Semiconductors

Chips do not decay if kept dry and static-safe. A salvaged stock can run for decades, and discarding salvaged electronics that would still work is a named failure.

• Store dry and cool, around 30–50% relative humidity, with little temperature cycling.
• Anti-static bags or conductive containers; handle on an earthed mat with an earthed wrist strap.
• Electrolytic capacitors dry out over years: reform old ones by raising the voltage slowly through a resistor before first use.
• Flash memory loses its data when unpowered for years: copy data to several media and print what matters.
• Keep datasheets and schematics on paper.

The honest strategy is threefold: preserve existing chips, keep vacuum tube manufacturing alive as an indigenous fallback, and treat semiconductor capability as a multi-generation goal.

Skipping the tube fallback means having no electronics at all when the scavenged stock runs out. Keep at least one tube line in production, with its people trained, even after semiconductors are available.

Assuming that understanding the physics implies being able to build it is the largest single misconception about rebuilding technology. The physics can be taught in an afternoon.

The barrier: semiconductor silicon must reach around 99.9999999% purity — nine nines. The dopants that make a device work are added at parts per billion, so any uncontrolled impurity at that level swamps the effect entirely. Everything smelted so far on this ladder would be catastrophic contamination. Contamination is invisible and destroys the work after the fact.

Germanium is the achievable laboratory target. It is recovered as a by-product of zinc ore processing and coal ash.

Zone refining: melt a narrow zone in a bar of germanium with a moving heater, and pass it slowly along the bar, many times. Impurities prefer the liquid, travel with the molten zone, and gather at one end, which is cut off. Then make a point-contact diode, and a point-contact transistor — two fine gold contacts pressed very close together on the germanium surface, as at Bell Labs in 1947. Measure its current gain.

Metallurgical silicon — around 98–99% pure — comes from quartz and carbon in an electric arc furnace at around 2,000 °C. Then to trichlorosilane, distilled to extreme purity, and decomposed back to polysilicon. Then grow a single crystal by the Czochralski method — pulling a rotating seed slowly from the melt — and slice it into wafers.

Trichlorosilane is flammable and reacts with moisture to release hydrogen chloride; this is chemical-plant work with the full safety practice of acids.

A single dust particle destroys the devices beneath it. Fabrication needs filtered air far cleaner than an operating theatre, ultra-pure water and gases, and people sealed in suits because humans shed particles constantly. Photolithography patterns features below the wavelength of visible light, with optics that are themselves a world-class manufacturing achievement. Then dozens of steps — deposition, etching, implantation, annealing, metallisation, test, packaging — each of which must succeed.

Realistic targets: discrete transistors and diodes for a small industrial society. Integrated circuits of any modern density are not reachable by a small society, and no individual brilliance changes that. Modern chip-making spans tens of thousands of people across a supply chain.

The verify for this step is an honest assessment, not a claim. Build a working discrete transistor, then write down, in the archive, exactly how far that is from an integrated circuit: purity achieved versus needed, feature size achieved versus needed, clean-room class, yield, and the people and plant each missing capability would take.

Publish it. The gap is the point of this step — and the reason the preservation of existing chips in entry 1 matters most.