So far, the potential to achieve very low cost has been emphasized. That is critical to making PV cost effective enough to be used economically. But there are other factors to the ‘TW Challenge.’ The main one is materials availability. To have 10-20 TW of PV energy installed by mid century, we need about 50-100 TW peak, due to the 15-25 % capacity factor of PV. To accomplish this by 2065, for example, would require very high initial growth, resulting in about 4000 GWp of annual PV production in 2065. This means that about 15 TW (not peak) of PV would be installed and producing electricity in 2065 (given an assumed average lifespan of 30 years). Figure 11.6 shows a growth rate for PV to reach 75 TWp installed in 2065.

The materials requirements to meet this challenge have been studied in an NREL FAQ (NREL, 2005 http://www. nrel. gov/ncpv/pvmenu. cgi? site+ncpv&idx=3&body=faq. html) by this author, which is summarized here.

We do not expect shortages of most basic materials (glass, steel, aluminum, and plastic) except perhaps copper if its current extraction growth rate falls. If this happens, the problem might be alleviated by changes in BOS design to use other conductors.

We do foresee possible availability issues for a few of the semiconductor materials. NREL, 2005 shows that only the CdTe and CIS technologies might be affected; the thin film silicon technologies are not limited, even those with germanium (there are major amounts of unused byproduct Ge in aluminum, coal, and zinc ore). Newer, 3G, technologies need to be studied and are not included in the FAQ or this discussion.


Figure 11.6 A physically reasonable world annual PV sales growth rate that would reach 75 TWp installed in 2065. PV has grown at over 30 % per year over the last six years (and over 50 % in 2004).

Given that CdTe and CIS are important low cost technologies, Table 11.20 summarizes the potential installed volumes of CIS and CdTe given an optimistic scenario of complete recycling and thinner layers (0.5 micron). Note that the thickness of today’s CIS and CdTe PV is about 1-3 microns. Record efficiencies in these materials (at 2-3 micron thicknesses or more) are 16.5 % for CdTe and 19.5 % for CIS. So far, nonoptimized 1 micron CIS cells have reached about 17 % at NREL (Ramanathan et al., 2005); and 0.9 micron CdTe cells have reached about 11 % at U. Toledo (Gupta, 2001; Gupta and Compaan, 2005); submicron cells are planned (even down to 0.25 micron), and some work is already being done (Ernst et al., 2003). To date this has not been an area of much research, because current module costs have not yet been optimized for semiconductor materials costs. However, as thickness has now been identified as a key criterion for TW production, NREL has recently shifted some funding into this area.

Table 11.20 gives CIS and CdTe production by 2065 by assuming: (i) all existing amounts (beyond current nonPV demand, itself assumed growing at 1 %/yr) are available for use, and within PV, (ii) there is 100 % materials use, and (iii) complete recycling. Photovoltaic materials are not used up the way fuels are; they can be fully reused in new devices. Most data in Table 11.20 is derived from Andersson, 2000 and Sanden, 2003 and USGS, 2003. The final amount in TWp (right column) is found by dividing the total cumulative amount of feedstock available between now and 2065 by the amount needed per TWp.

Amounts of tellurium or indium that could be mined as primary materials (not as byproducts) have not been included in Table 11.20. We do not know the potential size of such deposits. Especially for tellurium, a material with very small markets to date, it seems possible there could be significant unexploited deposits. Such deposits could change the whole picture of materials availability for the CIS and CdTe technologies.

Table 11.20 Potential installed TWp of CIS and CdTe in 2065 (with complete recycling)

Primary metal and its

assumed growth

Percent byproduct currently unused in primary metala

Cumulative amount in 2065 (MT) using assumed extraction growth rates

MT required per TWp

Maximum installed in 2065 (TWp)


Zn, 1 %//yr


100 000


17 TWp


Cu & coal, 1 %/yr




300 TWp


Cu, 1 %/yr


330 000

11 000

30 TWp

aIn all cases, most of the current byproduct is unused (Sanden, 2003); assumes 15 % efficiency, 0.5 micron layers. Future research may allow reducing layer thickness further, as well as higher efficiencies, both of which would reduce materials demand. No feedstock sources beyond those given in the table are considered (e. g., tellurium mines).

b Indium required in devices is reduced by 20 % replacement by Ga (as in existing devices); future designs may include even larger substitutions. Assumes 15% module efficiency and 0.5 micron thick layers. Future research may further reduce layer thickness and increase efficiencies, reducing materials demand.

Notes: The amounts in the table assume steady growth along historical lines in Cu and Zn extraction. Of these, Cu seems more vulnerable to slowing over the next few decades. Also, the unused byproduct amounts are very uncertain: they are based on extrapolating average Te and In levels in the primary ores. However, actually processing this material to extract a high percentage of Te and In will be an economic challenge. For example, only 60 %-80 % of the base metal content is extracted. In addition, the available byproduct will be unused early in the growth of PV but must remain available for future processing as demand increases; this is currently not a normal procedure in the mining industry.

Because such possibilities are unknown, we limit ourselves to the values in Table 11.20. Using a factor of five to reduce to TW (not peak) on the amounts in Table 11.20, CIS could contribute as much as 3.4 TW, and CdTe, as much as 6TW by 2065. This means that these technologies can each be considered capable of meeting the TW Challenge and effectively contribute to the reduction of climate change. The amounts are also a substantial fraction of the desired 10-20 TW amount (and of course, huge by any other measure; e. g., the size of US energy consumption is 3 TW). Possible additions from primary materials are not counted in this sum, so perhaps even more could be made. Further, a steady state should be attained around 2065 in which recycled modules and ongoing PV device improvements (thinner cells, higher performance) would stabilize the need for newly extracted materials after 2065. The need could actually decline. But to be prudent, we should not assume that CdTe and CIS will carry the entire load, alone, despite their potential economic leadership (and especially because CIS is still unproven in manufacturing).

The use of certain materials used to make thin film modules deserve a brief discussion. Although perceived as a problem by some, many studies show that no danger exists from mak­ing, using, or disposing/recycling CdTe modules (http://www. nrel. gov/cdte/; and especially Fthenakis, 2004). There also apparently are no issues in terms of market acceptance. The biggest market for CdTe has been Germany, a country sensitive to environmental and heavy metal issues.

The CIS technology has an echo of this problem due to the presence of selenium, also an element that is viewed with concern (though its recent use as a food supplement has ameliorated perceptions greatly). Other PV technologies usually have smaller, parallel problems that are less obvious – e. g., the Pb solder in x-Si technology, or the toxic/explosive gasses in thin film silicon. In fact, it is well accepted that no energy option, no matter how ‘green’, is totally without environmental impacts, especially on the TW scale. The best known and perhaps most rational measures of environmental impact, energy and CO2 paybacks, are favorable for thin films (NREL, 2005a) – about one year for the kind of large, thin film systems we are examining.

One other barrier often cited is the land area needed to supply TWs of PV. Actually, using the original Nate Lewis number of 125 000 TW of sunlight on the Earth’s surface can easily dissuade us of this concern. Assuming that this falls evenly on land and sea, this is about 36 000 TW falling on land. Assuming we need 20 TW of PV, and the PV systems only averaged 10 % sunlight-to-electricity conversion, that would be 0.55 % of the Earth’s land area for modules. Assuming a (module-system area) packing factor of 40 %, this requires 2.5 times more land, or 1.4 % of the land area. Today, 1.1 % of the US land area is used for national defense (bases and bombing ranges) and 0.04 % is used to raise Christmas trees. Not only is the use of 1.4 % of land for PV not a serious burden for converting our energy infrastructure to solar, it is a positive advantage of PV (as stated in detail in the FAQ NREL, 2005a) because no other nonCO2 resource except nuclear has anywhere near the same level of energy density/unit area and ubiquity. The above analysis completely ignores the reduction in land area requirements that would result from using PV on rooftops or other existing structures.

Updated: August 23, 2015 — 6:49 am