SimplySQS: An Automated and Reproducible Workflow for Special Quasirandom Structure Generation with ATAT

This paper introduces SimplySQS, an automated and reproducible online workflow that streamlines the generation of Special Quasirandom Structures (SQS) using the ATAT mcsqs module by eliminating manual input errors and successfully validating its utility through the accurate modeling of the Pb1-xSrxTiO3 perovskite system.

The Gist

Imagine you are a chef trying to bake the perfect "random" cake. You have a huge mixing bowl (a supercell) and you want to throw in flour, sugar, and eggs in a way that looks completely chaotic, just like a real random mixture.

In the world of materials science, scientists do this with atoms. They want to model disordered alloys—materials where different types of atoms are mixed together randomly, like a bowl of mixed nuts. To study these materials on a computer, they need to create a digital model called a Special Quasirandom Structure (SQS). This is a specific arrangement of atoms that looks random but is mathematically perfect for simulation.

The Problem: The "Manual Recipe" Nightmare

For years, creating these digital cakes has been like trying to bake a complex soufflé using a recipe written in a foreign language, with instructions like "add 3.4 grams of flour, but only if the moon is in the third phase."

The main tool for this, called ATAT (mcsqs), is powerful but old-school. It's like a high-end, professional oven that requires you to:

1. Write the recipe by hand on a piece of paper (manually typing code files).
2. Know exactly which buttons to press on the oven (command-line scripts).
3. Read the results from a cryptic display screen (interpreting raw data logs).

If you make a tiny typo in the recipe, the cake burns. If you forget a step, the oven doesn't start. This is frustrating for beginners and leads to mistakes even for experts.

The Solution: SimplySQS (The "Smart Kitchen" App)

The authors of this paper built SimplySQS, which is like a smart, interactive kitchen app that guides you through the whole process.

Instead of writing code, you just use a friendly website. Here is how it works, using our kitchen analogy:

1. Pick Your Ingredients (Structure Import): You upload a picture of your base cake (the crystal structure) or search for one in a digital pantry (databases like Materials Project).
2. Set the Recipe (Configuration): You tell the app, "I want 60% flour and 40% sugar." The app automatically figures out how to arrange them in the bowl so it looks random but follows the rules.
3. Press "Start" (Automation): The app writes the complex "recipe" (input files) and the "oven instructions" (execution scripts) for you. It even creates a single "All-in-One" script. Think of this as a magic wand: you hand this one file to the computer, and it does the entire baking process, checks the progress, and tells you when the cake is done.
4. Serve the Cake (Output): Once the computer finishes, the app takes the result and instantly converts it into a format you can use for other simulations (like VASP or LAMMPS), just like plating the cake perfectly.

The "Randomness Test" (The Taste Test)

One of the coolest features is a built-in "taste test." Sometimes, you might not need a perfect SQS; a simple random mix might be good enough. SimplySQS can quickly check your bowl of atoms and say:

• "Green Light": Your mix is already random enough! Save time, don't bother with the complex search.
• "Red Light": Your mix is too clumpy (not random enough). You definitely need the special SQS tool to fix it.

The Real-World Test: The "Pb-Sr-TiO3" Cake

To prove their app works, the authors used it to model a specific material called Pb1-xSrxTiO3 (a mix of Lead Titanate and Strontium Titanate). This material is special because it changes its shape (from a cube to a stretched box) depending on how much Strontium you add.

They used SimplySQS to generate a whole series of these "cakes" with different amounts of Strontium. Then, they used a super-smart AI (a machine-learning potential called MACE) to bake them and see how they changed shape.

The Result?
The app worked perfectly. The digital cakes matched real-world experiments almost exactly.

• When the Strontium was high, the material stayed a perfect cube.
• When the Strontium was low, it stretched into a box.
• The measurements were off by less than 1% compared to real life.

Why This Matters

SimplySQS is like giving a GPS to a driver who used to have to navigate by looking at paper maps and asking strangers for directions.

• For Beginners: It removes the fear of coding. You can learn about complex materials without needing to be a programmer.
• For Experts: It saves hours of typing and prevents silly mistakes.
• For Everyone: It makes science more reproducible. If you send your "magic wand" script to a friend, they can get the exact same result, no matter what computer they use.

In short, SimplySQS turns a difficult, error-prone chore into a smooth, automated, and fun process, allowing scientists to focus on the science rather than the software.

Technical Summary

1. Problem Statement

The generation of Special Quasirandom Structures (SQS) is a critical step in modeling disordered materials (alloys) under periodic boundary conditions. The ATAT (Alloy Theoretic Automated Toolkit) software, specifically its mcsqs module, is a widely used, established tool for this purpose. However, the current workflow presents significant barriers:

• Manual Complexity: Users must manually prepare complex input files (e.g., rndstr.in, sqscell.out), define chemical configurations, and manage supercell parameters.
• Error-Prone Process: The reliance on ad-hoc scripts and manual file handling introduces user-dependent errors, particularly regarding file format misunderstandings.
• Reproducibility Issues: The lack of standardized execution scripts makes it difficult to reproduce SQS searches across different computational studies.
• Steep Learning Curve: The command-line nature of ATAT poses challenges for researchers without strong programming backgrounds or experience with console environments.

2. Methodology

The authors developed SimplySQS, an automated, interactive, and reproducible workflow designed to streamline the SQS generation process using ATAT mcsqs.

• Platform Architecture:

  • Built in Python using the Streamlit framework for a browser-based, interactive interface.
  • Hosted online (simplysqs.com) but also available for local compilation.
  • Integrates Pymatgen and ASE for structure handling, Matminer for Partial Radial Distribution Functions (PRDF), and NumPy/Pandas for data management.
  • Supports structure import via file upload (POSCAR, CIF, LMP, XYZ) or direct API retrieval from databases (Materials Project, COD, MC3D).

• Core Workflow Features:

  1. Structure Input & Setup: Users define supercell sizes, element concentrations (global or per-sublattice), and cluster cutoff distances. The interface automatically calculates nearest-neighbor distances to assist in parameter selection.
  2. Automated Script Generation: The system generates all necessary ATAT input files and a comprehensive "all-in-one" bash script (monitor.sh). This script:
     • Executes the mcsqs search (single or parallel).
     • Monitors progress in real-time (objective function, runtime).
     • Automatically converts the final bestsqs.out output into standard formats (POSCAR, CIF, LMP, XYZ) compatible with DFT (VASP) and MD (LAMMPS) codes.
  3. Binary Concentration Sweep: A unique feature allowing users to generate SQSs across a continuous composition range (e.g., $Pb_{1-x}Sr_xTiO_3$) with a single script, automatically managing folders and parallelization for each concentration.
  4. Post-Processing & Visualization: The platform parses output logs to visualize objective function convergence, correlation values, and PRDFs.
  5. Randomness Assessment Module: A diagnostic tool that evaluates whether a random structure is statistically sufficient or if SQS optimization is required. It calculates a normalized randomness score (0–1) based on Warren-Cowley parameters for the first three coordination shells.
     • Score $\ge$ 0.95: Sufficiently random.
     • 0.85 – 0.95: Marginal (optimization likely beneficial).
     • < 0.85: Significantly ordered (SQS strongly recommended).

3. Key Contributions

• Automation & Reproducibility: Transforms the manual, error-prone ATAT workflow into a standardized, shareable process via the generated bash scripts.
• Lowered Entry Barrier: Provides a graphical interface that removes the need for command-line expertise, making SQS generation accessible to experimentalists and students.
• Integrated Ecosystem: Combines structure import, parameter definition, execution, and post-analysis in a single platform, eliminating the need to switch between different tools.
• Diagnostic Tool: Introduces a quantitative metric to determine a priori if SQS optimization is necessary for a specific system size and composition, saving computational resources.
• Educational Utility: Serves as a teaching tool for introducing concepts of random alloys and correlation functions without requiring coding skills.

4. Results

The authors validated SimplySQS using the $Pb_{1-x}Sr_xTiO_3$ (PSTO) perovskite system:

• Generation: A full series of SQSs ($0 < x < 1$) was generated using a single automated bash script. Each composition involved a 320-atom supercell ($4 \times 4 \times 4$ expansion) with 5 parallel search runs.
• Geometry Optimization: Structures were optimized using the MACE MATPES-r²SCAN-0 universal machine-learning interatomic potential (MLIP).
• Accuracy:
  • The workflow successfully reproduced the experimentally observed cubic-to-tetragonal phase transition near $x \approx 0.5$.
  • Cubic Region ($x > 0.5$): Simulated lattice parameters deviated from experimental values by < 0.3%.
  • Tetragonal Region ($x \le 0.5$): The $a$ parameter deviated by max 0.9%, and the $c$ parameter by max 3.4%. The authors attribute the larger deviation in the $c$ parameter to the universal potential's training data limitations regarding Pb-O bonding.
• Convergence: The objective function showed no change in the final three hours of the search, indicating convergence.

5. Significance

SimplySQS represents a significant step forward in the computational modeling of disordered materials by:

• Standardizing Best Practices: It enforces a reproducible workflow that minimizes human error in file preparation and execution.
• Democratizing Access: By removing the command-line barrier, it allows a broader range of researchers (including experimentalists and educators) to utilize high-fidelity SQS methods.
• Enhancing Efficiency: The "all-in-one" script and automated format conversion streamline the transition from structure generation to property calculation (DFT/MD).
• Future-Proofing: As a web-based application with open-source code (GitHub), it ensures long-term accessibility and community-driven improvement, addressing the common issue of software obsolescence in computational materials science.