Energy Balance of the Global Photovoltaic (PV) Industry - Is the PVIndustry a Net Electricity Producer?Michael Dale*,† and Sally M. Benson‡

†Global Climate and Energy Project, Stanford University, United States, and‡Energy Resources Engineering, Stanford University, United States

*S Supporting Information

ABSTRACT: A combination of declining costs and policy measuresmotivated by greenhouse gas (GHG) emissions reduction and energysecurity have driven rapid growth in the global installed capacity ofsolar photovoltaics (PV). This paper develops a number of uniquedata sets, namely the following: calculation of distribution of globalcapacity factor for PV deployment; meta-analysis of energyconsumption in PV system manufacture and deployment; anddocumentation of reduction in energetic costs of PV systemproduction. These data are used as input into a new net energyanalysis of the global PV industry, as opposed to device level analysis.In addition, the paper introduces a new concept: a model trackingenergetic costs of manufacturing and installing PV systems, includingbalance of system (BOS) components. The model is used to forecast electrical energy requirements to scale up the PV industryand determine the electricity balance of the global PV industry to 2020. Results suggest that the industry was a net consumer ofelectricity as recently as 2010. However, there is a >50% that in 2012 the PV industry is a net electricity provider and will “payback” the electrical energy required for its early growth before 2020. Further reducing energetic costs of PV deployment willenable more rapid growth of the PV industry. There is also great potential to increase the capacity factor of PV deployment.These conclusions have a number of implications for R&D and deployment, including the following: monitoring of the energyembodied within PV systems; designing more efficient and durable systems; and deploying PV systems in locations that willachieve high capacity factors.

■ INTRODUCTION

The amount of sunlight energy that flows through an area thesize of Connecticut (15,000 km2) each second is roughly equalto the average energy demand of the whole of human society(15 TW). Given the magnitude of the solar energy flow, thedirect conversion of sunlight to electricity via photovoltaic (PV)systems provides an opportunity for human society to meet asignificant amount of its energy needs from a low carbon,renewable source.The global installed capacity of PV grew at an average rate of

40% per year in the period 2000−2010.1−3 Global installedcapacity in 2010 was around 40 GW, split between six main PVtechnologies: single-crystal (sc-Si) and multicrystalline (mc-Si)silicon (which together comprise around 90% of the market),amorphous silicon (a-Si), ribbon silicon, cadmium telluride(CdTe), and copper indium gallium diselenide (CIGS) asshown in Figure 1. Growth has been particularly rapid (over100% per year in the period 2008−2010) in thin filmtechnologies, especially CdTe and CIGS.Because PV-generated electricity emits few greenhouse gases

(GHG) during its lifetime operation, governments worldwidepromote PV as a GHG emissions reduction strategy (otherreasons include energy security, health, and other environ-mental concerns). Many governments have policies that

support renewable energy technologies through mandatedproportions of renewable electricity generation, such asCalifornia’s Renewables Portfolio Standard (Senate Bill X1-02) of 33% by 2020;4 financial incentives for systeminstallation, such as the California Solar Initiative;5 orguaranteed prices paid on electricity generated and sold backto the grid, such as the Feed-In Tariffs Scheme (FITs) inGermany and the UK.6 Such policies are aimed at providing amarket for PV systems to stimulate production and drive downfinancial costs of PV deployment. The success of these policiesis typically judged by monitoring the financial cost of PVmodules and the installed capacity of PV systems, not by theirsuccess at reducing GHG emissions.Primary energy ‘consumption’, with associated GHG

emissions, occurs throughout the full life-cycle of a PV systemfrom extraction of raw materials, through panel and BOSmanufacture, system operation and maintenance, and finallydisposal. This life-cycle primary energy consumption is termedthe cumulative energy demand (CED).7 The time taken for a

Received: September 25, 2012Revised: February 15, 2013Accepted: February 26, 2013Published: February 26, 2013

Article

pubs.acs.org/est

© 2013 American Chemical Society 3482 dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−3489

system to ‘pay back’ its CED is termed the energy payback time(EPBT). Issues with definition and usage of EPBT arediscussed in refs 8−10.Energy technologies with long EPBT (due to high CED or

low capacity factor) may constrain industry growth and rapidtransition to a low GHG emission energy sector.11 A rapidlygrowing PV industry may even temporarily exacerbate theproblem of GHG emissions if operating at a net energy loss, i.e.consuming more energy each year in manufacturing andinstalling PV systems than is produced by systems in operation.Previous research has addressed the material constraints on

PV system deployment.12 There has been a long-standing myththat PV systems are unable to generate a positive net energyreturn over their lifetime.13 While it is clear that individual PVsystems can generate a positive net energy yield, the presentanalysis focuses on the industry-scale issue by addressing thequestion “Is the global PV industry a net electricity producer?”.If solar PV is to make an important contribution to the globalenergy system, the industry must eventually deliver moreenergy to society than it consumes, i.e. it must produce apositive net energy yield. When the PV industry is a smallproportion of the overall energy sector, the effects of its being anet energy consumer or provider are negligible. As the industrygrows the burden on the overall energy sector becomes greater.This burden may even stifle growth of the PV industry. Hence,there is a large benefit to the PV industry in reducing energeticcosts of PV deployment.This article first explores the general relationship between

energy inputs and energy production at the system level andthe energy balance of an industry with installed capacity of anumber of such systems. This relationship is then explored forthe specific case of the global PV industry by looking at theenergy consumed in manufacture and installation of PVsystems using an analogy of the CED metric.The PV industry is changing rapidly, both in terms of

growing installed capacity and decreasing energy inputs to PVsystem deployment. In order to track this evolution requires adynamic perspective that goes beyond the static measures often

employed in PV technology assessment. A model is presentedto track the change in electrical energy costs of PV systemdeployment and is applied to historic data for the PV industrybetween 2000 and 2010. This model is then used to forecastelectrical energy costs of future PV system deployment in orderto determine the gross and net electricity production of the PVindustry to 2020 to determine when the PV industry willbecome a net electricity provider. Finally, implications for R&Dand technology deployment are discussed.A number of useful definitions are outlined in Table 1.

■ ENERGY BALANCE AND INDUSTRY GROWTHInputs and Output from a Single Energy System and

an Energy Industry. Figure 2 depicts energy flows into andout of a single energy production device or system. Flows maybe in either primary energy or electricity but must be consistentbetween inputs and outputs. Construction begins at time, t−tc,requiring a total energy input to construction of Econ, assumedto flow at a constant rate, Econ. Once the project startsproducing energy it is assumed to produce a constant grossflow of energy at rate Eg over the whole lifetime tL. An energyflow, Eop, is required to operate and maintain the system. At theend of the system lifetime, some energy, Ed, is required fordecommissioning, again assumed to flow at a constant rate,Ed.

20 Two ratios used to characterize the scale of energy inputsand outputs at the system level are the EPBT and EROI (bothdefined in Table 1). To compute these ratios, energy flows arelevelized over the full lifecycle of the energy system, comparingenergy inputs - investment - with energy produced by thesystem. Turning instead to the industry-scale, both the timingand magnitude of energy inputs and output are important todetermine the energy balance of the whole industry. A rapidlygrowing industry may deploy new systems before previousdeployment has ‘paid off’ its initial investment, thereby drivingthe whole industry into deficit. Energy must be imported fromoutside the industry. The term energy subsidy is used to definethis energy input, to distinguish from the energy investment atthe system level. An industry composed of individual systems,each with a positive net energy contribution over their lifetimes,could still be a net energy consumer, when it grows rapidly.An energy production industry is based on sequential

installation of many of these energy production systems.Growth of the industry normally proceeds exponentially at rater. Grimmer (1981) defines the relationship between thefractional reinvestment, f [%], the EPBT [yrs] of a technology,and the growth rate, r [%/yr] of an energy production industrybased on deployment of that technology as f = rEPBR = r(CED)/(Eg).

21 Using this relationship, Figure S1 shows therelationship between EPBT, industry growth rate, and fractionalreinvestment. A fractional reinvestment, f = 100%, marks thebreakeven threshold. If f < 100%, then this means the industryis a net energy producer. If f > 100%, then the industry is a netenergy sink, i.e. consumes more energy than it produces. If anindustry is a net energy sink, then there are three means bywhich it may cross the breakeven threshold: (1) decrease theEPBT of system production by either reducing CED or byincreasing the energy output of the system; (2) decrease therate of growth; or (3) some combination of (1) and (2).If the industry grows too quickly, such that r ≥ (Eg)/(Econ),

the industry is a net energy sink and goes into deficit (see SISection Derivation of breakeven threshold conditions). Gross andnet energy output from this industry growing logistically up tosome limit are plotted in Figure 3. Rapid exponential growth of

Figure 1. Global cumulative installed capacity [GW] of PVdisaggregated by the major technologies: single crystal silicon (sc-Si), multicrystalline silicon (mc-Si), amorphous silicon (a-Si), ribbonsilicon, cadmium telluride (CdTe), copper indium gallium diselenide(CIGS), and other (mainly organic/polymer-based systems).Compiled using data from refs 1−3,14, and 15. Raw data for thisfigure are provided in Table S1.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−34893483

Table 1. Definitions

Figure 2. Energy inputs and outputs to a single energy productionsystem. Energy inputs (blue) are above the horizontal line, and energyproduction (yellow) is shown above the line. Figure 3. Energy inputs and outputs for an energy production industry

growing asymptotically to some upper limit. Gross output is shown asa bold line; net output is shown with the dashed line.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−34893484

the industry entails its consuming more energy in systemconstruction and deployment than it produces on an annualbasis, thus requiring an energy subsidy. Continued rapid growthdrives the industry further into deficit, as shown in Figure 3. Ifeither the industry growth rate or the CED of systems beingdeployed decrease sufficiently, the industry’s net energy outputwill pass through a minimum value, after which the industrymay then cross the breakeven threshold and thereafter produce apositive net energy output. This crossover happens in thebreakeven year. After some time, the industry produces enoughpositive net energy to “pay back” the energy subsidy requiredfor its early growth. The year in which this occurs is termed thepayback year.Financial and Energy Costs of PV Systems. Having

explored the general case, we now turn to the specific case forthe PV industry. Raugei et al. (2012) have compared the energyreturn on investment (EROI) of PV systems with fossil fuel-based electricity generating technologies. As shown in Table 1,the definition of EROI is the ratio of electricity production toprimary energy inputs. In order to make the resultscommensurate with the metrics developed in our analysis, wehave developed the electricity-return-on-investment (EeROI) asthe ratio of electricity outputs divided by the primary energyinputs converted into electrical energy equivalents assuming agrid conversion efficiency of ηgrid = 0.3. Raugei et al. have shownthat the EROI of electricity production from a single PV systemis at least as high as oil-fired electricity (EROI = 4−11, EeROI =12−35), and the EROI of CdTe, which has the lowest energeticrequirements for system manufacture and deployment of all PVtechnologies, is within the range of coal-fired electricity (EROI= 12−25, EeROI = 41−82).17 While the device level EROIshows that over its lifetime a PV system will produce between19 and 38 times the electrical energy equivalents needed toproduce it, this metric does not consider the implications ofhaving over 90% of the energetic cost “paid up front”.A breakdown of cost inputs into different stages in the

multicrystalline silicon (mc-Si) PV system production processindicates that one-third of financial costs occur in the last stageinvolving manufacture of the balance of system (BOS)components and system installation.22 This stage accounts foronly 13% of energy inputs. The majority of energetic costs(57%) involve the extraction and purification of poly siliconand the production of PV wafers,15,23 as illustrated in Figure S2.This indicates that not all financial cost reduction will result incommensurate energetic cost reductions.A number of such energy analyses have determined the CED

for the materials, manufacturing, and installation of PV modulesand installed PV systems, including BOS costs.7,11,23−49 Thesestudies were compared in a meta-analysis to allow comparisonbetween studies (see SI Section Meta-analysis for methodo-logical details). We convert all inputs to PV systemmanufacture into equivalents of electricity (discussed in SIsection Meta-analysis) to determine the cumulative electricitydemand (CEeD) [kWhe/Wp] which allows direct comparison ofenergy inputs into the PV system with electrical output fromthe system.Distribution of estimates for CEeD of the six PV technologies

that make up over 99% of the global PV installed capacity areshown in Figure 4. Underlying data are presented in SI PVCED.xlsx. Wafer technologies (sc-Si and mc-Si) have CEeDbetween 2 and 16 kWhe/Wp. In general thin film technologies(a-Si, ribbon, CdTe and CIGS) have lower CEeD, rangingbetween 1 and 7 kWhe/Wp.

Learning curves have been used extensively to track financialcosts in a variety of energy technologies.50 We present here ananalogous model tracking CEeD of PV system manufacturingand installation. A number of alternative models were evaluatedand are discussed in SI Section Alternative models of decreasingCEeD. The CEeD of the production of PV modules (leftcolumn) and systems (right column), per unit of peak capacityas a function of cumulative installed capacity [MWp], is shownon a log−log plot in Figure 5 for sc-Si, mc-Si, a-Si, and forCdTe. The curves demonstrate a clear reduction in the CEeD ofPV system manufacture and installation. Rates of decrease arecomparable to those for financial costs within the PV industry.

Cumulative Electricity Demand and Growth of the PVIndustry. The CEeD can be used to determine the electricitypayback time (EePBT) for a PV system. EePBT for a system isproportional to the CEeD of that system assuming a fixedcapacity factor. Using national data for PV installed capacityand electricity generation between 2005 and 2008,2,3 we havecalculated the distribution in capacity factor for the global PVindustry, shown in Figure 6. The median for the EIA data (darkblue) is 11.5%, and the median for the UN data (light blue) is12.5%. Average values are highly weighted toward the capacityfactor of Germany, which makes up nearly 40% of globalinstalled capacity.We combine rates of decrease in CEeD from Figure 5 with

EePBT and technology specific growth rates shown in Figure 1to determine the historical electricity balance of each of the PVtechnologies between 2000 and 2010. The growth trajectoriesof the PV technologies are depicted on a biannual basis inFigure 7, based on an average capacity factor of 11.5% forglobal PV deployments as a whole. Sizes of circles represent theinstalled capacity of each of the PV technologies.At the global average growth rate between 2000 and 2010 of

40%, for the PV industry as a whole to be a positive netelectricity provider, the EePBT must be below 2.5 yrs, which isequivalent to a CEeD of 2.5 kWhe/Wp. All of the PVtechnologies started at or below the breakeven threshold inthe year 2000. As their rates of growth increased over time, they

Figure 4. Distribution of estimates of CEeD [kWhe/Wp] of PV systemsfor a variety of technologies (sc-Si, mc-Si, a-Si, ribbon, CdTe, andCIGS) from sources.7,11,23−49 In general, thin film technologies have alower CEeD than wafer technologies. CdTe has the lowest CEeD of allof the technologies.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−34893485

all moved into the negative net electricity domain, with theexception of ribbon silicon, despite reduction in their CEeD.Our analysis found that the PV industry was a net electricity

consumer as recently as 2010, and in 2008 the PV industryconsumed 75% more electricity than it produced. Figure 7

shows that deployment of most PV technologies was operatingat a net electricity loss for most of the decade 2000−2010. As of2010 − the latest year of historical data − the only technologyproducing a positive net electricity yield was ribbon silicon. Thefractional reinvestment of CdTe in 2010 was similar to that ofsc-Si at around 150%, despite CdTe having a much highergrowth rate (160% compared with 63%). This highlights theobvious benefit of technologies with lower CEeD. Thisperspective does not incorporate the distribution in capacityfactor nor any uncertainty in the CEeD. To incorporate thesefactors, we undertook an analysis using a system dynamicsmodel outlined in SI Section Model Structure.The installed capacity of PV in the year 2000 was 1.4 GW,

rising to 22.9 GW in 2010.1 Using historical installed capacity(Figure 1) and projecting average growth rates from 2005 to2010 to estimate installed capacity beyond 2010, coupled withthe curves tracking reduction in CEeD (Figure 5), we modeledthe net electricity production of the PV industry to 2020, afteraccounting for electricity to construct and install new capacity.We assume a system lifetime of 25 years and zero operation andmaintenance and decommissioning electricity costs. In reality,operation, maintenance, and decommissioning costs accountfor less than 5% of total lifecycle electrical energy costs for PVsystems.7 As such, electrical energy inputs due to O&M activitywill be less than 1% of total annual electricity inputs to the PVindustry, since they are spread over the full 25 year lifetime ofthe PV system. Decommissioning electrical energy inputs lagconstruction inputs by the lifetime of the system, so remainnegligible over the modeled time period. This assumption isdiscussed in more detail in SI Section O&M and disposal costs.We also assume that the PV industry received until the year2000 a total electricity subsidy of 10 TWh from the existingelectricity system. We carried out a Monte Carlo analysis (fulldetails of which can be found in SI Section: Monte Carlosimulation) to test the sensitivity of the model outputs tovarying technology growth rates, CEeD and capacity factor ofPV installations. We assume a continuation of growth ratesfrom the period 2005−2010 and historical rates of reduction ofCEeD.

■ RESULTS AND DISCUSSIONFigure 8 shows the estimated gross and net annual power[TWh/yr] (top) and cumulative electricity [TWh] (bottom)output of the PV industry in both absolute terms (left axis) andannual output as a fraction of the 2010 global annual electricityproduction of 20,000 TWh/yr (right axis). Historical grosspower output [TWh/yr] is plotted as blue circles. The grossoutput (green line) represents the median value from thesensitivity analysis. The net output is plotted in different shadesfor both 50% and 100% confidence intervals about the median(dark red line).There is a 50% probability that the PV industry stopped

being a net consumer of electricity by 2011. Figure 8 (top)shows that, in the ‘worst case’ scenario, the PV industry doesnot become a net electricity producer until 2015.Figure 8 shows that the payback year has a 50% likelihood of

occurring between 2012 and 2015 with the probabilityincreasing to almost certainty by 2020. Thus, the PV industryshould “pay back” electricity consumed in its early growth onlyby the end of this decade.The CEeD of PV system production and installation has

decreased since the year 2000. With further reduced CEeD, thePV industry could achieve even greater net electricity yields and

Figure 5. Estimates of CEeD [kWhe/Wp] of PV modules (left column)and installed PV systems (right column) for a variety of technologies(sc-Si, mc-Si, a-Si, and CdTe) from sources7,11,23−49 plotted as afunction of cumulative production [MW] of each technology. Curvesare fitted to the data.

Figure 6. Distribution in global PV capacity factors using data fromrefs 2 and 3.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−34893486

thus have the potential to offset more GHG emissions. If thePV industry continues at its current growth rate, by 2020production could reach 3300 TWh/yr (approximately 17% ofcurrent global annual electricity production). If no decrease inCEeD of PV occurs, this would entail an electrical energy inputof around 2937 TWh/yr, representative of a fractionalreinvestment of 89%. If, however, the PV industry continuesthe historical rate of decreasing CEeD, the input would be 264TWh/yr, a fractional reinvestment of 8%, corresponding to anindustry-wide, mean CEeD of 0.2 kWhe/Wp. This is equivalentto the CEeD of some existing polymer PV systems.Consequently, it is important to continue to track and reducethe CEeD of the PV industry.A number of factors could decrease the fractional reinvest-

ment rate: reducing the growth rate of the PV industry,reducing the CEeD or increasing the capacity factor of PVsystem deployment. If the goal is a rapid transition to asustainable, low-carbon energy sector, then maintaining highPV industry growth rates is beneficial. Consequently, the PVindustry will benefit from further reductions in the CEeD anddeployment in areas with higher capacity factors.Technology development can reduce the CEeD of PV by

decreasing the material requirements per unit of capacity [kg/Wp] (or alternatively increasing the device efficiency per unit ofmaterial used [Wp/kg]) and by decreasing the energy required

to process the materials. R&D can decrease the amount andpurity of materials required in order to maximize the utilizationof a given material feedstock (e.g., silicon). There arefundamental thermodynamic limits to our ability to reducethe CEeD, such as the minimum chemical exergy that must beprovided in order to purify a material from its usual ore state,for example obtaining pure silicon from silica. It is thereforeimportant to simultaneously increase net energy production byalso increasing system efficiency and durability. A number ofnascent technologies exist to boost PV efficiency beyond theShockley-Queisser limit (which defines the maximum theoreti-cal efficiency of a single p-n junction solar cell at 34%) such asmultijunction and multiple-exciton cells, hot-carrier cells, up-conversion, or thermophotonics.51

The other major factor in how well the PV industry performsgoing forward is the capacity factor that new installations canachieve. Doubling the capacity factor, ceteris paribus, halves theEePBT. Deployment of PV in regions with high insolation andcapacity factors provides greater overall benefits for the energysystem. As such, policies aimed solely at rapid deployment inareas with low insolation, without increasing net electricity yield(by either decreasing CEeD or improving PV performance),might not be beneficial for the energy sector as a whole.In summary, a model has been developed to track the

evolution of electrical energy inputs to PV system deployment,

Figure 7. Growth rate [%/yr] as a function of CEeD) [kWhe/Wp] (top axis) and electricity payback time (EePBT) [yrs] assuming an 11.5% capacityfactor (bottom axis) for a number of fractional reinvestment rates [%] (diagonal lines). Red lines indicate negative net electricity yield, and greenlines indicate positive net electricity yield. The trajectories of CEeD (or EePBT) and annual growth rates for PV technologies are depicted on abiannual basis between 2000 and 2010. The size of the circles represents the amount of installed capacity [GW] of each technology on a logarithmicbasis.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−34893487

more specifically the change in CEeD as a function ofcumulative installed capacity of PV. The benefits of thismodel have been demonstrated by modeling both historicevolution of the PV industry between 2000 and 2010 and alsoby forecasting PV industry performance out to 2020.Judging the impact of a GHG emissions reduction strategy

should be based on the actual GHG emissions reductionachieved by PV deployment considering the full life-cycle ofemissions. Electrical energy inputs to and electricity productionby the PV industry should be monitored to judge the success ofsuch a strategy. This requires supplementing economic andmarket analyses with energy and material flow analysis. Theimportance of reducing CEeD and increasing net energy shouldalso be recognized as explicit goals in their own right.The electricity balance of the PV industry is critical to the

success of any GHG emissions reduction strategy based on PVdeployment. This analysis is, however, not limited to the PVindustry but applies equally to any rapidly growing alternativeenergy industry (e.g., wind, solar thermal energy conversion orenergy storage) as well as energy conservation strategies (e.g.,insulation, low-energy buildings, or fuel-efficient vehicles) thatrequire large, up-front investment.

■ ASSOCIATED CONTENT*S Supporting InformationTables S1−S3, Figures S1−S3, and text. This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*Phone: 650-725-8579. Fax: 650-723-9190. E-mail: mikdale@stanford.edu.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work was made possible with financial assistance fromGCEP. The authors are also indebted to Richard Sassoon,Charlie Barnhart, Adam Brandt, Mike McGehee, Steve Eglash,John Weyant, Dick Swanson, Roy Johnson, Mark Golden, andMark Schwarz at Stanford University for their invaluable advise.Thanks are also due to the reviewers for a number ofinteresting and useful suggestions.

■ REFERENCES(1) EPIA, Solar Generation 6; 2010.(2) EIA, International Energy Statistics. 2012. http://www.eia.gov/countries/data.cfm (accessed February 5, 2012).(3) UN, UN Energy Statistics Database. 2012. http://data.un.org/Explorer.aspx?d=EDATA (accessed February 5, 2012).(4) Renewables Portfolio Standard Senate Bill X1-02. http://www.cpuc.ca.gov/PUC/energy/Renewables/index.htm (accessed February5, 2012).(5) California Solar Initiative. http://www.cpuc.ca.gov/PUC/energy/solar (accessed February 5, 2012).(6) Sijm, J. The performance of feed-in tariffs to promote renewableelectricity in European countries; Energy Research Centre of theNetherlands ECN, 2002.(7) Amor, M. B.; Lesage, P.; Pineau, P.-O.; Samson, R. Candistributed generation offer substantial benefits in a NortheasternAmerican context? A case study of small-scale renewable technologiesusing a life cycle methodology. Renewable Sustainable Energy Rev. 2010,14, 2885−2895.(8) IEA, Photovoltaics Power Systems Programme - MethodologyGuidelines on Life Cycle Assessment of Photovoltaic Electricity. 2011.(9) Raugei, M. Energy pay-back time: methodological caveats andfuture scenarios. Prog. Photovoltaics 2012, DOI: 10.1002/pip.1249.(10) Fthenakis, V.; Kim, H. Photovoltaics: Life-cycle analyses. SolarEnergy 2011, 85, 1609−1628.(11) Mathur, J.; Bansal, N. K.; Wagner, H.-J. Dynamic energy analysisto assess maximum growth rates in developing power generationcapacity: case study of India. Energy Policy 2004, 32, 281−287.(12) Wadia, C.; Alivisatos, A. P.; Kammen, D. M. Materialsavailability expands the opportunity for large-scale photovoltaicsdeployment. Environ. Sci. Technol. 2009, 43, 2072−2077.(13) Richards, B.; Watt, M. Permanently dispelling a myth ofphotovoltaics via the adoption of a new net energy indicator.Renewable Sustainable Energy Rev. 2007, 11, 162−172.(14) DOE, 2008 Solar Technologies Market Report; 2010.(15) IEA, Photovoltaics Power Systems Programme - Life CycleInventories and Life Cycle Assessments of Photovoltaic Systems. 2011.(16) Botsaris, P. N.; Filippidou, F. Estimation of the energy paybacktime (EPR) for a PV module installed in North Eastern Greece. Appl.Solar Energy 2009, 45, 166−175.(17) Raugei, M.; Fullana-i Palmer, P.; Fthenakis, V. The energyreturn on energy investment (EROI) of photovoltaics: Methodologyand comparisons with fossil fuel life cycles. Energy Policy 2012, 45,576−582.(18) IEA/CIAB, Power Generation from Coal: Measuring andReporting Efficiency Performance and CO2 Emissions. 2010.(19) Peet, N. J. Energy Analysis: A Review of Theory and Applications;New Zealand Energy Research and Development Committee, 1986.(20) Herendeen, R. A.; Cutler, J. C. Net Energy Analysis: Concepts andMethods; Elsevier: New York, 2004; pp 283−289.

Figure 8. Annual electricity production (top) and cumulativeelectricity production (bottom). Gross electricity output (green line)represents the median across all simulations. The net electricity outputis plotted as median (dark red line). Confidence intervals of one andtwo quartiles about the median across all simulations are plotted indifferent colors.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−34893488

(21) Grimmer, D. P. Solar energy breeders. Solar Energy 1981, 26,49−53.(22) Swanson, R. M. Solar's Learning Curve Paves Way toCompetitive Costs. 2011. http://energyseminar.stanford.edu/node/387 (accessed February 5, 2012).(23) Stoppato, A. Life cycle assessment of photovoltaic electricitygeneration. Energy 2008, 33, 224−232.(24) Prakash, R.; Bansal, N. K. Energy analysis of solar photovoltaicmodule production in India. Energy Sources 1995, 17, 605−613.(25) Kato, K.; Murata, A.; Sakuta, K. An evaluation on the life cycleof photovoltaic energy system considering production energy of off-grade silicon. Sol. Energy Mater. Sol. Cells 1997, 47, 95−100.(26) Keoleian, G. A.; Lewis, G. M. Application of life-cycle energyanalysis to photovoltaic module design. Prog. Photovoltaics 1997, 5,287−300.(27) Alsema, E. A. Energy pay-back time and CO2 emissions of PVsystems. Prog. Photovoltaics 2000, 8, 17−25.(28) Frankl, P. Life Cycle Assessment of Renewables: Present Issues,Future Outlook and Implications for the Calculation of External Costs.2001.(29) Knapp, K.; Jester, T. Empirical investigation of the energypayback time for photovoltaic modules. Solar Energy 2001, 71, 165−172.(30) Mathur, J.; Bansal, N. K.; Wagner, H.-J. Energy andenvironmental correlation for renewable energy systems in India.Energy Sources 2002, 24, 19−26.(31) GEMIS, Global Emissions Model for Integrated Systems. 2002.(32) Gurzenich, D.; Wagner, H.-J. Cumulative energy demand andcumulative emissions of photovoltaics production in Europe. Energy2004, 29, 2297−2303.(33) Krauter, S.; Ruther, R. Considerations for the calculation ofgreenhouse gas reduction by photovoltaic solar energy. RenewableEnergy 2004, 29, 345−355.(34) Battisti, R.; Corrado, A. Evaluation of technical improvements ofphotovoltaic systems through life cycle assessment methodology.Energy 2005, 30, 952−967.(35) Fthenakis, V.; Alsema, E. Photovoltaics energy payback times,greenhouse gas emissions and external costs: 2004 - early 2005 status.Prog. Photovoltaics 2006, 14, 275−280.(36) Muneer, T.; Younes, S.; Lambert, N.; Kubie, J. Life cycleassessment of a medium-sized photovoltaic facility at a high latitudelocation. Proc. Inst. Mech. Eng., Part A 2006, 220, 517−524.(37) Mason, J. E.; Fthenakis, V. M.; Hansen, T.; Kim, H. C. Energypayback and life-cycle CO2 emissions of the BOS in an optimized3Au°5aA L’MW PV installation. Prog. Photovoltaics 2006, 14, 179−190.(38) Kannan, R.; Leong, K. C.; Osman, R.; Ho, H. K.; Tso, C. P. Lifecycle assessment study of solar PV systems: An example of a 2.7kW(p) distributed solar PV system in Singapore. Solar Energy 2006,80, 555−563.(39) Mohr, N. J.; Schermer, J. J.; Huijbregts, M. A. J.; Meijer, A.;Reijnders, L. Life cycle assessment of thin-film GaAs and GaInP/GaAssolar modules. Prog. Photovoltaics 2007, 15, 163−179.(40) Pacca, S.; Sivaraman, D.; Keoleian, G. A. Parameters affectingthe life cycle performance of PV technologies and systems. EnergyPolicy 2007, 35, 3316−3326.(41) Raugei, M.; Bargigli, S.; Ulgiati, S. Life cycle assessment andenergy pay-back time of advanced photovoltaic modules: CdTe andCIS compared to poly-Si. Energy 2007, 32, 1310−1318.(42) Ito, M.; Kato, K.; Komoto, K.; Kichimi, T.; Kurokawa, K. Acomparative study on cost and life-cycle analysis for 100aAL’MW verylarge-scale PV (VLS-PV) systems in deserts using m-Si, a-Si, CdTe,and CIS modules. Prog. Photovoltaics 2008, 16, 17−30.(43) Roes, A. L.; Alsema, E. A.; Blok, K.; Patel, M. K. Ex-anteenvironmental and economic evaluation of polymer photovoltaics.Prog. Photovoltaics 2009, 17, 372−393.(44) Fthenakis, V.; Kim, H. C.; Held, M.; Raugei, M.; Krones, J.Update of PV energy payback times and life-cycle greenhouse gasemissions. 2009.

(45) Raugei, M.; Frankl, P. Life cycle impacts and costs ofphotovoltaic systems: Current state of the art and future outlooks.Energy 2009, 34, 392−399.(46) Zhai, P.; Williams, E. D. Dynamic hybrid life cycle assessment ofenergy and carbon of multicrystalline silicon photovoltaic systems.Environ. Sci. Technol. 2010, 44, 7950−7955.(47) Nishimura, A.; Hayashi, Y.; Tanaka, K.; Hirota, M.; Kato, S.; Ito,M.; Araki, K.; Hu, E. J. Life cycle assessment and evaluation of energypayback time on high-concentration photovoltaic power generationsystem. Appl. Energy 2010, 87, 2797−2807.(48) Laleman, R.; Albrecht, J.; Dewulf, J. Life cycle analysis toestimate the environmental impact of residential photovoltaic systemsin regions with a low solar irradiation. Renewable Sustainable EnergyRev. 2011, 15, 267−281.(49) Held, M.; Ilg, R. Update of environmental indicators and energypayback time of CdTe PV systems in Europe. Prog. Photovoltaics2011, 19, 614−626.(50) McDonald, A.; Schrattenholzer, L. Learning rates for energytechnologies. Energy Policy 2001, 29, 255−261.(51) Guillemoles, J. Future concepts for photovoltaic energyconversion. Fundam. Mater. Energy Environ. Sustainability 2011, 238.

Environmental Science & Technology Article

dx.doi.org/10.1021/es3038824 | Environ. Sci. Technol. 2013, 47, 3482−34893489