Initial version of a quantum circuit editor.

Overview

Here’s the result so far:

The key detail is the two \( f(x) \) blocks. This circuit performs classical computing on the measured qubits mid-circuit to allow for additional insertion of mid-circuit gates.

Why care?

Quantum Error Correction! I’ll go more into detail later, but essentially to do QEC, we need a way to make mid circuit measurements and update the circuit for errors.

Why care about Quantum Error Correction? Quantum computers are super sensitive to noise (see the density operator post for more info. We need Quantum Error Correction to build usable Quantum Computers.

What went well?

• The circuit is built entirely out of HTML!
  • This is a little overkill… I probably should have gone with canvas, but it’s neat that you can build something like that with just HTML. I didn’t even use z-index. Everything is explicitly checked and added. (Lots of flexbox). Technically the LaTeX might be using canvas, not sure.
• Mid-circuit measurements works! This is something my last quantum simulator didn’t have. The ability to measure a single qubit during mid execution is pretty cool (and it was easier than I thought. Just tensor some projection state between a bunch of identity matrixes)
• For building gates like CCCNOTCCC, I used to have some pretty complex algorithm, but now that I’m one quantum smarter, it turns out I can do it a little simpler by being smart about the projection operators for the conditionals.
• Adding classical functions “works”:

c1.appendFunction((classicalBits: number[]): Moment[] => {
  let out = [];
  if (classicalBits[0]) {
    out.push(new Moment(XGate, [0], [], [], 2, 2));
  }
  if (classicalBits[1]) {
    out.push(new Moment(XGate, [1], [], [], 2, 2));
  }

  return out;
});

It’s a little ugly, but I’ll clean that up in the future. Each moment is a slice in time, with the gate, the gateIndexes, controlIndexes, classicalIndexes, and then circuit size.

What didn’t go well?

• Well, it went fine. But building circuits in HTML is a little bit of a pain. I’m holding off on adding barriers and muli-wire gates because I don’t want to touch it. But I will one day. I think overall I prefer HTML to canvas (for portability and interoperability)

Future

I have an ever increasing list of todo’s. But I think the main ones are:

• Start implementing different QEC schemes! I think I have all the pieces.
• Allow the user to turn on the noise models as seen in the RB playground to simulate errors
• Implement drag and drop (another thing I know how to do, I’m just dragging my feet on since it’s a pain)
• More than like 8 qubits is pretty slow. Probably need to use one of the GPU matrix libraries instead of rolling my own matrix math.
• Deploy!

Hopefully within a week or two people can use the circuit build to play with different error correction schemes!