Success in solar manufacturing: Challenges and opportunities

Spire calls itself the ‘turnkey solar factory company’, making machines to manufacture PV modules. Spire has been in the solar business for 30 years, and has believed in this business for a long time. We are based in Massachusetts, USA, with 250 employees and have helped our customers install about 50 turnkey factories all over the world. We’ve also supplied manufacturing equipment to most of the major PV module manufacturers as well as these startups. These days, we’re seeing a huge amount of interest from startups principally in places like India, China, Korea, the Pacific Rim and Europe – everyone wants to join the solar goldrush and is coming in aggressively and developing their own markets! We believe in the principle of local manufacturing – this stimulates local markets, and then the market worldwide

Recent market trends

PV is a dynamic market. The composite annual growth rate of PV systems in the world has been exceptional, at 40%–45% for a number of years, but only now is it becoming large enough to be classified as a major industry.

Back in the early1980s I remember writing to major corporations in the United States saying they should be in the photovoltaic manufacturing business – the worldwide market had gone to almost a megawatt and prices were down to US$8/W – here we are today, looking at something approaching 3 GW and prices in the region of $3/W.

The PV world is dominated by crystalline silicon production – including cast and ribbon silicon. Thin films are coming on, with a lot of manufacturing capacity for thin films added in the recent past, but still we find that manufacturing is dominated – at a level of 90% of the world market - by crystalline silicon.

We don’t see that changing radically. We see that crystalline silicon will continue to account for the bulk of solar manufacturing in the world for the next few years at least.

For a while, shortages of polycrystalline silicon ‘feedstock’ were holding back the industry. During this period, photovoltaics only used 5%–10% of world production of polysilicon. As time went on, solar grew to take up almost half of the world’s polysilicon production, and a shortfall in its supply began to limit the growth of the crystalline silicon manufacturers. During that time thin film technologies started to get a bigger toehold. But because of the shortfall, a tremendous amount of new capacity was announced and is now coming on line. Many major manufacturers have looked at this opportunity and concluded they want to be in the solar business – many with significant resources. They have found that polysilicon plants are a good entrance point into the industry. So while the existing major players such as Hemlock, Wacker and MEMC are expanding, at the same time many more new entrants are trying to get into the business. Projected capacity is expected to be greatly expanded over 2005. But even in addition to the eight established companies shown in Table 2 there are probably another 20 that are planning on getting into the polysilicon production process.

Producing polysilicon is comparatively easy, depending on the process used. Using the ‘BB’ process, such as MEMC uses, is more difficult. (Silicon made by the ‘BB’ process is desirable for pulling silicon ribbon, as done by some manufacturers.) The current price for polysilicon is as much as $300–400/kg. But with this new capacity coming on-stream, long-term contracts are now being signed at prices nearer $60–70/kg – and we shall see this will start to have a dramatic effect on the cost of PV modules in the future.

Investments – a rule of thumb

PV manufacture takes place in several stages. What happens first is we have to make polysilicon – this requires a large-scale factory that requires about $250 million capital investment. In brief, silicon ingots are cut to make wafers, which are made into cells, then combined into modules.

The cost of setting up a PV manufacturing plant depends at which stage you enter. To set up a polysilicon manufacturing plant, the minimum feasible size requires an investment in the range of $250 million. For a wafer production plant the minimum feasible size is 50 MW of wafers (50 MW output per year) – that’s likely to cost between $30 million and $40 million. A solar cell factory can be smaller – maybe 20 MW/year. It will typically require investment in the range of $10 million. A module plant can be viable as small as 10 MW/year, with investment of a couple of million dollars. So there exists an economy of scale going backwards up the chain, and many people enter the business by going into module manufacture as this is the lowest-capital-investment entry point into the PV manufacturing industry.

The manufacture of thin film PV, such as amorphous silicon, CIGS or cadmium telluride, typically starts with a sheet of glass on which the PV film is deposited. The interconnects are made on the deposited material to create a ‘laminate’. Turning that laminate into a module involves many of the same steps as production of a crystalline silicon module. We are working with several thin film producers on this ‘back-end’ process equipment and development.

The production of concentrator cells (such as using gallium arsenide) is more complex and the details depend on the collector design itself.

There has been an explosion of demand for PV equipment. Table 3 shows numbers from Prometheus (February 2008). If you look at the numbers that are in the highlighted column, you’ll see that for 2008 they estimate that solar PV production capacity will be over 3 GW worldwide, but with about 700 MW of new capacity coming online. If you consider new equipment that will meet the needs of 2008 going into 2009 as well, it is as much as 800 MW – so in their estimate as much as 800 MW of new manufacturing capacity is likely to be coming online during this year.

Prometheus breaks down the cost of a factory (and we think it’s fairly accurate) as follows:

• wafer equipment runs at about $0.60/W

• cell equipment is about $0.40/W

• module equipment about $0.30/W.

Totalled up, this means that we are adding to the manufacturing base at a cost of about $1.30/W. When that price is combined with the 800 MW growth number we get a capital equipment market with a value of $1 billion/year. (This ignores of course any capital investment associated with polycrystalline silicon factories, which will overwhelm the other numbers in 2008.) It’s clearly a large, and growing, market.

Manufacturing equipment and the importance of yield

Here it’s interesting to look back briefly. When we first started in this business, in the 1970s, factories were being installed with a production capacity of 1 MW/year. By the 1980s a factory would typically be 5 MW. In 2005 that had grown to 50 MW, and now we are looking at factories made of ‘building blocks’ of 100 MW in 2010. Some people talk about adding gigawatts – these gigawatt factories will be made up of 100 MW building blocks.

‘Turnkey’ module factories started some years ago, with basic machines:

• Most start with a cell test machine. Incoming cells from a manufacturer get sorted – they need to be matched because if the performance of one cell is poor it ‘drags down’ the rest of the cells in a module. So they have to be matched. These machines run about five million cells a year. The industry is looking at increasing throughput by a factor of ten.

• String assembler – which will solder several million cells a year. This throughput has to be increased by about ten times as well. So the industry is working on higher throughput on all of these kinds of machines.

• Laminators – these need to produce about 300,000 or 400,000 modules a year.

With all this equipment we are talking about incredibly high throughputs. Test machines are less of a problem in terms of throughput – however, it’s important that they are rock solid in terms of spectral intensity and uniformity, and standards. These test machines have to be put into place and made larger.

In terms of future directions, the current trends in manufacturing are towards:

• Thinner wafers (This is currently important as a means of reducing polysilicon use – but result in lower yield. Our view is that the poly supply issues will soon be resolved so thinner wafers will become less of an issue – but right now is important to consider)

• Use of back-side contacts isimportant to some manufacturers. If you put all the contacts on the back, the cell can become more efficient as the contacts are not blocking the sunlight (typically 5% of the sunlight is blocked by contacts on the front). Typically back-contact cells require very high quality silicon and thinner wafers in order to collect the carriers generated by the sun.

All of these machines need to have very high throughput. They need to be robust and reliable, with low downtime, low maintenance, and high yield.

Manufacturing cost analysis

We do a lot of analysis of manufacturing cost. Figure 2 may seem a little adventurous. It shows the sensitivity of the PV module cost to the cost of the polysilicon. The vertical axis shows the cost per watt, while the horizontal axis shows the cost of polysilicon in dollars per kilogram.

As mentioned earlier, we don’t think the price of polysilicon will go as high as $400/kg – here we’ve chosen the figure of $70/kg for the polysilicon as that’s where we think it’s going to be in the next few years. Taking that price you find out that the manufacturing cost for a module is about $2.11/W. The wafer cost is then just over a dollar ($1.08)/W. This shows the importance of the cost of polysilicon – it will make a big difference to the cost of modules in the future.

The other big driver – which should not be underestimated – is that of cell efficiency. In Figure 3 we have module cost on the vertical access, and cell efficiency is on the horizontal axis – up to 23%. We’ve chosen a (realistic) 16% cell – typically what you can achieve with a polycrystalline cell or a Czochralski cell – to achieve a module cost of $2.11/W.

This shows that if you can drive up the efficiency further (via back-contact cells and the like) then you can see that the module cost can be driven down considerably – to less than $2/W.

The yield is also an important parameter – in the model for analysis in Figure 4 we chose 85% from wafer to module (you can do better than that, but at 85% you are approaching the cost of the module.

When all these considerations are put together you come up with the following analysis for a 50 MW/year production line (see Figure 5):

• the wafer costs $1.08/W

• changing the wafer into a cell costs $0.40–45

• changing the cell into a module costs another $0.40–45

What’s easy to see is that the costs are driven by materials – polysilicon makes up a huge cost of the module. That’s why, as the cost of polysilicon comes down, we can expect the prices of modules to fall and the market to expand much faster than it has in the past few years

Plant depreciation is not a big issue – 10% percent or so in these factories. So when we talk about a gigawatt of new production capacity – or a billion dollars of new production capacity – with a 7-year lifetime the depreciation curve on a module basis is less than 10% of cost of the module.

The next generation factory – on site

When we talk about next generation factories, one of the concepts that we are pursuing is the ‘building block’ factory, a 100 MW/year module assembly factory. Within this model (see Figure 6), you start with glass sheets that come in and are prepared. And with cells that come in and are ‘plugged in’ to assemblers. The assemblers solder the cells together and then robots are used to pick up the cells on the sheet of glass and bring them to a busing area. The interconnected array of cells on the sheet of glass goes through to a lamination station, made up of multiple laminators. At 100 MW/year each laminator is producing about 25 MW of modules. Each assembler is running at about 25 MW of strings per year. The lamination process involves a heating cycle, applying EVA/vinyl acetate on both sides of the cell, a Tedlar back sheet surface. The heating cycle is carefully designed to avoid stress that may occur between cells and the glass itself.

Laminated modules are then brought out on conveyors, picked up by robots and taken to a taping and framing section. From there they are put on a test belt which takes them through a high-potential test and a simulation test – and finally it’s a completed module.

In a thin film line you do not of course need the string assembly portion, but you have the equivalent to a busing portion. And typically in a thin film line you use a simulator to test the thin film deposit before you run it into lamination. So to make thin film modules is not so very different.

One of the directions we see for the future is for such a building block factory of 100 MW that stands alone, manufactures at that rate, and stimulates a regional market. We believe in regional manufacturing to stimulate regional markets, which then has a major effect on the growth of the industry worldwide.

But one deployment scheme that’s worth considering is to take one 100 MW building block factory – as just described – and using it to move modules more rapidly into the utility system. First of all we’re talking about a 1 kW module – that’s four times bigger than today’s typical module. It would be about 3 metres x 12 metres in size, which might normally be hard to handle. In this case, however, the factory needs to be on a utility property where the modules can be deployed – so the modules virtually go out of the back door and onto the site.

This brings significant savings – first the cost savings for the module itself, which because of its bigger dimensions needs less – less framing material, less busing material – there are considerable economies of scale

But as the module is deployed on site you get the real additional cost savings, as the factory is built adjacent to the site for a very large utility power array – a solar farm, if you will – a 100 MW ‘breeder line’ placed on the edge of this site. (This might be a semicontaminated ‘brownfield brightfield’.) Over time you can really begin to see significant cost savings.

When you look at the model closely, and at the savings associated with deploying the modules in this ‘brightfield’ mode, you can see savings in labour needed for module/system assembly in the field, cost savings for major power wiring, for balance of systems, power conditioners. In all we think that the kind of number the utility systems can achieve is $2.87/W. This would then have a price in cents/kWh that becomes to be very interesting, especially when the appropriate incentives are available.

Future market

What about the future? We believe strongly in the market, and believe it’s going to continue to grow. I think there has been a little concern that maybe it won’t grow as fast as it has been – but there are new emerging markets all the time. The US market still has to take off, China has to get going, the UAE keeps announcing new cities (and it doesn’t take more than one such city inthat part of the world to absorb a lot of production). India is growing more rapidly than it has in the past, Australia is thinking about it, South America has some good activity going on.

So we believe the market will continue to grow. Whether it will continue to grow at a rate of 45%/year I’m not so sure, but this technology is coming. It’s permeating everybody’s power plants in the future. For 2009 and beyond we believe that the issue of silicon feedstock supply will be solved, and the dam will break on polysilicon. That will result in lower-cost modules, and will stimulate market growth worldwide. Crystalline silicon manufacture will expand, but that doesn’t mean the thin film won’t too – however we think that crystalline silicon is ready to come back in a very aggressive way in the next few years, because of the lower-cost polysilicon.

The market by 2010

By 2010 we might see:

• 8–10 GW/year manufacturing capacity worldwide

• 6 major manufacturers, each producing 1 GW or more

• 30 other manufactures each producing 100 MW or less

• modules selling at a price of $2.90/W are certainly achievable – and that allows margins for the growing industry

• 80% of PV will continue to be single and multicrystalline

• 20% of it will be thin-film – and other things. (We still have to see what market share can be achieved for concentrator systems – there’s some way to go there, so we don’t know quite what will happen.)

Let me close with a phrase from one of our ads – ‘push the button and you’re in the business’.

Roger G. Little is Chairman and CEO of Spire Corporation.

See the full presentation and listen to the Q&A at www.renewableenergyworld.com (find ‘webcasts’ under the ‘events’ tab)

Figure 1. Market dynamics – actual and projected. Source: PJC Poly Si Supply & Demand Analysis

Figure 2. Polysilicon – cost sensitivity. Module cost = $2.11/W at $70/kg poly-Si cost. Source: PJC Poly Si Supply & Demand Analysis

Figure 3. Cell efficiency sensitivity. Module cost = $2.11/W at 16% cell efficiency. Source: PJC Poly Si Supply & Demand Analysis

Figure 4. Process yield sensitivity. Module cost = $2.11/W at 85% cumulative yield. Source: PJC Poly Si Supply & Demand Analysis

Figure 5. 50 MW/year production costs. Source: PJC Poly Si Supply & Demand Analysis

Figure 6. 100 MW module assembly factory