One of the interesting things about qubits is that… well, one is not always sure where they are. As such, this is not surprising - quantum computing is still in its early development phase. Think back to the time when transistors were being researched, and bulky test equipment was used to determine if those transistors were actually functioning as expected. It’s kind of the same nowadays for quantum bits.
The idea is that a lot of tuning is needed to find out if we are actually “talking” to a qubit. I wrote “talking”, and probably whispering would have been a better description, considering that qubits are very sensitive to noise. But, in actual fact, “talking to qubits” is conventionally expressed as “control and read out”.
The idea is simple: First, you whisper to the qubit at a given frequency, and you see if it reacts by reading out the “qubit”. Sometimes it does, sometimes it does not, meaning that finding a qubit means doing a lot of tuning - and that’s what quantum physicist and experimentalists spend a large portion of their time on nowadays.
Credits: Qubit readout by the Josephson parametric oscillator
What is more interesting is that, once one qubit has been “found”, the game goes on, having to find one qubit among a set of qubits located in a processor, which we will refer to as a “QPU” or quantum processing unit.
Credits: Neils Bohr Institute
At first glance, it looks simple, but think about having to whisper to all of those qubits at the same time, without creating interference between the qubit signals. There is a lot to say here, but this goes beyond the scope of this blog. Should you be interested, you can start with the Neural Network-Based Frequency Optimization for Superconducting Quantum Chips paper.
What we need to remember here is that, before creating a quantum computer, we must first be able to manufacture QPUs with predictable performance in terms of Qubit control and readout. A lot is happening in this domain nowadays, so we can definitely expect this challenge to be solved within the next five years.
Meanwhile, there is something fantastic about the tuning work that needs to be done—and where AI/ML comes into play. That’s something we’ll deep dive into in the next article. So, stay tunned!