Life Cycle Greenhouse Gas Emissions of Thin-film Photovoltaic Electricity Generation

Systematic Review and Harmonization


  • Hyung Chul Kim,

  • Vasilis Fthenakis,

  • Jun-Ki Choi,

  • Damon E. Turney

Vasilis Fthenakis, Bldg.130, Brookhaven National Laboratory, Upton, NY 11973.


We present the process and the results of harmonization of greenhouse gas (GHG) emissions during the life cycle of commercial thin-film photovoltaics (PVs), that is, amorphous silicon (a-Si), cadmium telluride (CdTe), and copper indium gallium diselenide (CIGS). We reviewed 109 studies and harmonized the estimates of GHG emissions by aligning the assumptions, parameters, and system boundaries. During the initial screening we eliminated abstracts, short conference papers, presentations without supporting documentation, and unrelated analyses; 91 studies passed this initial screening. In the primary screening we applied rigorous criteria for completeness of reporting, validity of analysis methods, and modern relevance of the PV system studied. Additionally, we examined whether the product is a commercial one, whether the production line still exists, and whether the study's core data are original or secondary. These screenings produced five studies as the best representations of the carbon footprint of modern thin-film PV technologies. These were harmonized through alignment of efficiency, irradiation, performance ratio, balance of system, and lifetime. The resulting estimates for carbon footprints are 20, 14, and 26 grams carbon dioxide equivalent per kilowatt-hour (g CO2-eq/kWh), respectively, for a-Si, CdTe, and CIGS, for ground-mount application under southwestern United States (US-SW) irradiation of 2,400 kilowatt-hours per square meter per year (kWh/m2/yr), a performance ratio of 0.8, and a lifetime of 30 years. Harmonization for the rooftop PV systems with a performance ratio of 0.75 and the same irradiation resulted in carbon footprint estimates of 21, 14, and 27 g CO2-eq/kWh, respectively, for the three technologies. This screening and harmonization rectifies previous incomplete or outdated assessments and clarifies variations in carbon footprints across studies and amongst thin-film technologies.


Thin-film photovoltaic (PV) systems such as amorphous silicon (a-Si), cadmium telluride (CdTe), and copper indium gallium diselenide (CIGS) are expanding rapidly due to their low cost, ease of manufacturing, advancing conversion efficiency, and competitive sustainability indicators. These indicators are becoming crucial in assuring the public's acceptance of energy technologies since climate change is arguably the most significant threat facing our planet.

Recent studies show that PV technologies have very low environmental and human health impacts compared with those of conventional electricity generation (Fthenakis et al. 2008; Hondo 2005). A broad review of the literature, however, reveals several PV life cycle greenhouse gas (GHG) analyses with widely varying estimates. For example, reported life cycle GHG emissions of thin-film a-Si PV systems range from 11 to 226 grams carbon dioxide equivalent per kilowatt-hour (g CO2-eq/kWh) of electricity produced (Frankl et al. 2004; Yamada et al. 1995).1 A significant variation likewise exists in energy payback times for the same type of PV systems, ranging from 1.1 to 6.3 years across studies (Keoleian and Lewis 2003; Uchiyama 1997; Yamada et al. 1995). Such divergence reflects different assumptions on key parameters, for example, solar irradiation, performance ratio, and lifetime. Estimates also deviate because of the different types of installation possible, including ground mount, rooftop, and façade. Most importantly, assessments made from outdated information collected from antiquated PV systems are still cited in the literature and used for guiding policy analyses.

Besides screening out invalid or outdated studies, variability in published life cycle environmental studies can be significantly reduced by aligning system boundaries, parameters, and assumptions (Farrell et al. 2006), a meta-analysis approach called “harmonization” in the present study. By clarifying the central tendency and reducing the uncertainty of estimates, harmonization in this study aims to provide decision makers and interested audiences more accurate information and a balanced perspective on the life cycle GHG emissions from contemporary thin-film PV technologies.

The National Renewable Energy Laboratory (NREL), Columbia University, and Brookhaven National Laboratory (BNL) are engaged in a project for developing balanced comparisons of data and premises across these studies. The project team reviewed life cycle GHG analyses for all PV technologies, harmonizing them by enforcing identical system boundaries and assumptions on major parameters. In the current article we describe the processes for reviewing, screening, and harmonizing the life cycle GHG emissions from thin-film PV technologies (i.e., a-Si, CdTe, and CIGS).

This study reviews and harmonizes only the GHG emissions metric deemed central in making comparisons between different life cycles of electricity generation technologies. In a complete environmental assessment, other metrics such as energy payback time, toxicities, and resource uses need to be concurrently evaluated.

Harmonization Methodology

Life Cycle of Thin-Film Photovoltaics

The PV systems considered by this study comprise the grid-connected PV modules and the balance of system (BOS), which includes cables, inverters, and support structures for the modules. A BOS takes a different form in terms of equipment capacity and materials for ground-mount and rooftop installations, the two most common types. Systems mounted on building façades or with sun-tracking motors were not included in this study because life cycle GHG emissions studies are rare for the necessary BOS equipment. The life cycle of thin-film PV starts with raw materials acquisition, encompasses materials production, film deposition, module production, system assembly, and system operation, and ends with their disposal (figure 1). Also shown in the graph is the life cycle of the BOS, whose life cycle emissions will be added to those of the PV for a complete analysis and will be harmonized based on standard values. Note that the recycling stage of the thin-film PV life cycle was not included in the system boundary of this study, because thin-film installations are relatively new, and end of life has not been described in detail yet. Listed below are detailed processes during the life cycle stages of thin-film PV systems.

Figure 1.

The life cycle of thin-film photovoltaic systems with indications of focused stages. Dotted lines correspond to the system boundary for the harmonization study. BOS = balance of system.

  • 1Upstream processes
    • - Raw material acquisition: mining ores, extracting petroleum, and growing woods
    • - Materials production: alloying, purification, treatment, mixing, and polymerization
    • - Film deposition: chemical vapor deposition and vapor transport deposition
    • - Module production: contact formation, encapsulation, wiring, and assembly
    • - Module and BOS installation: installing module, inverter, and support structures
  • 2Operational processes
    • - Electricity generation: office use for utility-scale plant
    • - Maintenance: scheduled and unscheduled repair and maintenance
  • 3Downstream processes
    • - Decommissioning and disposal: demolition and transportation
    • - Recycling: collection, disassembly, shredding, and material separation

Literature Screening

Initial screening

We reviewed 109 studies on the life cycle environmental profile of thin-film PV electricity generation systems published through 2010. The studies were taken from journal articles, conferences, doctoral theses, and technical reports. During our first screening stage we examined the studies’ research methods to ascertain consistency with the standard life cycle assessment (LCA) framework. We screened out those studies that did not include the major life cycle stages or upstream material and energy flows. Studies conducted before 1980 were eliminated, as we deemed them outdated, and documents in the form of presentations, posters, and abstracts also were rejected as lacking sufficient documentation. Ninety-one life cycle environmental studies of thin-film PVs passed this first-stage screening process. Table S1 in the supporting information available on the journal's Web site presents a detailed breakdown of these studies. Most frequently studied is a-Si, at 51 times, followed by CdTe, at 37 times. We attributed this focus to the fact that these technologies have been manufactured and commercialized longer than other thin-film technologies. The total number of technology scenarios at this stage of the harmonization, 124, surpasses the number of studies, 91, because some studies examine multiple thin-film technologies or multiple scenarios for the same technology. Technologies reviewed but unspecified in table S1 in the supporting information on the Web include a-Si/nanocrystalline silicon (nc-Si), gallium arsenide (GaAs), gallium indium phosphide (GaInP), GaInP/GaAs, dye-sensitized cell, and quantum-dot CdSe.

Primary screening

More rigorous quality criteria were set during the second stage of screening for (1) completeness of reporting results and methods, (2) validity of the analysis methods, and (3) relevance to present-day technologies. The screening centered on the stages indicated in figure 1, that is, raw material acquisition, materials production, film deposition, PV module production, and operation. We established detailed subcriteria to facilitate the screening and to ensure consistent, transparent analyses:

  • 1Completeness of reporting results and methodsUnder this criterion we reviewed whether the studies included critical components of LCA, such as functional units, scoping, inventory analyses, and impact analyses. For our current harmonization we eliminated studies that did not examine the GHG emissions. In fact, a wide range of environmental metrics associated with thin-film PV technologies have been evaluated under the LCA framework, including risks, toxic emissions, primary energy, energy-payback times, land use, and water use. We did not consider such analyses, although many are recent and valid, because they did not investigate GHG emissions. The number of studies that included estimates of GHG emissions is 15, 13, and 7 for a-Si, CdTe, and CIGS, respectively (see table S1 in the supporting information on the Web).
  • 2Validity of the analysis methodsIn PV life cycle GHG emissions analyses it is essential to explicitly present the key parameters of analysis, that is, conversion efficiency, performance ratio, irradiation, and lifetime, along with the sources of the information, such as manufacturer, data collector, and age of the data. The guidelines of the International Energy Agency (IEA) (Alsema et al. 2009) detail such requirements for PV LCA.
  • 3Relevance to present-day technologiesWe rejected articles that do not represent modern technologies. To determine modernity we considered module efficiency, manufacturer, scale of production, and module design. In addition, studies based on a hypothetical manufacturing line, future projections, and conceptual modeling were screened out under this constraint. We considered only those investigations based on inventory data from real-world production lines, except those for pilot-scale productions that we deemed relevant. We accepted only the original sources of study results, meaning that we excluded studies that do not contain original investigations.

Our chosen metric for GHG emissions (G) is CO2-equivalent emissions per kilowatt-hour, which is derived as follows:


where W= GHG emissions from the life cycle of the PV system (g CO2-eq), I= irradiation (kilowatt-hours per square meter per year [kWh/m2/yr]), η= conversion efficiency, PR= performance ratio, LT= lifetime (years), and A= area of the module (m2).2 The major emissions considered as GHG emissions in these evaluations include CO2, methane (CH4), nitrous oxide (N2O), chlorofluorocarbons (CFCs), and perfluorocarbons (PFCs) converted to CO2 equivalents using the global warming potential on a 100-year time scale.

Several studies report emissions per square meter of module area or manufacturing capacity. Studies focusing on the environmental impact of processing thin-film PVs often express emissions in this form. Such estimates were converted to emissions per unit of electricity generation (i.e., kilowatt-hours) when sufficient information was given, information such as quantum efficiency and system efficiency, otherwise we discarded such studies. We also excluded studies that report “avoided GHG” emissions that are unconvertible to our functional unit. Finally, we omitted studies reporting normalized global warming indicators rather than presenting GHG emissions. Tables 1 through 3 list the studies that include life cycle GHG emissions. Table S2 in the supporting information on the Web presents those studies with “other” technologies that are not presented here. We note that these estimates are not harmonized and thus are inconsistent with each other in terms of system boundaries and technical parameters, such as performance ratio and lifetime expectancy, solar irradiation, and other assumptions.

Table 1.  Thin-film amorphous silicon (a-Si) photovoltaic life cycle environmental studies reporting greenhouse gas (GHG) emissions (15 studies, 26 scenarios)
Study g CO2-eq/kWh Solar irradiation (kWh/m2/yr) Module efficiency (%) PR Lifetime (years) Type Note
  1. Note: g CO2-eq/kWh = grams carbon dioxide equivalent per kilowatt-hour; kWh/m2/yr = kilowatt-hour per square meter per year; PR = performance ratio; N/A = not available; G = ground-mount; R = rooftop; kWp/yr = kilowatt-peak per year production capacity; MWp/yr = megawatt-peak per year production capacity; GWp/yr = gigawatt-peak per year production capacity.

Yamada et al. (1995) 2261,20080.7220GProduction scale = 0.01 GWp/yr
 1251,200130.7220GProduction scale = 1 GWp/yr
 1011,200160.7220GProduction scale = 100 GWp/yr
Uchiyama (1996a) 29N/A12.6 (cell)N/A30R3 kWp, production scale = 1 GWp/yr
Uchiyama (1996b) 29N/A12.6 (cell)N/A30R3 kWp; production scale = 1 GWp/yr
Uchiyama (1997) 29N/A12.6 (cell)N/AN/ARProduction scale = 1 GWp/yr
   8.6 (system)    
Martin (1997) 38N/AN/AN/A30GPower plant
Kato et al. (1998) 621,42780.8120R3 kWp; production scale = 10 MWp/yr
 481,427100.8120R3 kWp; production scale = 30 MWp/yr
 331,427120.8120R3 kWp; production scale = 100 MWp/yr
Kato et al. (2001) 581,43080.8120RProduction scale = 10 MWp/yr
 441,430100.8120RProduction scale = 30 MWp/yr
 301,430120.8120RProduction scale = 100 MWp/yr
Meier (2002) 391,8405.70.7430R8 kWp; building-integrated
Frankl et al. (2004) 43.41,74060.87520RRetrofit; Rome
 38.62,00060.8620RRetrofit; southern Spain
 62.31,20060.88520RRetrofit; central Europe
 36.91,74060.920RBuilding integrated; Rome
 291,74060.87520RIntegrated skylight roof; Rome
 10.91,74060.87520RIntegrated skylight roof substituting glass; Rome
Hondo (2005) 26N/A8.6N/A30N/AFuture case, 1 GWp/yr production
Pacca et al. (2006) 34.31,3596.30.9520RBuilding integrated case in Ann Arbor, MI
Pacca et al. (2007) 34.31,3596.30.9520RBuilding integrated case in Ann Arbor, MI
SENSE (2008) 311,7005.50.91220GGHG estimate for Rome
Ito et al. (2008) 572,0176.90.77130G100 MWp; Gobi desert
Dominguez-Ramos 271,82570.7830GInstalled in Spain
 et al. (2010)        
Table 3.  Thin-film copper indium gallium diselenide (CIGS) photovoltaic life cycle environmental studies reporting greenhouse gas (GHG) emissions (7 studies, 14 scenarios)
Study g CO2- eq./kWh Solar irradiation (kWh/m2/yr) Module efficiency (%) PR(%) Lifetime (years) Type Note
  1. Note: g CO2-eq/kWh = grams carbon dioxide equivalent per kilowatt-hour; kWh/m2/yr = kilowatt-hour per square meter per year; PR = performance ratio; N/A = not available; G = ground-mount; R = rooftop; MWp/yr = megawatt-peak per year production capacity; GWp/yr = gigawatt-peak per year production capacity.

Frankl et al. (2004) 43.41,74090.87520RRetrofit; Rome
 38.62,00090.8620RRetrofit; southern Spain
 62.31,20090.88520RRetrofit; central Europe
 36.91,74090.920RBuilding integrated; Rome
 321,74090.87520RIntegrated skylight roof; Rome
 20.51,74090.87520RIntegrated skylight roof substituting glass; Rome
Raugei et al. (2007) 951,700110.7520R 
SENSE (2008) 611,20011.50.91220G 
Ito et al. (2008) 38.52,017110.77630G100 MWp system in Gobi desert
Ito et al. (2009) 58.82,01710.10.77N/AG1 GWp system in Gobi desert
Ito et al. (2010) 441,702N/A0.78N/AG1 GWp system in Gobi desert
Dominguez-Ramos et al. (2010) 331,825100.7830GGerman production, installed in Spain

Figure 2 plots the estimates of GHG emissions from the listed studies. The values for CdTe and CIGS show a relatively narrower range than those for a-Si and “other” technologies, which may be partially related to the reporting years for each technology. The analyses of CdTe and CIGS were determined after 2000, while some estimates for a-Si were pre-2000, a fact that is linked to the history of a-Si technology. The median estimate of CdTe's emissions is the lowest, while that of CIGS is the highest. The maximum value for a-Si corresponds to the early estimates by Yamada and colleagues (1995), while the lower one represents the case of building-integrated PVs (BIPVs) with credits for glass substitution in work by Frankl and colleagues (2004). The maximum estimate for emissions in the life cycle of CdTe PVs describes a hypothetical installation case in a remote area (Ito et al. 2009), wherein 75% of the GHG emissions are from constructing the BOS, including the transmission lines, cables, foundation, and array support, that was designed for usage in Japan (earthquake region) (Ito 2010). The lower one corresponds to a rooftop system with 9% efficient modules in Europe (Fthenakis and Kim 2007; Raugei 2010).

Figure 2.

Box plot of GHG emissions reported in thin-film PV life cycle environmental studies published from 1995 to 2010. Figures in parentheses indicate the number of scenarios.

Further screening based on criteria 2 and 3 eliminated those technologies that are future projections (e.g., Hondo 2005; Uchiyama 1996a, 1996b, 1997), those based on hypothetical cases (e.g., Ito et al. 2008, 2009, 2010; Kato et al. 1998, 2001), and those lacking detailed parameters (e.g., Martin 1997; Meier 2002). In particular, under criterion 3 we accepted only studies based on life cycle inventory (LCI) data from actual production lines with modern relevance. Table 4 lists those nine studies, encompassing 12 cases, that fall under this classification.

Table 4.  Life cycle environmental studies based on actual production data
Technology Study Manufacturer Data year Production scale for data
  1. Note: a-Si = amorphous silicon; CdTe = cadmium telluride; CIGS = copper indium gallium diselenide; MWp/yr = megawatt-peak per year production capacity.

a-Si Pacca et al. (2006) United Solar2004Commercial, 28 MWp/yr
a-Si Pacca et al. (2007) United Solar2004Commercial, 28 MWp/yr
a-Si SENSE (2008) Free Energy Europe2003–2006Pilot
CdTe Fthenakis and Kim (2006) First Solar2005Commercial, 25 MWp/yr
CdTe Fthenakis and Alsema (2006) First Solar2005Commercial, 25 MWp/yr
CdTe Fthenakis and Kim (2007) First Solar2005Commercial, 25 MWp/yr
CdTe Fthenakis et al. (2008) First Solar2005Commercial, 25 MWp/yr
CdTe Fthenakis et al. (2009) First Solar2008Commercial, 716 MWp/yr
CdTe SENSE (2008) Antec Solar2003–2006Pilot
CdTe Raugei et al. (2007) Antec Solar2004Pilot
CIGS Raugei et al. (2007) Würth Solar2004Pilot
CIGS SENSE (2008) Würth Solar2003–2006Commercial, 15 MWp/yr

During the final stage of screening, we also considered the following: whether the product is a commercial one, whether the production line still exists, and whether the study references the same data from previous studies (e.g., Fthenakis and Kim 2006; 2007). At this stage we contacted the authors of these articles to verify if the technologies described in the analysis are relevant to modern practices. We confirmed that the a-Si and CdTe lines, detailed in SENSE (2008) and Raugei and colleagues (2007), were phased out after their studies, while the CIGS line is still operating at an expanded scale.

Table 5 lists those studies that passed the final screening, with a brief summary given in the following:

Table 5.  As-published and harmonized life cycle greenhouse gas (GHG) emissions (grams carbon dioxide equivalent per kilowatt-hour [g CO2-eq/kWh]) based on solar irradiation of 2,400 kilowatt-hours per square meter per year (kWh/m2/yr)
Technology Study As published Harmonized
Module System
Module System η LT I PR (G) PR (R) Otherb Allc (G) All (R) G R
  1. aThe average of three estimates.

  2. bAccounting for non-CO2 GHG emissions, using current global warming potential (GWP) values (Forster et al. 2007) and assuming 0.5% per year degradation of module efficiency (Alsema et al. 2009).

  3. cHarmonized by all the parameters previously listed, that is, η, LT, I, PR, and other.

  4. Note: a-Si = amorphous silicon; CdTe = cadmium telluride; CIGS = copper indium gallium diselenide; η= module efficiency; LT = lifetime; I = solar irradiation; PR = performance ratio; G = ground-mount; R = rooftop.

a-Si Pacca et al. (2006) 36N/A36242043463114152021
CdTe Fthenakis et al. (2008) 192616191319202112131617
CdTe Fthenakis et al. (2009) 12a19a121291213139101414
CIGS Raugei et al. (2007) 709567475066707632343638
CIGS SENSE (2008) 374337252642454022232627
  • 1Pacca and colleagues (2006). This study assesses the life cycle environmental impact of a hybrid installation of a-Si and multicrystalline Si PV systems on a rooftop in Ann Arbor, Michigan, USA. The materials and energy data for the manufacturing stage of a-Si PV were provided by United Solar, whereas those of multicrystalline Si PV were from the literature. The installed a-Si PV array facing the south with a 12° tilt angle receives a solar irradiation of 1,359 kWh/m2/yr in this location. The life cycle CO2 emissions from the a-Si PV module with 6.3% efficiency corresponded to 34.3 g/kWh over a 20-year lifetime. Note that this estimate takes into account an assumed degradation of module efficiency of 1.1% per year.
  • 2Raugei and colleagues (2007). This study investigated the environmental performance of thin-film PVs, including CIGS, under a European research project on the acceptability of advanced PV technologies (PVACCEPT). Data for CIGS were collected from a prototype batch line at Würth Solar, Germany (Raugei et al. 2007). It is noted that the GHG estimates of this study, 95 g CO2-eq/kWh under solar irradiation of 1,700 kWh/m2/yr, do not represent standard production; according to the author the electricity demand was probably overstated (Raugei 2010). Also, the higher glass demand (25 kilograms [kg]/m2) reflected a very high percentage of breakage in the prototype line (Raugei 2010; Raugei et al. 2007).3
  • 3SENSE (2008). A European Commission (EC) project, Sustainability Evaluation of Solar Energy Systems (SENSE), included an investigation of life cycle GHG emissions for a commercial standard line of CIGS at Würth Solar, Germany, with a 15 megawatt-peak production capacity per year (MWp/yr)4 (SENSE 2008). Under this project, a group of manufacturers and scientists assessed the life cycle environmental impact of thin-film PV technologies (i.e., a-Si, CIGS, and CdTe). Representing a standard line, and with a slightly higher conversion efficiency (11.5% versus 11%), the life cycle GHG estimate of CIGS PVs in this study, 43 g CO2-eq/kWh under solar irradiation of 1,700 kWh/m2/yr, is less than half of that by Raugei and colleagues (2007). However, since the current line produces 30 MWp/yr, the GHG emissions presented therein likely are not up-to-date (Held 2010).
  • 4Fthenakis and colleagues (2008). This study investigated the life cycle emissions of major commercial PV systems, multicrystalline Si, monocrystalline Si, ribbon Si, and CdTe, based on industry data from 2004 to 2006. For CdTe, the data were collected from the First Solar plant in Perrysburg, Ohio, USA, describing the operational conditions in 2005. The production capacity of the plant was 25 MWp/yr and the module efficiency was 9% at the time of this study. The GHG estimates were 26 and 21 g CO2-eq/kWh under solar irradiation of 1,700 kWh/m2/yr with a ground-mount installation, corresponding to the actual U.S. and hypothetical European production scenarios.
  • 5Fthenakis and colleagues (2009). This study is based on data collected from First Solar's plant in Perrysburg, Ohio, USA, and from the plant in Frankfurt-Oder, Germany, in 2008; it is an update of the Fthenakis and colleagues (2008) study that described the operational conditions in 2005. Reduced energy consumption in the production line resulted in lower GHG estimates of 18 to 20 g CO2-eq/kWh from the previous 26 g CO2-eq/kWh under the solar irradiation of 1,700 kWh/m2/yr. The improvement in efficiency of PV modules over this time also was significant (i.e., from 9% to 10.9%), which partially contributes to the reduction in GHG emissions between the two investigations (Fthenakis et al. 2009).

Harmonization Approach

For the LCA harmonization project5 as a whole, two levels of harmonization were devised. The more intensive and in-depth level envisions a process similar to that employed by Farrell and colleagues (2006) to harmonize the LCA results on ethanol, whereby analyses of life cycle GHG emissions are carefully disaggregated to produce a detailed meta-model enabling adjustment of parameters, realignment of system boundaries within and across life cycle phases, and review of all data sources for adequacy (Farrell et al. 2006). A less-intensive approach, which is adequate for a larger set of literature, could harmonize GHG emissions estimates at a more gross level for several influential performance characteristics and to common system boundaries. The former was chosen for harmonizing life cycle GHG emissions of thin-film PV technologies of which the qualified population is relatively small, and thus suitable for intensive analysis.

During the harmonization stage, we adjusted key parameters of the life cycle impact, such as module efficiency, lifetime, performance ratio, solar irradiation, and efficiency degradation. In addition, assumptions on the system's boundary were examined (e.g., types of BOS and frame). To obtain the life cycle GHG emissions of a complete system, as indicated in figure 1, the BOS components must be considered together with the PV module system, including inverters, cables, and mounting structures for ground-mounted BOS. The GHG emissions from rooftop BOS used in this harmonization were adapted from the latest information from the Crystal Clear project (de Wild-Scholten 2009), that is, 5 g CO2-eq/kWh under solar irradiation of 1,700 kWh/m2/yr, with 14% module efficiency and a performance ratio of 0.75. The same information for the ground-mounted BOS is taken from the analysis of the Tucson Electric Power (TEP) power plant in Springerville, Arizona, USA, where the GHG emissions correspond to 5.5 g CO2-eq/kWh with 12.2% module efficiency under an average solar irradiation of 1,800 kWh/m2/yr and a performance ratio of 0.8 (Mason et al. 2006). Emissions from the structural part of the BOS are adjusted according to the conversion efficiency of PVs because a high-efficiency module requires less structural material to produce a unit kilowatt-hour, in contrast to emissions from the inverter portion of the BOS, which are unchanged. Note that for harmonization we selected the frameless design of thin-film CdTe and CIGS PVs. Unlike crystalline Si modules that require an aluminum frame for structural stability, typically ∼3 kg/m2 of panel, CdTe and CIGS thin-film modules with a double-glass design do not necessarily require a frame. The current triple-junction a-Si module deposited on a stainless-steel substrate, manufactured by United Solar, uses an aluminum frame with a very thin profile, specifically, 15 g of anodized extruded aluminum per square meter of module, except for building-integrated applications (Pacca et al. 2006).

Below we list the reference parameters selected; they are the figures most accepted as reflecting current PV technologies. For module efficiency, the latest values in the reviewed literature are used.

  • 1Solar irradiation
    • - Southwestern Unites States (Phoenix, Arizona): 2,400 kWh/m2/yr
    • - Southern Europe: 1,700 kWh/m2/yr
  • 2Module efficiency
    • - a-Si: 6.3%
    • - CdTe: 10.9%
    • - CIGS: 11.5%
  • 3Degradation in efficiency: 0.5% per year (Alsema et al. 2009)
  • 4Performance ratio
    • - Ground mount: 0.8
    • - Rooftop: 0.75
  • 5Lifetime: 30 years
  • 6BOS data source
  • 7Global warming potential (GWP)
    • - Account for non-CO2 GHG emissions (CH4, N2O, CFCs, PFCs, and so on)
    • - Intergovernmental Panel on Climate Change (IPCC) 2007 values (Forster et al. 2007)


Table 5 shows the harmonized estimates based on the irradiation of the U.S. Southwest, where construction of major ground-mount PV power plants is in progress or on the way, which is 2,400 kWh/m2/yr. The cases with irradiation of 1,700 kWh/m2/yr can be found in table S3 in the supporting information on the Web. Harmonized estimates for each parameter as well as the combined harmonized values are presented. Figure 3 illustrates the harmonized and preharmonized data for the studies of ground-mount installation under 2,400 kWh/m2/yr. First, our harmonization greatly lowers the overall ranges of GHG estimates for the life cycle of thin-film PVs (e.g., from 12–70 to 9–32 g CO2-eq/kWh for modules and from 19–95 to 14–36 g CO2-eq/kWh for total ground-mount PV systems). The harmonization of rooftop BOS produced a similar range of 10 to 34 g CO2-eq/kWh for modules and 14 to 38 g CO2-eq/kWh for the total system. Note that if we exclude the earlier estimates of CdTe (Fthenakis et al. 2008) and CIGS (Raugei et al. 2007) from figure 2, the current harmonized estimates for the three thin-film PV systems are even lower, at 20, 14, and 26 g CO2-eq/kWh for a-Si, CdTe, and CIGS, respectively, for ground-mount applications under the reference conditions. The most significant decrease during harmonization (from 95 to 36 g CO2-eq/kWh) was that for the total system estimate of CIGS based on a 20-year lifetime and with an aluminum frame (Raugei et al. 2007). Simply extending the module's lifetime from 20 to 30 years reduces the module-only estimate of both a-Si (Pacca et al. 2006) and CIGS (SENSE 2008) by 30%. By additionally adjusting the degradation in efficiency from 1.1% to 0.5% per year, and increasing solar irradiation from 1,359 kWh/m2/yr in the original study to 1,700 kWh/m2/yr, the former estimate drops by ∼40%, although the performance ratio decreased from 0.95 to 0.75. The harmonization results based on an irradiation of 1,700 kWh/m2/yr is illustrated in figure S1 in the supporting information on the Web.

Figure 3.

Harmonization of GHG emission estimates that passed the final screening for ground-mount installation without an aluminum frame, based on solar irradiation of 2,400 kilowatt-hours per square meter per year (kWh/m2/yr). Module efficiency: CdTe, –10.9%; a-Si, –6.3%; CIGS, –11.5%; lifetime = 30 years and performance ratio = 0.8. The arrows show the effect of harmonization on individual technology scenarios.


Both the as-published and harmonized life cycle GHG emissions results indicate that the carbon footprint of thin-film PV technologies decrease significantly as the production capacity increases, reflecting technological advances in process and device designs. For example, between 2005 and 2008, First Solar's annual production capacity of CdTe PVs jumped from 25 to 716 MWp, and during the same period the module efficiency of CdTe PVs increased from 9% to 10.9% and the GHG estimate fell by ∼30% (Fthenakis et al. 2008, 2009). As of the first quarter of 2011, First Solar's CdTe PVs have a conversion efficiency of 11.7% (First Solar 2011). Scaling up a CIGS PV prototype to a 15 MWp commercial line for Würth Solar also corresponds to a significant (i.e., ∼50%) reduction in GHG emissions (Raugei et al. 2007; SENSE 2008). We also expect further reductions in GHG estimates for a-Si, as the capacity of United Solar has been expanding rapidly (178 MWp/yr as of 2009) and the data now available may be outdated (Energy Business Review 2010). Relatively small improvements in efficiency also occurred in a-Si PVs; the current efficiency of a-Si PV modules is 6.7%, compared with the 6.3% used in the most recent study we report herein.

In the harmonization process we allowed for variability in manufacturing locations and detailed system boundaries. The geographic location of the PV module plant affects the upstream grid mix, and consequently the GHG emissions factors per kilowatt-hour of electricity used for producing PVs. The estimates of Raugei and colleagues (2007) and SENSE (2008) assume the Union for the Co-ordination of Transmission of Electricity (UCTE) grid mix for electricity consumption, while those of Fthenakis and colleagues (2008), Pacca and colleagues (2006), and the U.S. case of Fthenakis and colleagues (2009) assume the average U.S. grid mix. The German cases of Fthenakis and colleagues (2009) use the German grid mix. The UCTE grid mix has a lower GHG emissions factor (510 g CO2-eq/kWh) than those for the United States (760 g CO2-eq/kWh) and Germany (660 g CO2-eq/kWh) (Frischknecht et al. 2007). These cannot be easily harmonized, as a detailed breakdown of energy use during the manufacturing stage was not provided in some studies. Harmonizing for a specific geographic location may be unnecessary and unrealistic as thin-film PV technologies are deemed difficult to replicate. Although the effect may be minor, the database for the same grid mix often varies across studies. For example, Raugei and colleagues (2007) employed the ETH-ESU database, while SENSE (2008) used the Gabi database for the same UCTE grid mix.

We note that harmonizing system boundaries of the manufacturing and end-of-life stages was not attempted because they were not clearly defined across studies or a detailed breakdown of these stages was not available. For example, the U.S. case discussed by Fthenakis and colleagues (2009) includes research and development (R&D)-related electricity uses in the system's boundaries, while other studies do not include it or do not specify if it is included. The recycling stage was included within the system boundary of SENSE (2008) analysis for CIGS and the German cases in the work of Fthenakis and colleagues (2009) for CdTe, while it was not included in the rest of the studies passing the final screening. However, harmonizing the end-of-life stage may affect the GHG emissions estimate only slightly; according to the SENSE (2008) analysis, this stage accounts for only 2.6% of the life cycle GHG emissions from CIGS PVs.


We reviewed 109 life cycle environmental studies on thin-film PVs. After rigorously screening the completeness, validity, and data quality of each study, we selected five studies as representative of the carbon footprint of modern thin-film PV technologies. We harmonized the major parameters of PV life cycle GHG emissions, including solar irradiation, performance ratio, and lifetime. The resulting latest estimates of GHG emissions are 20, 14, and 26 g CO2-eq/kWh for a-Si, CdTe, and CIGS, respectively, for ground-mount application under solar irradiation of 2,400 kWh/m2/yr, a performance ratio of 0.8, and a lifetime of 30 years. For the same technologies, the harmonized latest estimates for rooftop application under solar irradiation of 2,400 kWh/m2/yr and a performance ratio of 0.75 correspond to 21, 14, and 27 g CO2-eq/kWh. The screening and harmonizing described in this article significantly reduced the uncertainty of estimates of GHG emissions for thin-film PVs. In addition, harmonization allowed us to appraise the real variations of carbon footprint across device technologies, production scales, and the age of the data in thin-film PV life cycle GHG emissions analyses. In fact, the ranges of the estimates of GHGs from thin-film PVs were drastically narrowed through harmonization, that is, to ∼40% and ∼50%, respectively, for modules and total system rooftop application. Overall, this harmonization reduced the uncertainty and ambiguity of the reported values of the carbon footprint of these technologies, and contributed to rectifying previous incomplete or outdated assessments.


We would like to thank Garvin Heath, Pamala Sawyer, David Hsu, Patrick O’Donoughue, and Margaret Mann from the National Renewable Energy Laboratory for their crucial contribution in the stages of literature collection and screening, methodology development, and review.


  • 1

    Carbon dioxide equivalent (CO2-eq) is a measure describing the climate-forcing strength of a quantity of greenhouse gases using the functionally equivalent amount of CO2 as the reference. One gram (g) = 10−3 kilograms (kg, SI) ≈ 0.035 ounces (oz); one kilowatt-hour (kWh) ≈ 3.6 × 106 joules (J, SI) ≈ 3.412 × 103 British thermal units (BTU).

  • 2

    One square meter (m2, SI) ≈ 10.76 square feet (ft2).

  • 3

    One kilogram (kg, SI) ≈ 2.204 pounds (lb).

  • 4

    One megawatt (MW) = 106 watts (W, SI) = 1 megajoule/second (MJ/s) ≈ 56.91 × 103 British thermal units (BTU/minute).

  • 5

    Additional data and results of the project are available at

About the Authors

At the time this article was written, Hyung Chul Kim was an assistant scientist at Brookhaven National Laboratory in Upton, New York, and an associate research scientist at Columbia University in New York, New York. He is now a research scientist at the Ford Research and Innovation Center in Dearborn, MI, USA. Vasilis Fthenakis is a senior scientist at Brookhaven National Laboratory and a professor of earth and environmental engineering at Columbia University. Jun-Ki Choi is a Goldhaber distinguished fellow at Brookhaven National Laboratory. Damon E. Turney is an associate researcher at Brookhaven National Laboratory.