transversal gate

In plain terms

Protecting a quantum state is only half the job, as you also have to compute with it, and the obvious way of doing that destroys the protection. If a gate lets two qubits of the same encoded block interact, then one faulty component can corrupt both, and a code built to survive single errors is suddenly facing a double error it can't fix.

A transversal gate avoids the problem by never letting the qubits of a block meet. You apply the same small gate to each of the seven physical qubits of a block, one at a time, and the effect on the encoded qubit is the gate you wanted. Faults stay put, because each physical gate touches one qubit and can only spoil that one, so the block comes out of the operation with at most a single error and the recovery procedure handles it as usual.

Not every code admits transversal gates for every operation, and the Steane code is prized because it admits so many. Applying a Hadamard gate to all seven qubits performs a Hadamard on the encoded qubit, and applying a controlled-not gate between corresponding qubits of two blocks performs a controlled-not between the two encoded qubits. What no code can do is offer a transversal version of every gate, which is why fault-tolerant computers spend their budget on magic states for the one gate left over.

In CSD

The corpus proves the Steane code's transversal gates as operator identities and then composes them with its recovery theorem, which is what turns a claim about ideal gates into a fault-tolerance statement. Applying the Hadamard gate to each of the seven qubits conjugates the code projector to itself, exchanges the encoded bit flip and the encoded phase flip, and so implements the encoded Hadamard. The two-block controlled- not is proved in the forward direction, on encoded states rather than on operators, which needed a coset identity about the encoder rather than any amplitude calculation.

The fault model is then the honest one. A gadget is the ideal transversal gate followed by a single-qubit fault somewhere in the block, and the theorem says one recovery channel, fixed in advance and independent of where the fault landed, returns the encoded state. Composing that with the threshold theorem gives the circuit-level statement for these gates. Together the Hadamard, the phase gate and the two-block controlled-not cover the Steane code's Clifford generating set, so the gates a fault-tolerant machine performs directly are all present, and the remaining gate is the one magic-state distillation supplies.

Mathematically

hadTransversal is the seven-fold tensor power of the Hadamard matrix, built through permMat and the block tensor. hadTransversal_conj_pauliMat is the Clifford relation H X^u Z^v H = (-1)^(u dot v) X^v Z^u at the level of the whole register, and the Steane code's self-duality, proved as bdot_steaneA_steaneB together with the half-swap involution swapHalves, turns that into hadTransversal_conj_steaneProj: the transversal Hadamard fixes the code projector. Conjugating the encoded operators (hadTransversal_conj_logicalX, hadTransversal_conj_logicalZ) exchanges them, which is the encoded Hadamard.

steane_faultyHadamard_recovery states the fault-tolerant form: for the gadget that applies the transversal Hadamard and then an arbitrary single-qubit operator on one of the seven qubits, one recovery channel independent of the fault restores the encoded state, through the encoder form of the Knill-Laflamme conditions. The two-block controlled-not is cnotT_mul_pairEnc in SteaneTransversalCNOT.lean, proved from cosetAmp_shift, the fact that adding a vector of one coset carries the coset of a second onto the coset of their sum; its fault-tolerant form is steane_faultyCNOT_recovery. Honest scope: the syndrome-extraction circuit itself is not fault-tolerant here, as measurement gadgets and the extended rectangle are open rows.

The name

Transversal gates are the central device of Peter Shor's 1996 fault-tolerance paper, where acting qubitwise across a block is what makes error correction compatible with computation. Andrew Steane's seven-qubit code is self-dual, which is the property that makes its transversal gates unusually generous.

Bryan Eastin and Emanuel Knill proved in 2009 that no code with local error correction admits a transversal universal gate set, so the gap this vocabulary describes is permanent rather than a failure of ingenuity.

Related entries

• Steane codeSeven qubits protect one, as the Steane code was the first to correct any single- qubit error by using the same classical Hamming code twice, once for bit flips and once for phase flips.
• Knill-Laflamme conditionsQuantum error correction has an exact test, as a set of errors is correctable if and only if the code can't tell those errors apart, and that condition is a piece of algebra you can check before building any repair.
• threshold theoremThe threshold theorem says imperfect components suffice, as long as each one fails below a fixed rate you can make a quantum computation as reliable as you like by encoding it inside itself repeatedly, and the price is only a power of a logarithm.
• magic stateA magic state is the fuel of quantum advantage, a single special state that, consumed by otherwise classically trackable Clifford circuits, unlocks the one gate the Gottesman-Knill theorem can't simulate.
• Clifford plus TClifford plus T is the standard universal gate set, as the Clifford gates alone are classically simulable and adding the single eighth-turn phase gate makes every unitary reachable to any accuracy you ask for.