About Us

The Energy Challenge 2004 - Solar


Murray Duffin, Retired

Solar energy is our most abundant renewable resource. An analysis of insolation in the USA southwest shows that using only the 1% of the land area considered, that has a slope of <1% and >7kWh/sq. m./day insolation, concentrated solar (CSP) can provide power of approximately 30 Gwe. With effective storage (as for solar towers) this potential could provide at least 3x the total productive energy of today's economy.

Solar energy generation can be considered in 3 categories, solar thermal, solar photovoltaic (PV), and solar thermophotovoltaic (TPV). As TPV is still a far future technology, only the first two will be considered here. 2004 may have been the watershed year for the development of solar renewable energy, although that may not become obvious for several more years. While there has been little progress in installations in some years, technology has continued to improve, and with rising costs of coal oil and natural gas, interest in solar energy is now growing rapidly.

The new economic driver
North American production of natural gas is reported to have declined by more than 3% in 2003 vs. 2002, and based on reports by the major producers in America, production in 2004 seems to have declined at least 3% in the first half, and as much as 10 to 12% in the third quarter y-o-y. Hurricane damage can account for less than 40% of Q3 decline, so it seems that decline of mature fields is accelerating. Against earlier forecasts of natural gas prices below $5.00/Mbtu for H2 2004, recent cash prices have been above $7.00, at a time when demand is low, and storage is at record levels.

At the same time export demand for coal has caused prices to more than double, on average, for all but Powder River Basin coal; and we have seen oil prices rise 70% in a few months and are going into winter with heating oil stocks low and prices high. With possible brief respites, these trends appear irreversible.

Solar Thermal Overview
There are 2 major sets of solar thermal:
Direct heating and cooling
Electricity generation.

Each can be split into flat plate and concentrating (CSP) subsets although for electricity generation CSP is the economic choice for all but small-scale applications.

  • Direct heating and cooling
    The mature technology is water heating for home hot water, space heating and swimming pool heating. Flat plate technology is common, inexpensive and effective and has been used successfully with gradual growth for at least 50 years. Efficiency can be in the 25 to 50% range, depending on design. More recent technology that is growing rapidly, and is the preferred choice for larger installations like hotels and public swimming pools is evacuated tube heating systems which provide much more heat per unit area, remain effective even during light overcasts, and can reach 60% conversion efficiency.

    Recently compound parabolic concentrators (CPC) have begun to be industrialized, proving very effective in evacuated tube systems. The advantage of CPC is that, by concentrating sunlight it can raise liquid temperature in pressurized systems to >300 degrees F, enabling economic absorption chillers for cooling systems. CPC is also effective over a very wide angle of illumination, eliminating the need for a tracker in a concentrator system. The California Energy Commission retrofitted and optimized a 20 ton conventional double effect (2E) LiBr/water absorption chiller to be solar hot water driven, and have estimated that such a system can be supplied commercially for <$4500./ton, with a net reduction in electricity demand of 1.3kW/ton. Depending on hours/yr of operation and peak electricity costs, an economic payback of 4 to 8 years can be expected.

    Direct heating and cooling systems have the effect of displacing electricity, to use Amory Lovinss term, providing "negawatts" instead of megawatts, and generally at lower cost than increased generating capacity.

  • Electricity generation (CSP)

    Thermal power generation is being addressed in several ways—and for different sizes of installation:

    • Solar dish concentrators driving Sterling engine generators.
    • Trough concentrators heating a liquid to gas system driving a turbine generator.
    • Solar towers using large reflector (heliostat) arrays to heat molten salts which, through a heat exchanger, drive steam turbines.
    • Solar chimneys using rising air from a large ground level greenhouse to drive turbines at the base of a km high chimney.

CSP Details
Dish/Sterling systems tend to be aimed at tens of kW applications for grid connected distributed power, and reach conversion efficiencies near 30%. Cost of electricity is still high, though there is a wide range of estimates. Widespread use seems likely to be well in the future.

Trough concentrators get into the 100s of kW to tens of MW range, good for locally sighted factory power also at attractive efficiencies. The best-known examples are the SEGS series (now up to SEGS 9) in California. Recent projects have been commissioned in Nevada and Arizona. It has been estimated that a 100 mi. square in the Nevada desert could provide about 500 GWe, roughly equal to the USA installed electric power base. Up to now such applications have limited storage ability, so they are unsuitable to 24 hr. operation and dispatched power. CPC concentrators might overcome that drawback. These systems approach 14% efficiency today and are projected to get to 17% by about 2015.

Solar towers (Power Towers) are MW sized for utility type supply and have the advantage of retaining heat for 24-hour operation. Solar 1 was operated near Barstow Ca. in the 1980s as proof of concept. Solar 2, a 10 MWe upgrade of Solar 1 operated from 1992 to 1999, demonstrating the feasibility of storing heat for dispatchable power and 24 hr. operation. Solar Tres (17 MWe) has been planned for Spain, originally to go into operation by late 2003, but now delayed to 2006, seemingly by bureaucracy. Towers in the 100 MW range are projected. Solar towers are about 23% efficient in conversion of incident energy to electricity, and can realize up to 70% capacity factor. Current experience indicates a space demand of 10 acres/MW, with promise of at least a 20% reduction.

Solar chimneys4,5 are only theoretical so far, and seem to have captured most attention in Australia. They can be designed to heat water during the day to provide energy at night. Efficiency is estimated as 3%, but it seems likely that this can be at least doubled. Proposed designs have fresh air drawn into the heating area at ground level. Drawing in air near the tower top would augment generation with the sinking column of cooler air. In dry climates it should be possible to inject water vapor into intake air to further cool the descending air column. Current projected design is for 200 MW and requires about 23 acres/MW. Capital cost of $2.00/Wp is projected, but seems quite optimistic.

For utility scale electricity generation, the best choices today are trough concentrator and solar tower systems. An excellent 2003 analysis for trough concentrators2 (based on 2002 data and projections) considers a necessary competitive target price for electricity of $4.50/MBtu, assuming a floor fixed at that level by LNG. We now can be sure that LNG will not be a major factor for at least a decade, and even then will set a floor above $6.00/Mbtu. This analysis showed trough systems becoming competitive at 10 Gwe installed capacity and 6 cents/kWh. It now seems more likely that 7-8 cents/kWh will be good which can be reached at 5-6 Gwe installed. Another late 2003 report3, using well reviewed data and analysis developed independently be Sunlab and Sargent & Lundy gives present electricity costs of 10 – 12.6 cents/kWh now, going to 3.5-5.5 cents/kWh before 2020 for trough and tower systems. Growing fossil fuel shortages seem certain to accelerate progress relative to these studies.

Photovoltaic (PV)

Historically PV has been seen as much too expensive for widespread use, having been represented as "the energy of the future and always will be". 2003 saw a novel development that should change that conclusion. All of the pieces now seem to be in place for PV to breakthrough all the barriers of demand, cost and capacity that have been holding it back, but it seems that no one in the North American PV or electric utility industries has seen all the pieces yet, let alone put them together to make a picture

Recent NG demand growth is largely for electricity generation. From 1993 through 2003 nearly 300 GW of electrical generating capacity was installed in the USA, about 90% of which is NG fired, both to meet Clean Air Act requirements, and to add flexible capability to meet peak loads. Base load demand is estimated to grow at least 1.5%/yr (6 GW/yr), but seems to have shot up by at least 5% in 2004 vs. 2003. Therefore NG fired supply, intended for peaking, is being converted to base load supply, leaving a growing shortage of peaking capacity. Now declining NG supply means that peaking demand growth can no longer be met by adding new NG fired capacity. However peak demand coincides with peak insolation making PV an attractive alternative.

So, we have demand, at least if the cost is not too high. Can needed costs be met, and can there be adequate supply? In Renewable Energy World, Dec 2002, Auliche and Schulze (A&S) estimated worldwide polysilicon feedstock capacity for electronic grade (EG) silicon at 26,000 metric tons (MT)/yr, with production estimated at 14,000 MT/yr. With such a large excess capacity, poly suppliers have been happy to sell EG silicon for PV production at very attractive prices ($20.00-25.00/kg), enabling PV producers to lower their prices. As Maycock noted in Solar Today, Jan/Feb 2004, PV producers have sold cells and modules at cost, enabling very rapid industry growth in 2001-2003. System quotes as low as $4.00/Wp installed have been mentioned. Total world Si PV production in 2003 was about 0.7 GWp, having grown 32% worldwide while actually shrinking in the USA.

A&S estimated that about 2000 MT each of "off spec" and "non-prime" EG Si were supplied to the PV industry in 2000. At 17 MT/MWp that was enough to produce 235 MWp in 2000. Maycock shows 2000 production at 288 MWp, which implies another 1000 MT from capacity dedicated specifically for PV. With perhaps 8000 MT excess capacity in 2000 suppliers have had no incentive to add capacity. However, production of >700MWp in 2003 has surely consumed the excess capacity, even if price may not yet have been attractive for the poly producers. In parallel, while technology is reducing the share of off spec and non-prime Si being produced, microelectronics demand for Si is growing rapidly. As a result, in the last 12 months the price of poly has gone from $20.00-25.00/kg to >$30./kg and is projected to go to $40.00-60.00/kg. These price increases push bottom prices for PV installations back to the range of $6.00-7.00/Wp.

While there may still have been some stockpiles from prior years to work off in 2004, it is probably safe to say that PV growth will now be limited by poly capacity and price. To aggravate the situation, during the 2000-2003 period, poly producers experienced very low ROI, making it difficult now to attract the large increments of capital needed for conventional “Siemens process” poly production capacity. Unless there are dramatic technical advances, this condition is likely to persist for several years. John Schumacher has pointed out (Solar Today, Jan/Feb 2004), that breakthroughs are needed in both poly capacity capital and production costs, and in ways to get more collector surface per ton of poly. Fortunately, it seems that the technology now exists to meet both needs, and the only delay factor is time to recognition and industrialization.

Schumacher7 has already operated a "proof of concept" facility for a new poly process that has a capital cost about 40% of that for the Siemens process and projected product price of <$15.00/kg. Existing, possibly surplus, CZ pullers can be adapted to use the output of this new process with a probable increase in throughput at lower energy input, further lowering the cost of PV wafers.

In Dec. 2003, Origin Energy of Australia8, in conjunction with the Australian National University (ANU) announced a new "sliver cell"6 approach to making PV cells from Si wafers that is a classic example of "lateral thinking". Origin claims a 12x increase in collector surface per ton of silicon, and a 30x potential increase in Wp/wafer. My calculations do not confirm these claims, but taking all yield factors into account, they can probably get to >6x increase in collector surface/ton, which is still a sufficient breakthrough.

In 2004 ANU delivered a paper6 on sliver cells in concentrator applications, showing a 21% conversion efficiency at 20 suns. The Fraunhofer Institute has also worked with very thin silicon for PV and show 24% efficiency at 60 suns. Even at 20 suns and 6x yield/ton, poly scarcity ceases to be a restraint.

In writing a National Energy Policy “primer” for the House and Senate Energy Committees in 2001 (which regrettably, but not surprisingly, they totally ignored), I estimated that we would need the output of 50 large factories for 20 years to install enough collector surface at 20 suns to produce 10 quads of PV solar energy per year. The sliver cell will enable 5 quads in 20 years with only 4 factories. What seemed quite impractical in 2001, now appears quite feasible.

Sliver Cell Whole System Pluses
. ANU notes that the cells can readily be connected in series, reducing the need for protective diodes and eliminating the transformer from the inverter. In addition to lowering system cost, these changes would also improve conversion efficiency to a-c significantly, thus reducing the needed collector area for a given Wp. Taking all of these factors into account (Schumacher's poly + 6x surface increase/MT + elimination of diodes and transformer + light weight deriving from thin slivers + system efficiency) it seems likely that PV could get to an installed cost of $1.50/Wp before 2010. (ANU has estimated $1.80/Wp, but it’s not clear that they took all factors into account).

In a concentrator system, when used for peaking power in conjunction with a CCGT, the concentrator could also preheat water for the steam turbine stage, potentially increasing CCGT output by at least 3%, at no additional cost. If a 500 MW CCGT installation needed 100 MW for peaking, the extra 15 MW of thermal energy would lower the total investment per effective Wp to about $1.30. With regulated utility type financing (cost of money 3% above inflation) the resulting peaking electricity could be provided at a cost near 13 cents/kWh. Historic PV electricity cost estimates have typically been quoted (see the Wall Street Journal Special Report Sept. 2001) as 22 to 40 cents per kWh.

The average retail price of electricity in the USA in 2002 was 7 cents/kWh, and is surely higher now. Peak electricity price can be at least 3x, making conventional PV historically uncompetitive. (In some Calif. districts, base rates are 12 cents/kWh and conventional PV is marginally competitive for peak power now). At a base cost of 13 cents/kWh, even after markup for maintenance and OH, PV would be attractive for peaking supply across the southern tier. This base cost leaves room for attractive profit margins for everyone. I would expect NG fired power suppliers to start pushing very hard to have these technologies industrialized as rapidly as possible.


Solar thermal energy for hot water has long been attractive, and recent developments now make it attractive for air conditioning as well. Widespread use could reduce electricity demand in the USA by at least 10%, and this degree of reduction will probably become necessary as NG supply declines.

CSP for electricity production begins to look attractive with rising cost of fossil fuels and very long permitting and construction times for nuclear. The technology is now well-understood and poised for rapid development with corresponding cost reductions. We now need an intelligent National Energy Policy (NEP), with relatively modest subsidies to kick-start the needed development. We can be very confident of successful exploitation.

A major breakthrough in PV technology has now raised the potential of PV to the level of practicality. Production capacity is still a limiting factor. Lack of awareness is also a barrier. Again, an intelligent NEP is the key to further progress.

Reliance on imported fossil fuel energy, with its attendant cost, security risk and negative payments balance could realistically be overcome in less than 20 years, with a government driven “Apollo Program” for energy, focused on efficiency, conservation, renewables and nuclear. Renewable solar energy is now positioned to make its contribution.


1) http://www.energylan.sandia.gov/sunlab

2) http://www.eere.energy.gov/solar/pdfs/3solar_henryprice.pdf

3) http://www.energylan.sandia.gov/sunlab/PDFs/Assessment.pdf

4) http://www.sbp.de/de/html/projects/solar/aufwind/pages_auf/principl.htm

5) http://www.visionengineer.com/env/solar_flue2.shtml

6) http://solar.anu.edu.au/pages/publications2004.html

7) http://jcschumacher.com/

8) http://www.originenergy.com.au/news/news_detail.php?newsid=233&pageid=82

Readers Comments

Date Comment
Roger Arnold
Great article, Murray, with lots of specific numbers. That's always nice. Good references too.

I'm a little dubious about some of the statements about modules based on the sliver cell technology. While I certainly agree that it's a very promising technology for delivering much higher PV surface area per wafer, I don't understand why it should be any different from conventional PV in terms of module voltage or need for diode protection from shading. Conventional solar cells can be and normally are connected in series. Modules normally deliver a much higher voltage than the individual cells. Actually, I think they're connected in series-parallel networks that minimize shading losses. Isn't it the series connection that gives rise to the need for bypass diodes in the first place?

I also doubt that a higher module voltage would make much difference in the cost or efficiency of the inverter. Inverters are dominated by the cost of power switching silicon. Inductors and capacitors are comparatively minor components. (At least, that's my impression. I'm certainly no expert in that area.) It's true, though, that power switches are most efficient when switching a few hundred volts--the "sweet spot" for IGBT operation. And a few hundred volts is higher than conventional modules are wired to produce. But 48 volts is common, and I've never heard anybody complain that inverters driven from 48 volts lose much efficiency from too low an input voltage. They don't use transformers, BTW, to get 110 V on the AC outlet. I'm pretty sure that all modern high power inverters are formally switching DC - DC power supplies in buck-boost configurations. The "DC" output level is simply varied from +110V to -110V at 60 cycles per second.

Finally, I'm very doubtful of the suggestion that waste heat from concentrator modules could be used to pre-heat boiler water and improve the efficiency of a CCGT's steam turbine by 3%. High temperatures at the silicon junction badly degrade the performance of PV cells. For maximum efficiency from the PV system, you want the junctions cooled as closely as possible to ambient.

Though here's a thought: maybe one could use photonic array methods to give the concentrator two distinct focal distances: one for those near IR wavelengths most efficiently converted by silicon PV junctions, and another for all other wavelengths. Then it might be possible to have both a cool PV junction and a solar-thermal heat source. Wonder if that would work ...

One thing we most definitely agree on, however, is that the U.S. PV industry is being short-sighted in sticking with conventional silicon PV cells built from off-spec EG silicon. That's a recipe for no progress. Are they taking lessons from NASA?

Len Gould

Murray: Another excellent article. I hadn't heard of sliver solar before. Surprising how similar it is to spheral solar's approach http://www.spheralsolar.com/2_spheral-technology/2_manufacturing-process.asp (Subsidiary of ATS Automation) They've just opened a new manufacturing facility in Cambridge, Ontario. Has been very hard to get performance data on their products, though. If performance is anywhere reasonable, however, I'd guess their automation capabilities and the greater ease of automating the process should give them an edge. Interesting to follow.

If a person had to bet on which generation system will be cheapest (capital / peak kw at useful time) 25 years from now, it looks very risky to bet against solar PV with it's recently sustained record of 18%/yr cost reductions.

Todd McKissick

Murray: Nice article. I consider myself somewhat knowledgeable on this subject and still gained a substantial new understanding in some areas. Thanks for putting in the effort. I have some comments however.... Given CSP prices going from 10-12.6 cents/kWh now to 3.5-5.5 cents/kWh by 2020 and PV prices going from 22-40 cents/kW in 2001 to 13 cents/kW sometime(?)... Given that PV is very sensitive to Si availability and production limitations... Given that major technological advances in PV still only result in minor efficiency or production gains when CSP grows from a current 20+%... why is PV constantly over-promoted while CSP is only admitted that it "can" be viable? I notice that each of the 4 types of solar thermal received one paragraph of detail on them while PV development received 12. Research was quoted in numerous PV processes, giving details of advancements both past and future. Material availability, justifications for both cost and capacity, and many other details were mentioned. I must say that given the quality of your past articles and the level of detail given here for PV, I'm very disappointed in your investigation of the other solar options. In my view, when apples are compared to apples, solar thermal beats all other options in short and long term as well as installed and running costs. (I can't remember where I saw it, but Solar II was supposedly cost competitive a decade ago) I am not totally surprised by your assessment since you are simply following suit of every other organization on the net. This, I believe, is the exact problem. Solar thermal research seems to be the stepchild of the US government. It is the only RE source that has no private sector research going on. Basically, the government has deemed it only viable on a very large scale and everyone believes them. This has eliminated all grants in this area as well as scared off the private investment community. This has in turn stagnated the private sector development making it a self fulfilling prophecy. I would have liked to see you trade some print for this issue from maybe the issue of whether or not a new PV process can eliminate its diodes. Possibly the issue of using PVs excess heat to preheat boiler water could have been left out as well since that is already done by the exhaust of the boiler to get boilers' efficiency where it is today. It also seems unfair to justify PVs costs by pitting them directly against today's (over) 7 cents/kWh rate and its sometimes 12 cents/kWh rate, when CSP is only stated as becoming economic below 7-8 cents/kWh. Lastly, the potentially strongest point for CSP, which was only glazed over, is the fact that CSP is not only capable of 100% capacity with correctly implemented thermal storage but also 100% dispatchable. CSP also has the added capability of easily supplementing that thermal storage with bio/fossil fuels for emergency/critical backup. Make no mistake, I fully support the PV effort and applaud your diligence on that topic. However, in light of these issues, I would like to see comments on the limited focus of both private and public on solar thermal technologies. I have to wonder where it would be if it had one-tenth the research budget that PV gets.

Murray Duffin

Roger - If you quantify your doubts, I would welcome the resulting analysis. You'll probably end up with a number between 1.3 and 1.8 $/Wp and still have an attractive result. Len - I have also followed Spheral, but do not find technical analyses, and can't see how it can come close to sliver cells for reduced Si. Sliver cells can be 60 microns thick. I don't believe there is any automation technology to handle 60 micron Dia. spheres. Todd - I put more words on PV because this new breakthrough really changes the picture. I agree with you that thermal CSP has a larger role to play, but that is already well accepted, especially for base load. The real PV peaking potential needs to be understood also. Have you considered providing a paper on thermal? Murray

Len Gould
Murray: While you may be right about Spheral Solar's Si usage, there are other factors. According to a Netherlands study, http://www.chem.uu.nl/nws/www/publica/e2004-103.pdf, conventional crystal solar PV uses 15 g SI/wP, Spheral uses 9 g and Sliver uses 1.5 g. However the SI used by Spheral may actually be the beads produced by the fluidized bed system, which avoids a lot of expensive steps to full ingot. It is not clear with Sliver what their input source is or how expensive their manufacturing processes are, but it looks tricky, whereas spheral is already in demo volume production with a process they've been perfecting for 5 to 6 years now.

Kevin hagen

Thanks for your article and research it's a good view into the future of solar.

While the long term benefits of solar are interesting at the utility scale, I would argue that there are many more benefits for PV as a distributed generation technology. These include lower installed costs as part of other structures (e.g. Building integrated PV), reduced or eliminated transmission losses, and many benefits around user control. It’s also interesting that end users are willing to capitalize the generation investments.

As something of an industry outsider it is interesting to me that the utility world is very comfortable with the notation that countless tiny loads aggregate into huge demand, but somehow we discount the idea that small sources might aggregate into part of the solution.

As an example of present market status and perhaps market trends to come, consider peak power usage in southern California.

Currently peak retail power cost in many areas is 32 cents per KW/Hr. PV is routinely being installed in those areas at $7/Watt (before rebates, incentives, buy-downs, etc.). In the simplest calculation, with no incentives, this cost produces power over a 25 year life at 19.7 cents/KWHr for a 12 cent savings. (see footnotes for the math)

Before I get 100’s of posts about the long payback time or oversimplifying the math in this calculation, let me emphasize that the point here is to think retail, your customers do.

I work with corporate and institutional clients all the time who start with this basis and then add additional benefits such as hedging electricity costs and then turn the marketing dept loose to find some “soft stuff” and come away with the conclusion that solar PV is a good deal now.

Kevin Hagen Shuksan Energy www.shuksanenergy.com

Footnote on the math

1 KW DC nameplate of reasonably well installed PV delivers about .75 KW AC (figure can vary based on installation, equipment, etc from .7 to .8 and is based on experience and supported by work at Sandia labs and others for grid connected batteryless systems in the 5 to 50 KW class). Southern CA insolation taken at 5.2 solar hours. .75KW x 5.2 Hrs = 3.9 KWHrs/day. 25 year system life yields 35,587 KWHrs @ $7,000 = 19.7 cents/KWHr

Of course a real payback analysis is more sophisticated and would add cost of money, minor M&O expense, etc. However, it would also consider direct and indirect incentive programs, future electric cost increases and other factors.

Murray Duffin
Len - Thanks a million for the Alsema reference. I had not seen it before. He raises a couple of red herrings unfortunately. By 1999 my employer had phased out SF6 and reduced PFC use by about 80% with hopes of eliminating them also. We had also converted most production to lead free solder, and are probably past 100% conversion by now. As for health risks, the entire microelectronics industry has been using these chemicals in high volume production for years and years with zero health issues. Sliver cells not only have 1/6th the Si of spheres, but will undoubtedly have higher conversion efficiency. Murray
Chuck Simmons
Kevin --

Nice discussion of the fact that Solar competes with retail costs, not wholesale.

Murray --

I greatly enjoy your articles. I've been thinking about the extent to which polysilicon feedstock supply and prices will impact the rate of growth of the Solar PV industry. I have basically three comments:

1) The rule of thumb for the industry appears to be that polysilicon feedstock usage per watt drops about 10% per year. Thus, if 17MT/MW were used in 2000, we would currently use about 11MT/MW. (Photon magazine reported 13 MT/MW in 2003 which is inline with this estimate. [Google: photon 13 tons]).

2) Solar module and solar cell manufacturers are rapidly building new factories. It appears that in 2005, manufacturing capacities of cells and modules will be double what they are in 2004. It seems unlikely that so many manufacturers would expand production if they were not confident of increased supplies. I also notice that Elkem is substantially increasing its ability to manufacture silicon metal, implying that they expect strong growth in demand; and I speculate that the demand is coming from polysilicon producers.

3) There is evidence that polysilicon producers are expanding capacity. USGS estimates 2004 production at 25,000MT. Wacker has been expanding production and has plans to continue expanding production. SGS (Solar Grade Silicon) recently built 2400MT/year of capacity in Moses Lake, Washington. Tokuyama is addressing bottlenecks in their production lines to slightly increase capacity.

So, it looks like the industry, which has been expecting Solar PV to start driving demand for polysilicon for years, is prepared to handle the upcoming increases in demand.

David Katz
While this article reflects a wealth of information regarding solar I would like to add that Daylighting, both active and passive should also be considered solar energy. Why take the sun and convert it to electricity with low efficiencies and then bring it to a ballast that turns on a light bulb, when you can capture that sunlight in innovative ways and dim or turn off the adjacent light fixtures in real time, usually value peak time, by harvesting the daylight. Check out www.daylighting.com

Murray Duffin
Chuck, thanks for the input. PV would still be limited by poly capacity and price, but you are suggesting that the capacity will not severely limit growth and price might not rise as fast as I suggested. That is good news. Other - I just came across http://www.nrel.gov/docs/gen/fy04/36831d.pdf. Page 15 says that at $2.00/Wp installed for solar, electricity costs only 8 cents/kWh, so I must have done my calculation way wrong. At $1.80 /Wp system cost and perhaps $2.00 installed and 8 cents/kWh sliver cells are much more attractive than I thought, and Roger's caveats become unimportant.

Chuck Simmons
I did a bit more reading over the past 24 hours. The main method by which companies appear to be dealing with the poly-squeeze is vertical integration. Renewable Energy Corp (REC) owns a big chunk of Solar Grade Silicon as well as ScanWafer. This suggests that ScanWafer will pay a bit more than they want, and SGS will get a bit less than they want, but overall, REC will make money. SolarWorld and Wacker each appear to have similar set ups.

It sounds like Wacker and SGS think that development of slightly cheaper polysilicon manufacturing techniques using Fluidized Bed Reactors are on schedule for volume production in 2006 and 2007.

Meanwhile, the poly producers must be pretty happy that their factories are starting to run close to capacity and that they are making money again. The increased volumes will lower their costs and raise their profits.

It does sound like cell manufacturers will need to move to thinner wafers faster than they might have hoped, but development of equipment to handle 100 micron wafers has been ongoing for a few years.

Chuck Simmons
Just because I like to play with math, and nitpick... My apologies for being anal in advance. I want to walk through some of the math in terms of polysilicon production and capacity, starting with the paragraph "A&S estimated that about 2000 MT ...".

I think I found a copy of the referenced article at "http://www.jxj.com/magsandj/rew/2002_06/silicon_supply.html"

The article suggests that the capacity in 2001 (not 2000) was 26,000 MT, of which 14,000 MT were used, and 12,000 MT unused (not 8,000).

The article does then jump back to the year 2000 and state that 4,000MT were used to produce 235MW of crystalline solar cells. "www.nrel.gov/ncpv_prm/pdfs/papers/57.pdf" suggests that the remaining MW to reach Maycock's 288MW figure were non-crystalline amorphous or thin-film silicon and hence didn't consume an additional 1,000MT of poly. (The PDF uses a figure of 258MW of crystalline cells. Since Maycock estimates the 2001 chrystalline market share at 82%, the 235MW figure seems more consistent and the 258MW figure looks like a typo.)

In 2003, it appears that 740ish MW of PV were produced and that about 89% or 660MW were crystalline for a total solar PV consumption of around 9000MT of poly. [see "http://www.jxj.com/magsandj/rew/2004_04/pv_market_update.html" or "http://www.photon-magazine.com/akademie/"]. The USGS estimated overall 2003 polysilicon production at 23,100MT ["minerals.usgs.gov/minerals/ pubs/commodity/silicon/silicmyb03.pdf"]. So, there was probably a little bit of excess capacity (3,000MT) in 2003.

Going way out on a limb... for 2004, solar companies are reporting 80% increases in revenue, suggesting that 1300MW of PV are being produced this year. Assuming 90% crystalline and 11MT/MW, that's around 13,000MT. The semiconductor industry consumed 14,000MT in 2003, and probably about 16,000MT in 2004. We can find evidence for the following capacity additions in 2003/2004: Tokuyama added 400MT, Wacker added 800MT, SGS added 2000MT (solar only). These numbers (purely by coincidence:-) bring both estimated capacity and consumption to 29,000MT for 2004.

Of course, none of the minor corrections or additional estimates suggest any changes to Murray's basic conclusion: either someone needs to produce a lot more poly in 2005, or the Solar industry needs to use a lot less silicon in 2005.

Murray Duffin
Chuck - I do like people that quantify things, and you found some sources I didn't, which is all to the good. I'm glad my tentative conclusion seems to hold up. Now that poly and crystaline Si people have to start producing specifically for the PV market we will probably see some real progress on PV grade cost and Ein reduction. Murray

Todd McKissick

Can anyone show me how PV is a better choice than CSP? I'm looking for an apples to apples comparison that pits up-front dollars per annual kWh output. I'm not interested in KW since flat plate and concentrator systems have different daily sunlight amounts. I'm also not interested in how the loan amitorizes out for one when it's never the same rates used for the other. Loans are loans and they depend on the buyer's credit more than any other factor.

I would also be interested in captured area to grid efficiency to determine land use. The write-up on Sliver cells quoted 18% but then added another 75% factor for the ancillary drops. That drops it down to 13.5% overall efficiency which is the number that should be compared to the 23% for Power Towers. I remember Popular Science promising 30 years ago that we're ready to make the move from 18% to 31% in PV efficiency due to some "see through" cells breakthrough. Still waiting.

What about capacity? You quote that Power Towers can probably reach 70%. That was simply the design goal of only the second system ever built. Any capacity can be attained given a goal of that number before the plant is built. This also brings up storage. With no possibility of storage, PV (and wind) cannot be dispatchable. This is not the case for CSP. It lends itself easily to base load as well as peaking.

Then there's O&M. I think the insurance on PV systems has been missed as an operating expense. Certainly 10 acres of stowable mirrors would be cheaper to insure than 10 acres of expensive fixed mounted PV cells.

With all this talk about the supply of Si for mfg of PV systems, how about comparing the energy payback of these two also.

Murray: As to your comment about CSP not needing any words because "that is already well accepted, especially for base load", I completely disagree. The solar thermal direction is only being traversed by the government labs and very little private sector development is being done. This is the sole reason there have been no substantial developments lately. The DOE's '05 budget finally recognized it and only added $3M for further study. Let's see the PV industry get along on that money. How about wind? Both of these industries are to the point that with the renewable subsidy recently passed, they can be self sufficient. I think that public awareness needs to change and fast, toward solar thermal systems in order to ensure that they are not overlooked when the other two become 'accepted' as the only ones to make the cut.

PS: I'm not sure why all my earlier paragraphs got ran together, but there were supposed to be breaks in there. Sorry for the hard reading.

Copyright 2002-2004, CyberTech, Inc. - All rights reserved.