Paper presented at Second Photovoltaic World Conference, Vienna July 1998




Bent Sørensen

Roskilde University, Institute of Mathematics and Physics, Energy & Environment Group.

email:, www:, fax +45 4674 3020

P.O.Box 260, DK-4000 Roskilde, Denmark, Phone +45 4674 2028


ABSTRACT: The purpose of the work is to assess the social costs of photovoltaic systems, such that they can be compared with those of other energy solutions, including present fossil and nuclear technologies.

Keywords: Environmental effect - 1: life-cycle analysis - 2: Si-based -3



Life-cycle analysis: Identifying direct and indirect environmental and social impacts from a product or a system through all the phases of procurement, manufacture, use and disposal.

Life-cycle assessment: Evaluating the impacts found by life-cycle analysis, using monetisation or multivariate analysis with politically determined weights of different impact types.



The paper presents the results of using a state-of-the art methodology for life-cycle analysis and assessment (Kuem-mel et al., 1997) to multicrystalline and amorphous silicon solar cells, including direct and indirect impacts from mining to recycling of decommissioned cells, for current technologies as well as for projected future ones (characterised by smaller material inputs and larger scale production).

The task involved at each step in the cradle to grave sequence (raw materials mining over manufacture and use to disposal of any residues and dismantled equipment) is illustrated in Fig. 1. It illustrates the side-chains providing equipment and operating procedures, all of which may lead to impacts of concern.

Because of variations in current manufacturing methods and industry secrecy preventing precise quantitative estimates of chemicals used, one should expect studies based upon different assumptions, e.g. based on different concrete facilities, to exhibit a large spread in impact values. Additionally, several earlier studies include only a subset of impacts identified, and if impacts are monetised, the additional questions regarding valuation affect the results (e.g. greenhouse warming impacts caused by fossil fuels used in manufacture and resource extraction) (IEA PVPS Workshop, 1997).

The present model is based on recent assessment of IPCC (1996) impact data and it compares present production methods to the ones that would be needed to bring the price of PV cells down to what would become interesting for the bulk power sector. In this way some of the uncertainty of the present small scale production facility estimates can be avoided, and the results would be more relevant for decision support applications.

The result is that externality costs for applications of PV in average European conditions decline from present 3 ECU cents per kWh of power produced to about 0.4 cents, which is smaller than for all fuel-based power technologies and comparable to the externality cost of wind-produced power.



Even more than for wind power, the PV power system LCA is dominated by the manufacturing process. The manufacture of crystalline silicon panels involves the main steps of extracting quartz, fabricating first metallurgic and then semiconductor grade silicon, ingot growth, slicing and cell/module production, whereas for amorphous silicon panels, the wafer formation, cutting and surface doping steps are replaced by direct vapour deposition of doped gases onto a substrate. Different types of solar panels differ with respect to use of substrates, glass cover material and/or films. The cells may have surface textures and elaborate grooving patterns for conductors, and modules may be incorporating reflectors and in some cases bypass diodes and even inverters (which otherwise would be system components). Finally mounting of panels in arrays may involve dedicated support structures, or may be building-integrated.

On the system side, further transformer equipment may appear, as well as battery storage or backup devices in the case of stand-alone systems. Decommissioning and dismantling of the solar equipment is expected to follow recycling and reuse patterns emerging for the building industry in general, probably as a front-runner industry.

Assessing the impacts is helped by the experience with similar production process steps found throughout the microelectronics industry. The basic raw material for silicon cells is silicium dioxide (sand, quartzite). It is reduced to metallurgical grade silicon in arc furnaces. Both mining and reduction may produce dust (and hence risk of silicoses). The furnaces additionally produce carbon monoxide and a range of silicon-containing compounds, that appear as dust that might be inhaled e.g. during cleaning operations (Boeniger and Briggs, 1980). At current low penetration, the photovoltaic industry has used scrap material obtained inexpensively from the microelectronics industry, but in the future, solar grade material (much less expensive to produce than microelectronics grade) will be used.

The next step is production of silane (SiH4), in the case of amorphous cells, and for crystalline semiconductor grade multicrystalline silicon, for subsequent doping and growth of monocrystalline ingots. These are ground to cylindrical shape and sliced into wafers, which are then cleaned. Multicrystalline cells may be obtained by slicing ingots made of cast multicrystal silicon in a process similar to that of crystalline, or they may be formed by vapour deposition similar to the process for amorphous cells, albeit at considerably higher deposition temperatures. The material used for mono-crystalline cells is currently thick and expensive, also in input energy, but future production is expected to use thin-film materials, also for mono-crystalline cells.

The chlorosilane production involves hydrochloric acid, and the chlorosilanes themselves are both corrosive, skin and lung irritating as well as toxic. Workers are required to use protective clothing and face masks with filters. Further risks are posed by hydrogen-air mixtures present, which could ignite and explode. One such case has been reported (Moskowitz et al, 1994). For amorphous silicon, special precautions are needed for handling silane gas, as it ignites spontaneously. One solution is never to store larger quantities of silane gas, and to use special containers designed to avoid leakage even in case of strong pressure increases. Also, fabrication cells usually triclorosilane produced in a fluidized bed and subsequent-ly purified to sites are equipped with automatically releasing fire-extinguishing devices.

Vacuum growth of crystalline material may involve dispersal of oily aerosols, that have to be controlled by wet scrubbers and electrostatic filters (CECSMUD, 1982). Doping of p-type material may involve boron trichloride, which reacts with water vapour to form acids easily absorbed through the skin, or diborane, which is a strong irritant and flammable as well. The n-type doping at the top layer of a crystalline cell uses phosphorous diffusion of POCl3 or P2O5 in sealed environments, whereas the n-type doping of amorphous cells may involve phosphine (PH3), a highly toxic substance widely used in the semiconductor industry (Watt, 1993).

Grinding and cleaning of wafers produce silicon-containing slurry with remains of detergents used. An alternative is ribbon-growth, which avoid these problems (CECSMUD,1982). Amorphous cell manufacture also involves a number of cleaning agents. Etching of surface textures may employ a variety of techniques, selected on the basis of concern for recycling of chemicals and reduction of the use of toxic substances (Watt, 1993). Workers have to wear protective clothing, and high levels of ventilation are required. Drying uses liquid nitrogen and may be fairly energy intensive.

Screen printing of electric circuits involve possible work environment problems familiar to the microelectronics industry (caused by metal particles and organic solvents). Laser grooving involves the laser safety precautions for radiation and fires, and the application of coatings such as titanium oxide or silicium dioxide is considered relatively harmless. Cell testing and light soaking of amorphous cells (in order to avoid restructuring degradation) should be done in special rooms due to the risk of exposure to ultraviolet radiation. Personnel replacing bulbs should wear safety masks and gloves, if pressurised krypton lamps are used. Polymer coatings such as ethylene vinyl acetate (EVA) or polyvinyl fluoride (tedlar) may have some health impacts during their manufacture. If soldering is used in module assembly, fumes should be controlled.

The tendency is for increasing use of robots in the manufacturing process lines, leading normally to reductions in health impacts for remaining workers. The photovoltaic industry is aware of the problems of current pilot production lines and aims at controlling or replacing the chemicals identified as troublesome (Patterson, 1997).

There are few impacts during the operation of photovoltaic installations. Land use may be an issue for central plants, but not for building integrated photovoltaics. Albedo changes caused by the presence of the panels are not significantly different from those of alternative roof surfaces, in the case of roof-integrated cells, but could have climatic impacts in case of large, centralised solar plants located e.g. in desert areas. Reflections from panels located in cities could be annoying, and considerations of visual impacts will generally require careful architectural integration of panels. In some areas, cleaning of solar panel surfaces for dust may be required, and electronic control equipment such as inverters may cause radiofrequency disturbances, if not properly shielded. As mentioned for non-silicon cells, behaviour of panels during fires is an important consideration. Recycling of solar cell equipment has also been mentioned as a requirement (Sørensen, 1993).

For a silicon-based photovoltaic system integrated into a building, the LCA impact evaluation presented below shows modest negative impacts, most of which occur during the manufacturing phase, and substantial positive impacts in the area of impacts on the local and global society. The impacts during manufacture are to a large degree resulting from the use of fossil fuels for mining, manufacture and transport, according to the marginal approach taken in the references used. A comprehensive analysis of a renewable energy scenario would instead use the new energy system to determine indirect energy inputs, with substantially altered results as a consequence.

The source used for energy pay-back times and carbon dioxide emissions is Yamada et al. (1995). Other estimates give current energy pay back times for a-Si and c-Si (monocrystalline silicon) systems as 2-3 y and future ones below 1 year (Alsema, 1997; Frankl et al., 1997), while current greenhouse gas emissions have been estimated at 100-200 g CO2 equivalent per kWh of power produced, declining to some 40 g CO2-eq./kWh after year 2010 (Dones and Frischknecht, 1997).


The photovoltaic case is special, because the direct cost at present is far higher than that of the alternatives. If the cost is going down in the future, many of the impacts will also go down, because they are associated with material use or processes that will have to be eliminated or optimised in order to reach the cost goals. The lower cost estimate for 2010 is based on the stacked cell concept of Wenham et al. (1995). Yamada et al. (1995) quotes about 6.7 US cents/kWh for amorphous cell systems integrated into roofs. The cost per kWh produced obviously also depends on the location of the building.

The different spans of economic benefits from the power sold, exhibited in Table 1 below, are meant to reflect the differences in load-following capability of the three different types of plants. For a variable resource such as solar energy, there will be additional costs in case the penetration becomes large compared with the size of the grid system (say over 30%), because in that case additional equipment must be introduced to deal with the fluctuating power production (cf. power duration curves discussed in Sørensen, 1979).

Fig. 2 illustrates in schematic form the multi-pass nature of the assessment phase, into which the LCA analysis contributes valuable input.


  • Alsema, E and van Engelenburg, B., 1992. Environmental risks of CdTe and CIS solar cell modules. pp. 995-998 in Proc. 11th EC Photovoltaics Solar Energy Conference, Montreux. Harwood Academic Publ.

    Boeniger, M and Briggs, T., 1980. Potential health hazards in the manufacture of photovoltaic solar cells. Ch. 43 in Health Implications of New Energy Technologies (Rom and Archer, eds.), Ann Arbor Science Publ.

    CECSMUD, 1982. Sacremento Municipal Utility District 100 MW Photovoltaic Plant. Draft Environmental Impact Report, Californian Energy Commission, State Clearing House # 81111253, Sacremento.

    Dones, R and Frischknecht, R., 1997. LCA of PV systems: results of Swiss studies on energy chains. . Paper presented at IEA-PVPS workshop on "Environmental aspects of PV power systems", 9 pp., Utrecht.

    Frankl, P, Masini, A, Gamberale, M and Toccaceli, D., 1997. Simplified LCA of PV systems in buildings. Paper presented at IEA-PVPS workshop on "Environmental aspects of PV power systems", 14 pp., Utrecht.

    IEA PVPS Workshop, 1997. Photovoltaic Power System Group of the International Energy Agency: Workshop on Life-cycle analysis of PV systems, Utrecht, June.

    IPCC, 1996. Climate Change 95: Impacts, Adaptation and Mitigation, Cambridge UP.

    Kuemmel, B, Nielsen, S, and Sørensen, B., 1997. Life-cycle analysis of energy systems. 221 pp. Roskilde University Press.

    Moskowitz, P., Buchanan, W. and Shafarman, W., 1994. Lessons learned from a hydrogen explosion at a photovoltaic research facility. Brookhaven National Laboratory, preprint.

    Patterson, M., 1997. The management of wastes associated with thin film PV manufacturing. Paper presented at an IEA Workshop on Environmental aspects of PV power systems, Utrecht, 25-27 June. (5.pp).

    Srrensen, B., 1979. Renewable Energy, 683 pp., Academic Press, London.

    Srrensen, B., 1993. What is life-cycle analysis? pp. 21-53 in Life-cycle analysis of energy systems, Workshop Proceedings, OECD Publications, Paris.

    Sørensen, B., 1993a. Technology change: the actor triangle, Philosophy and Social Action, 19, pp. 7-12

    Watt, M., 1993. Environmental & Health Considerations in the Production of Cells and Modules. Centre for Photovoltaic Devices & Systems Report # 1993/02, University of New South Wales, Sydney.

    Wenham, S, Green, M, Edminston, S, Campbell, P, Koschier, L, Honsberg, C, Sproul, A, Thorpe, D, Shi, Z and Heiser, G., 1995. Limits to efficiency if silicon multilayer thin film solar cells, pp. 1234-1241 in 1994 First World Conf. on Photovoltaic Energy Conversion, Kona, vol. 2, IEEE

    Yamada, K, Komiyama, H, Kato, K. and Inaba, A., 1995. Evaluation of photovoltaic energy systems in terms of economic, energy and CO2 emissions. University of Tokyo, Preprint



    Figure 1 (above): Input and output streams for a particular life-cycle step (Kuemmel et al., 1997).

    Figure 2 (below): The actor triangle, a model of democratic planning, decision-making and continued assessment (Sørensen, 1993a).

    Table 1: Outcome of life-cycle analysis for Si-based PV cells (Kuemmel et al., 1997).

    Environmental impacts

    Releases from fossil energy used in the steps of the PV conversion cycle:

    CO2 (m-Si now and around 2010)

    ( a-Si now and around 2010)

    SO2 and NOx (m-Si now and around 2010)

    ( a-Si now and around 2010)

    Greenhouse effect from fossil emissions ( m-Si)

    ( a-Si, both either now or in 2010)

    ( if non-fossil energy is used in PV production)

    Mortality and morbidity from fossil air pollution

    described above (m-Si, now and 2010)

      ( a-Si, now and 2010)

    Land use

    Visual intrusion

    impact type: emissions


    75, 30

    44, 11

    0.3, 0.1

    0.2, 0.04
















    monetised value US cents/kWh






    2.4, 1.0

    1.4, 0.4

    or 0


    0.05, 0

    0.03, 0



    uncertainty & ranges*













    Social impacts

    Occupational injuries:

    1. From fossil fuel use ( m-Si now and 2010)

    ( a-Si now and 2010)

    2. From panel manufacture

    3. From construction and decommissioning (differential from using other building materials)

    4. From operation




    0.01, 0

    0, 0











    Economic impacts

    Direct costs (at present)

      ( around 2010)

    Energy payback time ( now and 2010, a-Si)

    Labour requirements ( now and 2010)

    Benefits from power sold ( penetration < 20%)




    3y, 0.5y

    40, 4 man-y/MW














    Other impacts

    Supply security (plant availability)

    Robustness ( technical reliability)

    Global issues ( non-exploiting)

    Decentralisation & choice

    Institution building ( grid required)















    NA= not analysed, NQ= not quantified. Values are aggregated and rounded (to zero if below 0.01 US cents/kWh).

    * (L,M,H): low, medium and high uncertainty. ( l,r,g): local, regional and global impact.

    ( n,m,d): near, medium and distant time-frame.

    The bold-face alternatives pertain to what we consider the most likely technologies for the future.