With 20 Qubits, the H1-1 Quantum Runs More Complex Algorithm

Quantinuum’s flagship quantum computer, the H1-1 is now running on 20 qubits. We sat down with Brian Neyenhuis, Quantinuum’s director of commercial operations to ask him about the 20-qubit upgrade, some of the technical details, and how this launch paves the way for scaling trapped-ion quantum computers in the future.

What are some of the key upgrades made to the H1-1 machine?

The biggest, or maybe the most notable, is that we expanded the number of fully connected qubits from 12 to 20. That is a significant increase and the most qubits we’ve added to an existing machine. Last year, we added two fully connected qubits to the 10 qubits H1-1 already had. It was a major accomplishment at the time. Now, that seems easy compared to this upgrade because for us, it is not as simple as adding qubits.

To add eight more qubits and maintain all-to-all connectivity, we upgraded the optics that deliver the light used to control our qubits. Previously, we were only delivering the light needed to complete quantum gates to three different regions of the trap, which we call gate zones. Now we can address all five zones in our trap simultaneously. This enables us to complete more single-qubit or two-qubit gates in parallel, which means users can run complex algorithms without experiencing a slowdown.

How does this compare to previous hardware upgrades?

This one was significantly more involved than previous upgrades. Although we didn’t modify the trap at the heart of the computer or the vacuum chamber and cryostat that enclose it, we redesigned the entire optical delivery system. This was necessary so as not to deliver light to more regions of the trap, but also to improve stability.

Increasing the size and complexity of the machine without improving the stability would be a recipe for disaster. Because we were able to improve the stability, we were able to add more qubits without sacrificing performance or key features our users expect such as all-to-all connectivity, high single and two-qubit gate fidelities, and mid-circuit measurement.

Why is the increase in zones significant?

The gate zones are where all the interesting quantum stuff happens. More zones allow us to run more quantum operations in parallel, allowing for faster, more complex algorithms.

What's the connection between more zones and more qubits?

Having more gate zones allows us to use more qubits in an efficient way.

Because we can do all these operations in five different locations in parallel, it finally makes sense to put more qubits into the trap. We could have loaded more qubits into earlier versions of the system, but without additional gate zones, it doesn’t make a lot of sense. In fact, doing that would create a bottleneck with qubits waiting for their turn to do a two-qubit gate, which

then slows down an algorithm. Now, we can do five quantum gates in parallel, which allows us to run more complex algorithms without sacrificing speed.

Twenty qubits are probably where this generation of traps ends. There is a possibility to add a handful more, but it feels like this is probably the most efficient number for these H1 systems due to layout of the trap. But future generations, some of which are already trapping ions in the lab today, will use even more qubits and with the same or better efficiency.

What is the “ion dance”?

In the QCCD architecture, trapped ions are easy to move around. By applying the right set of voltages to the trap—a small, electrode-filled device that holds qubits in place—we can arbitrarily rearrange the chain of qubits so any qubit can pair with any other and perform a quantum gate. So, you can think of any algorithm as a set of steps where we shuffle all the qubits to pair them up for the next set of gates, move them into the gate zones, and then shuffle them again to set them up for the next set of gates. The ions “dance” across the trap moving from partner to partner to execute a quantum circuit.

Some circuits, like quantum volume circuits, are densely packed, meaning that every possible pair wants to do a gate at each step of the circuit. Other circuits are very loosely packed, meaning you can only do a few gates in parallel before moving on to the next slice because you need to reuse one of those qubits with a different partner.

Although this dance may sound complicated, it makes it very easy to program our quantum computer. A user sends us a time-ordered set of gates without having to think about the layout of the qubits, and our compiler figures out how to pair up the appropriate qubits to make it happen. You don't have to worry about which ones are next to each other because any pair of qubits is equal to all the others. And, at any step, we can completely rearrange this chain and put any two qubits next to each other.

It’s like a square dance where someone calls out directions to the dancers.

Anything else in the works for Quantinuum’s hardware this year?

We will continue to work with our customers to improve our system performance and their overall experience. One of the reasons we have a commercial system now is to allow our customers to program their algorithms on a real machine. They're dealing with all the constraints of real quantum hardware. They're pushing on their algorithms while we're pushing on the hardware, to get the fastest iterations.

As they learn new things about their algorithm, we learn what the most important things are to improve. And we work on those. We are learning a lot from our customers, and they are learning a lot by running on our hardware.