QEMData

QEMData is an electromagnetic-simulation dataset for superconducting quantum chip experiments. It links parameterized chip layouts to physically grounded simulation artifacts and derived circuit metrics. The dataset is described in the paper "QEMData: A Realistic Electromagnetic Simulation Dataset for Superconducting Quantum Chip Experiments".

This Hugging Face repository is a compact example release containing paired JSON annotations and GDSII layout files for 1-qubit chip instances.

Dataset Summary

Each sample contains:

• a structured JSON record with metadata, simulation provenance, capacitance extraction results, qubit/readout metrics, and status information
• a matching GDSII layout file with the same filename stem

The full QEMData dataset described in the paper contains 1,180 valid chip instances spanning 1- to 8-qubit superconducting chip designs generated with Qiskit Metal and characterized using Ansys HFSS eigenmode analysis and Q3D capacitance extraction.

Current Repository Contents

Item	Count	Notes
JSON records	10	One JSON annotation per chip instance
GDSII files	10	One layout file paired with each JSON record
Qubit count	1Q	Compact example subset
Status	10 completed	All included JSON records report status = completed

Directory Structure

QEMData/
|-- README.md
`-- samples/
    |-- <sample_id>.json
    `-- <sample_id>.gds

The JSON and GDS files share the same <sample_id> filename stem. In each JSON file, meta.gds_path points to the paired GDS file using a repository-relative path.

Full Dataset Composition From The Paper

The paper summarizes the full QEMData scale as follows:

Qubit count	Samples	Layout type	RO	BC
1Q	500	Single node	1	0
2Q	300	Linear chain	2	1
3Q	200	Linear chain	3	2
4Q	100	2x2 grid	4	4
5Q	50	T-shaped	5	4
6Q	20	2x3 grid	6	7
8Q	10	2x4 grid	8	10

RO denotes the number of readout resonators, and BC denotes the number of bus couplers used for inter-qubit coupling.

Data Fields

QEMData uses a JSON-centered schema. The main groups in this example release are:

Group	Contents
meta	Sample identifier, unit conventions, timestamps, solver/export tags, and relative GDS path
resonators	Readout frequency, Kerr/nonlinearity values, mode list, quality factors, linewidths, external coupling, and warnings
qubits	Junction parameters, effective capacitance, qubit frequency, anharmonicity, dispersive parameters, energy scales, and T1/T2 estimates
q3d	Capacitance matrix, node names, and external coupling capacitance Cin_fF
chip	Frequently used aggregate metrics, including readout_f_GHz, readout_kappa_over_2pi_MHz, readout_Cin_fF, qubit_f_GHz, chi_MHz, g_over_2pi_MHz, chi_over_kappa, and T1/T2 estimates
status	Execution result such as completed or failure diagnostics

The full paper schema also describes optional groups such as position and coupling for multi-qubit releases.

Example Metrics In This Release

Ranges across the 10 included samples:

Metric	Unit	Min	Max	Mean
chip.qubit_f_GHz	GHz	4.6302	4.8179	4.7117
chip.readout_f_GHz	GHz	5.7329	6.2131	6.0352
chip.chi_MHz	MHz	4.7507	7.7812	6.1942
chip.readout_kappa_over_2pi_MHz	MHz	5.3492	6.0108	5.7690
chip.readout_Cin_fF	fF	51.3663	52.2030	51.8743
chip.g_over_2pi_MHz	MHz	206.5190	294.3444	239.9768
chip.chi_over_kappa	dimensionless	0.8024	1.4547	1.0798
chip.T1_qubit_est_us	us	0.6623	1.0702	0.8279
chip.T2_qubit_est_us	us	1.3246	2.1404	1.6559

Core Artifacts In The Full QEMData Release

The paper describes the full release as containing the following artifact types:

Artifact	Description
Layout parameters	Geometry knobs and device parameters used to instantiate Qiskit Metal layouts
Layout exports	GDSII geometry exports and optional renderings for visual inspection
Raw EM outputs	Eigenmode outputs, mode lists, and capacitance extraction results
Derived metrics	Chip-level quantities derived from raw outputs, such as readout frequency, linewidth, coupling capacitance, coupling strength, dispersive shift, and readout quality ratios
Index files	CSV-style dataset indices for filtering, querying, summary statistics, and benchmark construction

This example repository keeps the paired JSON and GDSII files only.

Supported Tasks

The paper defines several benchmark and workflow settings:

Task	Type	Inputs and targets
Downstream workflow evaluation	Use case	Use an EM-derived JSON record and tuning knobs to evaluate calibration/readout indicators
Design-to-metric prediction	Regression	Predict physical metrics such as readout frequency, linewidth, dispersive shift, qubit frequency, and readout-quality ratio from design/layout features
Specification feasibility	Classification	Predict whether a design satisfies metric thresholds, for example a minimum readout-quality ratio
Inverse design	Retrieval	Given a target metric vector, retrieve top-K candidate chip designs by metric proximity or cost

For this compact example release, the most direct uses are schema inspection, GDS/JSON loading tests, feature extraction, and small-scale design-to-metric prototyping.

Loading Example

from pathlib import Path
import json

root = Path("QEMData")
for json_path in sorted((root / "samples").glob("*.json")):
    with json_path.open("r", encoding="utf-8") as f:
        record = json.load(f)

    sample_id = record["meta"]["sample_id"]
    gds_path = root / record["meta"]["gds_path"]
    qubit_f = record["chip"]["qubit_f_GHz"]
    readout_f = record["chip"]["readout_f_GHz"]

    print(sample_id, gds_path, qubit_f, readout_f)

Citation

If you use this dataset, please cite the accompanying paper:

@misc{qemdata2026,
  title = {QEMData: A Realistic Electromagnetic Simulation Dataset for Superconducting Quantum Chip Experiments},
  year = {2026},
  note = {Dataset paper}
}

Notes

• GDSII files are binary layout artifacts, so this repository stores paired raw files rather than train.jsonl or Parquet splits.
• The license is marked as other until the final dataset license is specified.
• The current repository is an example subset; the paper's full release includes additional qubit counts, larger sample volume, CSV indices, and optional raw solver/rendering artifacts.