Working Doc

Purity Levels

Metallurgical grade silicon is 98% (one nine?) [1]
Solar grade silicon is 6-8N [2]
Microchip silicon (Electronic Grade Silicon) is nine nines pure, whereas power transistor silicon is 6 nines. [3]. Solar grade silicon is same purity as transistor grade - 6 nines. [4].
3 nines pure by acid leaching [5]
Czochralski silicon is 11 nines pure [6]
- Microchip silicon is nine nines pure, whereas power transistor silicon is 6 nines. [7]
Five nines achieved by electron beam degassing + Chochralski [8]. Interesting in that electron beam degas takes place of Siemens process. Result: cells only 13% good. But can go higher if PERC TopCon etc?
Acid leaching followed by multiple zone refining can reach 9 nines. [9]
3 nines achievable directly from molten salt electrolysis of silica? [10]. Hairy. [11]

Siemens process involves chemical vapor deposition of silanes, with metallurgical grade silicon as the feedstock. [12]
Fluidized bed reactor - similar feedstocks as Siemens, but is continuous, thus capable of lower energy use
3 possible processes: upgraded metallurgical, Siemens, Fluidized Bed - [13]
Upgraded metallurgical silicon requires about 25% less energy than standard (Siemens)
Fluidized bed reactor method uses 45-80 kWhr/kg of solar grade silicon produced [14]
Siemens process takes 70-90 kWhr/kg [15]
What about zone refining directly from metallurgical grade silicon? [16]
How about UMG by vacuum melting and directional solidification? [17]. Zone refining is a specific type of directional solidification. [18]
UMG can include zone refining, acid leaching, plasma treatment, vacuum melting, slag treatment. [19]
Zone refining takes several kw/kg [20]
Lots is in the research phase. Zone refining, solvent refining, and acid leaching is a winning combo. Promise of lower energy use than Siemens. See summary on p. 33 of this MGS refining thesis. [21]

Fluidized bed reactor, pure silicon granule producer [23]. Continuous version of Siemens process.

Si+HCl gets trichlorosilane + numerous impurities reacting with Cl. These are distilled in a column.

And this discusses poly si vs mono si [25]
embodied energy of 165-375 kWhr/kg [26]
- Good news - 65% of all energy goes into the wafers, as opposed to everything else (glass + All + EVA) according to same link
- Similar for amorphous silicon [27]
Compare to steel at 20 MJ/kg [28], and 200MJ/kg more than this for aluminum [29]
How many grams of silicon are needed for 1kW of PV panels? 2 kg [30], based on 2g/W.
5% of weight of PV is silicon [31]. 76% is glass, 10% is aluminum, and 10% is EVA.
Embodied energy of EVA is about 80 mega joules per kilogram [32]

600 kWhr for 2 kg of silicon
3kWhr per kg for glass, or 90 kWhr for 30 kg glass. [33]
About 40 kWhr per my of EVA/aluminum x 8 kg = 320 kwhr
Total 1010 kWhr
- if the low estimate of 165 kilowatt hours above is true and this is 65% of the energy required then we have around 300 kilowatt hours of energy for wafers, and total around 500 kWhr for 1 kW of new PV panels - meaning energy payback of around 1/4 year!
The case here would be easy for open sector PV production
365*6kWhr~2100 kWhr/yr
Payback is thus 1/2 yr with only the above.
This matches results from Fraunhofer 2021, close enough - .7-1.4 year payback time. [34]