High Volume Photovoltaic Cell Costs depend on production
technique. Large cost reductions are
possible within physical laws.
A
cost breakdown and suggestion for reducing costs.
M. Robert Showalter
January
5, 2004
A manageable
task is one in which the expected results can be easily identified;
success, failure, or completion of the task can be easily ascertained; the time
to complete the task can be easily estimated; and the resource requirements of
the task can be easily determined.
from "A Professional's Guide to
Systems Analysis", Martin E. Modell, 2nd. Ed. McGraw Hill,
1996.
The
task of removing energy as a fundamental constraint on human welfare isnt manageable
without much more specification. But sub-tasks that could contribute to that objective are
manageable tasks.
Defining the high volume mass production cost
of a basic Si photocell on the basis of clear assumptions and how that cost depends on component
costs is a manageable task.
Here is the basic structure of a generic
silicon photovoltaic cell
Suppose we assume that 20% efficiency, now available in the best low volume production silicon photocells, is achievable in high volume production. Let high volume be defined as a billion square meters per unit time (day, month, year).
(To match fossil fuel energy with PV would take about 100 billion square meters of 20% efficient PV material about 20,000 gigawatts of capacity. At a billion square meters/year, that would take a century to produce at a billion square meters/month 8.3 years at a billion square meters/week 2 years. )
It would be a manageable task for glass and automotive engineers to define the cost of assembling layers A, B, F, and an additional glass sealing layer below F at high volume. With ordinary high quality production engineering, that cost, for a 2 mm thickness assembly, would probably be around 1-2 $/square meter ( .5 - 1 cents/watt for a 20% efficient photocell. )
It would be a manageable task for glass, automotive, wire, and textile engineers to design a contact grid layer C with total losses (shadow losses plus N-Si layer conduction losses) under 2% for a high volume cost under .5 $/square meter (under .25 cents/watt for a 20% efficient photocell. )
If the N-type and P-type Si
layers are a total of 10 microns thick, silicon cost per meter squared will be
the cost of 23.3 grams of purified
silicon. ( .117 grams/watt ). At the price of metallurgical Si, that
would be about 2.5 cents/meter squared ( about .0125 cents/watt ). Silicon purification costs, and
shaping costs, are now many tens of dollars per square meter. The key to reducing photocell costs
is reducing these costs.
- - - - - - - - - - - - -
Summarizing assembly costs in high volume
Layers A, B, C, F - 1-1.5 $/meter squared - .5-.75 cents/watt.
Cost of Si, at metallurgical rates 2.5 cents/meter squared - less than .02 cents/watt.
PLUS the costs of purifying, doping, and shaping the Silicon which are now the dominant costs.
The electronics industry depends on the mass production of ultra-pure,
enormously uniform, very large silicon crystals. The science involved with silicon crystal formation has been
well understood for at least seventy years, and production systems for making
ultra-pure crystalline silicon have been evolving for half a century. Currently, the microphysical conditions
for producing very pure crystalline silicon occur routinely and reliably in
production systems with high unit costs.
These costs are relatively small parts of the total cost of the
semiconductors for which the silicon is made.
It is a technical
question whether these same microphysical conditions can be produced
in systems adapted for high volume mass production of crystalline silicon in
the thin sheet form that photocell production needs. The costs associated with
that high volume mass production will depend on technical details that are
partly dictated by the physical facts of silicon crystallization but mostly
depend on other technical facts.
Purification of Silicon to very
high purities is done by crystallization.
Impurities are left behind in the melt. Impurities are discarded in the melt. Impurity reductions of 1000 fold and more per crystallization
stage are standard. Si purification
to semiconductor purity could be accomplished in 3-5 recrystallization
stages. 90% or more of the silicon
input feed can be in the high purity crystalline output.
Processing the silicon has energy costs but these
energy costs are small the energy cost of 10 stages of
melting and recrystallization of Si (at 20 cents/kWh energy cost) would be an
additional 2.5 cents/meter squared of photocell less than .02 cents/watt . There is no scientific or
engineering reason that these recrystallization stages cannot be made to
occur in a steady-state, steady flow
automated process with energy costs not much higher than theoretical energy
costs.
Processing the silicon into the form needed for
photocells has other costs associated with producing the microphysical
conditions needed to produce the thin, doped crystalline layers of silicon
that high efficiency photocell production needs. These costs depend on the production system used. Different production systems can produce
identical products at very different costs.
I believe that following production engineering questions will
determine these costs in high volume production:
1. Is it possible to produce a steady state, steady flow multistage recrystallization process for silicon with good cleanliness, very well controlled heat transfer and precisely controlled material flows so that silicon of semiconductor purity ( impurities of parts per billion or less) can be produced from metallurgical silicon at an energy cost not much higher than the theoretical energy cost ? Or a batch recrystallization process meeting the same criteria?
2. Is it possible to pull thin, planar crystalline
silicon ( the geometry needed for
photocells ) directly off of a pool of liquid silicon on an automated
high volume basis? It seems certain that if this be done with pure silicon, it can
also be done for N-doped silicon, and P-doped silicon.
( Alternatively, is it possible to pull the planar crystals on another
automated basis, for example, by a modified Czochralski process, or by an
extrusion and recrysallization process? )
3. Can jobs 1 and 2 be done, and done at high
throughputs with low capital and operating charges per unit output? These
processes would have to operate just below the temperatures at which iron is
poured and cast, but with microphysical and microchemical conditions controlled
with much more precision than that typical of cast iron foundry practice.
I believe that the answers to questions 1-3 are all affirmative and
the jobs can be done well enough so that the high volume mass production
capital charges involved would be less than a cent per watt ( less than $2.00
per square meter.)
If these tasks can be done and done at that price it is reasonable
to estimate that high volume photocell production costs could be less than
$3.50/meter squared less than 1.75 cents/watt. At that low price, photovoltaic energy might compete strongly
with fossil fuel energy on a wholesale basis.
Answering questions 1-3 to perfect reliability would require actually building units that actually produced at full high volume production and testing them over time. That would cost many billions of dollars. But a great deal of the uncertainty involved can be addressed by design calculation and much more could be addressed by building prototypes. Feasibility of tasks 1-2 could be demonstrated by actually prototyping and testing units that did jobs 1 and 2 above. Full scale costs could be estimated reasonably well after this prototyping was done.
The task of removing energy as a fundamental constraint on
human welfare isnt manageable.
That task is too broad and too ill defined. But the tasks involved in design and prototyping to answer
the key questions about jobs 1 and 2 above are manageable tasks. Expected
results of these tasks can be clearly identified; success, failure, or
completion of the tasks can be clearly ascertained; the time to complete the
tasks can be estimated; and the resource requirements of the task can be
determined.
I believe that these design and prototyping jobs could be
completed or great progress could be made toward their accomplishment for
less than a million dollars.
Note: The subject matter discussed here involves matters that will be the subject of several patent applications by me.