Metallurgy & Alternate Recipes Design¶

Document Type: Design Document Version: 0.2 (Draft) Status: In Progress Last Updated: 2026-09-01 Vault destination: 01_Design Documents Phase: Phase 1 — Launch release Amends: 04_Factory & Resource System Design §Refined Resources (ref_metal_plate, ref_copper_wire sources) Related: [04_Factory & Resource System Design](<./04_Factory & Resource System Design.md>) · [25_Power System Design](<./25_Power System Design.md>) · [07_Pipe & Fluid System Design](<./07_Pipe & Fluid System Design.md>) · [05_World Generation & Planet Types](<../Content/05_World Generation & Planet Types.md>) · [08_Research Tree Design](<../Content/08_Research Tree Design.md>) · [01_Master GDD](<../01_Master GDD.md>)

1. Summary¶

This document covers how metal is made, and the alternate-recipe model that grew out of designing it. Silicon is here too: it is a metalloid rather than a metal, but it is reduced in an arc furnace by the same carbothermic process as iron (§5a). Two things live here:

The recipe model: how a single output can have several routes, and the rules that keep those routes from collapsing into one obvious answer.
The materials themselves: iron, copper and silicon in full, and the rule for every metal added later.

Power distribution items (transmission line, power cable, poles) came out of the same conversation but live in [25_Power System Design](<./25_Power System Design.md>) §4d, next to the pole radii and voltage-drop rules they depend on.

Every number in this document is a default, not a decision. They exist so the shape can be argued about with something concrete on the page. The balance pass will change most of them.

2. The recipe model¶

2.1 Recipes are indexed by output¶

A recipe belongs to the thing it produces. "The recipes for iron plate" means every recipe whose output is ref_iron_plate, and nothing else.

This sounds obvious and is easy to get wrong. A route like ore to liquid iron to plate is two recipes owned by two different outputs, not one recipe for plate. melt_iron belongs to fld_liquid_iron. Only cast_iron_plate belongs to the plate.

Getting this wrong makes the Recipe Explorer's "how do I make X" view incoherent, because a recipe would appear under an item it does not produce.

2.2 Alternate recipes: different inputs, same output¶

An alternate recipe produces an identical output from different inputs. The player chooses per machine, and the choice is visible and reversible.

This is not resource substitution. There is no mechanism by which copper stands in for iron in a recipe, because they are different metals and a recipe that wanted one would not work with the other. The family field on a resource is a reading aid, not a substitution mechanic (04_Factory & Resource System Design v0.7, Canonical §12a).

2.3 A good alternate pulls from a different raw branch¶

The raw resources divide into four branches:

Branch	Members
metal	iron ore, copper ore
mineral	coal, quartz, sand
biomass	wood
chemical	chemical gas, water

An alternate is only interesting when it moves the load onto a different branch. Two recipes that eat the same raws in slightly different ratios are a balance tweak wearing a costume. The point of an alternate is that a player who is ore-poor and coal-rich can still make the thing, so the map roll matters.

2.4 No route may be strictly better¶

If one route wins on every axis, it is not a choice, it is a patch note, and every other route becomes dead content the moment it unlocks.

Each route must give something up. The axes available:

Axis	Example
A different input branch	`reduce_iron_plate` spends coal to save ore
Power for materials	`leach_copper_plate` draws heavily, saves ore
Infrastructure for efficiency	the cast routes need a Foundry, a Caster and pipes
Pollution for throughput	see `[26_Pollution, Waste & Organic Systems Design](<./26_Pollution, Waste & Organic Systems Design.md>)`
Building tier	Advanced Press against Press Machine

The baseline route must survive. Smelting stays useful forever because it needs one cheap building and no second input. A remote outpost that cannot justify a fluid network still wants it.

2.5 Alternates scale with throughput, not rarity¶

Give alternates to the items produced in volume. Millions of iron plates justify three routes, because the player will actually compare them. A handful of a rare late-game alloy does not: the interesting decision there is obtaining it, not how it is processed.

Working default: at most three routes for a high-volume item, and one for everything exotic. Beyond three the Recipe Explorer stops being a tool and becomes a wall.

2.6 Routes are per metal, not a universal pattern¶

Iron's three routes are not a template to be recoloured for every metal. Each metal gets routes that suit that material, which is what stops later additions feeling like reskins. §6 gives the assignment rule.

3. The teaching rule¶

Recipes here follow the design principle stated in [01_Master GDD](<../01_Master GDD.md>):

Model the real process closely enough that the player learns why it works that way. Simplify freely, but never invert. A simplification may stay silent about something true; it may not teach something false. Where accuracy and fun conflict, fun wins, and the design goes quiet rather than lying.

Applied here, this is what decides which metal gets which route. Copper gets the chemical route because low-grade copper ore genuinely is why hydrometallurgy was developed. Giving iron the chemistry and copper the coal would play just as well and would teach a falsehood, so it is wrong.

The rule permits plenty of compression. Real copper goes ore, concentrate, matte, blister, cathode. The game does ore, plate. That is silence, not a lie.

4. Iron¶

Iron is the bulk metal: abundant, decent ore grade, cheap to process. It is a carbon problem, and its second route reflects that.

4.1 `ref_iron_plate`¶

Recipe	Inputs	Output	Time	Ore/plate	Building
`smelt_iron_plate`	2 iron ore	1 plate	2s	2.00	Smelter
`reduce_iron_plate`	3 iron ore + 1 coal	2 plate	4s	1.50	Smelter
`cast_iron_plate`	1 liquid iron	1 plate	1s	1.00 net	Caster

smelt_iron_plate is the baseline and never becomes obsolete. One cheap building, one input.

reduce_iron_plate is carbon reduction, the blast furnace. Coal is the reagent, not just the heat. Same plates per minute as smelting, 25% less ore, and it couples the plate line to coal. That coupling is the interesting part: coal also feeds process_fuel_cell, so the player competes with their own fuel chain.

Coal currently has no producer (OI-332). reduce_iron_plate cannot be built until that is resolved.

cast_iron_plate is continuous casting, and it is the payoff for infrastructure.

4.2 `fld_liquid_iron`¶

Recipe	Inputs	Output	Time	Building
`melt_iron`	5 iron ore	5 liquid iron	4s	Foundry, high PU

The output quantity is the dial that decides whether the fluid branch exists at all. Since casting is 1 liquid to 1 plate, ore per plate through the cast route is 5 / output:

Output	Ore/plate	Against `reduce` (1.50)
3	1.67	worse, the branch is dead content
4	1.25	17% better
5	1.00	33% better, chosen
6	0.83	45% better, smelting stops being worth keeping

Chosen 5, giving exactly 1 ore to 1 plate. It reads as a real milestone, and it justifies a Foundry, a Caster, pipes and the power bill.

Iron melts at 1538°C, so melt_iron draws more power than melt_copper.

5. Copper¶

Copper is the chemical metal. Real copper ore is dramatically lower grade than iron ore, often under 2% copper against iron ore's 30 to 65%, and that single fact drives everything below.

5.1 `ref_copper_plate`¶

Recipe	Inputs	Output	Time	Ore/plate	Building
`smelt_copper_plate`	5 copper ore	2 plate	3s	2.50	Smelter
`leach_copper_plate`	3 copper ore + 1 chemical solvent	2 plate	3s	1.50	Chemical Plant, high PU
`cast_copper_plate`	1 liquid copper	1 plate	1s	1.00 net	Caster

Copper's baseline is worse than iron's (2.50 against 2.00) and its improvement curve is steeper (60% total reduction against iron's 50%). That is the low-grade ore showing through, and it gives copper a real arc: painful when first met, transformed by chemistry.

leach_copper_plate is SX-EW: heap leach with acid, solvent extraction, electrowinning. Roughly a fifth of world copper production. It is power-hungry because the final step is electrical, and it consumes fld_chemical_solvent.

fld_chemical_solvent has no producer (OI-333 area). The chemical layer is not yet designed; leach_copper_plate is blocked on it.

Both metals reach 1.00 ore/plate fully built out. Mastering either chain converges on 1:1, which is a clean thing for a player to aim at.

5.2 `fld_liquid_copper`¶

Recipe	Inputs	Output	Time	Building
`melt_copper`	4 copper ore	4 liquid copper	3s	Foundry, lower PU than `melt_iron`

Copper melts at 1085°C against iron's 1538°C, so this is the cheaper melt. A player who notices is rewarded.

5.3 Low-grade ore, and why leaching survives¶

Without this section, cast_copper_plate (1.00 ore/plate) dominates leach_copper_plate (1.50), and leaching becomes dead content the moment the player can plumb, violating §2.4.

raw_copper_ore_low is ore that only the leach route can process. Smelting and melting cannot touch it. Its deposits are far more abundant than standard copper ore.

Recipe	Inputs	Output	Time	Building
`leach_copper_plate_low`	6 low-grade copper ore + 1 chemical solvent	2 plate	4s	Chemical Plant

This is the most true-to-life part of the design and the least widely known. Low-grade ore is exactly why hydrometallurgy exists: the rock is too poor to smelt economically, and leaching is the only process that can reach the metal in it. A player who builds the chemical route unlocks map deposits that were previously worthless, which is a real discovery rather than a number going up.

It also gives leaching a permanent niche that no other route can fill, which is what §2.4 requires.

Numbers here are the least examined in this document. The ratio between standard and low-grade ore, and the abundance multiplier on the deposits, need the world-generation pass before they mean anything.

5.4 `ref_copper_wire`¶

Recipe	Inputs	Output	Time	Building
`draw_copper_wire`	1 copper plate	2 wire	1.5s	Wire Extruder

One route only. Drawing wire is a mechanical operation with no interesting second chemistry behind it, and per §2.5 alternates belong where the real decisions are, which for copper is the plate step.

This replaces the current recipe, which is wrong. extrude_copper_wire today takes raw_metal_ore (substance iron) and produces copper wire, so copper is currently made out of iron (OI-336). The corrected chain adds a tier: ore to plate to wire, which lengthens the circuit board chain by one step. Circuit boards feed power cells, which are on the M1 ship path.

5a. Silicon¶

Silicon is a metalloid, not a metal, but metallurgical-grade silicon is made by carbothermic reduction in an arc furnace, which is the same process as §4.1's reduce_iron_plate and belongs in the same document.

Its identity is power. Copper is the chemical chain and silicon is the one that eats the grid.

5a.1 The raw material is quartz, not silicon¶

Silicon does not occur natively. What is mined is quartz or silica sand, which is silicon dioxide; the silicon has to be reduced out of it.

The current raw_silicon, named "Silicon Crystals" and extracted directly, therefore teaches something false, which §3 forbids. It becomes raw_quartz (OI-339). The name also fits the existing Crystal Extractor building better, since quartz genuinely is the crystalline form.

5a.2 `ref_silicon` (metallurgical grade)¶

SiO2 + C -> Si + CO. Carbon is the reagent, not the heat, exactly as in iron. Real silicon furnaces deliberately run a mix of coal, charcoal and wood chips, because charcoal is more reactive and far lower in ash, giving better recovery and cleaner silicon. That is the alternate, and it crosses raw branches as §2.3 requires.

Recipe	Inputs	Output	Time	Feedstock per silicon	Branch
`smelt_silicon_coal`	3 quartz + 2 coal	2 silicon	4s	1.50 quartz	mineral
`smelt_silicon_charcoal`	3 quartz + 2 charcoal	3 silicon	4s	1.00 quartz	biomass
`smelt_silicon_sand`	5 sand + 2 coal	2 silicon	5s	2.50 sand	mineral, abundant

All three draw heavily. Silicon smelting is one of the most electricity-intensive processes in real industry, and that belongs on the recipe, not the Refinery (OI-334).

Why none of the three is dominated (§2.4):

Coal is the baseline. Two inputs, no side chain, works the hour you find quartz.
Charcoal gives the best quartz yield because it really is the better reductant, and costs a whole wood-to-charcoal chain to get there. Coal against forest, decided by the map.
Sand has the worst ratio by a wide margin and the best availability: 05_World Generation line 128 lists Sand as Common, while quartz comes from discrete extractor sites. Its edge is that the feedstock is nearly free.

Sand is not low-grade copper ore repeated. Low-grade ore is locked to a single process and unusable by any other (§5.3). Sand is usable by an ordinary route at a worse ratio. One is exclusivity, the other is efficiency.

Honesty note. Real arc furnaces want lump quartz, roughly 10 to 100 mm, because the burden has to stay permeable for CO to escape; fine sand smothers it and is not used directly. The 2.50 ratio stands in for that loss. This is the edge of what §3 permits, and if it ever needs to be more honest the fix is an agglomeration step rather than a worse number.

5a.3 `ref_charcoal`¶

char_wood: 3 wood -> 2 charcoal, 3s, Chemical Plant

Wood pyrolysis. Gives raw_wood a second consumer, which today is only process_polymer.

Charcoal has an obvious future beyond silicon: historic blast furnaces ran on it, so it is a natural third route for iron plate whenever §4.1 wants one.

5a.4 `ref_silicon_wafer`¶

refine_silicon_wafer: 2 silicon + 1 chemical gas -> 1 wafer, 3s, Refinery, extreme PU

This compresses three real stages into one: Siemens-process purification to polysilicon, the Czochralski pull that grows a single crystal from a seed, and sawing the boule into wafers. That is compression, not falsehood: the document stays silent about the stages rather than claiming they do not exist.

One route only. The wafer already has three consumers (circuit boards, quantum processors, advanced science drives per 04_Factory & Resource System Design), so it clears the §2.5 bar for alternates, but the real second route (fluidised bed reactor instead of Siemens) differs in energy and granularity in ways that would not read as a decision in play. Better one honest route than two that feel identical.

5a.5 Two grades of silicon, deferred and why¶

Real silicon splits into solar grade (cheaper, less pure) and electronic grade (ultra-pure, for chips). It is a genuine industry distinction and it would make an excellent second axis.

It is not designed here because solar grade would have no consumer. Solar Panel is a building, and no construction cost exists anywhere: BuildingDefinition has no build-cost field and no document specifies one (OI-342). Adding ref_solar_silicon today would manufacture precisely the orphan this document exists to prevent.

Revisit once OI-342 resolves.

6. Adding a metal later¶

"Metal ore" is a generic term for every metal resource that arrives as ore, not a resource in its own right. The resources are raw_iron_ore, raw_copper_ore and so on (OI-335).

Real metallurgy gives a consistent rule for assigning a new metal its routes, so no metal has to be invented from scratch and none ships as a reskin:

Metal	Real process	Game axis
Iron	Carbon reduction, blast furnace	coal, cheap and dirty, early
Copper	Leach, solvent extraction, electrowinning	solvent and power, mid
Aluminium	Hall-Héroult electrolysis	enormous power, the grid-eating metal
Titanium	Kroll process	slow and reagent-hungry, late and rare
Nickel	Mond carbonyl process	gas-based, exotic
Gold	Cyanide leach	chemical, tiny volumes

Only iron and copper are in scope now. Per §2.5, a metal added later gets the full multi-route treatment only if it is produced in volume; exotic metals get one route.

Aluminium is worth holding in reserve. Real smelters are built next to dedicated power stations because the electrolysis is so electricity-intensive. If a metal is ever wanted whose entire identity is "this eats your grid", it is already waiting.

7. Open items raised by this document¶

ID	What
OI-334	Recipes have no power field, so per-recipe draw is unrepresentable
OI-335	`raw_metal_ore` splits into per-metal ores
OI-336	`ref_copper_wire` is currently made from iron ore
OI-337	`raw_copper_ore_low` as a new resource for the leach-only route
OI-339	`raw_silicon` is mined as elemental silicon; rename to `raw_quartz`
OI-340	New resources for the silicon chain: `raw_sand`, `ref_silicon`, `ref_charcoal`
OI-342	Buildings have no construction cost, anywhere in data or design

Also depends on OI-332 (coal has no producer) and the undesigned chemical layer that owes fld_chemical_solvent a producer.

8. Not yet designed¶

Two grades of silicon (solar against electronic), blocked on OI-342 (§5a.5).
The chemical layer. Owes a producer for fld_chemical_solvent.
Steel. Iron plus carbon is the obvious next alloy and is not designed.
Scrap and recycling. An obvious fourth route to plate, deliberately deferred.
Research gating. Whether alternates unlock through [08_Research Tree Design](<../Content/08_Research Tree Design.md>) or are available from the start is not decided.

Changelog¶

Version	Date	Change
0.2	2026-09-01	Silicon added (§5a). Quartz replaces mined elemental silicon, carbon source as the alternate axis (coal against charcoal), sand as the abundant low-ratio feedstock, charcoal from wood, and the wafer step. Two-grade silicon deferred: solar grade has no consumer until buildings have construction costs.
0.1	2026-09-01	Initial draft. Recipe model (indexed by output, alternates, no strictly-better route, alternates scale with throughput, routes per metal). Iron and copper in full with three routes each, both liquid metals, low-grade copper ore as the leach-only niche, corrected copper wire chain. Rule for assigning routes to metals added later.