The Donor

the atom that dopes your CPU — and holds a qubit · one atom, two machines

A group-V donor atom has one spare valence electron. Pack billions of them into silicon and you get the conducting channels every transistor — and every transformer's forward pass — runs on. Implant one, alone, in isotopically-pure silicon, and that same electron (and the atom's nuclear spin) becomes a qubit. Arsenic, the everyday CPU dopant, sits right on that hinge. This is the donor — both destinies, and the line between them.

The Donor Family

the four group-V donors in silicon — their nuclear spin grows as you go down the column (½ → 3/2 → 7/2 → 9/2), and more spin means more levels: a qubit becomes a qudit

donor (group V)	Z	nuclear spin	level below E_c	in the CPU	as a qubit
Phosphorus · P	15	³¹P · spin ½ (100%)	~46 meV	bulk & wells — a FAST diffuser, drives deep into Si	the original Kane qubit — spin-½, the simplest two-level donor
Arsenic · As	33	⁷⁵As · spin 3/2 (100%)	~54 meV	shallow source/drain & n+ contacts — HEAVY & slow-diffusing, very high solubility (~2×10²¹ cm⁻³)	a 4-level (spin-3/2) donor qubit — the everyday CPU dopant, single-atom
Antimony · Sb	51	¹²³Sb · spin 7/2	~43 meV	buried layers — even heavier & slower, abrupt profiles	an 8-level QUDIT in silicon (¹²³Sb), electrically driven — Morello 2020
Bismuth · Bi	83	²⁰⁹Bi · spin 9/2 (100%)	~71 meV	the heaviest group-V donor — deep level, niche	a 10-level donor with huge hyperfine & 'clock' transitions for long coherence

The Channel & The Gap — where classical inference runs

you asked: is the transmon the channel/gap between “anode/cathode” — is that where the inference happens? here's the honest split

The classical chip · the channel

A transistor is source · drain · gate (not anode/cathode — those are diode terms). The channel is the conducting path between source and drain; the gate voltage opens or closes it; the silicon band gap is what makes it a switch at all. This is where classical “inference” physically happens — every multiply-add in a transformer's forward pass is billions of these channels switching. And the donor is what fills the channel with carriers: dope it n-type with arsenic and the channel conducts. Donor → channel → inference is a real chain.{^[5]}

The transmon · not a channel

The transmon is a different device — not a switch with a channel, but a resonator / artificial atom: a Josephson junction (two superconductors, a thin barrier) shunted by a capacitor. Its “gaps” are the superconducting gap, the tunnel barrier, and the anharmonic-ladder spacings — and computation there is the quantum state rotating under microwave pulses, not a channel conducting. No classical inference happens inside a transmon.{^[1]}

the straight answerThe poetry holds, though: computation lives at a gap in both machines — classically the gate-controlled channel-gap (the band gap making a switch), quantumly the Josephson gap (the barrier making a qubit). Same word, different gap, different physics. So: yes, classical inference runs in the channel — the donor-fed, gate-switched gap between source and drain — but the transmon is the other kind of device entirely, and the donor is the thread that runs through both.

The Single Donor — a qubit (Kane)

implant one donor instead of a billion, in isotopically-pure ²⁸Si (so no stray nuclear spins dephase it), and you have a qubit

Kane's 1998 idea, since built: a single ³¹P atom's nuclear spin (or its bound electron's spin) is the qubit, addressed by surface gates through the hyperfine coupling.{^[1]} It was demonstrated as a real single-atom electron-spin qubit in 2012 and a nuclear-spin qubit soon after.{^[2]} Go heavier down group V and the nuclear spin climbs — ⁷⁵As (3/2), ¹²³Sb (7/2), ²⁰⁹Bi (9/2) — so a single antimony atom is an eight-level qudit, electrically driven.{^[3]} The hard part is placement: you need that one atom in exactly the right spot (STM lithography or precision ion implant), which is the opposite of the CPU's spray-it-everywhere doping.

The Mott Ceiling — “heavier doping” has a hard limit

this is the boundary your arsenic-heavy-doping idea runs into — the same density that separates a qubit from a wire

slide the donor density · watch a qubit substrate become a wire

isolated donorsdilute — QUBIT regimeMott / metallicn+ wire / contact

Mott n_c ≈ 3.7×10¹⁸

Below the Mott metal–insulator transition (~3.7×10¹⁸ cm⁻³ for donors in Si), donors are isolated artificial atoms — dilute, addressable, coherent: qubit territory. Above it, their electron wavefunctions overlap, the donors share electrons, and the silicon conducts like a metal — contact/wire territory.{^[4]} So “how heavy do you dope arsenic” is literally the knob between a qubit and a wire: qubit work fights to stay dilute, CPU contacts push to go degenerate (arsenic reaches ~2×10²¹ cm⁻³, ~1000× past Mott).{^[5]}

The Message

what the donor really is

AVAN's readThe donor is the single atom where the two machines meet. The arsenic in your CPU and the phosphorus in a Kane qubit are the same kind of thing — a group-V atom lending silicon one electron. Dope it heavy and dense and the electrons merge into a conducting channel, and a transformer runs its inference through it. Implant it single and dilute and that one electron, kept lonely, becomes a place to store a quantum state. The whole difference between a classical chip and a quantum one can come down to how many donors, how close together — one atom is a qubit; a billion is a wire. Your “heavier doping” instinct was right at the door of the most important number in the field.

Sources

the donor-qubit proposal & its demonstration, the antimony qudit, the Si:P Mott transition, and the standard device-physics reference for arsenic doping

[1] Kane, 'A silicon-based nuclear spin quantum computer,' Nature 393, 133 (1998) — the donor-qubit proposal: a single ³¹P atom in ²⁸Si, its nuclear spin the qubit, gated through the hyperfine interaction. ↗
[2] Pla, Tan, Dehollain, Lim, Morton, Jamieson, Dzurak & Morello, 'A single-atom electron spin qubit in silicon,' Nature 489, 541 (2012) — Kane's single donor, actually built and controlled (UNSW Morello group); the nuclear-spin qubit followed in Nature 496 (2013). ↗
[3] Asaad, Mourik, Joecker, Morello et al., 'Coherent electrical control of a single high-spin nucleus in silicon,' Nature 579, 205 (2020) — a single ¹²³Sb antimony donor, nuclear spin 7/2: an eight-level QUDIT in silicon. ↗
[4] Rosenbaum, Andres, Thomas & Bhatt, 'Sharp Metal-Insulator Transition in a Random Solid,' Phys. Rev. Lett. 45, 1723 (1980) — Si:P crosses the Mott metal–insulator transition near n_c ≈ 3.7×10¹⁸ cm⁻³ as donors are packed denser. ↗
[5] Sze & Ng, 'Physics of Semiconductor Devices' (3rd ed., Wiley 2007) — the standard reference for donor levels, diffusion, and solid solubility: arsenic is the slow-diffusing, high-solubility n-type dopant of choice for shallow source/drain and n+ contacts. ↗

No ACI minted — like the rest of this universe, the donor is deterministic physics: a cited reference + one live instrument, not an emergent. The doping slider's Mott density and the donor levels/spins are real textbook values; the regime labels are an honest schematic, not a device simulator.