The machine's pantry: which matter it holds on the shelf, why the shelf is small, why everything on it is deliberately boring, and why the shelf itself is an instrument rather than a bin. Size-free — the logic privileges chemistry and ratios, not litres.
To make almost anything, what do you keep on the shelf?
The intuition most people bring to a universal fabricator is a warehouse: one jar per thing you might build — a bin of cup-stuff, a bin of plate-stuff, a bin of spoon-stuff. That intuition is wrong, and it is wrong in the same way a kitchen proves it wrong. A good kitchen does not keep one cupboard per dish. It keeps flour, water, oil, salt, and a spice rack — a small set of dense, stable, boring staples — and the recipes live somewhere else, on paper. Any dish on demand, from a pantry you can walk past without noticing.
The Replicator's pantry is smaller than a kitchen's, because it works below recipes, at the level of atoms — and at that level the world is shockingly repetitive. A cup, a steak, a circuit board, and the shelf they sit on are overwhelmingly built from the same short list of elements: carbon, hydrogen, oxygen, nitrogen, silicon, a little iron and friends. Objects differ in arrangement, not ingredients. The machine assembles from the atoms' own chemistry, not from finished parts — so what it must keep is elements in bulk, not objects in miniature. Four boring staples cover most of the periodic demand: water (hydrogen and oxygen), charcoal (carbon), air (nitrogen, oxygen — and argon, free of charge), and rock (silicon, iron, aluminium, calcium and the rest of the crust's greatest hits). Add a spice rack of trace elements for the recipes that need a pinch of something rare, and the pantry is complete.
Two more things make this pantry unlike a warehouse. First, everything on the shelf is kept in its most boring chemical form — oxidized, dissolved, inert — because boring is what stable, safe, and dense look like in chemistry. The exciting forms (pure reactive metal, loose atoms) are made fresh at build time, the way a kitchen grinds coffee just before brewing. Second, the shelf is not passive. The storage chamber is a working instrument — a mirrored twin of the build chamber that sorts incoming matter, checks it, weighs it, and stages it for the next build, all while the chamber above is busy presenting the object. The pantry has hands.
That is the picture. The rest of this paper is the machinery under it: what exactly sits on the shelf and why chemistry forces those choices (§2), the spice rack (§3), the storage chamber as instrument (§4), how matter gets in and out and how the books are kept (§5), and how big the shelf must be — which turns out to be a question about what you build, not about the machine (§6).
The word "storage" names two entirely different things in this machine, and confusing them is the first error this paper exists to prevent:
| store | holds | substrate | owner paper |
|---|---|---|---|
| the pattern library | what objects are — blueprints, scans, drive chords, the machine's learned weights | digital / holographic — information | Memory & Compute |
| the matter stores | what objects are made of — elements in bulk, in storable chemical form | tanks, blocks, hoppers, bottles — atoms | this paper |
The split is exact and load-bearing. A pattern weighs nothing and can be copied; matter weighs exactly what it weighs and can only be moved — baryon conservation makes the machine a rearranger, never a creator (Principles). Everything about how patterns are held — why they are chords rather than bits, why the weights are the medium rather than a file — lives in Memory & Compute and is not restated here. This paper is about the atoms' side of the house: the shelf, its chemistry, and its instrument.
The taxonomy below is the canonical store inventory (moved here from Architecture §5, which keeps the loop topology around it). The governing rule is stated first because every row obeys it:
| subsystem | design | tier |
|---|---|---|
| bulk stores | water tank (H,O); solid graphite block (C — feedstock and screen chemistry are the same element, Architecture §2.4); air intake (N,O — never store what the atmosphere holds). Argon joins the air intake, not the cartridge: at 0.93% by volume it is the third most abundant atmospheric component, so the same rule covers it, and ordinary cryogenic air separation already yields it as a standard product | trivial |
| trace cartridge | ~20-element mineral pack (Si, Ca, Fe, Na, K, P, S…) — the recipe tail. Collapses to one "rock" bulk: silicate/regolith already carries Si, O, Al, Ca, Fe, Na, K, Mg in one stable solid, so a single mineral hopper replaces the slotted pack; the machine sorts specific metals out on demand (the EM m/q stage — Architecture §5). But that hopper is itself two stores — see the salt/ash split immediately below | trivial |
| mineral bulk — the salt/ash split | The mineral store divides in two, and the dividing property is solubility. Salt = the soluble fraction (alkali and alkaline-earth chlorides and carbonates), held as brine, riding the fluid loop — pumped, dissolved, addressable. Ash = the insoluble refractory residue (SiO₂, Al₂O₃, Fe₂O₃, CaO), riding the solids loop — augered, fluidized, sintered. The split is forced by chemistry, not chosen for tidiness: atomized metal cannot be stored in water. Per kg of metal in water, Li yields 1,762 L H₂, Al 1,359, Mg 1,006, Fe 657, Ca 610, Na 532 (Na and K ignite it) — and fine powder maximizes exactly the surface area that drives the reaction, so a metal-powder slurry stores hydroxide and rust, not metal. The resolution is that the metals reacting most violently with water are precisely those whose salts are most soluble (LiCl ≈ 830 g/L, CaCl₂ ≈ 745, MgCl₂ ≈ 540, NaCl ≈ 360), so the reactive tail stores pre-oxidized and dissolved — stable, pumpable, and already past the sort. Rule: store oxidized, never atomized. The former single "rock" hopper is the ash store; the brine is new | design rule |
| noble-gas cartridge | He, Ne, Kr, Xe only — a small pressure bottle, and the one place in the loop where a bottle beats everything. Dissolution in water is dead: at 1 atm and 25 °C water holds He 1.5 mg/L, Ne 9.1, Ar 56, Kr 210, Xe 578 — even xenon, the most soluble, is 0.06% by mass, and a plain 200-bar bottle beats water by ~3,900× (Xe, real-gas) to 21,000× (He). Use real-gas densities, not ideal-gas: at 200 bar and 20 °C, Peng–Robinson gives He 31 g/L, Ne 154, Ar 355, Kr 978, Xe 2,281 — the ideal-gas law understates xenon by 2.1× and krypton by 1.4×, because both are near enough to their critical points to be liquid-like under pressure. Atmospheric extraction is dead for these four too: 1 g needs 67 m³ of air (Ne), 256 (Kr), 1,166 (He), or 2,140 (Xe ≈ 714 kWh of air liquefaction per gram). But they are the ideal cartridge item — chemically inert, zero degradation, indefinite shelf life — and demand is narrow (sealed lamps, insulated glazing, some electronics), so a small bottle is proportionate. Clathrate hydrate is a near-miss worth recording rather than building: Xe hydrate is 56% xenon by mass and forms at only ~0.15 MPa at 273 K, but it needs refrigeration, does not beat pressurised storage, and fails where help is wanted, since Ar needs ~100 MPa and He and Ne will not form cages at accessible pressure at all. Condensed phases are real but not wanted at cartridge scale. All five liquefy and solidify (Xe: BP 165 K, MP 161 K, liquid 2,942 kg/m³, solid 3,540; Ar: BP 87 K, liquid 1,396; He is the exception — it does not solidify at 1 atm at any temperature, needing ~25 atm to freeze, because zero-point energy exceeds its van der Waals binding). Cryogenic liquid is how these ship in bulk — the density gain over a bottle is 7.8× (Ne), 4.0× (He), 3.9× (Ar), 2.5× (Kr) — but only 1.3× for xenon, and every one of them buys that gain with a cryoplant and continuous boil-off. At the gram-to-kilogram quantities this cartridge holds, ambient-temperature pressure storage wins on total system mass and has indefinite shelf life. Note also that xenon's critical temperature is 16.58 °C, below room temperature: at 20 °C it cannot be liquefied at any pressure and is stored as a dense supercritical fluid — which is exactly how xenon is held for ion thrusters | MEASURED components |
| container architecture | store count is a sorting-efficiency dial, not an element count. A container is a decision to pre-pay the element sort at resupply rather than on-demand at build. Because the stored form is stable oxidized compounds, a single mixed bulk of rock + graphite + ice is chemically stable cold — so the "one mass bulk" limit is blocked only by on-demand sorting energy, not by chemistry or safety. Ladder: slotted cartridge → brine + ash + water + carbon + free air, plus the noble-gas bottle (~4 stores and one cartridge — the instance floor, RH-1: Hardware Spec §4) → single oxidized bulk as in-field sorting efficiency climbs. Note that the salt/ash split is the one merge the ladder does not get to make for free: brine and ash can share a vessel cold, but re-separating them costs an evaporation, so they stay split until in-field sorting is cheap enough to pay it. The single bulk is the efficiency-limit endpoint (feedstock's "one performance, one block"), pulled closer by the recycle loop — repetitive make/unmake makes the bulk's element ratios track demand, shrinking the sort | design rule |
| stored form vs transport form | ionized/molten streams via CNT-membrane channels (Architecture §2.4) are transport phases; the stored form = densest chemically-boring stable phases (boring = safe = dense = pattern-free). Exotic dense-matter states are firewalled out of storage: Rydberg matter is low-density (wrong direction), and ultradense hydrogen H(0) is contested and, if real, a near-nuclear energy density — the opposite of "boring/safe." H(0)/RM belong (if anywhere) to the OPEN energy lane, never the feedstock bin. Ordinary condensed matter is dense enough | MEASURED components |
Provenance of the salt/ash split, stated in derivation order (2026-07-30). The partition above was reached from reactivity and solubility arithmetic — the H₂-per-kg and g/L figures in the row itself — and only afterwards noticed to reproduce the water / salt / ash triad of classical alchemy, which specifies exactly these three vessels and draws the same distinction we need: "pure ash has not emerged from a coagulation or deposition process, but from the (opposite) fire process" (Steiner, GA343, 1921) — i.e. precipitate and calcine are different routes to a solid mineral and therefore differ in solubility. That is a real distinction, correctly drawn, and the convergence is unsurprising once stated: both schemes classify matter by how it separates, and the separation verbs are few. It is recorded here as a taxonomy match, not a source — nothing in the row depends on it, and the ordering matters: the arithmetic came first and would stand alone.
Most of the periodic table is not pullable from the four bulk staples, and most of it is never needed — but the tail of real recipes (dopants, catalysts, phosphors, the odd gram of something exotic) is exactly what separates "makes most things" from "makes almost anything." The trace rack holds that tail: small sealed cartridges of elements that neither the air, the water, the carbon block, nor the rock hopper can supply, each stored — same rule as everywhere — in its most boring stable compound form. The noble-gas bottle of §2 is honorary rack membership: cartridge-scale, indefinite shelf life, narrow demand.
Cartridge geometry — the chart as parts organizer. The trace cartridge takes polar form: a circular magazine in which ring radius = period (octave) and angular position = group (tone) — the periodic chart in polar coordinates as a physical parts layout, with mechanically keyed slots (an element cartridge cannot seat at the wrong address). The instance realization — the cartridge circle loaded through the base door — is the Hardware Spec §4's to dimension.
The shelf has hands. Ruling R9
(REPLICATOR_CAP_STACK_RULINGS_2026-07-30.md; stack topology
owned by Architecture §3.2)
places the stores' working volume directly under the build volume: top,
middle, and bottom plate assemblies; build chamber above, storage
chamber below; the middle plate is double-faced — two cap stacks back
to back, the mirrored-pair logic promoted to an interior bulkhead,
radiating up into the build volume and down into storage.
The consequence this paper owns: the storage chamber is an active instrument, not a bin. Three standing duties:
Why a second chamber at all, rather than a hopper and a pump: the twin is the same plates. A same-size mirrored chamber means one bill of materials, one drive stack, one calibration — and a hidden volume where field-work can proceed while the build chamber presents the object. The machine is never idle; it is always either performing or preparing. Later options on the same logic: a calibration chamber, a dedicated disassembly/reclaim chamber. Deck line, quoted from the stack ruling: two chambers — one holds the object, one holds the matter it will become; between them, one plate.
Four interfaces connect the shelf to the rest of the machine.
Feedstock moves between chambers through the co-axial bores — ruling R10, owned in full by Architecture §3.3. Summary: three contrawound winding pairs stacked on one axis give six independent circuits, which is traveling-wave synthesis — a linear-induction elevator for conductive feedstock plus acoustic tube conveying for the rest. No belts, no augers, no moving parts: the lift is a field, not a mechanism. The bore shares duty with the optical sightline by aperture partition or time-multiplex.
The build mode's feed stage (Assemble §4) takes its input from here: feedstock leaves the storage chamber already sorted, characterized, and pre-heated (§4), rides the bore, and enters the build volume in whichever transport phase the build calls for — powder, droplet, vapor (the transport-phase ladder is Architecture §5's). The stores' contract with assemble is simple: nothing arrives unmeasured, and the reduction from stored (oxidized) form to build (reduced) form is paid en route, at build time — §2's rule read in the forward direction.
Dissolve's output comes back the same way (Dissolve §5): freed grains ride the trap pockets back, deeper unbind rungs ride the bore and the fluid loop as droplets or vapor — and everything re-enters storage in storable form, obeying the same salt/ash rule as resupply. Reclaimed mineral matter divides by solubility into brine and ash; reactive metals return pre-oxidized. Reclaimed stock passes the same incoming QC as bought stock — the storage chamber does not distinguish where atoms came from, only what they are.
Baryon conservation makes the inventory arithmetic exact: every gram in the object came from a store, and every gram returns to one. Stores grow by exactly one object-mass per derez. Cyclic element imbalances (replicate steak, derez plates → carbon surplus, iron deficit) are buffered by the trace rack and balanced by resupply — or, if ever demonstrated, by the transmutation sector as balancer of last resort (OPEN there, not here; that link is this paper's only one — feedstock chemistry is ordinary chemistry). Energy rides the same ledger in miniature: recombination energy released on re-forming simple stored molecules is captured by the harvest layer (Architecture §2.4), so derez is modest-net-cost, not free — tier: arithmetic and MEASURED components, as tagged at the source rows in §2.
Size-free, because the answer genuinely does not depend on the machine. The store volume is set by two distributions the machine does not choose:
Three consequences:
Register discipline: claims are tagged MEASURED (published or bench-verified art), DERIVED (arithmetic on measured quantities), or OPEN (registered bet with named falsification); taxonomy rows additionally carry their source tiers (trivial / design rule / arithmetic) verbatim. Nothing was upgraded in the move to this paper; where the source document carried a chip or tier, it is carried here unchanged.
Single-home note. This paper owns store-level detail: the taxonomy of §2, its chemistry rationale, the trace rack, the storage chamber's instrument duties, and the inventory bookkeeping. What it does not own: the matter-loop topology — the sorting energy ladder, the transport-phase ladder, disassembly and energy recovery — which stays with Architecture §5; the stack and bore rulings R9–R10 (Architecture §3.2–3.3); the reclaim narrative (Dissolve §5); pattern storage (Memory & Compute); and any specific build's dimensions (Hardware Spec §4). A change to what the machine keeps, or in what form, lands here first; the other papers then point at it.