Skip to content

Latest commit

 

History

History
29 lines (19 loc) · 2.52 KB

ReadMe.md

File metadata and controls

29 lines (19 loc) · 2.52 KB

$d=3 \quad X-$ memory rotated surface code

This repo implements Rust code to simulate the experiments for the $d=3$ rotated surface code with $X-$ memory checks from this paper by Evangelia Takou et al.. The algorithm in this code is from the paper's author who shared it with me via private correspondence. I mostly implemented it both as an exercise to better understand the paper for a journal club, understand coherent errors in DEMs as well as learn how to use Claude better 🤖

Acknowledgements: Massive thanks to the author E. Takou for explaining their paper to me. This code is more of a learning exercise I did based on conversations with the author.

[Rust] : Building the repo and simulating circuit with coherent errors

This is the primary code used to generate data files for the surface code. I also added some additional functionality for custom experiments that is not present in the Python code (see below). To set up:

  • Install Rust and Cargo using rustup
  • Within the directory, run cargo run --release
  • The Rust source code is within the /src/ folder, with tests in /tests/.

This should generate the x_ancilla_probs.csv file.

[Python] : Building the repo and simulating circuit with coherent errors

The Python code for the surface code was translated from Rust to Python using Claude.

  • Have a Python environment with Numpy and pytest installed.
  • All Python code and tests are under /python/ folder.

Building the DEM

Use the notebook in analysis notebooks/First step : building the DEM.ipynb as an entry point. This notebook I wrote reads the x_ancilla_probs.csv file, and builds the DEM by inferring the probabilities $p_{ij}$. The notebook also translates these probabilities into a effective coherent angle $\tilde{\theta}_{ij}$ using $p_{ij}=\sin^2 \tilde{\theta}_{ij}$. Note that for the example main.rs and .csv provided in this repo, $\theta=0.1\pi$ for the coherent errors, so we expect most $\theta_{ij}$ to be equal to $\theta$, except for the weight 4 checks where 2 data qubits contribute to one DEM edge. For those, we expect $\theta_{ij} = 2\theta$.

Comparison to Repetition Code

I also wrote a simulation for the $X-$ memory repetition code (Fig 1a of Takou et.al paper) in dem_repetition.ipynb. In this notebook, I do an actual circuit simulation by building a cirq circuit with data and ancilla qubits and using the results to construct the DEM.

Math Notes

For some math and derivations I did on why, see Math Notes.md.