Andrew Blakers and Klaus Weber
Centre for Sustainable Energy Systems
Engineering Department, Australian National University
The use of photovoltaic systems on a large scale in order to reduce fossil fuel consumption and greenhouse gas emissions requires that the energy associated with the construction, operation and decommissioning of PV systems be small compared with energy production during the system lifetime. That is, the energy payback time should be short. The energy intensity and cost of PV systems are closely related. At present the energy payback time for PV systems is in the range 8 to 11 years, compared with typical system lifetimes of around 30 years. About 60% of the embodied energy is due to the silicon wafers. As the PV industry reduces production costs and moves to the use of thin film solar cells the energy payback time will decline to about two years.
The use of photovoltaic systems on a large scale in order to reduce fossil fuel consumption and greenhouse gas emissions requires that the energy associated with the construction, operation and decommissioning of PV systems be small compared with energy production during the system lifetime. That is, the energy payback time should be short compared with the system lifetime.
A distinction needs to be drawn between energy consumption and carbon dioxide production associated with PV systems. Although there appears to be relatively limited reserves of oil and gas, coal is abundant. The most likely cause of a cessation of coal burning is not depletion of supplies, but rather unacceptable climatic consequences. In addition to energy payback time for PV systems, carbon dioxide payback times need to be considered. Both times must be short compared with the PV system lifetime if a large sustainable PV industry is to the established. In general, energy intensity and carbon dioxide intensity are closely related. Over the next few decades at least the energy used to construct PV systems will be derived primarily from fossil fuels. In the long term "solar breeding" will be possible, whereby energy for the production of PV systems will be derived from PV systems. This will reduce or eliminate carbon dioxide emissions associated with PV system manufacture.
By far the largest fossil fuel inputs for a photovoltaic system are associated with production and installation. Fossil fuel derived energy required for the operation and decommissioning of a PV system is trivial. Hydroelectricity and wind energy share this characteristic. Many studies have looked at energy inputs to PV systems. It is difficult to arrive at definitive numbers because production technology is constantly improving and because the fossil fuel intensity of various operations depends on production scale and production location.
An important conclusion can be drawn from the various studies. The cost of the various components of a PV system is well correlated with the energy content of that component. The reason for this is that PV is a material intensive technology, and the energy content of materials is reflected in their price. Thus the most expensive component of a conventional PV system, the silicon wafers, is also the most energy intensive component. It is clear that as the cost of PV systems declines, then so will the energy content.
Most PV systems are based on panels that comprise about 40 single or multicrystalline silicon wafers encapsulated behind glass using an EVA pottant material. An aluminium frame and a junction box complete the panel. Groups of panels are connected together on supporting structures that are mounted on buildings or in open fields. The cost of the silicon wafers amounts to about half of the cost of a PV panel. It is likely that thin film solar cells based on thin layers of crystalline silicon or alternative materials (amorphous silicon, copper indium diselenide, cadmium telluride) will challenge wafer based crystal silicon solar cells over the next decade. The result will be a substantial reduction in the energy and carbon dioxide intensity of PV systems.
This study focuses on crystalline silicon PV panels. At present, crystalline silicon wafer panels have 85-90% of the world market. There is a high likelihood that the dominance of crystal silicon solar cells will continue for many years to come because of the abundance and non-toxicity of silicon, the high and stable efficiency of silicon solar cells, the ability to share R&D, infrastructure and human resources with the IC industry and its present market acceptance and dominance. Although solar cells based on silicon wafers are likely to eventually be replaced over the next decade by thin film solar cells, there is every chance that the thin film solar cells will in fact be fabricated from thin films of crystalline silicon rather than from other materials.
The main goal of the Epilift [Description 1997] project is to create efficient solar cells using thin crystalline silicon cells that use 10% of the amount of silicon used in conventional wafer-based silicon solar cells. The energy content of the crystal silicon in an Epilift PV panel will be greatly reduced. Amorphous silicon panels and Epilift panels will have similar embodied energy per square metre. [Press release 1997, Third Generation 2000]
The various studies of the embodied energy of currently produced PV systems that have been performed over the years make a range of different assumptions. Their results range over a factor of two for the energy payback time of a PV system. However, the predictions for the energy payback time of future PV systems are all small (a few years). This is because the availability of PV systems with a sufficiently low cost to be deployed in large quantities will require the elimination of expensive components, which also tend to be the energy intensive components (e.g. silicon wafers, aluminium frames and expensive support structures).
In this study, estimates at the high end of embodied energy in PV systems have been adopted. However, the likely future elimination of the energy intensive components of PV systems means that most future estimates converge at rather small numbers for embodied energy.
In addition to energy embodied in PV materials, the energy embodied in the machines used in manufacturing need to be taken into account. Similarly, the energy used to manufacture the machines that are used to manufacture the production machinery needs to be taken into account, and so on. However, in an energy intensive product such as a PV panel the energy embodied in the materials far exceeds the energy embodied in the production machinery, and the latter can be neglected for practical purposes. Indirect energy, such as for heating, lighting, office equipment and transport is a significant overhead and must be included.
Electricity is related to primary energy (usually fossil fuels) through the average electrical energy conversion efficiency of the electricity industry, which is assumed to be 38%. Since the output of a PV panel as well as most of the energy inputs are in the form of electricity, electricity (in kilowatt hours, kWh) has been used as the basic energy unit in this study. Megajoules and kilowatt-hours are related by 1 kWh = 3.6 MJ.
The lifetime of a PV system is assumed to the 30 years. Many manufacturers now offer 20-year guarantees, and PV panels might last 40 to 50 years in non-maritime locations.
The energy output to of a PV panel depends on its location. In this study the location is assumed to be Sydney. The PV panels are assumed to be mounted on a fixed frame facing north and tilted at the latitude angle. The average irradiation on the panels is 6,935 MJ per square metre per year (1,926 kWh/m2/year) . In Port Hedland in NW Australia, which is an excellent site by world standards, the average irradiation on the panels is 2,494 kWh/m2/year. Sydney has good insolation compared with most places in Europe (30-70% larger).
The actual electrical output of the PV system is the irradiation multiplied by the system efficiency. The single crystalline silicon solar cells are assumed to have an efficiency of 14% under standard testing conditions. The cell efficiency (after encapsulation) during actual operation (elevated temperature, reduced insolation intensity, dirty glass etc) is reduced to 11%. The panel will have an output of 100 W/m2 under 1 kW/m2 illumination (allowing for a cell packing factor of 91%). In Sydney the annual average energy production of a panel will be 193 kWh/m2/year before system electrical losses.
Electrical losses caused by the inverter, transformer and electrical resistance are assumed to be 15%. Electrical losses in the State distribution network are assumed to be 7%. The overall electrical efficiency of the balance of systems (BOS) components is thus 79%. This reduces the annual average energy production of a panel to 153 kWh/m2/year.
Production and installation of a PV system can be divided into three sectors:
Silicon wafers are produced from electronic grade silicon (EG-Si), which in turn is produced from metallurgical grade silicon (MG-Si). Metallurgical grade silicon is used in large quantities in the steel and other industries. It is created by the carbothermic reduction of silicon dioxide (quartz, sand), a process in which coal, coke and woodchips are heated together with silicon dioxide. The carbon strips the oxygen from the silicon dioxide to create carbon dioxide and silicon. This process produces carbon dioxide both directly as part of the chemical reaction and indirectly since the reactor must be heated electrically. The energy content of the carbon could have been used to produce electricity. This forgone energy production is included in these calculations.
Metallurgical grade silicon has a typical purity of about 98%. This must be upgraded to around 99.9999999% in order to meet the requirements for the semiconductor industry. The purification process is generally accomplished via the Siemens process. Silicon is reacted with HCl to produce trichlorosilane, which is then the decomposed with the aid of hydrogen at 1200 degrees to produce highly pure electronic grade polycrystalline silicon (EG-Si) and silicon tetrachloride (SiCl4). There is 3.5 kg of silicon embodied in the SiCl4 for each kg of EG-Si produced. The SiCl4 is used to produce pigment, quartz fibre etc. The energy content of the SiCl4 is ascribed to these other industries rather than to the solar cell industry. The yield of the purification step is assumed to be 95%.
The next step is to melt the EG-Si in a Czochralski crystal puller at 1400 degrees and slowly crystallise the silicon to form a single crystal ingot of silicon. Alternatively, the EG-Si can be melted and crystallised by the casting process to make a large grained multicrystalline silicon ingot. It is assumed that only 80% of the EG-Si loaded into the Czochralski crystal puller is used, with the losses mostly due to the removal of the top and tail of the ingot (which has a lower purity). The yield of the Czochralski process itself is assumed to be 90%, giving a total yield of 72%.
The final step in wafer production is to slice the ingot into wafers. This is done with a multiwire saw and abrasive slurry. 40 to 50% of the ingot is lost as sawdust in this process. The ingot is typically sliced with a pitch of 0.5 to 1 mm to produce wafers with a thickness of 0.3 to 0.5 mm. The wafer thickness is assumed to be 350 microns and the kerf loss is assumed to be 300 microns, giving a yield of 54%.
It could be argued that the widespread use of off-spec EG-Si and wafers from the IC industry by PV manufacturers substantially reduces the energy payback time associated with the wafers, since the energy content of the EG-Si and wafers could be ascribed to the IC industry. However, this is inappropriate since the IC industry does not actually make use of this silicon. The end user (in this case the PV industry) should be charged with the energy content of the silicon, although perhaps with some sharing of the energy content with the IC industry. In some cases, off-spec IC wafers are remelted in a Czochralski crystal growth process. In other cases they are cleaned up and used directly for solar cell substrates. In the former case the energy content of the resulting wafers will be quite high, since they will have gone through two Czochralski growths. For the sake of simplicity, it is assumed that the energy content of PV wafers calculated under the assumptions described in preceding paragraphs holds: a simple flow of silicon from quartz to Czochralski ingot..
Cell fabrication entails a sequence of high temperature diffusion, oxidation, deposition and annealing steps. Following metallization, the cells are connected into strings with copper tabs. Panel formation entails the lamination of the cells behind glass with EVA and Tedlar using heat and pressure. A junction box is mounted on the back of the panel. In most cases, an aluminium frame is placed around the panel perimeter. The yield of the cell fabrication and encapsulation process is assumed to be 90%.
The aluminium frame represents a significant fraction of the panel's embodied energy, but it is not required with some panel mounting systems. Determination of the energy content of aluminium is difficult, since it depends on the fraction that is recycled. Determination of the CO2 content is also difficult. Aluminium is often made using hydroelectricity, which has a small greenhouse impact (neglecting methane and carbon dioxide emissions from hydroelectric reservoirs, which could, in fact, be significant even in comparison with a coal fired power station). Thus aluminium production in Tasmania will have lower carbon dioxide intensity than aluminium production in Queensland.
The wafer area is 110 cm2. After trimming to make a pseudosquare solar cell the area is 100 cm2. The density of silicon is 2.3 gm/cm3. Thus the mass of each solar cell is 8.1 gm. These cells are encapsulated to make a PV panel with a packing factor of 91% (i.e. 91% silicon and 9% open space between the cells).
The balance of system (BOS) comprises wiring, power electronics, foundations, support frames, transport and installation. Of these, the support frames and foundations are by far the most energy intensive. In a system installed in an open field, the foundations are typically concrete while the support frames are steel. Both of these materials are energy and carbon dioxide intensive. In a system installed on a building roof, the foundations can generally be dispensed with. In addition, if the PV array forms part of the roof structure then the energy embodied in the displaced roof components can be set against the embodied energy in the PV array. Thus the energy payback time for the BOS components is considerably smaller for roof-mounted systems than for systems deployed in open field. Systems deployed in open field will generally have smaller inverter and electrical resistance losses, will have unimpeded access to sunshine and will often be in sunnier regions than cities. On the other hand, distribution losses will be higher.
It is difficult to estimate the energy savings possible by displacing roofing materials with PV panels, because many different types of roofing materials are in common use. For example, an aluminium facade could be replaced with a PV facade, saving large amounts of energy because aluminium is an energy intensive material. On the other hand, clay tiles or coated steel have relatively low embodied energy. The replacement of roofing materials with PV panels can introduce the problem of panel overheating, because the panel will be less able to shed heat from the back surface. On the other hand, if use is made of the heat generated by the PV panel, which is typically two to three times as much as the electrical energy output of the panel, then the energy payback time of the PV system will be very low.
A square metre of PV panel will require 90 solar cells with a total mass of 725 gm and an area of 0.9 m2. Allowing for the yield of the Czochralski growth (72%), ingot slicing (54%), cell fabrication (90%) and cell trimming (90%) processes, a total mass of 2,300 gm of EG-Si is required. A mass of 2,400 gm of MG-Si is required (excluding the silicon that is incorporated into the SiCl4).
Energy requirements for each step are assumed to be as follows:
The total energy requirement to produce a PV panel is 1,060 kWh/m2. In Sydney the useable panel output will be 153 kWh/m2/year, giving an energy payback time (EPBT) for the panel of 6.9 years. After mounting in an open field or on a roof the EPBT will be 11.5 or 8.3 years respectively. These energy payback times are well short of the likely system lifetime of 30 years. Figure 1 shows the energy payback time of the various process steps for a roof-mounted panel. Production of the silicon wafers accounts for 60% of the total energy payback time.
The energy payback time of a PV system in 10 years time is likely to be far lower than the current figure. Large-scale deployment of PV systems will force costs down, and hence also the energy content of PV systems. The following assumptions have been used to derive the likely energy payback time for PV systems in the year 2010:
After mounting on a roof the EPBT will be two years. The PV module alone has an EPBT of 1.7 years. The PV module is likely to have a similar embodied energy to a currently produced amorphous silicon module, in which the silicon thickness is only a few microns. The efficiency of the crystalline silicon panel will be much higher. The energy payback time in Sydney of an amorphous silicon module (United Solar UPM-880) is estimated to be 2.5 and 1.5 years with and without an aluminium frame respectively, assuming that the efficiency can be improved from 5% to 9% .
The cost of a PV systems will have dropped sharply by 2010, although not as sharply as the embodied energy because material costs will be less important in future PV systems. Figure 2 shows the energy payback time of the various process steps for a roof-mounted panel. Embodied energy in the balance of systems now dominates system embodied energy.
Australian electricity production is predominantly by means of coal fired power stations. The Australian Coal Association (ACA) and the Australian Gas Association (AGA) have recently produced environmental assessments of the production of electricity from coal and gas [6,7]. Both studies emphasise the importance of a complete life cycle analysis. In the case of natural gas, fugitive emissions of methane are included. On the other hand, the attraction of direct space heating and heating of water with gas, in terms of overall greenhouse emissions, is pointed out. In the case of coal, fugitive emissions such as methane from coal mines are included. On the other hand, the ability to use waste products to reduce net greenhouse emissions is analysed. For example, fly ash can be used as an extender in cement production. This 'displacement credit' could potentially reduce the greenhouse intensity of existing coal fired electricity power stations by around 10% if all the fly ash were to be used. The ACA and AGA studies are in good agreement with each other with respect to the greenhouse intensities of the production of electricity from coal and gas.
The ACA study also examined renewable energy greenhouse gas intensities. The carbon dioxide equivalent intensity of a PV system in Australian can be estimated from the energy payback time as follows. The national average carbon dioxide equivalent intensity for electricity production is approximately 0.98 kg of CO2 per kWh . The embodied energy in the PV system can be divided by the system lifetime (30 years) to calculate the equivalent greenhouse gas emissions per year. The data is shown in figure 3. The conclusions of the ACA study for photovoltaics are slightly more optimistic than the conclusions of this study, but do not take into account likely reductions in EPBT of PV systems over the next decade.
The most important parameter in the determination of the energy payback time of a PV system is the thickness of the crystalline silicon layer and the method of mounting the panels. Other important parameters are:
Fossil fuel use during PV system operation and decommissioning is negligible. Virtually all of the fossil fuel energy and carbon dioxide production associated with PV systems arises from the initial production and installation of the system.
Typical energy payback time at present is around 7 years. Mounting and installation of the system adds a further 1 to 4 years, depending upon whether it is on a roof or in an open field. This gives a total energy payback time for a PV system of 8 to 11 years.
Future PV panels that use thin films of crystalline silicon or other materials will have greatly reduced energy payback times. Such panels will be required if cost targets for large-scale production are to be met. The expected energy payback time will be in the vicinity of two years. The PV industry is enjoying production growth rates of about 30% per year, which is likely to bring substantial investment into the industry. It is expected that thin film PV panels will be in widespread use by around 2010.