A spreadsheet was developed for the estimated component costs of thin film modules of various types. This was done in two parts – so-called commonalities or BOM of most thin film modules; and the unique aspects of each design, being mostly the semiconductors that convert sunlight to electricity (nonBOM). An initial production level of 25 MWp/yr was assumed, which is like the capacity of existing facilities or those that are planned by new start-ups. Then similar estimates were developed for future, larger scale production levels, in which both economies of scale and technical progress were estimated.

In most cases, thin film manufacturing bifurcates into: (i) approaches that take advantage of high quality processes that cost more but produce the best films or (ii) less expensive processes that produce less efficient devices. Another bifurcation is between glass and flexible substrates, and this can have implications at the system cost level (discussed below). In addition, some technologies have low semiconductor materials costs almost independent of how wasteful the approach; others are very sensitive to materials utilization rates. List 11.1 shows the primary set of thin film approaches examined here and the breakdown along these lines, with some simplifications where possible. We do not know which of these various strategies will be the best. Nor will any of them with only current technology meet the TW Challenge – significant R&D still needs to be done. But many of them have the potential if fully developed.

The technologies will be analyzed in these broad categories to capture the impacts of each difference in cost and efficiency. In no cases are actual company costs used, since these are not known and could not be used if they were. In addition, by retaining some flexibility, more

List 11.1 Characteristic thin film designs



Selected example companies

present and future variations could be examined rather than being limited to actual process lines and device designs. It will be established that this approach does not much affect the accuracy or ruggedness of the estimates. In fact, in all cases, there is substantial overlap that underpins comparisons, e. g., efficiency, substrates, and BOM. The differences can be smaller than the similarities.

In addition to the above, some newer thin film technologies were also investigated, but in a more rudimentary fashion. These technologies are 5-20 years behind developmentally those in Table 11.1, and the purpose of examining them is to see if they have potential to be even better if all research challenges are overcome. However, meeting the TW Challenge does not require that any technologies beyond those given in Table 11.1 (and existing x-Si) be developed (but no doubt, some will be, reducing risks even further).

Various sources were used for these estimates. The most critical challenge was to get the right process steps and the right set of materials. If process steps or materials are missing, then the analysis can be off substantially. Literature, presentations, parallel work at NREL, and private communications were the source of the process steps. However, it is almost certain that a few, idiosyncratic processes were missed, as these are the least likely to be publicly discussed. Shunt curing and photolithography are examples of processes that may be overlooked (but actually, were not). Steps like these are not major costs, and they vary among approaches. To capture this type of missing information, a small additional ‘miscellaneous’ category was included for all options.

There is a further check on near term cost analysis: both the amorphous silicon and CdTe technologies are in production at the 25 MWp/yr level, and several companies publicly an­nounced that they are profitable (e. g., First Solar and UniSolar, May 2005). Thus one can deduce that their costs are within a reasonable range of those of x-Si for existing produc­tion. Similarly, many companies publish partial accounting of their costs (e. g., total capital
investment for new plants). Both First Solar and UniSolar did this recently ($2.5/Wp for UniSo­lar and $1.2/Wp for First Solar, from press releases). Others have as well. Another example of an ‘easy’ cost to estimate is semiconductor materials. The feedstock costs can be found (e. g., Appendix 1 of Keshner and Arya 2004) and the layer thicknesses estimated from published articles on cells. The fraction of material wasted is important, since this can vary greatly. All in all, many aspects of production can be estimated in a like manner.

Estimates included processing equipment and the materials, and other associated costs such as labor, energy, and rent. The BOM costs were easiest, since they are usually commodity materials such as glass, metal, plastic, or ethylene vinyl acetate (EVA). In those cases, quotes were taken from the internet, from vendors, or from private communications. In many cases, common sense estimates were made based on these inputs. No number represents an absolute, since vendors vary, prices vary with volume, and costs change. They are educated estimates, and that’s all they should be viewed as. Where they are cost drivers, care was taken to assure that they are reasonable.

The nonBOM portion was more challenging, but estimates are possible from publications; and trends are relatively easy to discern. Handling and other miscellany were included. Main­tenance was taken as 4 % of the initial capital cost for all technologies. Capital cost per unit output was assumed to be based on a capital recovery factor of 15 % per year.

Analysis was also done of future costs, based on expected technical progress. For example, costs can be adjusted for thinner layers and improved process waste percentages. Based on an increase in throughput (from faster film fabrication, from thinner layers, and wider substrates, etc.), capital, maintenance, energy, rent and other costs can be adjusted. This type of analysis was done to project future cost trajectories.

In several cases the author received private communications of in depth breakdowns of entire thin film processes from private sector sources. But these were always given within a framework of confidentiality, to be used in a generic, nonattributable way. That is why they cannot be referenced; and the particulars of processes are not broken down in parallel to any single company’s approach. Yet detailed numbers exist that back up the results given here, and often the author’s estimates agree quite closely or are higher than numbers mentioned in various venues (conferences, workshops, group conversations).

In all cases, some common sense had to be used. No single cost number is usable alone; experience and insight must be applied. A synthesis was made.

Are the estimates off? Definitely, but likely within reason. Are the process sequences off? Probably not much, except for the proprietary aspects. It is the author’s belief that the estimates in this chapter for Table 11.1 thin films will track real world costs within about 20 %. For others (so-called third generation (3G) thin films), the uncertainty is much greater and will be noted where possible.

A simple, but important arithmetic relationship underlies the analysis. Most PV costs are given in dollars per watt peak ($/Wp). This is fine for the end user (especially if it is a system price), but it hides the nature of the technical challenges, especially in thin films. Two compo­nents go into a cost in $/Wp: the output or efficiency of the device, and its manufacturing cost per unit area. By combining them you get a cost in $/Wp. However, costs and efficiency vary greatly among thin films. There are many thin film options with very low potential costs (especially, po­tentially low nonBOM costs), but often they have inadequate efficiencies, resulting in high $/Wp costs. Their challenge is to raise their performance. But this is often overlooked when people merely highlight their low nonBOM manufacturing cost. Conversely, there might be thin films with high efficiency that have high area costs, and the balance determines their competitiveness.

The actual relationship is very simple: the dollars per watt cost can be found simply by dividing the manufacturing costs per unit area (say $/m2) by the output of the same area (which for 1m2 is 1000 Wp/m2 times the efficiency). The same relationship works at the module level: the module cost (in $/module) divided by its output (Wp/module) is its $/Wp cost. Obviously, the same relationships show how to go the other way: if one knows the $/Wp cost and either the efficiency (or unit output) or the area cost, one can calculate the missing parameter. The simple relationship is as follows:

$/Wp = (Cost/unit area)/(output/unit area) (11.1)

Unit area can be the module area; or, as in this chapter, the cost per square meter. Output per square meter is 1000 Wp/m2 times the efficiency (or it is the module’s nameplate rating at **STC). Thus a $10/m2 area cost for a 10 % module is $0.1/Wp; and for a 5 % module, it is $0.2/Wp. A $1/Wp cost for an 8 % module implies an $80/m2 manufacturing area cost; but for a 12 % module, it’s $120/m2. This is all arithmetic using the relation above. Because thin film technologies vary across a wide spectrum in both cost and efficiency, this is the way the analysis must be done to reveal underlying issues.

Updated: August 23, 2015 — 4:11 am