UMG-Si Production

About 10 kWhr/kg, and 10 kg/hr at 100kW process scale. [1]. Heated by induction, with graphite crucible as susceptor (starts the melt)

Melt

Process Step	Description	Typical Temp (°C)	Energy (kWh/kg Si)	Notes (100 kW scale effects)
Feedstock Selection & Blending	Crushing, sorting, magnetic separation, blending of MG-Si	Ambient	0.05–0.15	Mostly mechanical energy; minor relative contribution
Primary Melting & Slag Refining	Melt MG-Si (~1414°C), add CaO/SiO2 slag, impurity oxidation & partitioning	1450–1550	3.5–6.0	Dominant energy load; small furnaces have high radiant/convective losses
Gas Refining / Vacuum Refining	Removal of B, P via O2, H2O, H2, or vacuum evaporation	1500–1600	1.5–3.0	Inefficient gas utilization at small scale; vacuum pumps add parasitic load
Directional Solidification	Controlled solidification to segregate impurities (top cut removal)	1400 → 1200	2.0–4.0	Long cycle times → major heat loss; poor insulation penalizes small systems
Crushing, Classification, QC	Break ingot, remove impurity-rich zones, size grading, analysis	Ambient	0.2–0.5	Includes mechanical comminution and some analytical overhead
Optional Repeat Refining Passes	Re-melt and re-solidify for higher purity	1450–1550	2.0–5.0	Highly variable; depends on target purity (solar vs near-electronic grade)
Total (single pass, no repeats)	—	—	7.25–13.65	Typical practical range for small-scale UMG-Si
Total (with 1 repeat pass)	—	—	10–18+	Required for higher purity (approaching SoG-Si)

Melt + Vacuum

Subsystem	Description	Low-Cost Build ($)	Practical Build ($)	High-End Build ($)	Notes
Vacuum Chamber & Lid	Stainless steel chamber, flanges, seals (Viton/metal), viewport, structural frame	8,000–15,000	15,000–30,000	30,000–60,000	Must handle radiant heat from ~1500°C melt; water-cooled walls recommended
Feedthroughs & Ports	Electrical (coil), thermocouple, pressure gauge ports, gas inlet, vacuum port	1,500–3,000	3,000–8,000	8,000–15,000	Includes ceramic/metal vacuum feedthroughs and high-temp insulation interfaces
Roughing Pump	Rotary vane or dry scroll pump for initial pumpdown (atm → ~1e-2 bar)	2,000–4,000	4,000–8,000	8,000–15,000	Oil pumps cheaper but require filtration; dry pumps cleaner but more expensive
Roots Booster Pump	Increases pumping speed in mid-vacuum range	5,000–8,000	8,000–15,000	15,000–25,000	Critical for reducing cycle time at batch scale
High Vacuum Pump (Diffusion)	Achieves ~1e-3–1e-5 mbar for impurity evaporation	6,000–10,000	10,000–20,000	20,000–40,000	Lowest cost option; requires cooling and oil management
OR High Vacuum Pump (Turbo)	Cleaner alternative to diffusion pump	10,000–15,000	15,000–30,000	30,000–60,000	Preferred for contamination-sensitive silicon processing
Cold Trap / Baffles	Condenses SiO, P, and other vapors before pump	2,000–5,000	5,000–10,000	10,000–20,000	Prevents pump contamination; essential for longevity
Vacuum Gauges & Instrumentation	Pirani + Penning/cold cathode gauges, controllers	1,500–3,000	3,000–8,000	8,000–15,000	Dual-range measurement required for process control
Valves & Plumbing	Gate valves, foreline valves, bellows, piping	2,000–5,000	5,000–10,000	10,000–20,000	Must be vacuum-rated; include isolation and safety interlocks
Cooling System Upgrade	Water cooling for chamber, pumps, baffles	3,000–7,000	7,000–15,000	15,000–30,000	Often underestimated; diffusion pumps especially require stable cooling
Controls & Integration	PLC, interlocks, sequencing (pumpdown → heat → refine → vent)	2,000–5,000	5,000–15,000	15,000–40,000	Prevents operator error and protects pumps/furnace
Total (Diffusion Pump System)	—	25,000–45,000	45,000–90,000	80,000–150,000	Most cost-effective configuration
Total (Turbo Pump System)	—	30,000–55,000	60,000–120,000	100,000–200,000	Cleaner operation, higher capital cost

Overall Process

#	Process Step	Main Purpose	Principal Impurities Targeted	What Happens in This Step	Typical Hardware / Equipment	Output / Decision Point
1	Feedstock selection and blending	Start with the cleanest practical MG-Si feed and reduce lot-to-lot variability	Gross variation in Fe, Al, Ca, Ti and other metallic impurities	Incoming MG-Si is sorted by source and chemistry, oversized or contaminated pieces are rejected, lots may be blended to get a more consistent starting composition	Feed bins, jaw crusher, screens, magnetic separator, scale, mixer/blender, sample splitter	A characterized and blended MG-Si feedstock with baseline assay
2	Primary crushing and sizing	Prepare feed for uniform melting and handling	No major chemical removal; this is a physical prep step	MG-Si lumps are broken into a size range that melts consistently and supports representative sampling	Jaw crusher, roll crusher or hammer mill, screens, sieves, bins	Sized feed with known particle range
3	Initial assay / feed characterization	Establish the starting impurity inventory	Fe, Al, Ca, Ti, Mn, Cu, Ni, B, P, and others depending on process goals	Representative samples are taken and sent for elemental analysis so the refining recipe can be chosen intelligently	Sample prep tools, outsourced ICP-OES / ICP-MS / GDMS or in-house assay tools	Baseline composition for process planning
4	Primary melting	Convert solid feed into a controlled molten bath	Not primarily a purification step by itself	Silicon is charged to a crucible and melted, often by induction with a graphite or SiC susceptor/crucible arrangement	Induction furnace, power supply, crucible, coil, cooling system, temperature instrumentation	Stable molten silicon bath ready for refining additions
5	Slag refining	Remove boron and many oxidizable metallic impurities by partitioning into slag	Especially B; also Al, Ca and other metallic impurities depending on slag chemistry	A reactive slag is added to molten silicon; impurities preferentially transfer from the silicon melt into the slag phase, which is later separated	Induction melt furnace, slag addition system, crucible, skimming tools or tapping arrangement	Refined melt plus spent slag; decision on whether impurity removal was sufficient
6	Slag separation / melt cleanup	Physically remove the impurity-bearing slag phase	Slag-borne impurities captured in prior step	The impurity-enriched slag is skimmed, tapped off, or otherwise separated from the silicon melt to prevent recontamination	Skimmer, tilt-pour hardware, ladle, refractory tools	Cleaner molten silicon after slag treatment
7	Vacuum refining or gas refining	Remove volatile impurities from the melt, especially phosphorus	Especially P; in some routes also helps with volatile species generated from B-removal chemistry	The molten silicon is held under vacuum or exposed to reactive gas treatment so volatile impurity species evaporate or react and leave the melt	Vacuum chamber, roughing pump, Roots booster, diffusion or turbo pump, gauges, valves, cold trap; or gas injection / plasma system depending on route	Lower-phosphorus melt; decision on hold time or repeat pass
8	Optional repeat melt-refining pass	Increase purity when one pass is insufficient	Residual B, P, and metals	The melt may go through another slag-refining and/or vacuum-refining cycle if assay data shows impurity levels are still too high	Same furnace and refining hardware as above	Incremental purity improvement at the cost of energy, time, and yield
9	Directional solidification	Segregate remaining impurities during controlled freezing	Many metallic impurities and some residual P, depending on segregation behavior	Silicon is solidified in a controlled thermal gradient so many impurities concentrate in the last-to-freeze regions rather than uniformly throughout the ingot	Directional solidification furnace, mold/crucible, controlled cooling and gradient system	Refined ingot with impurity gradient from clean region to dirty tail/top
10	Cropping / cut-off of contaminated regions	Remove impurity-rich portions of the ingot	Impurities concentrated in tail, top, edge, or last-solidified regions	The most contaminated parts of the ingot are cut off and rejected or recycled to an earlier refining step	Saw, splitter, handling table, labeling and traceability tools	Cleaner ingot sections retained for SoG-Si feedstock
11	Secondary crushing and classification	Convert refined ingot into the desired chunk or granular form for downstream use	No major chemical removal; physical finishing step	Accepted ingot sections are crushed and classified into the size specification needed for storage, sale, or further crystal growth	Crusher, grinder, sieve shaker, bins, packaging tools	Sized refined silicon product
12	Final assay and quality control	Verify whether material meets the solar-grade target	B, P, metallic impurities, and sometimes O and C	Samples from multiple ingot locations and final product fractions are analyzed; results determine accept, reprocess, blend, or downgrade decisions	Sample prep station, outsourced or in-house ICP-OES / ICP-MS / GDMS / SIMS / FTIR as appropriate	Qualified SoG-Si batch or decision for additional refining
13	Packaging and traceable batch release	Preserve quality and maintain lot traceability	Recontamination risks from handling and storage	Accepted refined silicon is labeled, packaged, and documented with assay data and process history	Clean containers, labels, inventory system, batch records	Solar-grade silicon feedstock ready for downstream crystal growth or sale

Diagnostics

Instrument	What it is	What it does	Main use in solar-silicon refining	Typical commercial cost	Rough open-source cost	Open-source feasibility	Notes
ICP-OES	Inductively Coupled Plasma Optical Emission Spectroscopy. A liquid sample is nebulized into a very hot argon plasma, and the light emitted by excited atoms is measured at characteristic wavelengths.	Multi-element elemental analysis, typically after dissolving the silicon sample in acid.	Measures many metallic impurities such as Fe, Al, Ca, Ti, Mn, Cu, Ni, etc. Good for routine batch tracking and process control.	New: about $50k–$250k typical market range; used: often about $20k–$150k.	$20k–$80k for a serious open build; $80k–$150k if aiming for robust automation and better optics.	Medium	Most realistic advanced spectroscopy tool to open source. Hard parts are RF plasma source stability, optics, detector calibration, acid-resistant sample handling, and software. Commercial ICP-OES systems are positioned as routine elemental analyzers for busy labs.
ICP-MS	Inductively Coupled Plasma Mass Spectrometry. Similar plasma front-end to ICP-OES, but instead of measuring light, it feeds ions into a mass spectrometer.	Ultra-trace elemental analysis in solution, often down to ppb, ppt, or lower depending on element and matrix.	Measures very low impurity concentrations. Important when impurities are too low for comfortable ICP-OES quantification, and for difficult trace-element work.	New: about $50k–$500k typical market range; used: about $15k–$150k. Used Agilent 7850 listings are around $90k.	$250k–$1M+ and likely multi-year development.	Very low	Open-sourcing a useful ICP-MS is far beyond a normal shop build. The plasma interface, ion optics, ultra-high-vacuum mass analyzer, RF electronics, contamination control, detector chain, and software are all difficult.
GDMS	Glow Discharge Mass Spectrometry. A solid conductive or semiconductive sample is sputtered directly in a glow discharge source, and the sputtered ions are analyzed by a high-resolution mass spectrometer.	Direct solid analysis of high-purity conductive and semiconductive materials, often to ppb levels, with minimal wet chemistry.	Extremely useful for bulk purity analysis of silicon and other high-purity solids without dissolving the sample first.	Commonly quote-only commercial systems; practical expectation is high six figures to low seven figures new.	$500k–$2M+ and likely a specialized team effort.	Extremely low	This is one of the best-fit methods for high-purity bulk solids, but one of the least realistic to open source. Thermo positions GD-MS specifically for direct analysis of high-purity conductive and semiconductive materials at ppb levels.
SIMS	Secondary Ion Mass Spectrometry. A focused primary ion beam sputters the surface, and the emitted secondary ions are mass analyzed.	Ultra-sensitive surface and depth-profile analysis of dopants and impurities; can map concentration versus depth.	Used for very low-level dopants such as B and P, and for depth profiling near surfaces, diffusion layers, and contamination gradients.	Commonly quote-only commercial systems; practical expectation is high six figures to several million dollars new. Used older CAMECA IMS 7F systems do appear on the used market.	$1M–$5M+ and a major long-duration development program.	Essentially not realistic	Fantastic tool, but not a practical open-source first target. Primary-ion optics, vacuum system, mass analyzer, detectors, stage control, and calibration are all very hard.
FTIR	Fourier Transform Infrared Spectroscopy. Measures how a sample absorbs infrared wavelengths to infer bonding and composition.	Identifies molecular bonds and some light-element-related features; in silicon work it is commonly used for oxygen, carbon, and some bonding-state information depending on sample prep and mode.	Screening for oxygen/carbon-related information, contamination, and some material-ID tasks. More of a support instrument than the primary impurity workhorse.	New: about $15k–$150k typical market range; used: about $7k–$60k.	$5k–$25k for a credible open build; $25k–$60k for better mechanics, interferometer stability, and software.	High	This is the most open-sourceable instrument on this list. Many subsystems are comparatively accessible: IR source, Michelson interferometer, detector, sample holder, and transform software.

Towards Open Source

UMH-Si takes 75% less energy (120 kWhr/kg vs 30) to produce wafers.

It competes on cost.

There is a need for a standardized, reproducible, openly documented process stack for UMG-Si PV.

And is less polluting. [2]