Flattening Quantum Circuits: “Planus!”
Quantum Art is redefining all-to-all connectivity.
He didn’t want to do it, but Harry Potter is an old friend of The Quantum Dragon, and fire-breathing dragons can be rather persuasive. You see, after reading “Quantum Art Validates a Scalable Path to Fault-Tolerant Quantum Computing Using Multi-Qubit Gates,” The Quantum Dragon wanted to focus on something that’s not obvious from the title: significant circuit compression.
One Step
The approach is different from others you may have read about. Quantum Art’s trapped ion qubits enjoy all-to-all connectivity, which is not unique, but the “magic” here is being able to apply multi-qubit gates, any ion to any ion, all at the same time. If you want to use 50 qubits, for example, they can be entangled in one step through the simultaneous execution of 1,200 gates.
This is in stark contrast to superconducting quantum computers, in particular. You put the zeroth qubit into superposition and then entangle it with a neighboring qubit. Depending on connectivity, those 2 qubits can then be entangled with one neighboring qubit each. Further depending on connectivity, those 3-4 qubits can then be entangled with one more neighboring qubit each. Let’s gloss over error rates and coherence for a moment, and simply note how many steps this requires compared to only one for Quantum Art.
Harness the Beast
The “beast” should refer to The Quantum Dragon, but he’s in so much agony right now that he’s allowing Quantum Art to apply the term to noise. Part of what makes it possible to do what Quantum Art is doing is that they’re paying attention to how noise propagates during execution. Since the noise channel follows the gate structure, they know where the noise is going, they control how it propagates, and then they mitigate it with the right decoders.
Decompose and Flatten
Not every gate can be executed in one step. The familiar Toffoli gate, aka controlled-controlled-NOT gate, aka CCNOT, is normally decomposed into a dozen or so timesteps, which can be greatly worsened by poor qubit connectivity. While Quantum Art can’t do anything about the sequence of operations, it can do something about parallelization. A Quantum Art Toffoli may require only 3 timesteps in comparison.
The key is preserving performance while compressing circuits by orders of magnitude and while working with error correcting codes.
Executive Summary
There’s quite a bit to unpack here:
- Entangling all qubit pairs simultaneously reduces n entangling layers of a quantum circuit to only 1 layer. Compression is to the maximum level that physics allows.
- Harnessing the “beast,” the noise, helps to mitigate it.
- Parallelization results in fast multi-qubit gates and fast circuits.
- The architecture transfers information, not ions, so there are no shuttling-related delays.
- Whereas superconducting qubits are better known for speed and trapped ions are better known for coherence, this approach closes the gap at the circuit level as the SWAP gates required for limited connectivity slow superconducting systems down.
- The “magic” is a combination of architecture (solving advanced nonlinear equations for fast gates) + control (multi-tone) + compiler (multi-qubit gates).
- Having more degrees of freedom allows Quantum Art to do tricks, like the aforementioned schmooshing of the Toffoli gate.
- This isn’t all! For massive circuits, they say more can be done.
- This approach is aligned with quantum error correction codes.
For more information on Quantum Art’s architecture, check out “The Grand Staircase of Quantum Computing.”
