photovoltaics: life-cycle analyses - brookhaven national laboratory

Available online at www.sciencedirect.com

www.elsevier.com/locate/solener

Solar Energy 85 (2011) 1609–1628

Photovoltaics: Life-cycle analyses

V.M. Fthenakis a,b,*, H.C. Kim a

a Center for Life Cycle Analysis, Columbia University, New York, NY, USAb Photovoltaic Environmental Research Center, Brookhaven National Laboratory, Upton, NY, USA

Available online 23 February 2010

Communicated by: Associate Editor Yogi Goswami

Abstract

Life-cycle analysis is an invaluable tool for investigating the environmental profile of a product or technology from cradle to grave.Such life-cycle analyses of energy technologies are essential, especially as material and energy flows are often interwoven, and divergentemissions into the environment may occur at different life-cycle-stages. This approach is well exemplified by our description of materialand energy flows in four commercial PV technologies, i.e., mono-crystalline silicon, multi-crystalline silicon, ribbon-silicon, and cadmiumtelluride. The same life-cycle approach is applied to the balance of system that supports flat, fixed PV modules during operation. We alsodiscuss the life-cycle environmental metrics for a concentration PV system with a tracker and lenses to capture more sunlight per cell areathan the flat, fixed system but requires large auxiliary components. Select life-cycle risk indicators for PV, i.e., fatalities, injures, and max-imum consequences are evaluated in a comparative context with other electricity-generation pathways.� 2009 Elsevier Ltd. All rights reserved.

Keywords: Photovoltaics; Life-cycle analysis; Life-cycle assessment; Environmental and health effects; Energy payback times

1. Introduction

Photovoltaics (PV), made from semiconducting materi-als, convert photons into electricity. When sunlight hitsthese materials, photons with a certain wavelength triggerelectrons to flow through the materials to produce directcurrent (DC) electricity. Commercial PV materials includemulti-crystalline silicon, mono-crystalline silicon, amor-phous silicon, and thin film technologies, such as cadmiumtelluride (CdTe), and copper indium diselenide (CIS). Atypical PV system consists of the PV module and the bal-ance of system (BOS) structures for mounting the PV mod-ules and converting the generated electricity to alternatecurrent (AC) electricity of the proper magnitude for usagein the power grid.

0038-092X/$ - see front matter � 2009 Elsevier Ltd. All rights reserved.

doi:10.1016/j.solener.2009.10.002

* Corresponding author. Address: Photovoltaic Environmental Re-search Center, Brookhaven National Laboratory, Upton, NY, USA.

E-mail address: [email protected] (V.M. Fthenakis).

Life-cycle analysis (LCA) is a framework for consideringthe environmental inputs and outputs of a product or pro-cess from cradle to grave. It is employed to evaluate theenvironmental impacts of energy technologies, and theresults are increasingly used in decisions about R&D fund-ing and in formulating energy policies. Informational pub-lications for decision-makers in the European Community(European Commission, 2003) and in Australia (AustralianCoal Industry Association Research Program (ACARP),2004) indicated that photovoltaics have relatively highenvironmental impacts compared with other technologies,e.g., greenhouse-gas (GHG) emissions of 180 g CO2-eq./kWh in Germany, and 100 g CO2-eq/kWh in Australia,10 and 2.5 times higher than the GHG emissions of thenuclear-fuel cycle for each country, and 45% and 23% ofthose of combined cycle (CC) natural-gas power generationin the same country.

These impacts reflect the fossil-fuel-based energy used inproducing the materials for solar cells, modules, and sys-tems; however, the data used in these studies were outdated

http://dx.doi.org/10.1016/j.solener.2009.10.002

mailto:[email protected]

http://dx.doi.org/10.1016/j.solener.2009.10.002

slattuca

Typewritten Text

slattuca

Typewritten Text

BNL-96354-2011-JA

Nomenclature

a-Si amorphous siliconAC alternate currentBOS balance of systemc-Si crystalline siliconCdS cadmium sulfideCdTe cadmium tellurideCIGS copper indium gallium selenideCIS copper indium diselenideDC direct currentDNI direct normal irradianceEPBT energy payback timeESP electrostatic precipitatorsFBR fluidized bed reactorGaAs gallium arsenideGHG greenhouse gasGWP global warming potentialHCl hydrogen chlorideHCPV high-concentration PVHF hydrogen fluorideLCA life-cycle analysis (or assessment)LCI life-cycle inventory

LPG liquefied petroleum gasmc-Si multi-crystalline siliconmono-Si

mono-crystalline siliconNAICS North American Industry Classification SystemNG natural gasNOx nitrogen oxidePM particulate matterPR performance ratioPSA probabilistic safety assessmentPSI Paul Scherrer InstitutePV photovoltaicsRMP Risk Management ProgramSiH4 silaneSiHCl3 trichlorosilaneSOx sulfur oxideTeO2 tellurium dioxideTPE thermoplastic elastomerUCTE Union for the Coordination of Transmission of

ElectricityVTD vapor transport deposition

1610 V.M. Fthenakis, H.C. Kim / Solar Energy 85 (2011) 1609–1628

and some assumptions made were invalid. In this paper wesummarize the results of PV life-cycle analyses based oncurrent data for three silicon and one thin-film technolo-gies, emphasizing basic metrics including energy paybacktimes (EPBTs), GHG emissions, criteria pollutant emis-sions, toxic metal emissions, human injuries and fatalities.

2. Previous studies

Previous life-cycle studies reported a wide range of pri-mary energy consumption for Si-PV modules. Alsemareviewed such analyses from the 1990s and found consider-able variance between investigators in their estimates ofprimary energy consumption (Alsema, 2000). Normalizedper m2, the researchers reported 2400–7600 MJ of primaryenergy consumption for mc-Si, and 5300–16,500 MJ formono-Si modules. Besides uncertainties in the data, heattributed these differences mainly to the assumptionsand allocation rules that each author adopted for modelingthe purification and crystallization stages of silicon. Inthose days, solar cells were mostly tailored from off-specproducts of electronic-grade silicon not directly fromsolar-grade silicon, so that multiple allocation rules mightwell be applied to the energy and material inputs for eachgrade of silicon; currently, only 5% of solar cells are fromoff-spec electronic-grade silicon (Alsema, 2000; Alsemaand de Wild-Scholten, 2005). Selecting only those processsteps needed to produce solar-grade silicon, Alsema’sown estimates were 4200 and 5700 MJ/m2 for mc-Si andmono-Si modules, respectively (Alsema, 2000). These val-

ues correspond to an energy payback time (EPBT) of 2.5and 3.1 years, and life-cycle greenhouse-gas emissions of46 and 63 g CO2-eq./kWh for mc-Si PV with 13% efficiencyand mono-Si with 14% efficiency, respectively, underSouthern European (Mediterranean) conditions: insolationof 1700 kWh/m2/year, and a performance ratio of 0.75.The balance of system (BOS) components, such as amounting support, a frame, and electrical componentsaccount for additional �0.7 years of EPBT, and �15 gCO2-eq./kWh of GHG emissions.

Meijer et al. (2003) more recently assessed a slightlyhigher energy expenditure of 4900 MJ/m2 to produce anmc-Si module. They assumed that the 270-lm thick SiPV with 14.5% cell efficiency was fabricated from elec-tronic-grade high-purity silicon, which entails greaterenergy consumption. Their corresponding EPBT estimatefor the module was 3.5 years excluding BOS components;i.e., higher than Alsema’s earlier determination of 2.5years. The increase stems mainly from the low level of inso-lation in the Netherlands (1000 kWh/m2/year) comparedwith the average for Southern Europe (1700 kWh/m2/year), and, to a less degree, from the higher energy estima-tion for silicon (Alsema, 2000; Meijer et al., 2003). Jungb-luth reported the life-cycle metrics of various PV systemsunder environmental conditions in Switzerland in 2000(Jungbluth, 2005). He considered the environmentalimpacts for 300-lm thick multi- and mono-Si-PV modulewith 13.2% and 14.8% conversion efficiency, respectively.Depending on which of the two materials he evaluated,and their applications (i.e., fac�ade, slanted roof, and flat

Raw Material

Acquisition Material

ProcessingManufactur-ing

Decommis-sioning

Treatment /Disposal

Use

Recycling

M, Q

E

M, Q M, Q M, Q M, Q M, Q

E E E E E

M, Q: material and energy inputs E: effluents (air, water, solids)

M, Q

E

Fig. 1. Flow of the life-cycle stages, energy, materials, and wastes for PV systems.

V.M. Fthenakis, H.C. Kim / Solar Energy 85 (2011) 1609–1628 1611

roof), he arrived at figures of 39–110 g CO2-eq./kWh ofGHG emissions and 3–6 years of EPBT for the averageinsolation of 1100 kWh/m2/year in that country. Heassumed that the source of silicon materials was 50% fromoff-grade silicon and 50% from electronic-grade silicon,which is distant from the composition of the current(2005) PV supply (Alsema and de Wild-Scholten, 2005;Jungbluth, 2005).

There are fewer life-cycle studies of thin-film PV tech-nologies; evaluations of the life-cycle primary energy con-sumption of amorphous silicon ranged between 710 and1980 MJ/m2 (Alsema, 2000). The differences are largelyattributed to the choice of substrate and encapsulationmaterials. The lowest estimate, made by Palz and Zibetta(1991), considered a single glass structure, while the highestone by Hagedorn (1992) was based on a double-glass con-figuration to protect the active layer (Alsema, 2000; Palzand Zibetta, 1991). For CdTe PV, Hynes et al. (1994) basedtheir energy analysis on two alternative technologiescurrent at that time. The first employed non-vacuumelectro-deposition of a 1.5-lm absorber layer (CdTe), inconjunction with chemical-bath deposition of the 0.2-lmwindow layer (CdS); the second method deposited bothof these layers i.e., �5-lm thick absorber layer and �1.7-lm thick window layer by thermal evaporation yielding.Their primary energy estimate for the first technologywas 993 MJ/m2, and that for the second was 1188 MJ/m2. Kato et al.’s (2001) energy estimates were pertinentto the scale of annual production; they suggested thatenergy consumption will decline as the scale of productionrises; they cited values of 1523, 1234, and 992 MJ/m2 forframeless modules with annual capacities of 10, 30, and100 MWp,1 respectively. However, these earlier estimatesfall far short of describing present-day commercial-scaleCdTe PV production that, unlike previously, now encom-passes many large-scale production plants.

1 Peak power.

3. Life cycle of photovoltaics

The life-cycle stages of photovoltaics involve (1) the pro-duction of raw materials, (2) their processing and purifica-tion, (3) the manufacture of modules and balance of system(BOS) components, (4) the installation and use of the sys-tems, and (5) their decommissioning and disposal or recy-cling (Fig. 1).

Production starts with mining the raw materials (i.e.,quartz sand for silicon PV; Zn- and Cu-ores for CdTePV), and continues with their processing and purification(Fig. 2) (Fthenakis et al., 2008). The silica in the quartzsand is reduced in an arc furnace to metallurgical-grade sil-icon, which must be purified further into “electronic-grade”

or “solar-grade” silicon, typically through a “Siemens”process. Crystalline silicon modules typically are framedfor additional strength and easy mounting. The recentLCAs for crystalline silicon are based on life-cycle inven-tory (LCI) data provided, collectively, by eleven Europeanand US photovoltaic companies participating in the Euro-pean Commission’s Crystal Clear project. The data setswere published in separate papers by Alsema and deWild-Scholten (2005) and by Fthenakis and Alsema (2006).

The life-cycle inventories of the minor metals used inthin-film PVs such as Cd, In, Mo, and Se, are closelyrelated to the production cycle of base metals (Zn, Cu).The allocations of emissions and energy use between theformer and the latter during mining, smelting and refiningstages are described elsewhere (Fthenakis et al., 2009).Fthenakis (2004) described the material flows of cadmium(Cd) and emissions from the entire life-cycle stages of cad-mium telluride (CdTe) PV. The life cycle starts with theproduction of Cd and Te that are byproducts, respectively,of smelting Zn- and Cu-ores (Fig. 2). Cd is obtained fromthe Zn waste streams, such as particulates collected in air-pollution-control equipment, and slimes collected from Zn-electrolyte purification stages. Cadmium is further pro-cessed and purified to meet the four or five 9 s purityrequired for synthesizing CdTe. Te is recovered andextracted after treating the slimes produced during electro-lytic copper refining with dilute sulfuric acid; these slimes

(a) Silicon PVs (b) CdTe PVs (frameless)

Metallurgi-cal grade Si

Solar grade Si

Ribbon

Mono-Si Crystal

Multi-Si Ingot

Mono-Si Wafer

Multi-Si Wafer

Cell

Module

Frame

PV System

BOS

Quartz Mining

CdTe powder

Thin film CdS/CdT

Module

Encapsulation

CdS Powder

Cd Te

Cu Ores Zn Ores

PV System

BOS

Fig. 2. Detailed flow diagram from the raw material acquisition to manufacturing stage of PVs (Fthenakis et al., 2008).


also contain Cu, and other metals. After cementation withcopper, CuTe is leached with caustic soda to produce asodium-telluride solution that is used as the feed for Teand TeO2. Additional leaching and vacuum distillationgives Cd and Te powders of semiconductor grade (i.e.,99.999%).

4. Life-cycle inventory

4.1. Modules

The material and energy inputs and outputs during thelife cycles of Si PVs, viz., ribbon-Si, multi-Si, mono-Si,and also thin-film CdTe PV, were investigated in detailbased on actual measurements from PV production plantsbetween 2004 and 2006. Alsema and de Wild-Scholtenrecently updated the life-cycle inventory (LCI) for the tech-nology for producing crystalline silicon modules in Wes-tern Europe under the framework of the Crystal Clear

Table 1Materials and energy inputs for PV systems to produce 1 m2 of module includ

Category Inputs Ribbo

Components (kg) Cell materials 0.9Glass 9.1Ethylene vinyl acetate 1.0Others 1.8

Consumables (kg) Gases 6.1Liquid 2.2Others 0.01

Energy Electricity (kWh) 182Oil (l) 0.05Natural gas (MJ) 166

project; a large European Integrated Project focusing oncrystalline silicon technology, co-funded by the EuropeanCommission and the participating countries (Alsema andde Wild-Scholten, 2005; de Wild-Scholten and Alsema,2005). Fthenakis and Kim reported the LCI data for CdTethin-film technology taken from the production data fromFirst Solar’s plant in Perrysburg, Ohio, United States (Fth-enakis and Kim, 2005). Table 1 presents the simplified life-cycle inventories (LCI) for 2006, compiled from the datafrom eleven European and two US plants along with valuesin the literature of Alsema and de Wild-Scholten (2005)and Fthenakis and Kim (2005).

The typical thickness of multi- and mono-Si PV is 270–300 lm, and that of ribbon-Si is 300–330 lm; 72 individualcells of 156 cm2 (125 cm � 125 cm) comprise a module of1.25 m2 for all Si-PV types. The conversion efficiency ofribbon-, multi-, and mono-Si module is 11.5%, 13.2%,and 14.0%, respectively. On the other hand, as of 2006,First Solar’s 25-MWp plant manufactures frameless,

ing process loss, updated for 2006 (excluding the frame for Si modules).

n-Si Multi-Si Mono-Si CdTe

1.6 1.5 0.0659.1 9.1 19.21.0 1.0 0.61.8 1.8 2.0

2.2 7.8 0.0016.8 6.6 0.674.3 4.3 0.4

248 282 590.05 0.05 0.05308 361 –

Table 2Mass balance of major components for the 3.5 MW Tucson ElectricPower generating plant in Springerville, AZ, based on 30 years ofoperation.

Balance of system Mass (kg/MWp) % of total

BOS

PV support structure 16,821 10.3Module interconnections 453 0.3Junction boxes 1385 0.8Conduits and fittings 6561 4.0Wire and grounding devices 5648 3.4Inverters and transformers 28,320 17.3Grid connections 1726 1.1Office facilities 20,697 12.6Concrete 76,417 46.6Miscellaneous 5806 3.5Total 163,834 100.0

Framea 18,141

a Based on 12.2% rated efficiency for mc-Si module.


double-glass, CdTe modules of 1.2 m by 0.6 m, rated at 9%photon-to-electricity conversion efficiency with �3-lmthick active layer.

The data for Si PVs extend from the production stage ofsolar-grade Si to the module manufacturing stage, whilethose for CdTe PV correspond to the deposition of theCdTe film and the module’s manufacturing stages. Themetallurgical-grade silicon that is extracted from quartz ispurified into solar-grade polysilicon by either a silane(SiH4) or trichlorosilane (SiHCl3)-based process. Theenergy requirement for this purification step is the mostimportant demand for crystalline Si-PV modules, account-ing for 45% of the primary energy used for fabricatingmulti-Si modules. Two technologies are currentlyemployed for producing polysilicon from silicon gases:the Siemens reactor method and the fluidized bed reactor(FBR) method. In the former, which accounts for themajority (�90% in 2004) of solar-grade silicon productionin the US, silane- or trichlorosilane-gas is introduced into athermal decomposition furnace (reactor) with high temper-ature (�1100–1200 �C) polysilicon rods (Aulich and Schu-lze, 2002; Woditsch and Koch, 2002; Maycock, 2005). Thesilicon rods grow as silicon atoms in the gas deposit ontothem, up to 150 mm in diameter and up to 150 cm in length(Aulich and Schulze, 2002). The data on Si PVs in Table 1are based on averages over standard and modified Siemensreactors. The former primarily produces the electronic-grade silicon with a purity of over nine 9 s, while the latterproduces the solar-grade silicon with a purity of six to eight9 s, consuming less energy than the former (Williams,2000). The scenario involving the scrap silicon from elec-tronic-grade silicon production is not considered as themarket share of this material accounts for only 5% in2005 (Rogol, 2005).

4.2. Balance of system (BOS)

Little attention has been paid to the LCA studies of thebalance of system (BOS), and so inventory data are scarce.Depending on the application, solar cells are either roof-top- or ground-mounted, both operating with a proper bal-ance of system (BOS). Silicon modules need an aluminumframe of 3.8 kg/m2 for structural robustness and easyinstallation, while a glass backing performs the same func-tions for the CdTe PV produced in the US (Alsema and deWild-Scholten, 2005; Fthenakis and Kim, 2005). For arooftop PV application, the BOS typically includes invert-ers, mounting structures, cable and connectors. Large-scaleground-mounted PV installations require additional equip-ment and facilities, such as grid connections, office facili-ties, and concrete.

A recent analysis of a 3.5 MWp mc-Si installation at theSpringerville Generating Station in Arizona affords adetailed materials- and energy balance for a ground-mounted BOS (Table 2) (Mason et al., 2006). For thisstudy, Tucson Electric Power (TEP) prepared the BOS billof materials- and energy-consumption data for their mc-Si-

PV installations. The life expectancy of the PV metal sup-port structures is assumed to be 60 years. Inverters andtransformers are considered to last for 30 years, but partsmust be replaced every 10 years, amounting to 10% of theirtotal mass, according to well-established data from thepower industry on transformers and electronic compo-nents. The inverters are utility-scale, Xantrec PV-150 mod-els with a wide-open frame, allowing failed parts to beeasily replaced. The life-cycle inventory includes the officefacility’s materials and energy use for administrative, main-tenance, and security staff, as well as the operation of main-tenance vehicles. Aluminum frames are shown separately,since they are part of the module, not of the BOS inven-tory; there are both framed and frameless modules on themarket.

de Wild-Scholten et al. (2006) studied two classes ofrooftop-mounting systems based on a mc-Si-PV systemcalled SolarWorld SW220 with dimensions of1001 mm � 1675 mm, 220 Wp: they are used for on-roofmounting where the system builds on existing roofingmaterial, and in-roof mounting where the modules replacethe roof tiles. The latter case is credited in terms of energyand materials use because roof tile materials then are notrequired. Table 3 details the LCI of several rooftop-mount-ing systems, cabling, and inverters. Two types (500 and2500 W) of small inverters adequate for rooftop PV designwere inventoried. A transformer is included as an electroniccomponent for both models. The amount of control elec-tronics will become less significant for inverters with highercapacity (>10 kW), resulting in less material use per PVcapacity (de Wild-Scholten et al., 2006).

5. Energy payback times and greenhouse-gas emissions

5.1. Energy payback time

The most frequently measured life-cycle metrics for PVsystem environmental analyses are the energy payback time

Table 3LCI of balance of system (BOS) for rooftop mounting (de Wild-Scholten and al., 2006).

On-roof In-roof

Phonix, TectoSun Schletter, Eco05 + EcoG Schletter, Plandach 5 Schweizer, Solrif

(a) Mounting system (kg/m2)

Low alloy steel 0 0 0 0Stainless steel 0.49 0.72 0.28 0.08Aluminum 0.54 0.97 1.21 1.71Concrete 0 0 0 0Frame 3.04 0 0 0

Helukabel, Solarflex 101, 4 mm2, DC Helukabel, NYM-J, 6 mm2, AC

(b) Cabling (g/m2)

Copper 83.0 19.9Thermoplastic elastomer (TPE) 64.0 0.0PVC 0.0 16.9

Philips PSI 500 (500 W) Mastervolt SunMaster 2500 (2500 W)

(c) Inverters (g)

Steel 78 9800Aluminum 682 1400Copper 2Polycarbonate 68ABS 148Other plastics 5.4Printed circuit board 100 1800a

Connector 50Transformers, wire-wound 310 5500Coils 74Transistor diode 10Capacitor, film 72Capacitor, electrolytic 54Other electric components 20

a Including electric components.

2 Global warming potential, indicator of the relative radiative effect of asubstance compared to CO2, integrated over a chosen time horizon 25.Intergovernmental Panel on Climate Change, Climate Change 2001: TheScientific Basis, 2001, Cambridge, UK: Cambridge University Press.


(EPBT) and the greenhouse-gas (GHG) emissions. Energypayback time is defined as the period required for a renew-able energy system to generate the same amount of energy(either primary or kWh equivalent) that was used to pro-duce the system itself.

Energy Payback TimeðEPBTÞ¼ ðEmat þ Emanuf þ Etrans þ Einst þ EEOLÞ=ðEagen � EaoperÞ

where Emat, primary energy demand to produce materialscomprising PV system; Emanuf, primary energy demand tomanufacture PV system; Etrans, primary energy demandto transport materials used during the life cycle; Einst., pri-mary energy demand to install the system; EEOL, primaryenergy demand for end-of-life management; Eagen, annualelectricity generation in primary energy term; Eaoper, an-nual energy demand for operation and maintenance in pri-mary energy term.

Calculating the primary energy equivalent requiresknowledge of the country-specific, energy-conversionparameters for fuels and technologies used to generateenergy and feedstock. The annual electricity generation(Eagen) is represented as primary energy based on the effi-ciency of electricity conversion at the demand side. Theelectricity is converted to the primary energy term by the

average conversion efficiency of 0.29 for the United Statesand 0.31 for Western Europe (Dones et al., 2003; FranklinAssociates, 1998).

5.2. Greenhouse-gas emissions

The greenhouse-gas (GHG) emissions during the life-cycle stages of a PV system are estimated as an equivalentof CO2 using an integrated time horizon of 100 years; themajor emissions included as GHG emissions are CO2

(GWP2 = 1), CH4 (GWP = 23), N2O (GWP = 296), andchlorofluorocarbons (GWP = 4600–10,600) (InternationalPanel on Climate Change, 2001). Electricity and fuel useduring the PV materials and module production are themain sources of the GHG emissions for PV cycles.Upstream electricity-generation methods also play animportant role in determining the total GHG emissions.For instance, the GHG emission factor of the averageUS electricity grid is 40% higher than that of the average

g C

O2-

eq./k

Wh

0

10

20

30

40

50

Module Frame BOS

Ribbon-Si, 11.5%

European production

Multi-Si, 13.2%

European production

Mono-Si,14.0%

European productionCdTe*, 9%

US production

CdTe, 8%

European production

Fig. 3. Life-cycle GHG emissions from silicon and CdTe PV modules,wherein BOS is the balance of system, that is the module’s supports,cabling and power conditioning (Alsema et al., 2005; Fthenakis et al.,2004; Wild-Scholten et al., 2005; Fthenakis et al., 2005; Raugei et al.,2007). Unless otherwise noticed, the estimates are based on rooftop-mountinstallation, Southern European insolation, 1700 kWh/m2/year, a perfor-mance ratio of 0.75, a lifetime of 30 years. *Based on ground-mountinstallation, average US insolation of 1800 kWh/m2/year, and a perfor-mance ratio of 0.8.


Western European (UCTE) grid although emission factorsof fossil-fuel combustion are similar, resulting in higherGHG estimates for the US - produced modules (Doneset al., 2003; Franklin Associates, 1998).

With material-inventory data from industry, Alsemaand de Wild-Scholten (2005) demonstrated that the life-cycle primary energy and greenhouse-gas emissions ofcomplete rooftop Si-PV systems are much lower than thosereported in earlier studies. Primary energy consumption is2300, 3700, and 4200 MJ/m2, respectively, for ribbon-,multi-, and mono-Si modules. Fthenakis and Alsema alsoreport that the GHG emissions of Si modules correspond-ing to 2004–2005 production are within a 30–45 g CO2-eq./kWh range, with an EPBT of 1.7–2.7 years for a rooftopapplication under Southern European insolation of1700 kWh/m2/year and a performance ratio3 (PR) of 0.75(Figs. 3 and 4, respectively) (Alsema and de Wild-Scholten,2005; Fthenakis and Alsema, 2006). We note that in theseestimates, the BOS for rooftop application accounts for4.5–5 g CO2-eq./kWh of GHG emissions and 0.3 years ofEPBT. These calculations were based on the electricitymixture for the current production of Si, within the contextof the Crystal Clear project. For CdTe PV, we estimate theenergy consumption 1200 MJ/m2, based on the actual pro-duction data of the year 2005 from the First Solar’s25 MWp plant in Ohio, United States, close to the earlystudies reviewed (Hynes et al., 1994; Kato et al., 2001).The greenhouse-gas (GHG) emissions and energy paybacktime (EPBT) of ground-mounted CdTe PV modules underthe average US insolation condition, 1800 kWh/m2/year,were determined to be 24 g CO2-eq./kWh and 1.1 years,correspondingly. These estimates include 6 g CO2-eq./kWh of GHG and 0.3 year of EPBT contribution fromthe ground-mounted BOS (Fthenakis and Kim, 2005).On the other hand, Raugei et al. (2007) estimates a lowerprimary energy consumption, �1100 MJ/m2, and therebyless GHG emissions and lower EPBT than ours, basedon the data of the year 2002 from the Antec Solar’s10 MWp plant in Germany (Figs. 3 and 4). The source ofthe difference between the two estimates is yet to bestudied.

We note that this picture is not a static one and expectthat improvements in material and energy utilization andrecycling will continue to improve the environmental pro-files. A recent, major improvement is a recycling processfor the sawing slurry, the cutting fluid that is used in thewafer cutting (Alsema et al., 2006). This recycling processrecovers 80–90% of the silicon carbide and polyethyleneglycol which used to be wasted. As we show in Fig. 4, recy-cling silicon carbide and polyethylene glycol from siliconslurries decreases the EPBTs of these technologies by10%. On the other hand, any increases in the electric-con-version efficiencies of the modules will entail a proportional

3 Ratio between the ideal and actual electricity output.

improvement of the EPBT. Fig. 5 compares these emissionswith those of conventional fuel-burning power plants,revealing the considerable environmental advantage ofPV technologies. The majority of GHG emissions are fromthe operation stage for the coal, natural gas, and oil fuelcycles while the material and device production accountsfor nearly all the emissions for the PV cycles. The GHGemissions from the nuclear-fuel cycle are mainly relatedto the fuel production, i.e., mining, milling, fabrication,conversion, and enrichment of uranium fuel. The detailsof the US nuclear-fuel cycle is described elsewhere (Fthena-kis and Kim, 2007a).

6. Criteria pollutant and heavy metal emissions

6.1. Criteria pollutant emissions

The emissions of criteria pollutants during the life cycleof a PV system are largely proportional to the amount offossil fuel burned during its various phases, in particular,PV material processing and manufacturing; therefore, theemission profiles are close to those of the greenhouse-gasemissions (Fig. 6). Toxic gases and heavy metals can beemitted directly from material processing and PV manufac-turing, and indirectly from generating the energy used atboth stages. Accounting for each of them is necessary tocreate a complete picture of the environmental impact ofa technology. An interesting example of accounting forthe total emissions is that of cadmium flows in CdTe andother PV technologies, as discussed next.

Yrs

0.0

0.5

1.0

1.5

2.0

2.5

3.0

Module Frame BOS

Ribbon-Si, 11.5%

European production

Multi-Si, 13.2%

European productioin

Mono-Si, 14.0%

European production

Slurry recyclingMono-Si, 14.0%

European productionCdTe,* 9%

US production

Multi-Si, 13.2%

European production

Slurry recyclingCdTe, 8%

European

production

Fig. 4. Energy payback time for silicon and CdTe PV modules, wherein BOS is the balance of system, that is the module supports, cabling and powerconditioning (Alsema et al., 2005; Fthenakis et al., 2004; Wild-Scholten et al., 2005; Fthenakis et al., 2005; Raugei et al., 2007; Alsema et al., 2006). Unlessotherwise noticed, the estimates are based on rooftop-mount installation, Southern European insolation, 1700 kWh/m2/year, a performance ratio of 0.75,and a lifetime of 30 years. *Based on ground-mount installation, average US insolation of 1800 kWh/m2/year, and a performance ratio of 0.8.

g C

O2-

eq./k

Wh

0

200

800

1000

1200Fuel Production Transportation Operation Materials

Coal (Kim and Dale 2005)

Natural gas (Kim and Dale 2005)

Petroleum (Kim and Dale 2005)

Nuclear, baseline (Fthenakis and Kim 2007)

PV, CdTe (Fthenakis and Kim 2005)

PV, mc-Si (Fthenakisand Alsema 2006)

24 24a 37b

1210

760

880

Fig. 5. Comparison of GHG emissions from PV with those from conventional power plants (abased on the average US insolation, 1800 kWh/m2/year anda performance ratio of 0.8; bbased on the Southern European insolation,1700 kWh/m2/year and a performance ratio of 0.75).


6.2. Heavy metal emissions

6.2.1. Direct emissions

Cadmium is a byproduct of zinc and lead, and is col-lected from emissions and waste streams during the pro-duction of these major metals. The largest fraction ofcadmium, with �99.5% purity, is in the form of a spongefrom the electrolytic recovery of zinc. This sponge istransferred to a cadmium-recovery facility, and is furtherprocessed through oxidation and leaching to generate anew electrolytic solution. After selectively precipitatingthe major impurities, cadmium of 99.99% purity is recov-ered by electrowinning. It is further purified by vacuumdistillation to the five 9 s purity required for CdTe PV

manufacturing. The emissions during each of these stepsare detailed elsewhere (Fthenakis, 2004). They total up to0.02 g per GWh of PV-produced energy under SouthernEuropean condition (Table 4). On the other hand, cad-mium emissions during the lifespan of a finished CdTemodule are negligible; the only conceivable pathway ofrelease is if a fire broke out. Experiments at BrookhavenNational Laboratory that simulated real fire conditionsrevealed that CdTe is effectively contained within theglass-to-glass encapsulation during the fire, and onlyminute amounts (0.4–0.6%) of Cd are released. The dis-solution of Cd into the molten glass was confirmed byhigh-energy synchrotron X-ray microscopy (Fthenakiset al., 2005).

(a)

mg/

kWh

0

20

40

60

80

100Module Frame BOS

Ribbon 11.5%

Multi-Si13.2%

Mono-Si14.0%

CdTe 9%

(b)

0

20

40

60

80

100

120

140

160

180

mg/

kWh

Ribbon 11.5%

Multi-Si13.2%

Mono-Si14.0%

CdTe 9%

Fig. 6. Life-cycle emissions of (a) NOx, and (b) SOx emissions from siliconand CdTe PV modules, wherein BOS is the balance of system, that ismodule supports, cabling and power conditioning. The estimates are basedon rooftop-mount installation, Southern European insolation, 1700 kWh/m2/year, a performance ratio of 0.75, and a lifetime of 30 years. It isassumed that the electricity supply for all the PV system is from the UCTEgrid.

Table 4Direct, atmospheric Cd emissions during the life cycle of the CdTe PVmodule (allocation of emissions to co-production of Zn, Cd, Ge and In).

Airemissions(g Cd/tonneCda)

Allocation(%)

Airemissions(g Cd/tonneCda)

mgCd/GWhb

Mining of Zn ores 2.7 0.58 0.016 0.02Zn smelting/refining 40 0.58 0.23 0.3Cd purification 6 100 6 9.1CdTe production 6 100 6 9.1PV manufacturing 3 100 3 4.5Operation 0.3 100 0.3 0.3Disposal/recycling 0 100 0 0

Total 15.55 23.3

a Tonne of Cd produced.b Energy produced assuming average Southern European insolation

(i.e., 1700 kWh/m2/yr), 9% electrical-conversion efficiency, and a 30-yearlife for the modules.


6.2.2. Indirect emissions

Coal and oil-fired power plants routinely generate Cdduring their operation, as it is a trace element in both fuels.According to the US Electric Power Research Institute’s(EPRI’s) data, under the best/optimized operational andmaintenance conditions, burning coal for electricityreleases into the air between 2 and 7 g of Cd/GWh (ElectricPower Research Institute (EPRI), 2002). In addition, 140g/GWh of Cd inevitably collects as fine dust in boilers, bag-houses, and electrostatic precipitators (ESPs). Further-more, a typical US coal-powered plant emits per GWhabout 1000 tons of CO2, 8 tons of SO2, 3 tons of NOx,and 0.4 tons of particulates. The emissions of Cd fromheavy-oil burning power plants are 12–14 times higher thanthose from coal plants, even though heavy oil containsmuch less Cd than coal (�0.1 ppm), because these plantsdo not have particulate-control equipment. Cadmiumemissions also are associated with natural gas and

nuclear-fuel life cycles because of the energy used in theassociated fuel processing and material productions (Doneset al., 2003).

We accounted for Cd emissions in generating the elec-tricity used in producing a CdTe PV system (Fthenakisand Kim, 2007b). The assessment of electricity demandfor PV modules and BOS was based on the life-cycle inven-tory of each module and the electricity input data for pro-ducing BOS materials. Then, Cd emissions from theelectricity demand for each module were assigned, assum-ing that the life-cycle electricity for the silicon- andCdTe-PV modules was supplied by the UCTE grid. Theindirect Cd emissions from electricity usage during the lifecycle of CdTe PV modules (i.e., 0.24 g/GWh) are an order-of-magnitude greater than the direct ones (routine andaccidental) (i.e., 0.02 g/GWh).

The complete life-cycle atmospheric Cd emissions, esti-mated by adding those from the electricity and fuel demandassociated with manufacturing and materials productionfor various PV modules and balance of system (BOS),are compared with the emissions from other electricity-gen-erating technologies (Fig. 7) (Fthenakis et al., 2008).Undoubtedly, displacing the others with CdTe PV mark-edly lowers the amount of Cd released into the air. Thus,every GWh of electricity generated by CdTe PV modulescan prevent around 5 g of Cd air emissions if they are usedinstead of, or as a supplement to, the UCTE electricitygrid. Also, the direct emissions of Cd during the life cycleof CdTe PV are 10 times lower than the indirect ones dueto electricity and fuel use in the same life cycle, and about30 times less than those indirect emissions from crystallinephotovoltaics. Furthermore, we examined the indirectheavy metal emissions in the life cycle of the three silicontechnologies discussed earlier, finding that, among PV tech-nologies, CdTe PV with the lowest energy payback timehas the fewest heavy metal emissions (Fig. 8) (Fthenakiset al., 2008).

ribbon-Simulti-S

imono-Si

CdTeHard coal

Lignite

Natural gas OilNuclear

Hydro

UCTE avg.

g C

d/G

Wh

0

1040

50

0.8 0.9 1.00.3

3.1

6.2

0.2

43.3

0.5 0.03

4.1

Fig. 7. Life-cycle atmospheric Cd emissions for PV systems from electricity and fuel consumption, normalized for a Southern European average insolationof 1700 kWh/m2/yr, performance ratio of 0.8, and lifetime of 30 years. A ground-mounted BOS is assumed for all PV systems (Fthenakis et al., 2008).


7. Concentrator PV systems

7.1. Overview

The primary purpose for establishing concentrating PVsystems is to capture the maximum amount of sunlightand focus it on to a small cell area, thereby ultimatelyreducing the cost of high-efficiency solar-cell materials.For example, the commercialization of III–V multi-junc-tion solar cells that originally were designed for spaceapplication is expanding via combination with concentra-tors (Swanson, 2003; Peharz and Dimroth, 2005). As theoptics of concentrator modules capture only direct solarillumination, they must move to track the sun’s trajectoryduring the day. Single- or two-axis trackers with hydraulic

Arsenic Chromium L

Emis

sion

s (g

/GW

h)

0

5

10

15

20

25

30

ribbon-Si, 11.5% multi-Si, 13.2% mono-Si, 14.0% CdTe, 9% (frame

Fig. 8. Emissions of heavy metals from PVs due to electricity use, based on Uinsolation of 1700 kWh/m2/year, performance ratio of 0.8, and lifetime of 30described by Mason et al. (2006).

drives, light sensors, and controllers are the commonestconfiguration.

We conducted an LCA for one of the most promisingconcentrator designs, the Amonix high-concentrator PV(HCPV) system (Fig. 9) (Kim and Fthenakis, 2006). Thissystem consists of units called MegaModules mounted ontwo-axis trackers. Each MegaModule consists of 48 blocksof sub-module units, rated at 4.8 kWp-ac each under850 W/m2 of direct normal irradiance (DNI). A refractiveconcentration technology based on acrylic Fresnel lensesachieves an effective concentration ratio of 250. Each pieceof thin acrylic plate with 24 lenses, 4-mm thick and anti-reflection coated, is mounted on each sub-module. A totalof 1152 single-junction, single-crystal silicon cells, each1.2 cm2, are mounted at the focal points of a MegaModule.

ead Mercury Nickel

less)

CTE grid mix. Emissions are normalized for Southern European averageyears. Each PV system is assumed to include a ground-mounted BOS as

Fig. 9. Details of MegaModule in Amonix HCPV system, and complete 24 kW systems in APS STAR center in Phoenix (www.amonix.com).


This AM-10e single-crystal solar cell operates at 26.5%efficiency (Garboushian et al., 1997). Arizona Public Ser-vice (APS) has installed and tested several Amonix HCPVsystems with capacities of around 500 kWp, the largestbeing the 168 kWp installation at their PV plant near Pres-cott, Arizona.

The major components of this 5-MegaModule, 24 kWp

system include frames, optics, single-crystal Si cells, heatsinks, tracker, foundation, hydraulic drive, motor, con-troller, inverter, sensor, power meter, and anemometer.While the aperture area is 182 m2, the total area of theconcentrator is 230 m2 including frame areas. The trackerconsists of a 5.5 m high pedestal, a torque tube and ahydraulic drive with a controller. The concrete foundationused in the STAR test facility in Phoenix, from where thedata for this analysis were taken, is 5.5 m deep and 1.2 min diameter. Each 24 kWp MegaModule system has a30 kW Xantrex inverter and multiple MegaModules shareone transformer.

The direct beam solar radiation for concentrating collec-tors with a two-axis is, on average, 2480 kWh/m2/year inPhoenix, AZ (NREL, 1994). Performing with 16% ratedAC conversion efficiency, this system ideally should pro-duce 72 MWh (2480 � 0.16 � 182 m2) in a year under stan-dard conditions (25 �C ambient temperature, 1 m/s windspeed). However, the 24 kWp Amonix HCPV in Phoenix,Arizona, generated a net of 49.2 MWh in 2005, a sizeable30% difference (Herbert et al., personal communication).Losses associated with soiling, alignment, shadows between

units and other objects, temperature, equipment failuresexplain the discordance between the ideal and realperformance.

7.2. Materials and energy balance

7.2.1. Materials and components production

The two-axis tracker and its foundation account for 58%of the 24 kWp Amonix concentrator system. The concen-trator module (MegaModule), which includes frames,inside structure, Fresnel lenses, solar cells, and heat sinks,accounts for the next most significant contribution (41%).Steel is the material predominantly used for the HCPV sys-tem’s heavy structures and equipment in the tracker andconcentrator modules (Fig. 10). We calculated the energydemands and greenhouse-gas emissions during the manu-facturing and assembly of the components that constituteAmonix HCPV systems, based on commercial and publicdatabases (Table 5). Those processes encompassed areinjection molding, contour shaping, wire drawing, pipedrawing, section-bar extrusion, solar-cell manufacturing,electronics manufacturing, and transportation of materialsand parts.

7.2.2. Assembly and installation

The concentrator modules are assembled in manufactur-ing plant and transported to the field where other compo-nents are gathered, installed, and welded. After drilling ahole 1.2 m in diameter and 5.2 m deep, a 50-ton crane

http://www.amonix.com

Steel69%

Aluminum2%

Concrete24%

Plastics4%

Others1%

Fig. 10. Breakdown of materials used for the 24 kW, 21.3 metric tonAmonix HCPV system.


and a crew of four people place, assemble, and install theentire structure within half a day (Herbert et al., personalcommunication).

7.2.3. Maintenance and operation

We assumed that a monitoring and administrative officeis built, as in a real, commercial application. Office energyuse was estimated from a study on a PV BOS in the TucsonElectric flat-plate PV plant, Springerville, Arizona (Masonet al., 2006). A 2-hp pump runs intermittently for about anhour per day to track the sun during daytime under normaloperations. It pulls about 1.4 kW, which translates into adaily parasitic energy use of 1.4 kWh (Herbert et al., per-sonal communication).

7.2.4. End-of-life

The disposal of the Amonix HCPV components involvesdismantling, transporting, and shredding them. Weassumed that dismantling the components would requirethe same amount of energy as that used to install them.Also, we assumed that heavy trucks would transport the

Table 5Life-cycle primary energy demand and greenhouse-gas emissions (Kimand Fthenakis, 2006).

Stage Energy (%) GHG (%)

Components manufacturingMegaModule 58 54Controller 0.5 0.3Tracker 33 37Electrical 2.3 1.8Transportation 3.2 3.5Office 0.3 0.4Subtotal 97 97

Assembly and installation 0.1 0.2Operation and maintenance 1.1 0.9End-of-life disposal 1.7 1.9

Total primary energy = 817 GJ; total GHG = 56,000 kg CO2-eq. for a24 kWp system.

dismantled components to a disposal facility 100 km away.Shredding and separating them will take 0.34 MJ/kg ofenergy (Staudinger and Keoleian, 2001). We consideredthe energy source for dismantling, transporting, and shred-ding those components to be diesel fuel.

7.3. Life-cycle metrics

Table 5 breaks down the primary energy demand duringthe life cycle of the 24 kW Amonix HCPV. The componentmanufacturing stage dominates the demand, accountingfor 97% of the primary energy. Considering the 1.4 kWh/day of parasitic energy, the actual energy generated fromthis system is 48,700 kWh/year in Phoenix, which repre-sents 605 GJ of primary energy avoided in using the pri-mary to electric energy-conversion factor of 0.29 for theUS average grid mix (Franklin Associates, 1998). Thisresults in an EPBT of 1.3 years.

7.3.1. Greenhouse-gas emissions

The greenhouse-gas (GHG) emissions during the life-cycle stages of the 24 kWp Amonix HCPV are estimatedas an equivalent of CO2 using an integrated time horizonof 100 years; they include CO2, CH4, N2O, and chloroflu-orocarbons. Unlike fixed, standard PV configurations inwhich the emissions mostly are linked with manufacturingthe solar cells, the tracking and concentrating equipmentcontributes the majority of the GHG emissions from thisHCPV system. After normalizing for the electricity gener-ated in Phoenix, 48,700 kWh/year, this system generates38 g CO2-eq./kWh during its life cycle.

7.4. Discussions

Peharz and Dimroth recently investigated a FLATCONsystem that uses Fresnel lenses with a concentration rate of500, with III–V multi-junction solar cells (Peharz and Dim-roth, 2005). With an output power of 6 kWp, this systemincorporates 333 individual triple-junction modules thathave a DC conversion efficiency of 26%, which is 40%higher than the efficiency of the Amonix modules. Theinvestigators calculated that the EPBT of the FLATCONsystem was about 0.7-0.8 years under operating conditionsin Tabernas, Spain with direct solar radiation of 1900kWh/m2/yr. This value corresponds to an EPBT of 0.5-0.6 years when adjusted to 2480 kWh/m2/yr of direct solarinsolation with the 2-axis tracker observed in Phoenix(Table 6). This low EPBT is mainly attributed to the highefficiency of the triple-junction III–V solar cells. However,the stated annual electricity generation of the FLATCONsystem was not measured in the field; instead, it was esti-mated assuming an extremely low 6% loss of efficiency incables and inverters, and no heating loss (Peharz and Dim-roth, 2005). Furthermore, the authors did not fully accountfor losses in efficiency during field operation that stem fromhigh wind, dust, maintenance, tracking errors, and cell-focus errors. After adjusting for them, the advantage that

Table 6Comparison of life-cycle parameters and performances across PVtechnologies (normalized for Phoenix insolation).

PV system AmonixHCPV,24 kWp

current

AmonixHCPV,24 kWp,future

Mono c-Siground-mount*

CdTeground-mount**

Module DC efficiency (%) 18 18 14 9System loss (%) 32 17 20 20Insolation (kWh/m2/yr) 2480a 2480a 2370b 2370b

EPBT (years) 1.3 1.1 1.8 0.8GHGc (g CO2-eq./kWh) 38 31 35 18

* Adapted from Alsema et al. (2005) and de Wild-Scholten and Alsema(2005).** Adapted from Fthenakis et al. (2005).

a Direct normal insolation with two-axis tracker.b Longitude optimal insolation for flat-plate collectors.c One hundred years of integrated time horizon.


a FLATCON system has over an Amonix system in termsof EPBT seemingly involves cell efficiency. With minimaladjustments, the Amonix system can be adapted to mountIII–V solar cells. While this would reduce the EPBT, thedecrease would not be proportional to the increased elec-tricity generation since manufacturing multi-junction III–V solar cells involves many energy-intensive processes dif-ferent from those in making single-crystal solar cells (Mei-jer et al., 2003).

The Amonix HCPV exhibits an advantage over theflat, fixed single-crystal silicon solar cell in terms ofEPBT, but has a poorer performance for GHG emissions.This contrast between them mainly is related to the differ-ent energy sources for materials production. The citedstudies of silicon solar cells assume that high-purity poly-crystalline silicon, feedstock material for crystalline siliconingots, is produced by a combined hydropower and gas-fired cycle (Alsema and de Wild-Scholten, 2005; deWild-Scholten and Alsema, 2005). The Amonix HCPVincorporates a considerable number of steel structuresthe production of which consumes much energy fromburning coal (Franklin Associates, 1998). However, bycomparison, its life-cycle impact in terms of these twometrics is higher than that for thin-film CdTe PV at thesame location (Table 6).

These findings must be interpreted cautiously as theAmonix systems are not produced or installed in a scalecomparable to flat panel, Si- and CdTe-PV systems.Major components for the Amonix system includingthe Megamodules, trackers, and electrical equipmentare yet to be optimized. The data for producing the sin-gle-crystal Si material for this system were taken fromthe literature, and the accuracy of the amount of electric-ity generated year round currently is being validated.Therefore, the present analysis may represent the life-cycle performance of a prototype, near commercial-ver-sion concentrator system, as distinct from the commer-cial PV modules.

8. Life-cycle risk analysis

8.1. Risk classification

Perhaps the greatest potential risks of the PV fuelcycle are linked with chemical usage during materialsproduction and module processing (Hirschberg et al.,1998; Fthenakis, 2003). Although rare, accidents likeleakage, explosion, and fire of toxic and flammable-sub-stances are the major concerns to be addressed to ensurepublic acceptance of PV technology. But little is knownabout their frequency and scale in terms of human fatal-ities, injuries, and economic losses during the PV fuelcycle. In its burgeoning stage, the PV industry has notexperienced major accidents of the same scale as otherenergy sectors. Furthermore, the stages of the PV fuelcycle are closely connected to the semiconductor pro-cesses that characterizing PV risks, i.e., equitably allocat-ing risks to the PV and semiconductor industries posesmany difficulties.

The risks associated with energy technologies can beclassified into four types based on their scale, frequency,and the severity of harmful events (Fthenakis et al., 2006).

(1) Risk during normal operation-risks of which the con-sequences are typically accepted, for example, GHGemissions, toxic chemical emissions, routine radioac-tive emissions, chemical/radioactive waste, andresource (fuel, water) depletion.

(2) Risk of routine accidents-risk of accidents with highfrequency and low consequences, for example,small-scale leakage of chemicals, small-scale explo-sion/fires, transportation accidents, and small-scaleradioactivity release.

(3) Risk of severe accidents-risk of accidents with low fre-quency and high consequences, for example, coremeltdown, collapse of dam, and large-scale fire/explosion.

(4) Risk difficult-to-evaluate-risks subject to, and some-times reinforced by, perception. These include terror-ist attacks on reactor/used fuel storage, geopoliticalinstability, military conflicts, energy security/nationalindependence, and nuclear proliferation.

The first type of risk is characterized as routine undernormal conditions as opposed to accidents, and theirimpact usually is limited by safety measures often estab-lished by regulation. This type of risk usually is deter-mined by analytical tools, such as life-cycle analysis(LCA) and impact pathway analysis. The second andthird types cover anomalous events or accidents. Thegeneral public is more concerned about the third typeof risk, low-probability catastrophic events, than the sec-ond type, high-probability less severe accidents. Thefourth type of risk is often associated with the public’sperception, and the probability and consequences are dif-ficult to evaluate (Haimes, 1998).


8.2. Risks of accidents in the PV life cycle

The greatest potential risks of the PV life cycle involvethe toxic and flammable chemicals used for producing PVmaterials and for manufacturing modules (Fthenakis,2003). Fthenakis and Kim undertook a life-cycle risk anal-ysis based on accident records from a national database,the Risk Management Program (RMP). The study focusedon the stages of handling and using the chemicals whilemaking solar-cell materials and the modules (Fthenakiset al., 2006). The risks of potential accidents during instal-lation, operation, and disposing/recycling PV modules maybe negligible, as then, chemical usages are little orunnecessary.

According to the Clean Air Act Amendments of 1990,the US Environmental Protection Agency (EPA) is to pub-lish regulations and guidance for facilities that use extre-mely hazardous substances to prevent chemical accidents.The US EPA developed the Risk Management ProgramRule to implement the section 112 (r) of the amendments.This rule requires the facilities that produce, use, or handletoxic or flammable chemicals of concern in quantities overspecific thresholds to submit a Risk Management Plan(RMP). Regulated substances and their threshold quanti-ties are listed under section 112 (r) of the Clean Air Actin 40 CFR Part 68. The RMP should include a 5-year acci-dent history detailing the cause of accidents and the dam-ages to the employees, environment, and local public.Other RMP components include a hazard assessment forthe most-likely and the worst-case accident scenarios, aprevention program covering safety precautions and main-tenance, monitoring, and employee training measures, andan emergency response program that describes emergency

Table 7Incident records and estimated production for substances used in PV operatio

Substance Source Total averUS produc

Toxic

Arsine (AsH3) GaAs CVD 23Boron trichloride (BCl3) p-Type dopant for epitaxial silicon NABoron trifluoride (BF3) p-Type dopant for epitaxial silicon NADiborane (B2H6) a-Si dopant �50Hydrochloric acid (HCl) Cleaning agent for c-Si 3500Hydrogen fluoride (HF) Etchant for c-Si 190Hydrogen selenide (H2Se) CIGS selenization NAHydrogen sulfide (H2S) CIS sputtering >110Phosphine (PH3) a-Si dopant NA

Flammable

Dichlorosilane (SiH2Cl2) a-Si and c-Si deposition NAHydrogen (H2) a-Si deposition/GaAs 18,000 m3

Silane (SiH4) a-Si deposition 8Trichlorosilane (SiHCl3) Precursor of c-Si 110

Sources: Williams (2000), EPA (1992), Census of Bureau (2002), Board on Env(2003), Agency for Toxic Substances and Disease Registry (2004), and Pichel

a Number excludes incidents in petroleum refineries (NAICS 32411).

health care, employee training measures, and proceduresfor informing the public and response agencies (Environ-mental Protection Agency, 2002). In 1999, around 15,000relevant facilities in the United States first submitted theiraccident records during the previous 5-year period, alongwith other documents. The current (2005) updated RMPsmostly cover accidents that occurred since mid-1999 for17,000 country-wide facilities. The program must be resub-mitted every 5 years and revised whenever there is a signif-icant change in usage.

Fthenakis et al. estimated the risks of the chemicals usedin the PV life cycle, i.e., accidents, fatalities, and injuriesduring the production, storage, and delivery of the flamma-ble and toxic chemicals listed in 40 CFR Part 68, based onthe current and past histories in the RMPs reported (Fthe-nakis et al., 2006). Table 7 shows the number of accidents/death/injuries between 1994 and 2004 for substancesinvolved in PV materials and module production. Sincethere are very few data on the production/consumptionof trichlorosilane (SiHCl3), estimates are given based onpolysilicon production and capacity data. About 2 kg ofpolysilicon are needed to produce a module of 6 � 12 cellsof 125 mm � 125 mm, and an input of 11.3 kg of trichlo-rosilane is required to generate 1 kg of polysilicon (Alsemaand de Wild-Scholten, 2005; Williams, 2000). No accidentswere reported in the RMP database for phosphine anddiborane in these years.

We note that the number of occurrences shown inTable 7 represents the total for all the US facilities thatproduce, process, handle, and store these chemicals inquantities over their specified thresholds. These statisticsmay include data irrelevant to the PV industry. For exam-ple, the majority (61%) of accidents/death/injuries attrib-

n and regulated under the RMP rule.

age (1994–2000)tion (1000 tonne/yr)

Number of incidents (employees and public)in the RMP database (1994–2004)

Incidents Injuries Deaths

2 1 00 0 01 1 00 0 0

28 12 1165 (57)a 209 (70)a 1

4 17 040 47 10 0 0

2 0 057 65 45 2 0

14 14 2

ironmental Studies and Toxicology (Best) (2003), U.S. Geological Surveyand Yang (2005).

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 99-04

Inju

ries/

GW

yr

10-6

10-5

10-4

10-3

10-2

10-1

100

101

Fig. 11. Estimated injury rates of c-Si per GWyr electricity produced in the United States based on the year of accidents for HF, HCl, and SiHCl3(insolation = 1800 kWh/m2/year, performance ratio = 0.8).


uted to hydrogen fluoride (HF) occurred in petroleumrefineries (NAICS4 32411), although refinery use accountsfor only 6% of the US consumption (EPA, 1992). The pet-rochemical industry uses 100% HF under high pressure inlarge, multi-component units (e.g., alkylation units withmany pipes, fittings, valves, compressors, and pumps) fromwhere two-phase releases may occur, whereas the PVindustry uses only aqueous solutions of HF (typically49 wt%) in etching baths. Moreover, the usage of HF andhydrogen chloride (HCl) in the US PV industry accountsfor less than 0.1% of the total in the United States. There-fore, we excluded HF use in the petroleum industries andthe corresponding accidents from our analysis of risks inthe PV fuel cycle.

To normalize the rate of accident/death/injuries, the fig-ures are divided by the amount produced for those chemi-cals in the United States. The risks of each substance usedin a 25 MWp/year scale PV industry are determined basedon the amount of materials in Table 2. Then, the number ofaccidents/death/injuries per GWyr of electricity generatedis determined as a risk indicator of chemical accidents inthe PV fuel cycle based on the average US insolation of1800 kWh/m2/year and a performance ratio of 0.8 (i.e.,20% system loss).

Fig. 11 depicts the rates of accidents/death/injuries forc-Si per GWyr electricity produced based on the averageUS insolation, 1800 kWh/m2/year. The last column withthe most recent incident data (i.e., data submitted from1999 onwards) may better represent the current evolutionof the fast-growing PV industry. In general, for most chem-icals, the numbers of accidents reported for the secondcycle in the RMP (second half of 1999–2004) are lowerthan those reported in the first cycle (1994 to the first halfof 1999). Specifically, only one accident involving silane

4 North American Industry Classification System.

was reported during the second cycle of the RMP, whereasfour accidents occurred during the first. Likewise, the num-ber of accidents during the second RMP cycle involving tri-chlorosilane fell from 13 to 1 between 1994–1999 and 1999–2004. This across-the-board reduction of incidents likelyrepresents improved safety in the whole US PV industry.For c-Si-PV modules, SiHCl3, the feedstock of polysiliconpresents greater risks than other chemicals (Fig. 12). Onthe other hand, limited statistics suggest that silane posesthe greatest risk in a-Si module manufacturing. The fatali-ties in Fig. 12 are related to HF. However, this risk is notbased on the real events in PV facilities but is a virtual risk,derived from accident rates in other industries and theamount consumed in the PV industry. The real risk ofHF in the PV industry, however, is likely to be lower thanthis current estimate due to the different characteristics ofprocesses across industry sectors.

8.3. Comparison with other energy technologies

Fig. 13 compares fatality and injury rates across conven-tional electricity technologies and PV technologies. The fig-ures for the conventional fuel cycles were extracted fromthe compilation of severe accident records of the GaBEproject by the Paul Scherrer Institute (PSI), Energy-relatedSevere Accident Database (ENSAD) from 1969 to 2000(Hirschberg et al., 2004). Although such a direct compari-son may not be entirely appropriate, the data imply thatthe expected risks in the PV fuel cycle are lower than inother technologies, as we explain below. Thus, there areseveral caveats in directly comparing our estimate of PVrisks and the risks estimated by the PSI’s investigators.First, Hirschberg et al.’s assessments (2004) are based onsevere accidents only, and small-scale accidents areignored; the former are defined as events with at least 5fatalities, 10 injuries, $5 million of property damage, or

HCl HF SiHCl3 c-Si (total) a-Si

Inci

dent

/GW

yr

10-6

10-5

10-4

10-3

10-2

10-1

Death Injuries

Fig. 12. Estimated incident rates by chemicals used between 1999 and2004 (insolation = 1800 kWh/m2/year, performance ratio = 0.8). The rateof a-Si covers 1997–2004.


200 evacuees. On the other hand, only 20 out of the 318incidents in Table 7 can be classified as severe accidentsunder the same definition. Therefore, the PSI’s values forrisk of conventional energy technologies may be underesti-mates. Also, we expect the safety records of the evolvingPV industry to improve with time, whereas those for themature conventional energy technologies are less likely tochange. The risk estimate for the nuclear cycle in Fig. 13relies on the probabilistic safety assessment (PSA) toolfor the OECD countries rather than historical evidencedue to insufficient experience related to the nuclear acci-dents. Although PSA is a logical approach to identify plau-sible accidents, their rates and consequences, it is notproven to provide absolute risk estimates due to its limita-

Coal, OECD

Coal, non-OECDOil, OECD

Oil, non-OECD

NG, OECD

NG, non-OECD

LPG, OECD

LPG, non-OEHy

Inci

dent

s/G

Wyr

10-6

10-5

10-4

10-3

10-2

10-1

100

101

102

Fatalities Injuries

Fig. 13. Comparison of fatality and injury rates across electricity-generatioperformance ratio = 0.8 was assumed. The incident rates for coal, oil, NG, LPGet al. (2004) (NG, natural gas; LPG, liquefied petroleum gas; � module type isNational Laboratory (BNL) based on US data (Fthenakis et al., 2006)).

tions associated with data uncertainty; for example, risks ofunexpected events or those triggered by human errors areimpossible to fully account for (Linnerooth-Bayer andWahlstrom, 1991; Nuclear Energy Agency, 1992). For thesimilar reason, the PSI’s estimate for the PV cycle reliedon data of the chemical industries handling similar sub-stances used in the PV cycle (Peter et al., personalcommunication).

We also examined the maximum consequences of eachenergy technology. People are rarely neutral about risk;decision-makers, risk analysts, and the public are moreinterested in unforeseen catastrophes, such as bridges fall-ing, dams bursting, and nuclear reactors exploding thanin adverse, but routine events, such as transportation acci-dents. Comparing low-probability/high-damage risks withhigh-probability/low-damage events within one expectedvalue frame often distorts the relative importance of conse-quences across technology options. Therefore, describingthe maximum consequence potential makes sense (Haimes,1998). Fig. 14 shows the maximum fatalities: figures for thefossil fuel and nuclear cycles are from a database (Hirsch-berg et al., 1998) and for PV from literature (Fthenakiset al., 2006; Hirschberg et al., 2004). From a scale of con-sequence perspective, PV technology is remarkably saferthan other technologies. We anticipate that the maximumconsequence will remain at the same level shown inFig. 14 unless there is a significant change in PV productiontechnologies.

The nature of risks varies with different energy technol-ogies. The PSI analysis focuses on fuel mining, fuel conver-sion, power-plant operation, and transportation of fuels.On the other hand, fuel-related or power-plant-relatedrisks are non-existent in the PV fuel cycle. Instead, our

CD

dro, OECD

Hydro, non-OECD

Nuclear, OECD

Nuclear, non-OECDPV, PSI*

PV, c-Si, BNL**

PV, a-Si, BNL**

n technologies. The average US insolation = 1800 kWh/m2/year and a, hydro-, nuclear-, and PV technologies given by PSI are from Hirschbergnot specified and injuries are not estimated; �� estimates of Brookhaven

Coal Oil NG LPG

Nuclear (Chernobyl)

Nulcear (except Chernobyl)PV, PSI

PV, BNL

Fata

litie

s

100

101

102

103

104

105

434

3000

100

600

125

2

9000-33000

100

Fig. 14. Maximum fatalities from accidents across energy sectors. The number for Chernobyl includes latent fatalities. The incident rates for coal, oil, NG,LPG, hydro, nuclear, and PV given by PSI are from Hirschberg et al. (1998) (NG, natural gas; LPG, liquefied petroleum gas).


analysis of PV risks focuses on accidents associated withfeedstock materials as well as process consumables, i.e.,upstream risks, which the PSI analysis does not include.

8.4. Limitation of study

As discussed, the hazards of trichlorosilane when mak-ing modules of c-Si, and of silane for a-Si modules havedominated concerns over other chemicals due to the largeamount required and their flammability. Their limitednumber of incident records in the RMP database preventsour accurately measuring the safety of the PV industry.Since death records are very rare for these chemicals, therisk of fatalities is highly sensitive to a single incident. Onthe other hand, injury rates are relatively stable againstthe incremental number of injuries. This illustrates thatthe scale of the PV industry in terms of capacity andemployees still is small so that such analyses are inadequateto directly compare with other technologies. PV technologystill is in the early stage of commercial application, and,with time, Risk Management Programs complied throughexperience eventually should stabilize the number of abnor-mal incidents. Other technologies, such as nuclear power,experienced a similar period during the early years of com-mercialization (Hirschberg et al., 1998).

Although PV technology is at an early stage comparedwith other energy systems, it is rapidly growing withinthe context of EU policy to increase the contribution ofrenewable energies to the total EU energy mix (Directive2001/77/EC of the European Parliament and of the Coun-cil, 2001; European Commission, 1997). Accordingly, thereare strong efforts to boost the PV sector, and therefore, it

should be treated more and more as an energy sector,avoiding comparisons with the chemical or semiconductorindustry. Our analysis of risk in the PV life cycle is at alevel lower than many other technologies. Given this situa-tion, the PV industry should help by sharing informationon risk, which, at the same time, could surely give PV morecredibility and transparency.

9. Outlook

Alsema et al. (2006) recently outlined the majorimprovements in materials and energy consumption as wellas conversion efficiencies which expected to be realizedwithin a few years in the crystalline-Si PV sector. Theyforecast that the efficiency of ribbon-, multi-, and mono-Si module will improve to 15%, 17%, and 19%, respectively,in near future, in accordance with the target established bythe Crystal Clear project. A fluidized bed reactor (FBR)currently being deployed will be able to reduce the energydemand for polysilicon by 70–90% from the popular Sie-mens process although it is unconfirmed if this new reactortype is capable of producing the same high-purity polysil-icon as the latter (Alsema et al., 2006; Fthenakis andKim, 2007a). At the same time, Si wafers will become thin-ner: 150 lm for multi- and mono-Si and 200 lm for rib-bon-Si. Also, the energy demand during the casting ofmulti-crystalline ingots and the Czochralski growing of Simono-crystal will fall as much as 3-fold (Alsema et al.,2006). The corresponding EPBT and GHG emissions arepresented in Fig. 15.

We discuss the future of life-cycle GHG emissions fromCdTe PV in a separate paper (Fthenakis and Kim, 2007a).

(a)

Ribbon-Si Multi-Si Mono-Si CdTe*

Yrs

0.0

0.5

1.0

1.5

2.0

2.5

3.0

Current Future

(b)

Ribbon-Si Multi-Si Mono-Si CdTe*

g C

O2-

eq./k

Wh

0

10

20

30

40

Fig. 15. Future forecast for (a) EPBT and (b) GHG emissions from thelife cycles of PV modules (frame and BOS are not shown). Unlessotherwise noticed, the estimates are based on the Southern Europeaninsolation, 1700 kWh/m2/year, a performance ratio of 0.75, and a lifetimeof 30 years (*Based on the average US insolation of 1800 kWh/m2/yearand a performance ratio of 0.8).


The US manufacturer of CdTe PV predicts: (1) a linearincrease in electrical-conversion efficiency from the current9% to 12% by 2010; (2) a reduction of electricity require-ments by about 25% within a couple of years through opti-mization of the deposition processes in CdTe lines; (3)about 20% of the manufacturing requirement will be satis-fied via on-site solar electric generation. For a plant manu-facturing 25 MWp/year of 12% efficient PV modules, thelatest scenario will require a 1.3 MWp PV installation on2.7 acres, which is area available on the roof and parkingof the facility. Then, the EPBT would fall to 0.4 yearsand the GHG emissions to 10 g CO2-eq./kWh for the lifecycle of installed CdTe PV under the average US insola-tion, 1800 kWh/m2/year (Fig. 15).

10. Conclusions

This review offers a snapshot of the rapidly evolving life-cycle performances of photovoltaic (PV) technologies andunderlines the importance of timely updating and reportingthe changes. During the life cycle of PV, emissions to theenvironment mainly occur from using fossil-fuel-basedenergy in generating the materials for solar cells, modules,and systems. These emissions differ in different countries,

depending on that country’s mixture in the electricity grid,and the varying methods of material/fuel processing. Thelower the energy payback times (EPBT), that is the timeit takes for a PV system to generate energy equal to theamount used in its production, the lower these emissionswill be. Under average US and Southern Europe condi-tions (e.g., 1700 kWh/m2/year), the EPBT of ribbon-Si,multi-crystalline Si, mono-crystalline Si, and CdTe systemswere estimated to be 1.7, 2.2, 2.7, and 1.0–1.1 years, corre-spondingly. The EPBT of CdTe PV is the lowest in thegroup, although electrical-conversion efficiency was thelowest; this was due to the low energy requirement in man-ufacturing CdTe PV modules. We also report the potentialenvironmental impacts during the life cycle of a 24 kWAmonix HCPV system which is being tested for optimiza-tion. The estimated EPBT of this system operating in Phoe-nix, AZ, is about 1.3 years and the estimated GHGemissions are 38 g CO2-eq./kWh. The EPBT of the Amonixmono-Si HCPV is shorter than that of a flat-plate mono-Si-PV ground-mount system, whereas GHG emissions arehigher.

The indirect emissions of Cd due to energy used in thelife cycle of CdTe PV systems are much greater than thedirect emissions. CdTe PV systems require less energyinput in their production than other commercial PV sys-tems, and this translates into lower emissions of heavymetals (including Cd), as well as SO2, NOx, PM, andCO2 in the CdTe cycle than in other commercial PVtechnologies. However, regardless of the particular tech-nology, these emissions are extremely small in compari-son to the emissions from the fossil-fuel-based plantsthat PV will replace.

The greatest potential risks of the PV fuel cycle arelinked with the use of several hazardous substances,although in quantities much smaller than in the processindustries. In an effort to measure the potential risks asso-ciated with the PV fuel cycle in comparison with other elec-tricity-generation technologies, we used the US EPA’sRMP accident records that cover the entire major USchemical storage and processing facilities. Our analysisshows that, based on the most recent records, the PV fuelcycle is much safer than conventional sources of energyin terms of statistically expected, and by far the safest interms of maximum consequence. On the other hand, inves-tigations are under way into the land and water uses duringthe PV fuel cycles in a comparative context, which is a rel-atively unknown territory of LCA (Koehler, 2008; Fthena-kis et al., 2009).

Apparently, the PV industry is striving for cost savingssimultaneously for advanced performance, which largelytranslates into life-cycle energy savings and emissionsabatement. The conversion efficiency, material usage, andproduction energy efficiency of both Si and CdTe PVs areimproving rapidly, thereby we also expect the risks associ-ated with the life cycles of PV systems to be reduced. Fre-quent updates of these analyses are necessary to follow thisevolution.


References

Agency for Toxic Substances and Disease Registry, 2004. ToxicologicalProfile for Hydrogen Sulfide.

Alsema, E.A., 2000. Energy pay-back time and CO2 emissions of PVsystems. Progress in Photovoltaics: Research and Applications 8, 17–25.

Alsema, E., de Wild-Scholten, M., 2005. Environmental impact ofcrystalline silicon photovoltaic module production. In: MaterialResearch Society Fall Meeting, Symposium G: Life Cycle AnalysisTools for ‘‘Green” Materials and Process Selection, Boston, MA.

Alsema, E.A., de Wild-Scholten, M.J., Fthenakis, V.M., 2006. Environ-mental impacts of PV electricity generation – a critical comparison ofenergy supply options. In: 21st European Photovoltaic Solar EnergyConference, Dresden, Germany.

Aulich, H.A., Schulze, F.-W., 2002. Crystalline silicon feedstock for solarcells. Progress in Photovoltaics: Research and Applications 10, 141–147.

Australian Coal Industry Association Research Program (ACARP), 2004.Coal in a Sustainable Society.

Board on Environmental Studies and Toxicology (BEST), 2003. AcuteExposure Guideline Levels for Selected Ariborne Chemicals, vol. 3.The National Academic Press.

Census of Bureau, 2002. Current Industrial Reports – Inorganic Chem-icals. Available from: <http://www.census.gov/manufacturing/cir/index.html>.

de Wild-Scholten, M., Alsema, E., 2005. Environmental life cycleinventory of crystalline silicon photovoltaic module production. In:Material Research Society Fall Meeting, Symposium G: Life CycleAnalysis Tools for ‘‘Green” Materials and Process Selection, Boston,MA.

de Wild-Scholten, M.J., Alsema, E.A., ter Horst, E.W., Bachler, M.,Fthenakis, V.M., 2006. A cost and environmental impact comparisonof grid-connected rooftop and ground-based PV systems. In: 21stEuropean Photovoltaic Solar Energy Conference, Dresden, Germany.

Directive 2001/77/EC of the European Parliament and of the Council,2001. Promotion of Electricity Produced from Renewable EnergySources in the Internal Electricity Market, 27 September, 2001.

Dones, R. et al., 2003. Sachbilanzen von Energiesystemen. Final ReportEcoinvent 2000, vol. 6. Swiss Centre for LCI, PSI.

Electric Power Research Institute (EPRI), 2002. PISCES Data Base forUS Power Plants and US Coal.

Environmental Protection Agency, 2002. RMP Program Overview.<http://yosemite.epa.gov/oswer/Ceppoweb.nsf/content/RMPover-view.htm/>, October 11th, 2002 (cited 22.02.07).

EPA, 1992. Hydrogen Fluoride Study. Report to Congress. Office of SolidWaste and Emergency Response, US Environmental ProtectionAgency, Washington, DC.

European Commission, 1997. Communication from the Commission“Energy for the Future: Renewable Sources of Energy”. White Paperfor a Community Strategy and Action Plan.

European Commission, 2003. External Costs. Directorate-General forResearch, Office for Official Publications of the EC: Luxembourg,EUR 20198.

Franklin Associates, 1998. USA LCI Database Documentation. PrairieVillage, Kansas.

Fthenakis, V., 2003. Overview of Potential Hazards. In: Markvart, T.,Gastaner, L. (Eds.), Practical Handbook of Photovoltaics: Funda-mentals and Applications. Elsevier.

Fthenakis, V.M., 2004. Life cycle impact analysis of cadmium in CdTe PVproduction. Renewable and Sustainable Energy Reviews 8 (4), 303–334.

Fthenakis, V., Alsema, E., 2006. Photovoltaics energy payback times,greenhouse gas emissions and external costs: 2004–early 2005status. Progress in Photovoltaics: Research and Applications 14,275–280.

Fthenakis, V.M., Kim, H.C., 2005. Energy use and greenhouse gasemissions in the life cycle of CdTe photovoltaics. In: Material Research

Society Fall Meeting, Symposium G: Life Cycle Analysis Tools for‘‘Green” Materials and Process Selection, Boston, MA.

Fthenakis, V.M., Kim, H.C., 2007a. Greenhouse-gas emissions from solarelectric- and nuclear power: a life-cycle study. Energy Policy 35, 2549–2557.

Fthenakis, V.M., Kim, H.C., 2007b. CdTe photovoltaics: Life cycleenvironmental profile and comparisons. Thin Solid Films 515, 5961–5963.

Fthenakis, V.M. et al., 2005. Emissions and encapsulation of cadmium inCdTe PV modules during fires. Progress in Photovoltaics: Researchand Applications 13, 713–723.

Fthenakis, V.M. et al., 2006. Evaluation of risks in the life cycle ofphotovoltaics in a comparative context. In: 21st European Photovol-taic Solar Energy Conference, Dresden, Germany.

Fthenakis, V.M., Kim, H.C., Alsema, E., 2008. Emissions from photo-voltaic life cycles. Environmental Science & Technology 42, 2168–2174.

Fthenakis, V., Wang, W., Kim, H.C., 2009. Life cycle inventory analysisof the production of metals used in photovoltaics. Renewable andSustainable Energy Reviews 13, 493–517.

Fthenakis, V., Kim, H.C., 2009. Land use and electricity generation: a life-cycle analysis. Renewable and Sustainable Energy Reviews 13, 1465–1474.

Garboushian, V., Roubideaux, G., Turner, G., Gunn, J.A., 1997. Long-Term Reliability Concerns Resolved by Third Generation IntegratedHigh-Concentration PV Systems, 26th IEEE Photovoltaic SpecialistConference. Anaheim, CA.

Haimes, Y.Y., 1998. Risk Modeling, Assessment, and Management,fourth ed. John Wiley & Sons.

Hagedorn, G.,1992. Kumulierter Energieaufwand von Photovoltaik undWindkraftanlagen Technischer Verlag Resch KG, Munich, Germany.

Herbert, H., personal communication. Arizona Public Service.Hirschberg, S., Spikerman, G., Dones, R., 1998. Severe Accidents in the

Energy Sector, first ed. Paul Scherrer Institute, Switzerland.Hirschberg, S. et al., 2004. Sustainability of Electricity Supply Technol-

ogies under German Conditions: A Comparative Evaluation. PaulScherrer Institute, Switzerland.

Hynes, K.M., Baumann, A.E., Hill, R., 1994. An assessment of theenvironmental impacts of thin-film cadmium telluride modules basedon life cycle analysis. In: IEEE 1st World Conference on PhotovoltaicConversion (WCPEC), Hawaii.

Intergovernmental Panel on Climate Change, 2001. Climate Change2001: The Scientific Basis. Cambridge University Press, Cambridge,UK.

Jungbluth, N., 2005. Life cycle assessment of crystalline photovoltaics inthe Swiss ecoinvent database. Progress in Photovoltaics: Research andApplications 13 (8), 429–446.

Kato, K. et al., 2001. A life-cycle analysis on thin-film CdS/CdTe PVmodules. Solar Energy Materials & Solar Cells 67, 279–287.

Kim, H.C., Fthenakis, V.M., 2006. Life cycle energy demand andgreenhouse gas emissions from an Amonix high concentrator photo-voltaic system. In: IEEE Fourth World Conference on PhotovoltaicEnergy Conversion, Hawaii.

Koehler, A., 2008. Water use in LCA: managing the planet’sfreshwater resources. International Journal of Life Cycle Assess-ment 13, 451–455.

Linnerooth-Bayer, J., Wahlstrom, B., 1991. Applications of ProbabilisticRisk Assessments: The Selection of Appropriate Tools. Risk Analysis11 (2), 239–248.

Mason, J.E., Fthenakis, V.M., Hansen, T., Kim, H.C., 2006. Energypayback and life-cycle CO2 emissions of the BOS in an optimized 3.5MW PV installation. Progress in Photovoltaics: Research and Appli-cations 14, 179–190.

Maycock, P.D., 2005. National Survey Report of PV Power Applicationsin the United States 2004. International Energy Agency.

Meijer, A. et al., 2003. Life-cycle assessment of photovoltaic modules:comparison of mc-Si, InGaP and InGaP/mc-si solar modules. Progressin Photovoltaics: Research and Applications 11 (4), 275–287.

http://www.census.gov/manufacturing/cir/index.html

http://www.census.gov/manufacturing/cir/index.html

http://yosemite.epa.gov/oswer/Ceppoweb.nsf/content/RMPoverview.htm/

http://yosemite.epa.gov/oswer/Ceppoweb.nsf/content/RMPoverview.htm/


NREL, 1994. Solar Radiation Data Manual for Flat-Plate and Concen-trating Collectors. National Renewable Energy Laboratory, Golden,CO.

Nuclear Energy Agency, 1992. Probabilistic safety assessment: an analyt-ical tool for assessing nuclear safety. In: NEA Issue Brief. Availablefrom: <http://www.nea.fr/html/brief>.

Palz, W., Zibetta, H., 1991. Energy pay-back time of photovoltaicmodules. International Journal of Solar Energy 10, 211–216.

Peharz, G., Dimroth, F., 2005. Energy payback time of the high-concentration PV system FLATCON. Progress in Photovoltaics:Research and Applications 13, 627–634.

Peter, B., personal communication.Pichel, J.W., Yang, M., 2005. 2005 Solar Year-End Review & 2006 Solar

Industry Forecast. Piper Jaffray & Co..Raugei, M., Bargigli, S., Ulgiati, S., 2007. Life cycle assessment and energy

pay-back time of advanced photovoltaic modules: CdTe and CIScompared to poly-Si. Energy 32, 1310–1318.

Rogol, M., 2005. Silicon and the solar sector: mapping a new world. In:Presentation at the Second Solar Silicon Conference, 11 April, 2005,Munich, Germany.

Staudinger, J., Keoleian, G.A., 2001. Management of End-of-LifeVehicles (ELVs) in the US. Center for Sustainable Systems, Universityof Michigan.

Swanson, R.M., 2003. Photovoltaic concentrators. In: Luque, A., Hege-dus, S. (Eds.), Handbook of Photovoltaic Sciences and Engineering.John Wiley & Sons.

U.S. Geological Survey, 2003. Mineral Commodity Summary: Arsenic.Williams, E., 2000. Global Production Chains and Sustainability: The

Case of High-Purity Silicon and its Applications in IT and Renewable

Energy. Institute of Advanced Studies, The United Nations University,Tokyo, Japan.

Woditsch, P., Koch, W., 2002. Solar grade silicon feedstock supply for PVindustry. Solar Energy Materials & Solar Cells 72, 11–26.

Vasilis Fthenakis is Senior Scientist and the Head of the PhotovoltaicsEnvironmental Research Center at Brookhaven National Laboratory ina joint appointment with Columbia University where he is a Professorof Earth and Environmental Engineering and the Founder and Directorof the Center for Life-Cycle Analysis. He is the author of 200 publi-cations, member of the Editorial Boards of Progress in Photovoltaicsand the Journal of Loss Prevention. He also a Fellow of the AmericanInstitute of Chemical Engineers, a Fellow of the International EnergyFoundation, and member of several energy and sustainability expertpanels.

Hyung Chul Kim received his bachelor’s and master’s degrees from theInorganic Materials Engineering at Seoul National University, and Ph.D.from the School of Natural Resources and Environment at the Universityof Michigan, Ann Arbor. In 2005, he joined the PV EnvironmentalResearch Center at the Brookhaven National Laboratory where heworked as a post doctoral research associate. Since May 2008, he has beenworking as an Associate Research Scientist in the Center for Life-CycleAnalysis at Columbia University. His research interests center on the LCAof products and energy systems including automobile, refrigerator,photovoltaics, and nuclear power. He also serves as a member of IEA’sTechnical Committee for Task 12, Environmental Health and SafetyIssues of PV.

http://www.nea.fr/html/brief

photovoltaics: life-cycle analyses - brookhaven national laboratory

Documents