SQuADDS/SQuADDS_Layouts

<center> <img src="https://github.com/LFL-Lab/SQuADDS/blob/master/docs/static/images/squaddslogodarkname.png?raw=true" width="80%" alt="SQuADDS Logo" /> </center>

SQuADDS Layouts - versioned GDS artifacts for superconducting quantum hardware

SQuADDS Layouts is the geometry-artifact companion to SQuADDS_DB, the Superconducting Qubit And Device Design and Simulation Database. It provides checksum-verified GDS files, stable geometry identities, and machine-readable geometry metadata so a simulation result can be traced to the exact layout that produced it.

• —Homepage: https://lfl-lab.github.io/SQuADDS/
• —Repository: https://github.com/LFL-Lab/SQuADDS
• —Simulation database: SQuADDS/SQuADDS_DB
• —Paper: https://arxiv.org/pdf/2312.13483.pdf
• —Point of contact: Sadman Ahmed Shanto

Contents

The current release contains 24,106 reproducible superconducting-device layouts.

The incomplete q3d_cap_00317 export without a simulation record, matrix, or GDS artifact is intentionally excluded. The generalized dataset uses north and south as canonical terminal names; top and bottom simulation fields are retained in SQuADDS_DB as compatibility aliases.

Layer convention

Three of the four families now share one geometry convention, matching the GeneralizedCapNInterdigital sweep that has always used it:

The etch is expressed as a hole in the ground plane, not as its own layer, so 1/11 is absent for these families. metadata/layer-semantics-v1.json (schema 1.3.0) is the machine-readable contract and lists exactly which layers each family carries.

CavityClawRouteMeander still uses the earlier convention, 1/10 conductors with a 1/11 etch layer and no ground plane or ports. It is a resonator rather than a two-terminal capacitor, so assigning it two lumped ports is a physics decision that has not been made yet. The semantics file states this per family; consumers should read it rather than assume a uniform layout.

Breaking change in this release

TransmonCross and CapNInterdigitalTee were regenerated directly from Qiskit Metal qgeometry to adopt the convention above. layout_id is a content hash of the polygons, so it changed for all 2,827 affected identities.

• —design_id and source_id are unchanged and remain the stable join keys. SQuADDS_DB.get_layout_ref and every published study join on these.
• —GeneralizedCapNInterdigital and CavityClawRouteMeander rows are untouched.
• —The previous geometry remains available by pinning the prior dataset revision.

The correction matters. The old CapNInterdigitalTee export omitted positive CPW path, leaving its conductor at roughly half its true area, and the old TransmonCross export carried no ground plane at all. Cross-component transfer studies improved substantially once all three families were described in the same reference frame.

Dataset Configurations

layout-manifest

metadata/manifest.parquet is the authoritative index. Every record includes:

• —layout_id: content-derived geometry identity.
• —artifact_id: SHA-256 checksum of the exact GDS file.
• —design_id: component-aware identity derived from design options.
• —source_id, simulation_repo, and simulation_config: links to the source simulation record.
• —gds_path: repository-relative raw artifact path.
• —bbox_um, layers, polygon_count, and cell_count: geometry metadata in micrometers.

geometry-features-v1

metadata/geometry-features-v1.parquet contains stable numerical features derived from the manifest without changing the raw GDS artifact. It includes bounds, area, aspect ratio, polygon and cell counts, plus sparse per-layer area and polygon statistics. This configuration is the reproducible input contract for future embedding models; it is not itself a learned embedding.

Access

Use the SQuADDS Python API to resolve a layout from a SQuADDS_DB row, download the checksum-verified GDS file, and inspect it on demand:

from squadds import SQuADDS_DB

row = ...  # one SQuADDS_DB simulation row
gds_path = SQuADDS_DB.download_gds(row)

Or use LayoutClient directly when you have a layout, design, or source ID:

from squadds.layouts import LayoutClient

client = LayoutClient()
layout = client.find(source_id="exp6/cap_0000")
summary = client.summary(layout)

Raw artifacts are organized as:

raw/GeneralizedCapNInterdigital/<campaign>/<instance>.gds
raw/CapNInterdigitalTee/generated_from_cap_matrix/<instance>.gds
raw/CavityClawRouteMeander/generated_from_cavity_claw_eigenmode/<instance>.gds
raw/TransmonCross/generated_from_transmon_cross_cap_matrix/<instance>.gds

The layer and port convention is documented in metadata/layer-semantics-v1.json (schema 1.3.0); see Layer convention above. Install the SQuADDS optional GDS dependency group (uv sync --extra gds) to inspect geometry locally.

Provenance and Contributions

The generalized-capacitor simulation data was contributed by Saikat Das of the Levenson-Falk Lab at USC. CapNInterdigitalTee artifacts are reproducibly generated from their paired SQuADDSDB design definitions using Qiskit Metal. The cavity and TransmonCross artifacts are likewise generated from their paired SQuADDSDB design options. Their GDS files contain functional conductor and etch geometry while omitting the finite simulation-domain ground plane.

New layouts should include a matching SQuADDS_DB simulation row, a stable source identifier, exact artifact checksum, component name, geometry metadata, and documented layer semantics. This keeps raw geometry, simulations, derived features, and future embeddings independently versioned but linkable.

Citation

If you use SQuADDS Layouts in your research, cite SQuADDS:

@article{Shanto2024squaddsvalidated,
  doi = {10.22331/q-2024-09-09-1465},
  url = {https://doi.org/10.22331/q-2024-09-09-1465},
  title = {{SQ}u{ADDS}: {A} validated design database and simulation workflow for superconducting qubit design},
  author = {Shanto, Sadman and Kuo, Andre and Miyamoto, Clark and Zhang, Haimeng and Maurya, Vivek and Vlachos, Evangelos and Hecht, Malida and Shum, Chung Wa and Levenson-Falk, Eli},
  journal = {{Quantum}},
  volume = {8},
  pages = {1465},
  year = {2024}
}

License

This project is licensed under the MIT License.