Tech notes/using adda with python

<h2>How to simulate the plasmon resonance using discrete dipole approximation</h2> <p>Many years ago, I used ADDA to calculate plasmon resonances of gold nanorods while they were etched by potassium cyanide (DOI: <a href="https://doi.org/10.1039/C6CP01679K">https://doi.org/10.1039/C6CP01679K</a>). I wrote the code in Python and is available here (<a href="https://github.com/aquilesC/plasmon_adda">https://github.com/aquilesC/plasmon_adda</a>). </p> <p>This tutorial explains the architecture and design decisions behind the <code>plasmon-adda</code> package. It is intended for users who want to understand how to automate ADDA simulations using Python.</p> <h2>Architecture Overview</h2> <p>The project is structured as a standard Python package to ensure modularity, ease of installation, and reusability.</p> <pre class="codehilite"><code>plasmon_adda/ ├── src/ │ └── plasmon_adda/ │ ├── __init__.py # Package marker │ ├── cli.py # Command-Line Interface entry points │ ├── simulation.py # Core logic for single simulations │ ├── batch.py # Logic for parallel batch processing │ ├── plotting.py # Logic for data visualization │ └── indexes.py # Material properties (Refractive Indices) ├── pyproject.toml # Package metadata and dependencies └── README.md # General usage instructions </code></pre> <h2>Core Components</h2> <h3>1. Material Properties (<code>indexes.py</code>)</h3> <p><strong>Design Goal</strong>: Encapsulate complex refractive index data handling.</p> <p>Instead of hardcoding values or parsing files repeatedly, we use a <code>RefractiveIndex</code> base class. This class handles:</p> <ul> <li>Reading raw data files (Wavelength vs Real/Imaginary Index).</li> <li>Interpolating the data using <code>scipy.interpolate</code> to allow querying the refractive index at <em>any</em> arbitrary wavelength.</li> </ul> <p><strong>Extensibility</strong>: To add a new material (e.g., Silver), you simply subclass <code>RefractiveIndex</code> and point it to the new data file.</p> <pre class="codehilite"><code class="language-python">class SilverIndex(RefractiveIndex): def __init__(self): super().__init__('METALS_Silver_Johnson.txt') </code></pre> <h3>2. Adaptive Simulation (<code>simulation.py</code>)</h3> <p><strong>Design Goal</strong>: Efficiently resolve spectral peaks without wasting computation on flat regions.</p> <p>The <code>run_simulation</code> function implements an adaptive step-size algorithm ("Smart Search"): 1. <strong>Initialization</strong>: It starts at a given wavelength (e.g., 550nm). 2. <strong>Step Adjustment</strong>: It calculates the slope ("difference") of the Extinction Efficiency (<code>Q_ext</code>) between previous steps. - <strong>Fast Change (Peak)</strong>: If the slope is steep, the step size decreases (down to 2nm) to capture fine details. - <strong>Slow Change</strong>: If the slope is flat, the step size increases (up to 15nm) to speed up the process. 3. <strong>Peak Detection</strong>: It tracks if it has passed a peak (slope sign change) and stops the simulation 100nm after the peak is confirmed.</p> <p><strong>ADDA Integration</strong>: We use the <code>subprocess</code> module to call the binary <code>adda</code> tool. This separates the physics engine (ADDA) from the control logic (Python). - <strong>Command Construction</strong>: Arguments like <code>-shape</code>, <code>-size</code>, and <code>-m</code> (refractive index) are dynamically built. - <strong>Output Parsing</strong>: We capture standard output (<code>stdout</code>) and robustly parse the key metrics (<code>Q_ext</code>, <code>C_ext</code>, etc.).</p> <h3>3. Parallel Batch Processing (<code>batch.py</code>)</h3> <p><strong>Design Goal</strong>: Scale simulations across multiple CPU cores.</p> <p>Since ADDA simulations for different geometries are independent, they are "embarrassingly parallel." - We use <code>concurrent.futures.ProcessPoolExecutor</code> to spawn a pool of worker processes. - Each worker runs an instance of <code>run_simulation</code> for a specific configuration (e.g., a specific Length). - This drastically reduces total computation time compared to a sequential loop.</p> <h3>4. Command-Line Interface (<code>cli.py</code>)</h3> <p><strong>Design Goal</strong>: Make the tools accessible from the terminal without modifying code.</p> <p>We use <code>argparse</code> to define user-friendly commands. - <strong>Entry Points</strong>: <code>pyproject.toml</code> maps these functions to commands like <code>plasmon-run</code>. - This allows you to run simulations on a remote server or cluster just by typing a command, making it easy to script and automate.</p> <h2>How to Extend</h2> <p>If you want to simulate different shapes or materials: 1. <strong>Modify <code>indexes.py</code></strong>: Add your new material class. 2. <strong>Update <code>simulation.py</code></strong>: - Change <code>AuIndex()</code> to your new material class. - Update the <code>cmd</code> list to use different shapes (e.g., change <code>-shape capsule</code> to <code>-shape sphere</code>).</p> <h2>Conclusion</h2> <p>This architecture separates concerns: <strong>Data</strong> (indexes), <strong>Logic</strong> (simulation), <strong>Scaling</strong> (batch), and <strong>Interface</strong> (CLI). This makes the code easier to test, debug, and extend.</p>