Friday, February 5, 2016

Semiconductor manufacturing - part 2, solar panels

 This section will discuss perhaps the simplest of semiconductor products, solar photovoltaic cells. I'm not going to dive too deeply into the physics, but rather focus on the mechanics of making these devices.

 First we will look at bandgap and why it matters. Then I'll start with thin films, cover single wafers then discuss multi-junction cells.

 I think the takeaway here is that semiconductor manufacturing uses dangerous chemicals, high-voltage devices, vacuum and heat. This is not an activity that should be done in a habitat. A manufacturing facility for these devices should be physically separated from the habitat and equipped with isolated life support systems. Even so, basic devices are within reach of simple equipment that does not rely on gravity to function. Some types of panels can be manufactured with equipment no more advanced than the refining equipment used elsewhere in space as a prerequisite.

 Details after the jump.
(part 1) (part 3) (part 4)

 Bandgap is a term for the energy required for an electron in the valence level of an atom to jump up to the conduction level. Only electrons in the outer shell (conduction level) of an atom are available for electrical interaction. Conductors have little to no energy gap, so many electrons can come and go easily by jumping between shells. Insulators have large gaps, so electrons cannot jump shells without a lot of encouragement. Semiconductors fall in between, though usually closer to conductors.
 Two forms of energy that can cause an electron to jump are heat (as phonons) and light (as photons). Solar PV cells use light energy to knock electrons around, then take advantage of the two semiconductor types to get those electrons to travel in a common direction as current. This is most efficient when the energy of a photon is just enough to knock an electron loose; not enough and the energy becomes heat, but the extra energy for a too-powerful photon also becomes heat.
 Each type of semiconductor has a specific bandgap. This is the lower limit of photon energy it can use. It can be tuned somewhat with dopants, or for alloy semiconductors by varying the amount of each material. For sunlight, the best single bandgap is about 1.3 electron-volts with a maximum theoretical efficiency of about 33%. Gallium arsenide at 1.43 eV and silicon at 1.11 eV are both fairly close to this and both make good PV material in the range of about 20-24% real-world efficiency.
 Improving efficiency beyond this limit is possible if you use multiple materials (multiple junctions, hence multijunction cells). The outermost layer has a high bandgap, absorbing UV and higher visible light. Intermediate layers (if any) collect lower and lower energies, while the base layer has a very low gap and collects infrared photons from the sun and from the other layers above it. As will be seen below, this is complex and expensive so most uses today are in spacecraft where the improved efficiency outweighs the added cost.


 Thin films are made as described in the previous post. For PV cells this is normally by plasma-enhanced CVD. First a substrate material is chosen, usually glass but sometimes plastic or steel, then cleaned and polished to electronic standards. If the substrate is not conductive then a layer of metal (molybdenum and aluminum are common) is plated or sputtered to serve as the back contact. If the substrate might diffuse contaminants into the semiconductor then there may be a barrier layer of cobalt, ruthenium, tantalum, tungsten nitride or titanium nitride. This layer should be as reflective as possible so that any light that is not absorbed by the cell will travel back through it for a second chance at absorption. Next the absorber (the active p-type semiconductor substance) is deposited. Next is the front contact, made of a transparent conducting oxide like ITO (tin-doped indium oxide) or AZO (aluminum zinc oxide). This layer is normally added with magnetron sputtering, a form of PVD, and is n-type. The interface between this conductive layer and the semiconductor beneath it form a junction; in some designs the underlying semiconductor is doped to n-type at the surface rather than relying on the conductor to form the junction. Above this, a network of very fine aluminum wires may be applied to collect current from the transparent conductor layer. If the cells are narrow enough this is not necessary, but large-area films may require these wires to reduce resistance across the cell. The final coat is an anti-reflective layer of silicon oxide or titanium oxide, though there are a variety of other options.
 This process requires a vacuum reaction chamber for the pe-CVD and magnetron sputtering. Vacuum pumps, metered gas dispensers, a DC or RF plasma discharge device, infrared heating coil. sputter targets and a magnetron complete the production hardware. Test equipment is needed to check the thickness and electrical properties of films, usually a spectroscopic reflectometer and/or an xray imager plus a multimeter and leads. Most operations can be automated, but the equipment must be thoroughly cleaned periodically to remove buildup of deposited materials and reduce contamination.

 All of that may sound complex and intimidating, but hobbyists have made thin film silicon on glass solar cells using scavenged equipment. This is a little rougher than high school science fair projects simply because of the high-voltage equipment, but the process is quite tolerant of variation and defects. Efficiencies improve if the manufacturing process is more precise. With modification, some of the refining equipment used at a mining base could be used to perform most of these steps if the substrate is heat-resistant


 Wafers are prepared as described in the previous post. Construction is similar to thin films in many ways. The base layer is p-type, with the surface made n-type by applying a dopant (ion implantation or CVD/PVD followed by diffusion) or depositing a layer of n-type material (typically CVD). Some cells use a conductive oxide layer while others use fine wire networks to collect charges. An antireflective coating is applied to the top. The back side of the wafer gets a conductive metal layer. Once complete and tested, wafers are cut into individual cells with a diamond saw, cleaned, then mounted to a structural panel. Cells are wired together to form a full panel. A sheet of glass covers the front surfaces and protects the cells from damage (including radiation).
 This process also requires a vacuum reaction chamber with associated pumps, gas dispensers and CVD equipment, and may require an ion gun for implantation. The entire process must be scrupulously clean and will require a cleanroom and dangerous chemicals (sulphuric, hydrochloric and hydrofluoric acids, 30% peroxide, sodium hydroxide and ammonium hydroxide). A diamond saw and polishing table are also required.

 Wafer processing is actually complex and dangerous. Of particular concern are the highly-corrosive cleaning compounds used to prepare surfaces. Monosilicon cells and gallium-arsenide cells produced this way are superior to thin films, but not by enough to justify the added processing on their own. Facilities to manufacture cells like this will be general purpose and used for making other semiconductor devices like LEDs and microprocessors.


 Lastly, multijunction cells. In simplest form this can be adding a second layer of semiconductor to an existing process, like adding microcrystalline silicon to an amorphous-silicon thin film. This is called a tandem cell, and can even be produced by layering and wiring two physically separate cells together. The high-end monolithically integrated cells are wafer-based and have a complex series of 18 or more layers that create three or four junctions. These layers are deposited by metal-organic CVD as epitaxial growth (meaning the layers have the same crystal structure and orientation as the substrate). For use on spacecraft, the very high-end monolithic cells are preferred as they give the highest efficiencies available at low mass.
 Equipment for MJ cells is not different from the basic thin-film or wafer equipment depending on technique. The wafer process requires indium, gallium and germanium and also uses arsine and phosphine gases. Finished wafers are cut into cells, mounted, wired and covered by front glass.


  1. It was necessary for someone to write on this topic and you grabbed the heart of millions by taking the initiative which is enough to bring the evolution. Thanks for bringing it in light.
    thin ceramic substrates

  2. It is an amazing collection of beautiful thoughts. I loved the quoted examples which made me to read the post multiple times. Nice work.

    substrate dicing

  3. I really like this site, I would like to see new content on the site, and hope to be able to inform me when I'll be back.

    machining alumina ceramic