Duke University and IonQ Develop New Types of N-Qubit Gates

Diagram of Duke/IonQ Setup for Implementing N-Qubit Gates. Credit: IonQ

Quantum gates are typically either single qubit gates or 2-qubit gates. The most common 2-qubit gates are things like CNOT, CONTROL-Z (CZ), or a Mølmer-Sørensen gate commonly used in ion trapped based computers. These gates operate such that the output state is a function of the both the input qubits. However, sometimes one wants to use something more complex, such as a three input Toffoli Gate (also called CCNOT) which will invert the state of the third qubit only if the states of both the first and second qubits are a logical "1". The Toffoli gate can be quite useful for implementing the classical AND logic function and can be used to implement such things as a classical digital adder circuit.

However, it is not quite so easy to implement a three input Toffoli gate because it needs to be synthesized from smaller single qubit and 2-qubit gates as shown in the picture below. This is a problem because each level adds to the error rate and the resulting error rate of the total circuit will be much worse.

Synthesis of a Toffoli Gate (Left) with 15 Individual One and Two Qubit Gates (Right)

What Duke and IonQ have developed is a way to implement various types of N-Qubit gates, including the Toffoli gate, natively in one level. Using these gates can potentially simplify and speed up the function while simultaneously offering a reduced error versus the synthesis approach shown above. And this approach is not just limited to three qubits. It can potentially be used on all the qubits in the ion trap chain. This can potentially be a significant benefit to programmers, depending upon the particular algorithm they are implementing.

Duke and IonQ claim their approach is only possible with their particular ion trap architecture because of a multi-qubit communication bus they are using. Additional information on this can be found in a news release provided by IonQ which is available here and also in an arXiv preprint paper which has been posted here.

February 10, 2022