Subscriber access provided by Universiteit Utrecht
Environmental Science & Technology is published by the American Chemical
Society. 1155 Sixteenth Street N.W., Washington, DC 20036
Emissions from Photovoltaic Life Cycles
Vasilis M. Fthenakis, Hyung Chul Kim, and Erik Alsema
Environ. Sci. Technol., 2008, 42 (6), 2168-2174 • DOI: 10.1021/es071763q • Publication Date (Web): 06 February 2008
Downloaded from http://pubs.acs.org on February 2, 2009
More About This Article
Additional resources and features associated with this article are available within the HTML version:
Links to the 1 articles that cite this article, as of the time of this article download
Access to high resolution figures
Links to articles and content related to this article
Copyright permission to reproduce figures and/or text from this article
Emissions from Photovoltaic Life
V A S I L I S M . F T H E N A K I S , *, † , ‡
H Y U N G C H U L K I M ,†A N D E R I K A L S E M A§
PV Environmental Research Center, Brookhaven National
Laboratory, Upton, New York, Center for Life Cycle Analysis,
Columbia University, New York, New York, and Copernicus
Institute of Sustainable Development, Utrecht University,
Heidelberglaan 2, 3584 CS Utrecht, The Netherlands
Received July 17, 2007. Revised manuscript received
December 19, 2007. Accepted January 4, 2008.
Photovoltaic (PV) technologies have shown remarkable
progress recently in terms of annual production capacity and
life cycle environmental performances, which necessitate
gas emissions, criteria pollutant emissions, and heavy metal
emissions from four types of major commercial PV systems:
multicrystalline silicon, monocrystalline silicon, ribbon silicon,
and thin-film cadmium telluride. Life-cycle emissions were
determined by employing average electricity mixtures in Europe
and the United States during the materials and module
production for each PV system. Among the current vintage of
PV technologies, thin-film cadmium telluride (CdTe) PV
emits the least amount of harmful air emissions as it requires
the least amount of energy during the module production.
However, the differences in the emissions between different
PV technologies are very small in comparison to the emissions
from conventional energy technologies that PV could displace.
As a part of prospective analysis, the effect of PV breeder
was investigated. Overall, all PV technologies generate far less
life-cycle air emissions per GWh than conventional fossil-fuel-
based electricity generation technologies. At least 89% of
air emissions associated with electricity generation could be
prevented if electricity from photovoltaics displaces electricity
from the grid.
The production of energy by burning fossil fuels releases
many pollutants and carbon dioxide to the environment.
Indeed, all anthropogenic means of generating energy,
including solar electric, create pollutants when their entire
life cycle is taken into account. Life-cycle emissions result
smelting, production, and manufacturing facilities. These
emissions differ in different countries, depending on that
country’s mixture in the electricity grid, and the various
methods of material/fuel processing.
Previous life-cycle studies reported a wide range of
primary energy consumption for PV modules. Alsema
in their estimates of primary energy consumption. In those
off-spec products of electronic-grade silicon and various
for each grade of silicon; also solar cells were much thicker
than the current ones (1). Meijer et al. evaluated 270-µm-
thick Si PV with 14.5% cell efficiency fabricated from
electronic-grade high-purity silicon (2). They estimated
energy payback time (EPBT, the time it takes for a photo-
voltaic (PV) system to generate an amount of energy equal
to that used in its production) for the module only of 3.5
kWh/m2/yr). Jungbluth reported the life-cycle metrics of
various PV systems (2000 vintage) under average insolation
in Switzerland (1100 kWh/m2/yr) (3). He estimated green-
house gas (GHG) emissions in the range of 39–110 g CO2-
equiv/kWh and EPBT of 3–6 years.
There are a few life-cycle studies of thin-film PV tech-
by Keoleian and Lewis (4–7). The CdTe studies were based
on R&D data and hypothetical production lines since
presented that the EPBT of the frameless module of double-
junction amorphous silicon with 5% efficiency is 4.6 years
in Detroit, MI and 2.2 years in Phoenix, AZ. This study is not
different photovoltaic rooftop installations, namely ribbon-
Si, multicrystalline Si (multi- or mc-Si), monocrystalline Si,
and thin-film CdTe systems. Their corresponding EPBTs,
under the average Southern European insolation of 1700
kWh/m2/yr, were 1.7, 2.2, 2.7, and 1.1 years. The EPBT of
CdTe PV was much lower than that of the other systems
although its electrical-conversion efficiency was the lowest
in the group (i.e., 9% for CdTe vs 11.5% for ribbon, 13.2% for
multicrystalline Si, and 14% for monocrystalline Si).
facet in assuring the acceptability of any particular one. In
this paper, we update the greenhouse gas emissions, and
present the first comprehensive assessment of emissions of
criteria pollutants and heavy metals, from cradle to gate, of
four commercial PV systems based on the most recent data
(i.e., 2004–2006): ribbon-silicon, multicrystalline silicon,
monocrystalline silicon, and thin-film cadmium telluride.
These are, for the most part, indirect emissions associated
in the life cycle of photovoltaics. Direct emissions of heavy
are also included, whereas liquid and solid waste are for the
most part being recycled, and were not considered in this
study. The choice of electricity and fuel sources plays an
important role in determining the total emissions. In this
context, we investigated a scenario of PV breeding where PV
supplies a fraction of the electricity required in manufacturing.
* Corresponding author tel: 631-344-2830; fax: 631-344-7650;
†Brookhaven National Laboratory.
Environ. Sci. Technol. 2008, 42, 2168–2174
21689ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 200810.1021/es071763q CCC: $40.75
2008 American Chemical Society
Published on Web 02/06/2008
2. Life Cycle of Silicon and Thin-Film CdTe
The life cycle of PV systems starts with the mining of quartz
sand (silicon PV) or metal ore (CdTe PV). The silica in the
quartz sand is reduced in an arc furnace to obtain metal-
lurgical grade silicon, which has to be purified further into
“electronic grade” or “solar grade” silicon (Figure 1). The
hydrogen (H2) gases is heated to 1100–1200 °C for growing
silicon rods, or the “modified Siemens” process in which
silane (SiH4) and hydrogen (H2) gases are heated to ∼800 °C
for the same, resulting in lower energy consumption (9).
A detailed life-cycle inventory (LCI) of crystalline silicon
modules for polycrystalline silicon feedstock purification,
crystallization, wafering, cell processing, and module as-
sembly with the current status of technology (2004-early
2005) was recently completed within the “CrystalClear”
European Commission project (10). The sources of LCI data
for this project include 11 commercial European and U.S.
photovoltaic module manufacturing companies supple-
mented by numbers from the literature. Depending on the
follows, along with tailoring the wafers. Cell manufacturing
and subsequent module assembly, which is virtually equal
for each module, concludes their life cycle (Figure 1). Each
module assembly typically consists of seventy-two 0.125 m
× 0.125 m solar cells with silver contacts in front and back
sides. Ethylene-vinyl acetate and glass sheets encapsulate
the PV module to provide protection from the physical
elements during operation. Crystalline silicon modules
typically have aluminum frames for additional strength and
Fthenakis (11) described the material flows of cadmium
(Cd) and emissions from the entire life-cycle stages of
cadmium telluride (CdTe) PV. The life cycle starts with the
namely particulates collected in electrostatic precipitation
and bag house and slimes collected from Zn electrolyte
during electrolytic copper refining, which also contain Cu,
Se, and other metals. The slimes are treated with dilute
sulfuric acid to extract Te. After cementation with copper,
CuTe is leached with caustic soda to produce a sodium-
telluride solution that is used as the feed for Te and TeO2.
Cadmium is further processed and purified either through
leaching and vacuum-distillation, or through electrolytic
distillation, to produce the 99.999% purity required for the
synthesis of CdTe. Tellurium is also further purified by the
above-mentioned methods. CdTe is produced from Cd and
Te via proprietary processes.
We analyzed in detail the life-cycle inventory of CdTe PV
obtained at a CdTe PV manufacturing plant in Perrysburg,
are 1.2 m × 0.6 m with an electricity conversion efficiency
of 9% (The efficiency of the modules produced by this plant
(CdTe) absorber layer and cadmium sulfide (CdS) window
layer in First Solar’s production scheme are laid down by
vapor transport deposition (VTD), based on subliming the
powders and condensing the vapors on glass substrates. A
stream of inert carrier gas guides the sublimed dense vapor
cloud to deposit the films on glass substrates at 500–600 °C
with a growth rate over 1 µm/s (13). Depositing layers of
common metals followed by series of scribing and heat
metals/elements are used in the back contact layers.
Table S1 in the Supporting Information of this paper
presents the material compositions of silicon- and CdTe-
particularly of the frameless CdTe module where two panes
eliminates the need for an aluminum frame which accounts
The use of CdTe powder per m2of thin film CdTe module
of cell materials of the former is ∼3 µm compared with
ribbon-, and multi-Si PVs, and (b) thinfilm CdTe PVs. Detailed descriptions of the life cycles are available elsewhere (10, 12, 13).
1. Simplified process-flow diagrams from mining to system manufacturing stages, namely cradle-to-gate for (a) mono-,
VOL. 42, NO. 6, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2169
270–300 µm for silicon modules. The CdTe module also
requires smaller amounts of gases, liquids, and other
consumables than does a silicon module.
3. GHG and Criteria Pollutant Emissions
We estimate the emissions GHG, SO2, and NOxduring the
PV life cycles. Together with the heavy metal emissions
assessed later in this paper, these emissions comprise the
main hazards to the environment and human health from
energy use and materials extraction during the PV life cycle.
These emissions are normalized by the electricity generated
during the life cycle of PV. The major parameters for the life
cycle, i.e., lifetime electricity generation of a PV system,
include conversion efficiency (E), solar insolation (I), per-
formance ratio (PR), and lifetime (L). The total lifetime
electricity generation (G) per m2of PV module is calculated
as follows: G ) E × I × PR × L. We consistently use, for our
own analysis, the Southern European average insolation of
1700 kWh/m2/yr, a performance ratio of 0.8, and a lifetime
of 30 years.
Alsema and de Wild report that the GHG emissions of Si
kWh range, with an EPBT of 1.7–2.7 years for a rooftop
application under Southern European insolation of 1700
geographically specific production of Si (Figure 2, Case 1).
Fthenakis and Kim (12) recently investigated the GHG
emissions and EPBT of CdTe PV modules, based on U.S.
production and insolation conditions (insolation ) 1800
FIGURE 2. Life-cycle emissions from silicon and CdTe PV modules. BOS is the Balance of System (i.e., module supports, cabling,
and power conditioning). Conditions: ground-mounted systems, Southern European insolation, 1700 kWh/m2/yr, performance ratio of
0.8, and lifetime of 30 years. Case 1: current electricity mixture in Si production-CrystalClear project and Ecoinvent database. Case
2: Union of the Co-ordination of Transmission of Electricity (UCTE) grid mixture and Ecoinvent database. Case 3: U.S. grid mixture
and Franklin database.
2170 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 2008
kWh/m2/yr; performance ratio 0.8; lifetime 30 years). Their
estimates were 24 g CO2-equiv/kWh of GHG emissions, and
1.1 yrs of EPBT for ground-mounted installations. In the
them for constant solar irradiation, performance ratio, and
electricity mixture. Figure 2, Case 2 shows emissions cor-
responding to upstream electricity for the average grid
of Transmission of Electricity, UCTE), and Figure 2, Case 3
shows the same for the average U.S. grid mixture. The most
grid and Franklin for the U.S. grid mix, are employed for the
energy and emission factors (14, 15).
The production of polycrystalline silicon is the most
energy-consuming stage of the silicon module’s life cycle; it
accounts for 45% of the total primary energy usage in the
multi-Si module life cycle (10). Electricity demand during
emissions from Case 1 which is based on the electricity mix
of CrystalClear project, are lower than those from Cases 2
and 3, mainly because of the higher portion of hydropower
used by the producers of polycrystalline solar grade silicon
(Table S2 in the Supporting Information). For the same
reason, the emission estimates based on the UCTE grid
mixture (Case 2) typically are lower than those based on the
U.S. grid (Case 3) (i.e., the former is a cleaner fuel mix).
The life-cycle emissions from mono-Si PV are greater than
substantial energy during the ingot growing process (by
Czochralski crystal pulling).
4. Heavy Metal Emissions
We followed the direct and indirect (due to energy use)
emissions of heavy metals (arsenic, cadmium, chromium,
lead, mercury, and nickel) during the life cycles of the four
PV technologies we studied. The CdTe PV can emit Cd both
directly and indirectly whereas the crystalline Si PV stages
would emit such only indirectly.
4.1. Direct Cd Emissions. Fthenakis (11) compiled the
PV modules based on 30 years of module lifetime, 9%
efficiency, and the average U.S. insolation of 1800 kWh/m2/
smelting, and purification of the element and the synthesis
of CdTe are 0.015 g/GWh. The total direct emissions of
cadmium during module manufacturing are 0.004 g/GWh
(11). Emissions during accidental releases (e.g., fires) are
extremely small, if any. Such emissions could add to the
total of 0.02 g/GWh. The latter have been investigated
microprobes (16). Cd emissions from the life cycle of CdTe
modules (Table S3 in the Supporting Information) are
estimated to be 90–300 times lower than those from coal
power plants, which are 2–7 g Cd/GWh (17).
4.2. Indirect Cd Emissions due to Electricity and Fuel
of electricity used in producing a PV system. Electricity
generation by fossil fuels creates heavy metal emissions as
those are contained in coal and oil and a fraction is released
in the atmosphere during combustion. The electricity
demand for PV modules and BOS were investigated based
input data for production of BOS materials. Then, Cd
assigned, assuming that the life-cycle electricity for the
such as natural gas, heavy oil, and coal for providing heat
and mechanical energy during materials processing, for
climate control of the manufacturing plant, and for the
cycle of PV modules. The dominant sources of such indirect
Cd emissions were found to be the use of coal during steel-
making processes and the use of natural gas during glass-
use are indirect, from the boiler materials and from the
electricity supply needed in the boiler, not from the burning
of gas itself.
The complete life-cycle atmospheric Cd emissions were
estimated by adding the Cd emissions from electricity and
fuel demand associated with manufacturing and materials
production for PV module and Balance of System (BOS).
These are shown in Figure 3. The results show that CdTe PV
displacing other electricity technologies actually prevents a
GWh electricity generated by CdTe PV module can prevent
around 4 g of Cd air emissions if used instead of or as a
the indirect emissions due to the electricity and fuel use in
FIGURE 3. Life-cycle atmospheric Cd emissions for PV systems from electricity and fuel consumption, normalized for a Southern
Europe average insolation of 1700 kWh/m2/yr, performance ratio of 0.8, and lifetime of 30 yrs. Ground-mounted BOS (18) is assumed
for all PV systems; comparisons with other electricity generation options.
VOL. 42, NO. 6, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2171
the same life cycle, and about 30 times lower than those
and fuel usage of PV systems are shown in Figure 4. The
calculated emission factors are the products of electricity
and fuel usage during the life cycle of PV modules and the
metal emission factors in Cases 1–4 are based, respectively,
on the following grid mixtures and databases: Case 1,
Ecoinvent database and the grid mixture of the CrystalClear
project in which electricity mix of gas-fired combined cycle
and hydropower was used for production and purification
for medium-voltage electricity of the UCTE grid (14); Case
3, Franklin database for the U.S. average grid mixture (15);
and Case 4, emission factors of a recent study by Kim and
Dale for the U.S. grid mixture (19). The last one adopts the
DEAM LCA database and the eGRID model from the U.S.
data sources vary greatly, with the factors quoted by Kim
The CdTe PV module performs the best, and replacing the
those atmospheric heavy-metal emissions.
than the Cd emissions depicted above, direct, heavy-metal
emissions from materials processing have not been deter-
mined by the present study for several reasons. First,
emissions during processing highly depend on the selection
of the system’s boundary, and therefore, the allocation
because unwanted copper scraps containing Cd as an alloy
additive are mixed and melted with aluminum scrap during
recycling. In this case, allocation can be avoided (according
to ISO guideline) if the Cd emissions are assigned to the
primary copper alloy production. Moreover, the amount of
unabated emissions may significantly decrease with tech-
nological progress and stricter regulatory standards. For
arc furnace based on one database is 15 times higher than
that cited from another database (1.5 mg/kg of steel from
composition of the metal, in other words, the amount of
impurities mixed with matrix metal, often decides heavy-
and 150 times more chromium than unalloyed steel (14).
Therefore, estimating heavy-metal emissions directly from
materials processing, i.e., from mining, smelting, and pu-
rification, entails large inherent uncertainties.
metal emissions from electricity and fuel, with the direct
emissions from material processing based on heavy-metal
emission factors from the Ecoinvent database (Figure 5).
Direct heavy-metal emissions from copper, lead, and steel
alloying processes together with aluminum recycling that is
unrelated to electricity or fuels have been estimated for the
multi-Si PV module. It is shown that the electricity con-
sumption is the most important source of heavy-metal
of direct Pb emission from material processing is related to
solar glass manufacturing, which accounts for about 80% of
FIGURE 4. Life-cycle atmospheric heavy-metal emissions for PV systems (normalized for Southern European average insolation of
1700 kWh/m2/yr, performance ratio of 0.8, and lifetime of 30 yrs). Each PV system is assumed to include a ground-mounted BOS as
described by Mason et al. (18). The four types of PV modules and corresponding efficiencies are ribbon-Si 11.5%, multi-crystalline Si
13.2%, monocrystalline 14%, and CdTe 9%.
2172 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 2008
such Pb emission. However, this result may be an overes-
timation since Ecoinvent database assumes a construction-
grade glass for the solar glass, although the glass used in PV
to solar glass as an additive (20). In fact, the lead emission
factor of glass manufacturing in the ETH-ESU database (the
a factor of 250 (14, 21). For the above reasons estimates of
the energy-related emissions. Further work is required to
improve reliability of estimating such emissions.
5. PV Breeder
of increased PV penetration in the “quality” of the energy
mixture used in PV production. At the limit, all electricity
used in PV manufacturing can be generated by onsite or
nearby PV. In this section, we explore potential benefits of
returning electricity generated by PV to the PV fuel cycle. As
the electric power generated by PV is variable to insolation,
an electrical storage system will be needed to fully meet the
electricity demand for the PV production; with the reality of
today’s electricity grid (22), around 30% of the electricity
system. In Figure 6, we illustrate the effect of incrementing
electricity supply from a PV breeder scheme (i.e., PVs that
supply electricity to the PV life cycles) of multicrystalline
(mc)-Si and CdTe PVs. For mc-Si, around 250 kWh per m2
while producing the same area of CdTe requires 59 kWh of
electricity (10, 12). If the considered PV breeder system
supplies 30% of the electricity required in each Si PV
production process, i.e., silicon, wafer, cell, and module, as
well as in CdTe PV production process, 6 and 2 g/kWh of
grid mix (Figure 6). A recent study demonstrates that large-
(23), which could enable a 100% electricity supply for PV
manufacturing. This would reduce around 50% of life-cycle
exercise by Pacca et al. resulted in greater GHG reductions,
i.e., 68% and 82% for multi-Si and amorphous-Si PV,
respectively (24). The greater reductions are related to the
higher CO2emissions from the background electricity being
replaced in their study, i.e., the U.S. average grid mix, which
is 45% more carbon-intensive than the UCTE grid (14, 15).
A PV breeder system could directly supply a large part of the
electricity generation and production related-parameters,
cell/film thickness are also advancing in parallel and would
also result in reduced emissions.
Using data compiled from the original records of twelve PV
manufacturers, we quantified the emissions from the life
cycle of four major commercial photovoltaic technologies
and showed that they are insignificant in comparison to the
emissions that they replace when introduced in average
grid electricity with central PV systems presents significant
environmental benefits, which for CdTe PV amounts to
89–98% reductions of GHG emissions, criteria pollutants,
installations, such pollution reductions are expected to be
networks are reduced, and part of the emissions related to
the life cycle of these networks are avoided. It is interesting
that emissions of heavy metals are greatly reduced even for
the types of PV technologies that make direct use of related
compounds. For example the emissions of Cd from the life
cycle of CdTe PV are 90-300 times lower than those from
coal power plants with optimally functioning particulate
in CdTe PV than in crystalline Si PV, because the former use
less energy in their life cycle than the later. In general, thin-
emissions of heavy metals, SOx, NOx, PM, and CO2. In any
case, emissions from any type of PV system are expected to
be lower than those from conventional energy systems
provide the benefits of significantly curbing air emissions
harmful to human and ecological health. It is noted that the
as efficiencies and material utilization rates increase and
this kind of analysis needs to be updated periodically. Also,
ules. UCTE grid mix and Ecoinvent database are used for heavy-
metal emission factors of electricity, fuel, and materials.
5. Breakdown of heavy-metal emissions for PV mod-
when using a PV breeder that supplies electricity for PV pro-
duction. Insolation of 1700 kWh/m2/yr, performance ratio of 0.8, and
lifetime of 30 yrs are assumed. BOS is not included. In 2005, the
total (100%) electricity corresponded to 250 kWh/m2for mc-Si
and 59 kWh/m2for CdTe PV.
6. Greenhouse gas emissions profile for PV modules
VOL. 42, NO. 6, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2173
future very large penetrations of PV would alter the grid
composition and this has to be accounted for in future
The crystalline silicon research was conducted within the
and cadmium telluride research was conducted within the
AC02-76CH000016 with the US Department of Energy. We
thank M. de Wild-Scholten (ECN), A. Meader (First Solar), T.
Hansen (Tucson Power Electric), J. Mason (Solar Energy
Campaign, NY), Terry Jester (Shell Solar), and a number of
industry experts who contributed to the collection of
their help: Deutsche Cell, Deutsche Solar, Evergreen Solar,
First Solar, HCT Shaping Systems, Isofoton, Photowatt and
Shell Solar (currently Solar Word). Links to the industry-
improvements are listed in citation 27.
Supporting Information Available
contaminants. This material is available free of charge via
the Internet at http://pubs.acs.org.
(1) Alsema, E. A. Energy Pay-back Time and CO2Emissions of PV
Systems. Prog. Photovolt: Res. Appl. 2000, 8, 17–25.
cycle assessment of photovoltaic modules: Comparison of mc-
Appl. 2003, 11, 275–287.
(4) Palz, W.; Zibetta, H. Energy pay-back time of photovoltaic
modules. Int. J. Sol. Energy. 1991, 10, 211–216.
(5) Hynes, K. M.; Baumann, A. E.; Hill, R. An Assessment of the
Environmental Impacts of Thin Film Cadmium Telluride
Modules based on Life Cycle Analysis. Presented at the IEEE 1st
World Conference on Photovoltaic Conversion (WCPEC), HI,
1994; pp 958–961.
Energy Mater. Sol. Cells 2001, 67, 279–287.
(7) Keoleian, G. A.; Lewis, G. M. Application of life-cycle energy
analysis to photovoltaic module design. Prog. Photovolt: Res.
Appl. 1997, 5, 287–300.
greenhouse gas emissions and external costs: 2004-early 2005
status. Prog. Photovolt: Res. Appl. 2006, 14, 275–280.
(9) Aulich, H. A.; Schulze, F. Crystalline silicon feedstock for solar
cells. Prog. Photovolt: Res. Appl. 2002, 10, 141–147.
(10) Alsema, E.; de Wild-Scholten, M. Environmental Impact of
Crystalline Silicon Photovoltaic Module Production. Presented
at Materials Research Society
PV production. Renew. Sust. Ener. Rev. 2004, 8, 303–334.
(12) Fthenakis, V. M.; Kim, H. C. Energy Use and Greenhouse Gas
Emissions in the Life Cycle of CdTe Photovoltaics. Presented at
Processing of Thin-Film CdTe PV Module, Phase I Annual
Report; National Renewable Energy Laboratory: Golden, CO,
(14) Ecoinvent Centre. Ecoinvent data v1.1. Final reports ecoinvent
2000 No. 1–15; Swiss Centre for Life Cycle Inventories: Düben-
(15) USA LCI Database Documentation; Franklin Associates: Prairie
Village, KS, 1998.
(16) Fthenakis, V. M.; Fuhrmann, M.; Heiser, J.; Lanzirotti, A.; Fitts,
PV modules during fires. Prog. Photovolt: Res. Appl. 2005, 13,
Research Institute: Palo Alto, CA, 2002.
(18) Mason, J. M.; Fthenakis, V. M.; Hansen, T.; Kim, H. C. Energy
3·5 MW PV installation. Prog. Photovolt: Res. Appl. 2006, 14,
States Electricity System. Int. J. LCA 2005, 10, 294–310.
A. Implications of European Environmental Legislation for
Conference, Barcelona, June 6–10, 2005.
(21) Frischknecht, R., et al. Öko-inventare von Energiesystemen, 3rd
ed.; ETH-ESU: Switzerland, 1996.
(22) Denholm, P.; Margolis, R. M. Evaluating the limits of solar
2007, 35, 2852–2861.
(23) Zweibel, K.; Mason, J.; Fthenakis, V. A Solar Grand Plan. Sci.
Am. 2008, (Jan.), 64–73.
(24) Pacca, S.; Sivaraman, D.; Keoleian, G. A. Parameters affecting
the life cycle performance of PV technologies and systems.
Energy Policy 2007, 35, 3316–3326.
(25) Silicon PV Material Inventory Data. http://www.ecn.nl/docs/
(26) Cadmium Telluride Material Inventory Data. http://www.
clca.columbia.edu (to be posted).
(27) de Wild-Scholten, M. J.; Alsema, E. A. Environmental Life Cycle
Production–Status 2005/2006; ECN: Petten, March, 2007.
Symposium, Boston, Nov.
2174 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 2008