The Exercise
The unitary of a quantum circuit is the product of the matrices of its gates. Several of the standard gates are a power of times a matrix whose entries are with and integers: the Hadamard gate is times a matrix of entries , and the phase gate needs no factor. The exercise is to show that this form survives multiplication, so that a circuit of such gates has a unitary for some natural number , with of the same kind, and that the gate, whose matrix holds , does not have it. Circuits that use the gate keep the form once the entries of may also contain the powers of .
Two Rings of Numbers
The Gaussian integers are the complex numbers with and integers.✦ The primitive eighth root of unity is (1)✦ whose square is . The cyclotomic integers are the complex numbers (2)✦
The Gate and the Scaled Form
The gate is the gate on one qubit whose matrix in the computational basis is (3)✦
Let be a subring of the complex numbers. A complex matrix has a scaled form over if there are a natural number and an matrix with entries in such that (4)✦
The Claim
The library states the exercise as three theorems.
Theorem 1. Let be square complex matrices of one size, each with a scaled form over . Then there are a natural number and a matrix of the same size, every entry of which is for some integers , such that (5)✦
Theorem 2. Let be square complex matrices of one size, each with a scaled form over . Then there are a natural number and a matrix of the same size, every entry of which is for some integers and , such that (6)✦
Theorem 3. The gate has no scaled form over : whatever the natural number and the matrix with entries in , (7)✦
What Is Checked
The claims are stated over the scaled form (4) of a matrix over a subring of the complex numbers, taken over the Gaussian integers and over the cyclotomic integers (2), which are built on the eighth root of unity (1); the third is about the gate (3).
About these constructions the library proves that both sets of numbers are subrings of the complex numbers: each contains and and is closed under sums and negatives, and products stay in the set by (8) for integers , and by (9) for Gaussian integers .