Partain, 1995). Modules of this material are now
commercially available in small volumes with e$ciency
Many semiconductors made from compounds can ab- up to 12% demonstrated in pilot production. This is not
sorb light more strongly than the elemental semiconduc- far behind what is achieved with standard crystalline
tors, silicon and germanium, for reasons that are well silicon wafer modules.
understood but quite subtle (Kazmerski, 1997) (silicon Apart from the use of cadmium and even more limited
and germanium are `indirecta rather than `directa ban- known resources of indium than tellurium, an often
dgap materials). This means compound semiconductor quoted limitation of this technology is `manufacturabil-
cells can be thin but still operate e$ciently. Most com- itya. This is often interpreted as meaning it is di$cult to
pound semiconductors, when formed in polycrystalline diagnose problems in production with this material, since
form, have poor electronic properties due to highly del- the di!erence between good and bad material is not
eterious activity at grain boundaries between individual su$ciently well understood to allow di!erentiation and
crystalline grains in the material. A small number main- control during the various manufacturing steps.
tain good performance in polycrystalline form for rea-
sons that are not usually well understood. These are the 5.4. Thin-xlm polycrystalline silicon cells
candidates for thin-"lm polycrystalline compound
semiconductor solar cells. As previously mentioned, silicon is a weak absorber of
One such semiconductor is the compound cadmium sunlight compared to some compound semiconductors
telluride (CdTe). Technically, it is an ideal material, giv- and even to hydrogenated amorphous silicon. Early at-
ing properties suitable for making reasonable solar cells tempts to develop thin-"lm solar cells based on the poly-
even with relatively crude material deposition ap- crystalline silicon did not give encouraging results since
proaches (such as electrodeposition, chemical spraying, the silicon layers had to be quite thick to absorb most of
and so on). The junction in these cells is again between p- the available light.
and n-type material, but for the latter, a di!erent com- However, in early 1980s, understanding of how e!ec-
pound semiconductor, cadmium sulphide, gives best re- tively a semiconductor can trap weakly absorbed light
sults. CdTe cells have been used mainly in pocket into its volume greatly increased (Green, 1995). Due to
calculators to date, but large area, moderate performance the optical properties of semiconductors, particularly
modules have also been demonstrated (Kazmerski, 1997). their high refractive index, cells can trap light very e!ec-
The main concern with this technology is the toxicity of tively if the light direction is randomised, such as by
the materials involved, even though very small amounts striking a rough surface, once it is inside the cell. Optical-
are used in the modules. At the very least, this would ly a cell can appear about 50 times thicker than its actual
 M.A. Green / Energy Policy 28 (2000) 989}998 995

Fig. 9. One approach for high-temperature preparation of silicon thin-

"lm solar cells.
Fig. 10. Prototype thin-"lm polycrystalline silicon-on-glass module
(photograph courtesy of Paci"c Solar Pty. Ltd., Sydney).
thickness if this occurs. Such `light trappinga removes
the weak absorption disadvantage of silicon.
Work on polycrystalline thin-"lm solar cells is pro-
ceeding in two areas. A variety of `high temperaturea
approaches such as suggested by Fig. 9 are being ex-
plored. There generally involve either high-temperature
deposition of silicon onto a substrate or melting the
silicon after deposition, to obtain large grain size in the
"nal "lm. Although preparation details are sketchy, the
thin-"lm silicon product available from the US company,
Astropower, is the most developed representative of this
class of approach. In this case, the silicon is deposited
onto an expansion-matched ceramic substrate. The "nal
material consists of millimetre sized grains and is similar
in appearance to multicrystalline silicon wafers. Small
area cell performance in the 16}17% range has been
demonstrated (Kazmerski, 1997), similar to that from
such cells on moderate-quality multicrystalline silicon.
The performance of large arrays of such cells has also
been similar.
The other type of approach is a `low-temperaturea
approach, generally based on amorphous silicon techno-
logy. One approach is to deposit the silicon in amorph- Fig. 11. Nanocrystalline dye sensitised solar cell.
ous form and then crystallise it by heating for prolonged
periods at intermediate temperatures. This `solid-phase based on an amorphous silicon precursor (Paci"c Solar,
crystallisationa approach has produced cells of quite 1999).
reasonable performance (Baba et al., 1995). Another ap-
proach has involved changing the amorphous silicon 5.5. Nanocrystalline dye cells
deposition conditions, to produce a nanocrystalline
phase of silicon. The potential of this approach was A completely di!erent thin-"lm approach is based on
highlighted by early results with the `micromorpha solar the use of ruthenium-based organic dyes [8,9]. Dye mol-
cell (Wyrsch et al., 1998). More recently, cell e$ciency ecules are coated onto a porous network to titanium
above 10% has been con"rmed with this approach dioxide particles and immersed in an electrolyte (Fig. 11).
(Yamamoto et al., 1999). There are plans to use such cells In a process bearing some relationship to photosynthesis,
as the lower cell in a tandem con"guration with an light absorbed by the dye photoexcites a electron into the
amorphous silicon upper cell, with commercial product titanium dioxide which completes the circuit through the
targeted for 2002 (Yamamoto et al., 1999). Also targeted external load and the electrolyte.
for commercialisation in the same timeframe is a poly- Interestingly, the dye only absorbs a band of photon
crystalline silicon on glass product shown in Fig. 10, energies, rather than all photons of energy above the
996 M.A. Green / Energy Policy 28 (2000) 989}998

Fig. 12. Multigap-cell concepts: (a) spectrum-splitting approach; (b) tandem-cell approach.

bandgap, as is the case for normal cells. This gives rise to highest bandgap cells uppermost. In the limit of an in"-
prospects for unique devices, such as transparent win- nite number of such cells, the limit on conversion e$cien-
dows that convert the infrared wavelengths while letting cy is increased to 70%, and further to 87% if focussed
visible light through. sunlight is used (Marti and Araujo, 1996) (the di!use
component of sunlight would have to be wasted in the
6. Longer-term developments latter case, giving a lower e!ective e$ciency in practice).
Three-cell stacks are now in production for high-perfor-
As outlined above, photovoltaic technology is on the mance cells for spacecraft and to improve the stability of
verge of a major transition from `"rst-generationa silicon amorphous silicon cells (Section 5.2.2).
wafer-based technologies to a `second generationa thin- Are there other ways of obtaining such performance
"lm product. This transition should proceed at an accel- that may be simpler in practice? A few other approaches
erating pace over the coming decade. Once this transition have been suggested. These should become experi-
is complete, is there anywhere else for the technology to mentally more feasible with on-going improvements in
go? material science over the coming decades.
Although there will be evolutionary improvements in One class of device is based on `hot carriersa (Ross and
both "rst- and second-generation cell performance, the Nozik, 1982). Although the solar cells stay cool, the
e$ciency of standard cells is restricted to less than 33% charge carriers in such devices ideally reach distributions
by quite fundamental considerations (Marti and Araujo, typical of much higher temperatures by avoiding energy
1996). Basically, standard cells are quantum converters, loss by collisions with the atoms making up the cell,
ideally converting one sunlight photon to an electron in a fundamental loss mechanism in standard cells. E$cien-
the load. This alone limits e$ciency to about 44%. An cies similar to the in"nite tandem cell are possible in this
additional loss arises since the cell can supply this current case. Another approach is to use not just excitations
at a voltage somewhat less than the potential corre- between two bands of carrier energy as in a normal cell,
sponding to its energy bandgap. but excitations in materials designed to have more than
However, the thermodynamic limit on the conversion two bands of allowed energy participating in the process
of sunlight into electricity is a more impressive 93%. Is it (Luque and Marti, 1997). Again, e$ciencies approaching
possible for solar cells to come closer to this limit? the in"nite tandem cell case are feasible, in principle.
The answer seems to be `yesa, although new ideas are
probably required to make this a reality. One idea that is
well established is based on splitting sunlight into di!er- 7. Policy implications
ent wavelength bands and sending to cells of bandgap
optimised for each band (Fig. 12(a)). Fortunately, this The history of the evolution of solar cell costs versus
splitting can be achieved more simply merely by stacking accumulated production volume is shown in Fig. 13
cells on top of one another as in Fig. 12(b), with the (Green, 2000). The straight line "tted to the data shows
 M.A. Green / Energy Policy 28 (2000) 989}998 997

Note to be outdone, the US Government announced

a `million roofa program in mid-1997, targeting this
number of systems also for 2010. At the end of 1997, the
European Union also agreed to a similar target for 2010,
with half the systems to be installed inside Europe and
half outside. Since then, individual countries such as the
Netherlands, Germany, Italy and Australia have an-
nounced targets for photovoltaic roofs.
These programs generally involve a three-way subsidy
between the government, home-owner, and the local util-
ity. Government subsidies are usually large in the early
stages of the program (50}70%) but reduce with time as
system costs drop (currently 33}1/3% in the Japanese
program). This allows more systems to be subsidised for
Fig. 13. Average selling price of photovoltaic modules versus total the same sum as the programs develop.
accumulated shipments (the straight line corresponds to an 80% `learn- Even with this government subsidy, it would still be
ing curvea). cheaper for the home-owner to buy power from the grid.
Those participating are usually attracted by the oppor-
tunity to do something positive for the environment,
rather than waiting for governments or utilities to take
an 80% learning curve, i.e. the behaviour expected if cell concerted action.
costs decreased to 80% of their original for each doubl- The systems operate without energy storage by selling
ing in production volume. This straight line can be seen excess power generated in the daytime to the utility
to be quite a good "t to the data actually observed. operating the local grid. At night, power #ows in the
Almost all the data shown apply to wafer based techno- opposite direction. In some parts of the world, utilities
logy, with thin-"lm technology yet to make its impact. are required by legislation to pay the same for electricity
With on-going improvements in cost expected from supplied by a photovoltaic system as they charge for
both second- and third-generation technology, the type electricity sent the other direction. Other utilities do this
of `learninga behaviour shown in Fig. 13 is expected to voluntarily. By adopting such `net meteringa. utilities
be maintained well into the future, or even accelerated also help improve the viability of these systems by e!ec-
when new technology makes its appearance. tively providing free storage for the solar generated elec-
If this scenario is accepted, the key to reducing photo- tricity. Utilities are already well placed to provide such
voltaic costs lies in increasing the quantity sold. This storage, since `spinning reservesa and stand-by units are
encourages both e$ciencies in production and the invest- always available to meet a sudden increase in demand or
ment in the development and commercialisation of new the unscheduled outage of a large, conventional power
generations of technology. However, at present costs, station. It is not until photovoltaics were supplying more
photovoltaics are competitive mainly in remote areas. than about 15% of the electricity in any particular net-
Markets here are small in the developed world and po- work that any signi"cant change in operation of the grid
tentially enormous in the developing world, but largely network would be required.
inaccessible due to lack of access to "nance. As a result of these residential programs and others
Fortunately, a market-pull mechanism seems to have likely to be stimulated by them, there is expected to be
been recently put into place that seems likely to provide a rapidly increasing demand for photovoltaics over the
healthy growth and cost reductions over the coming coming decade. This should continue driving down the
decade. This is by the subsidisation of rooftop mounted costs of photovoltaic systems as suggested by Fig. 13, to
systems in urban areas of the developed world (Green, the level where systems should be attractive with minimal
2000). subsidies. It is not di$cult to imagine that, at some stage
Interest in this rooftop application started in the USA in the future, all new housing might require its own
in the late 1970s. Japan took over the running in the photovoltaic system to make it greenhouse neutral. Re-
mid-1980s with the construction of a test bed for over 200 duced costs will progressively open up other applications
residential systems on Rokko Island in 1986. After sev- such as large centralised photovoltaic power stations
eral years of evaluating technical issues related to their providing bulk electricity to the grid in direct competitive
grid connection, a subsidised installation program was to conventional, but less benign, large-scale generators.
launched by the Japanese government in 1993. By the If this prognosis is correct, appropriate policies are
end of 1999, over 20,000 rooftop photovoltaic systems those that encourage the success of these urban rooftop
had been installed in urban areas of Japan, with a mass- photovoltaic systems, while simultaneously encouraging
ive 1.5 million being targeted for 2010. the development and commercialisation of improved
998 M.A. Green / Energy Policy 28 (2000) 989}998

photovoltaic technology. In schemes where premiums for References

sustainable energy generation are available, it is impor-
tant for photovoltaic development that `portfolioa rather Baba, T., Shimam, M., Matsuyama, T., Tsuge, S., Wakisaka, K., Tsuda,
than `least-costa strategies be adopted. In portfolio ap- S., 1995. 9.2% e$ciency thin-"lm polycrystalline silicon solar cell by
a novel solid phase crystallization method. Conference Proceedings,
proaches, some portion of the funds generated by the 13th European Photovoltaic Solar Energy Conference, Nice, Octo-
premium would be assigned to photovoltaically gener- ber, pp. 1708}1711.
ated electricity. Even though it is presently not directly Berger, L.I., 1997. Semiconductor Materials. CRC Press, Boca Raton,
competitive with electricity from other sustainable sour- FL.
ces such as wind or biomass, this strategy would recog- Bruton, T.M., Luthardt, G., Rasch, K.-D., Royl, K., Dorrity, I.A.,
Garrard, B., Teale, L., Alonso, J., Ugalde, U., Declerq, K., Nijs, J.,
nise the greater long-term potential of photovoltaics to Szlufcik, J., Rauber, A., Wettling, W., Vallera, A., 1997. A study of
impact energy generation and use. the manufacture at 500 MWp p.a. of crystalline silicon photovoltaic
modules. Conference Record, 14th European Photovoltaic Solar
Energy Conference, Barcelona, July/July, pp. 11}16.
8. Conclusion Green, M.A., 1982. SOLAR CELLS: Operating Principles, Technology
and System Applications. Prentice-Hall, New Jersey.
Green, M.A., 1995. Silicon Solar Cells: Advanced Principles and Prac-
As we enter the new millennium, photovoltaics are tice. Bridge Printery, Sydney.
poised to make a more signi"cant impact upon energy Green, M.A., 2000. Power to the People. UNSW Press, Sydney.
use. Kazmerski, L., 1997. Photovoltaics: a review of cell and module
Although still dominating the marketplace, `"rst gen- technologies. Renewable and Sustainable Energy Reviews 1,
erationa technology based on silicon wafers is starting to 71}170.
Luque, A., Marti, A., 1997. Increasing the e$ciency of ideal solar cells
be challenged by `second generationa thin-"lm techno- by photon induced transitions at intermediate levels. Physical Re-
logy. This has the advantage of much lower material view Letters 78, 5014}5017.
costs and of being better suited to high-volume manufac- Marti, A., Araujo, G.L., 1996. Limiting e$ciencies for photovoltaic
ture. energy conversion in multigap systems. Solar Energy Materials and
There is also scope for a third generation of techno- Solar Cells 43, 203}222.
Paci"c Solar Pty. Ltd., 1999. Annual Review.
logy, based on principles not yet fully developed, o!ering Partain, L. (Ed.), 1995. Solar Cells and Their Applications. Wiley, New
prospects for signi"cantly enhanced e$ciency at some York.
stage further into the future. Accompanying the rapidly Perlin, J., 1999. From Space to Earth: The Story of Solar Electricity.
growing demand for photovoltaics and such on-going Aatec Publications, Ann Arbor, MI.
improvements in technology, photovoltaic costs have Riordan, M., Hoddeson, L., 1997. Crystal Fire. Norton, New York.
Ross, R., Nozik, A.J., 1982. E$ciency of hot-carrier solar energy con-
been steadily decreasing and are expected to do so well verters. Journal of Applied Physics 53, 2813}3818.
into the future. Stachel, J. (Ed.), 1998. Einstein's Miraculous Year. Princeton University
The application driving market growth at the present Press, New York.
is in urban residential rooftop systems. This is expected Wyrsch, N., Torres, P., Goetz, M., Dubail, S., Feitknecht, L., Cuperus,
to continue to be the most important commercial ap- J., Shah, A., Rech, B., Kluth, O., Wieder, S., Vetterl, O., Stiebig, H.,
Beneking, C., Wagner, H., 1998. Development of inverted microm-
plication over the coming decade, although dependent on orph solar cells. Conference Proceedings, 2nd World Conference on
subsidies for its viability. The required level of subsidy is Photovoltaic Solar Energy Conversion, Vienna, July, pp. 467}471.
expected to decreases over the coming decade, with this Yamamoto, K., Yoshimi, M., Tawada, Y., Okamoto, Y., Nakajima, A.,
application fully economic by the end of the decade. 1999. Cost e!ective and high performance thin "lm Si solar cell
Policies that support this market pull mechanism while towards the 21st century. Tech. Digest, 11th International Photo-
voltaic Science and Engineering Conference, Sapporo, September,
encouraging the development and commercialisation of pp. 225}228.
improved photovoltaic technology are believed to be the Zweibel, K., Green, M.A. (Eds.), 2000. Millennium Issue, Prog. Photo-
most appropriate. voltaics 8, 1}195.