Binomial code

Binomial bosonic codes encode information in finite superpositions of Fock states weighted by binomial coefficients. They satisfy quantum error-correction conditions for selected orders of photon loss, gain, and dephasing while retaining finite mean energy.

Codewords occupy number states spaced by S+1 and split even and odd binomial terms between logical states. Common designs take S=L+G with truncation large enough for loss L, gain G, and dephasing order D; explicit Fock support and normalization identify the code more precisely than the label.

The designed error set contains finite powers of annihilation, creation, and number operators. Under continuous-time Lindblad noise, finite-order correction is exact or approximate in a short-time expansion and does not mean that arbitrarily many photon jumps are correctable.

One mode commonly encodes one logical qubit, but meaningful costs are maximum occupation, mean photon number, Fock-support width, and nonlinear control. Shorthands such as the 0-2-4 code should be accompanied by codewords because conventions may target different error sets.

Number-dependent phases, selective pulses, and SNAP-like control implement logical operations within finite Fock support. Gates that leave the designed support or mix uncorrected orders require leakage detection or recovery; “single mode” does not automatically define transversality.

Measure photon number modulo S+1, or an equivalent rotation syndrome, to identify jump classes and apply conditional recovery. Optimal decoding can also use jump timing, analog readout, and priors over different error sequences.

Dispersive syndrome ancillas, selective pulses, and reset create backaction and phase diffusion. A fault-tolerance analysis must show that one ancilla or control fault does not generate a high-order oscillator error outside the correctable set.

Superconducting-cavity experiments have demonstrated binomial codewords, syndrome measurements, and correction gain. Engineering records should include preparation and recovery pulses, mean energy, the break-even baseline, and any syndrome postselection.

Compared with cat codes, binomial codes have finite Fock support and an algebraically selected finite-order error set; unlike GKP, they do not use an infinite periodic phase-space lattice. These representation differences determine energy, gate, and syndrome resources.

Do not interpret L, G, and D as a qubit-code distance; they specify orders in a designed operator error set. Comparisons require the same loss channel, correction cadence, energy constraint, and assumptions about ideal recovery.