Positronium Sensing

Repository for the manuscript Superconducting antiqubits achieve optimal phase estimation via unitary inversion.

This repository provides a transparent and reproducible data-processing and analysis pipeline, starting from the released probability datasets (CSV/HDF5) and ending with the figure exports used in the manuscript.

This repository is organized for readers who want to reproduce the analysis end-to-end from the included CSV/HDF5 artifacts:

• notebooks/ contains the step-by-step analysis notebooks (numbered in recommended run order).
• data/ contains the minimal experimental probability datasets used by the notebooks.
• data_analysis/ contains derived alpha, P, FI CSV products produced by the notebooks.
• si_figures/ contains supplementary-information figure exports produced by the notebooks.
• main_text_figures/ contains figure exports intended for the manuscript main text (for example, Fig. 3).
• figures/ contains small static assets embedded in notebooks/documentation.

Analysis overview

The major advantage demonstrated in this work relies on platform-specific unitary inversion (which reverts the gyromagnetic ratio). While the singlet state is utilized, the platform-specific unitary inversion is the key enabler.

Terminology used throughout the repo

• t: experimental time samples (ns in the released CSVs).
• α (alpha): rotation angle (radians) obtained from a global fit, used as the analysis x-axis.
• P(α): probability of a measurement outcome as a function of α.
• FI(α): (classical) Fisher information computed from P(α) using local fits.

Data processing flow

%%{init: {'theme': 'neutral'} }%%
flowchart TD
    Raw["Raw experimental readout<br/>probability distribution"]

    Raw -->|Corrections for<br/>readout fidelity| ReadoutCorrected["Readout-fidelity-corrected data"]

    Raw --> RawData["Raw data"]

    ReadoutCorrected -->|Corrections for<br/>entangling gate fidelity| GateCorrected["Gate-fidelity-corrected data"]

Loading

The notebooks work with three related probability “styles” for the singlet dataset:

• readout-fidelity-corrected data (_readout_corrected in filenames): corrected for measurement assignment errors.
• raw data (_raw in filenames): raw uncorrected measurement probabilities (obtained by undoing readout correction).
• readout-and-gate-corrected data (_readout_and_gate_corrected in filenames): singlet curves corrected for both readout and entangling-gate errors (provided as derived CSVs and used for figure assembly).

Data-processing workflow (high level)

%%{init: {'theme': 'neutral'} }%%
flowchart LR
  subgraph Inputs
    D_csv["CSVs in `data/`<br/>(time samples and populations)"]
    D_h5["HDF5 in `data/`<br/>(ADC demo + integration windows)"]
  end

  subgraph "FI Analysis Pipeline (Notebooks 03–06)"
    L["Load probability curves P(t)"]
    A["Map time → rotation angle α<br/>(global decaying-cosine fit)"]
    F["Estimate local derivative dP/dα<br/>(sliding-window fits in α)"]
    I["Compute classical Fisher information<br/>FI(α) = (dP/dα)² / (P(1−P))"]
    X["Export `data_analysis/*_alpha_P_FI_*.csv`"]
    P["Assemble manuscript/SI panels<br/>(common α grid + mean across axes)"]
    O_si["Write `si_figures/` exports"]
    O_main["Write `main_text_figures/` exports"]
  end

  subgraph RWD[" "]
    direction TB
    WTitle["Readout Integration Weights Demo<br/>(Notebooks 07 + A01)"]
    style WTitle fill:transparent,stroke:transparent,color:#111

    subgraph RWDFLOW[" "]
      direction LR
      W1["Windowed downsampling<br/>(rectangular integration windows)"]
      W2["Compute integration weights<br/>(Linear Discriminant Analysis)"]
      W3["Write `si_figures/integration_weights*`"]
      W1 --> W2 --> W3
    end
    style RWDFLOW fill:transparent,stroke:transparent
  end

  D_csv --> L --> A --> F --> I --> X --> P --> O_si
  P --> O_main
  D_h5 --> W1

  RC["Optional: raw extraction<br/>(apply readout confusion matrices)"]
  GC["Optional: gate-corrected singlet curves<br/>(used for figure assembly)"]
  RC -.-> L
  GC -.-> L

Loading

Contents

I. Experimental data (data/)

See data/README.md for file descriptions.

II. Notebooks (notebooks/)

Run the notebooks in numeric order:

1. notebooks/01_z_gate_curves_fig2cd.ipynb
2. notebooks/02_ac_stark_shift_sweeps_fig2ef.ipynb
3. notebooks/03_singlet_readout_corrected_fi_analysis.ipynb
4. notebooks/04_singlet_raw_fi_analysis.ipynb
5. notebooks/05_singlet_gate_corrected_fi_analysis.ipynb
6. notebooks/06_combined_si_figures_and_fig3.ipynb
7. notebooks/07_readout_integration_weights_demo.ipynb
8. notebooks/08_readout_fidelity_correction_ibu_demo.ipynb

Optional training-set notebook:

• notebooks/A01_integration_weights_demo_training_data.ipynb

Entrypoint notebook (overview + links):

• positronium_sensing.ipynb

Previews

Main Text Figures

• Fig 2c,d: Z-gate rotation curves

• Fig 2e,f: AC Stark shift sweeps

• Fig 3: Singlet vs. Entanglement-free

Supplementary Information Figures & Demos

• Rectangular windows for calculating the integration weights

• Integration weights

• Readout fidelity correction

Requirements

See requirements.txt.

Running in Binder

Click the badge above to launch Binder directly into positronium_sensing.ipynb.

Running locally

git clone https://github.com/murchlab/positronium-sensing.git
cd positronium-sensing
pip install -r requirements.txt

License

This project is released under the MIT License. See LICENSE.

Cite this work

@article{positronium2025,
  title         = {Superconducting antiqubits achieve optimal phase estimation via unitary inversion},
  author        = {Song, Xingrui and Borjigin, Surihan Sean and Salvati, Flavio and Wang, Yu-Xin and Yunger Halpern, Nicole and Arvidsson-Shukur, David R. M. and Murch, Kater},
  journal       = {arXiv},
  eprint        = {2506.04315},
  primaryClass  = {quant-ph},
  year          = {2025},
  doi           = {10.48550/arXiv.2506.04315},
  url           = {https://arxiv.org/abs/2506.04315}
}