Solar panels now pay back the energy used to make them in little more than a year

How much electricity do we get back from the large amounts of energy invested in making solar panels? An impressively detailed paper from researchers at the University of Utrecht provides some answers to this crucial question. In short, conventional PV modules made next year will achieve ‘energy payback’ in not much more than a year.

Some of the press commentary on the new article in Nature Communications focuses on this benign impact of solar and the scope for continuing improvements in energy use. Other writers took a very different tack. Instead of focusing on the rapid payback on the energy invested in the manufacturing processes of today, the journalists chose to concentrate on the much higher inputs in the past. This makes solar look less good. The headline on Ben Webster’s article in the London Times was ‘Solar panels less green than you think, say experts’. An anti-renewables US website’s headline was ‘Solar power actually made global warming worse, says new study’.

One of the aims of the Utrecht paper was to give us an estimate of when the annual global expenditure of energy on making panels fell below the amount of electricity produced by all the solar PV ever previously made. This calculation showed that the extremely rapid growth rate of solar, combined with the previously huge energy costs of making PV, meant that it wasn’t until about 2011 that solar PV ceased to be in overall yearly energy and carbon debt. The Times article got very confused about this point, implying that this meant that an individual panel made early than this date didn’t typically generate enough electricity to pay back the energy invested in making it. No, the Utrecht conclusion was that in any individual year until about 2011 the manufacturing energy use of the whole PV industry was greater than the electricity output of panels already on roofs and in fields. This is a very different point.

 The ‘energy return on energy invested’ debate for PV continues to rage. Those, like me, who believe solar will become the dominant world energy source at some not-to-distant point tend to believe panels to pay back their energy reasonably quickly. If, at the other extreme, you are deeply sceptical of renewables and want fossil fuels such as coal to provide most power, then proving solar PV is energy efficient assumes great importance. UK journalists such as Matt Ridley sit firmly in the camp of fossil fuel advocacy and voice the view that panels installed in the UK, for example, never pay back their embodied energy.

Mr Ridley can call on just one recent academic paper to support his view. This particular piece of research suggested a very long payback time indeed for panels installed in high latitudes. One of the reasons that the Utrecht work is so important is that it summarises all the research on energy payback, including 40 separate assessments of the energy payback on solar manufacturing. 39 of these assessments show very different figures indeed to those supported by Matt Ridley.


More generally, I wanted to get the some of the many additional Utrecht conclusions into the public arena. So I wrote to the lead authors Atse Louwen and Professor Wilfried van Sark to ask five specific questions on issues not directly reported in the study but which can inferred from their work.

The full Q+A follows. It is unamended by me (except to add a brief explanation of some technical terms). Many thanks indeed to Atse and Wilfried for the extraordinarily rapid and full response. (By the way, I think they are being very cautious in their assessment of how much electricity that the average solar panel generates in the course of year. This makes their assessment conservative.  Payback is probably faster than they say). 

1.       Please would you tell me approximately how much energy it typically takes to make 1 megawatt (peak) of solar modules today? And roughly how much did it take 5 and 10 years ago? 

Today, according to the experience curve we used, it would take for production of 1 megawatt-peak of average (~40% mono, ~60% poly) solar PV systems (so modules and inverters, installation, mounting structure) roughly 18.4 TJ of primary energy. Considering the locations of this installed capacity the average yield of 1 MW globally would be conservatively estimated to be 1200 kWh/kWp. This would correspond to an energy payback time of ~1.4 years. 5 years ago the figures would have been 25.9 TJ/MW (EPBT of ~2.0 years).10 years ago this would have been 36.9 TJ/MW (EPBT of ~2.8 years)

(My note: EPBPT is ‘energy payback time’, TJ is ‘terajoule’. 1 TJ is equal to about 278 megawatt hours).

2.       How long will it typically take a polycrystalline panel made on 1st January 2017 (and installed on the same date) to generate enough electricity to repay the energy used in its manufacture?

A complete PV system based on polycrystalline panels, made in 2017, would need 15.8 MJ of primary energy per watt-peak. This corresponds to an EPBT of roughly 1.2 years (for global average yield)

3, Solar PV production has historically grown at 45% a year over the last decades, according to your estimates. If growth were to continue at this rate, and the reductions in the energy required in manufacture of silicon panels also falls at the same pace as they have done historically, when would a panel made on 1st January 2020 reach ‘energy payback’? Perhaps 45% is too high a figure to use for future growth rates; when would the energy payback be on a panel made on 1st January 2020 be if future growth runs at just 20% a year?

Indeed over the period 1975-2015 the average annual growth rate (or CAGR, compound annual growth rate) was indeed 45%, but the last years this figure has already been a little bit lower, slightly below 30%. Forecasts are again a bit lower, we guess 20% would be a decent estimate. In our study we use projections for future development of capacity that are around 20% (slightly lower). With this in mind, an average PV system in 2020, in our model, would have an energy demand of 16.7 MJ per watt-peak, corresponding to an EPBT of about 1.3 years. Note this is higher than for a poly panel in 2017, as the share of mono systems is increasing and these have a slightly higher energy demand for production.

 4.       In their comments on your research, some journalists have focused on one aspect of your work. They quote your conclusion that in the ‘Increasing PR (performance ratio) scenario, (energy) debt was likely already repaid in 2011 for both CED (energy) and GHG emissions’ In other words, they say, until 2011 solar panels had a net adverse effect on carbon in the atmosphere. Your conclusion seems to arise because solar PV has been growing so rapidly that in any single year energy use in module manufacturing would have exceeded the total electricity generation of all previously produced modules. Are you able to confirm that if, say, PV had only grown at 20% per annum, then net energy debt would have been repaid earlier than 2011? In other words that it is precisely the very high rate of growth in PV that means energy debt increased until 2011?

This probably true indeed, but if solar growth would have been lower, then the reduction in EPBT would also have been lower, as it is a result of experience during production and as such is a function of the cumulative production of PV capacity. However, generally speaking, net energy is consumed when growth rates are larger than (1/PBT). As growth rates were in the past on average 45%, sometimes higher, and EPBT has dropped below (1/0.45 = 2.2 years) only recently, it is likely that if growth rates were constrained to 20% the break-even point would have occurred sooner. 

However, in terms of experience curves, the investments you need to make (in this case in terms of energy and GHG emissions) to bring the technology down to a certain environmental “cost” level, are always more or less the same, whether you take 20 or 40 years to make these investments. So the faster growth and temporary adverse effects now result in a faster increase of the positive effect, so to say. 

5.       Lastly, on the basis of the literature search you carried out during the research for your article, what do think is the current ‘Energy Return on Energy Invested’ (ERoEI) of a polycrystalline panel?

The ERoEI of a global average polycrystalline based system would according to our figures be about 19.8. For NW Europe, taking our home town Utrecht as an example with an estimated annual yield of 875 kWh/kWp (which is an average, actual yield for the Netherlands that my colleagues measured using data from thousands of PV systems), this would be 15.2. 

This ERoEI has been debated, also recently, even to the point were critics state that the ERoEI of systems in N Europe is smaller than 1.0. According to everything we have seen in literature and in our own research, this is just not true. A recent example of a study that states this ERoEI to be smaller than 1 is that by Ferroni and Hopkirk in Energy Policy  but upon review of this study we, and many researchers in the field, found that the authors severely overestimate the energy required to produce PV systems, and underestimate their electricity yield, among other issues with this paper. A rebuttal paper written by a large number of colleagues (not us) in the field has already been submitted to the journal, which we hope will be published soon. 

(My note: The Ferroni and Hopkirk paper is the one always cited by anti-renewables commentators).

Several readers have pointed out that the Utrecht paper uses estimates of production energy that suggest a much longer energy payback than the authors propose. My reasoning as to why I think the Utrecht paper is nevertheless correct is appended as responses to these comments below. Thank you to the commenters for raising this issue, which I should have spotted myself.