RydSFF

Repository to run real experiments, simulated experiments, and purely numerical calculations for QuEra's Aquila device for our paper (arXiv:https://arxiv.org/abs/2604.24854)

This code is not yet refined enough to contain a tutorial for using AWS Braket, so if you have any questions about that in particular or want to discuss other parts of the code, please contact me at oscar.scholin@physics.ox.ac.uk or orsa2020@mymail.pomona.edu.

version requirements:

For AWS, we enforce Python 3.10.20 (this is handled in the S3 environment creation during hybrid job; otherwise we run into errors submitting programs to Aquila). For classical backend simulation, we use Python 3.12.2.

File descriptions:

Experimental data plotting

• main_expt_plot.py
  • I use a hash code to reference the input configuration to bloqade -- this is denoted by files with the name structure task_{uid}.json, which are stored inside dir_main/tasks/* subdirectory. Tasks have corresponding stem_tasks which record the overall structure of the hamiltonian but not the specific h_i values or seq_ls that are actually executed.
  • Task and stem ids can be generated using _get_task
  • run_for_N_protocol is the lowest level of the plotting functions -- it calls process_rbp.py after reading the tasks
  • time_disorder_protocol is the next-highest level plotting function we use that actually creates the figures
  • the main function to call that specifies all the tasks and experimental timestamps to pass to time_disorder_protocol is main_expt_figs which takes as arguments opt=='sim' for simulated bloqade and opt=='expt' for experimental tasks.
  • to run parallelised T2star simulations use main_T2star_parallel and then compile_T2star to recombine the data into the main directory

AWS:

• task_submission_main.py - schedule_run_all_chunks is the main, high level function to run experiments on Aquila. automatically checks to avoid submitting during the calibration period so all data can be run together - First in an AWS environment, run option=="RUN" with _FORCE_TRY_HYBRID == True to initialise hybrid job and submit quantum tasks via braket. Then, run option=="SAVE" but _FORCE_TRY_HYBRID == True while still in the AWS environment to fetch the .json and return results as .npy. The directory now may be zipped and downloaded to your local machine for further processing. - requirements_revised.txt is used to setup the
• helpers
  • cancel_all_tasks.py
    • terminates all actively queued tasks in amazon braket
  • cancel_task_arn.py
    • terminates specific hybrid task given arn
  • make_tasks_table.py
    • compile results of quantum tasks submitted to braket
  • parse_majd_calib.py
    • compile QuEra calibrations and compare against our own
  • manual_download.py
    • download the S3 bucket containing the hybrid task result
  • chain_benchmark.py, diagnose_driver.py are helper functions for the Rabi oscillation calibration which is called in run_for_N_protocol

Middle level functions

• master_params_rbp.py
  • called by task_submission_main.py: contains funcs to create task uids (read_expt_task hashes the input parameters to the Hamiltonian and other input configs; the hashing is done by QuEraToolbox.expt_file_manager.py), and to execute these tasks on simulated backend or on AWS hardware (execute_bloqade_task)
• process_rbp.py
  • process_bitstrings is the main function of this file: tasks a list of h_ls and seq_ls extracted from the task by functions in main_expt_plot (run_for_N_protocol); first calls the respective functions to gather simulated or experimental bitstrings and then in lines 1965 - 1974 will helper functions to actually calculate the purity (est_purity). Includes support for T2* uncertainty (bloqade simulator and qutip) and T2 dephasing noise (qutip only).
    • get_all_qutip_probs is the main function for qutip emulation tasks
    • get_all_single_hams_rand is the main function for bloqade (emulation and simulation) tasks; calls get_single_ham_rand for individual h_i list, which calls
  • numerical computes numerically the second order renyi entropy including the time dependent ramp structure of the experiment but excluding the randomised measurements

QuEraToolbox -- low-level processing and execution helpers

• random_bp_prep.py
  • lowest level file to run experiments (or emulate) on Aquila. (!!!!!)
  • expt_run is the main function of this file: data is either loaded if the corresponding files exist (either the .json from Braket or the .npy from our processing) or the program is compiled using compile_program_oneU which gets the initial ramp up, plateau, ramp down and compile_rand_seq which intersplices the randomised phi-flip sequence.
  • random_bp_qutip.py emulates this function but also includes option for T2 modeling
• expt_file_manager.py
  • hashes an input dict with json-friendly keys and values through the object ExptStore
• hamiltonian.py
  • defines the Aquila hamiltonian in QuTip with helpers for ramp structures -- main function is the function factory drive_main
• helper_rbp.py
  • contains the functions to estimate purity using the randomised measurement toolbox formula (est_purity) and apply readout error to numerical probability vectors (apply_readout_channel)

Illustrative figures

• fig_supplemental_signatures.py
  • all figures shown in the Supplemental Information, which focus on the ergodicity to localisation transition
• fig1_single_qubit_chain.py
  • illustrative figure to show the growth of entanglement for Fig 1 in paper
• fig2_randmeas.py
  • Fig 2 in the paper to illustrate the behavior of the $phi$-quench gates at single qubit level.
• appendix_spectrum.py
  • computes the median eigenvalue as a function of the mean local detuning

Data:

DOI 10.5287/ora-dow5oapxe contains all the data needed to reproduce the figures in the main text of the paper. Due to upload size constraints we have to separate the main directory paper_main_data into two .zip files: data.zip and paper_main_data_rest.zip.

Inside data/, the following files can be found:

• full_expt_{task_id}_{timestamp}: experimental data from Aquila for the hash identifier task_id (see tasks/task_{task_id}.json) and the integer timestamp
• bloqade-sim-no-rc_{task_id}_{0}: simulated data for the task task_id for the noiseless emulation of experimental protocol
• avg_{uid}.npy and result_{uid}.npy: simulated data for the task task_id for noisy emulation of experimental protocol including readout error and $T_2^*$ effects. Note, the meaning of the uid can be determined by accessing the corresponding {uid}.json file in paper_main_data_rest/combos/

In paper_main_data_rest are subdirectories that specify: - the tasks/* for all tasks and stem tasks needed for the experiments, as called in main_expt_plot.py - the combos/* which explain the input dict to generate the hashes - the supporting mdata and ndata for the plots, - and the results/* themselves

expt_tasks.zip contains the expt_tasks directory needed to run main_T2star_parallel