About SMAD

This page provides a technical overview of the software stack, HPC environment, screening workflows, and the policies around using and citing SMAD in scientific work.

Software stack

SMAD is built as a modular scientific platform. The current instance runs on:

  • Backend: FastAPI (Python) for high‑performance REST and WebSocket APIs.
  • Database: SQLite with SQLAlchemy ORM for structured materials data.
  • Remote access: Paramiko‑based SSH bridge for interacting with HPC login nodes.
  • Frontend: Bootstrap 5 with custom CSS for a lightweight, responsive UI.
  • Automation & AI: Pluggable Python workflows for optimization, surrogate models, and active learning.

You can extend this stack with additional databases, schedulers, or AI libraries according to the needs of your project or institution.

HPC & infrastructure

SMAD is designed to sit on top of institutional HPC clusters or cloud‑based resources. Typical deployments include:

  • Login nodes accessed via SSH for job submission and monitoring.
  • Schedulers such as SLURM, PBS, or similar batch systems.
  • Compute partitions with CPU‑only and GPU‑accelerated nodes.
  • High‑performance storage for wavefunctions, checkpoints, and derived datasets.

This instance exposes an experimental cluster console for lightweight commands; production workflows should be encoded as automated pipelines rather than manual terminal usage.

Screening and optimization workflows

SMAD supports multi‑stage superconducting materials discovery campaigns. A typical workflow is:

  1. Curate candidate structures (from databases or generative models).
  2. Submit large‑scale electronic‑structure or model Hamiltonian calculations on the cluster.
  3. Aggregate results into the SMAD database with metadata and provenance.
  4. Train surrogate models to predict \(T_c\), stability, and other target properties.
  5. Use active learning or Bayesian optimization to propose new candidates.
  6. Iterate until convergence on high‑value superconducting families.

The exact codes (e.g. DFT, DMFT, Monte Carlo) and optimization strategies can be customized for your project and should be documented in your workflow configuration or lab documentation.

News & updates

Use this space to track major changes to the SMAD platform and your discovery campaigns.

  • [YYYY‑MM‑DD] Initial deployment of SMAD on the superconductivity cluster.
  • [YYYY‑MM‑DD] Added automated screening workflow for layered cuprates.
  • [YYYY‑MM‑DD] Integrated new AI model for predicting critical temperatures.

Replace these placeholder entries with your real release notes, campaign milestones, or announcements for collaborators.

Publications using SMAD

List peer‑reviewed articles, preprints, and conference contributions that used SMAD or data generated by SMAD.

  • [Author et al.] Title of superconducting materials paper, Journal / arXiv:XXXX.XXXXX (Year).
  • [Author et al.] High‑throughput search for unconventional superconductors, Conference / Workshop (Year).

You can link each entry to a DOI, arXiv record, or institutional repository.

Terms of use

SMAD is intended for research and educational purposes. Access may be restricted to members of a specific institution, collaboration, or project.

  • Data responsibility: Users are responsible for verifying and interpreting results.
  • Compliance: All usage must comply with institutional HPC policies and software licenses.
  • Security: Do not store sensitive personal data; SSH access should use keys and strong practices.
  • Attribution: Publications must acknowledge SMAD and relevant funding sources.

Replace or extend this text with your official legal terms, institutional policies, or license statements.

How to cite SMAD

If you use SMAD in a publication, please cite it so that others can trace the computational infrastructure behind your results. A suggested citation format is:

Text citation: “This work used the SMAD (Superconducting Materials Automated Discovery) platform for high‑throughput screening and analysis of superconducting materials.”

BibTeX (example):

@misc{smad_platform,
  title        = {SMAD: A Platform for Automated Discovery of Superconducting Materials},
  author       = {Your Name and Collaborators},
  year         = {2026},
  howpublished = {\url{https://smad.live}},
  note         = {High-performance computing and AI-assisted materials discovery platform}
}

Update authors, year, and other fields once a formal software paper or institutional report is available.