.. _multistage_optimization : ----------------------- Multistage Optimization ----------------------- The Multistage Optimization workflow in QpiAI-Pharma demonstrates the power of chaining computational tasks to achieve sophisticated molecular analysis. This workflow combines two complementary computational approaches to balance efficiency with accuracy in molecular structure determination. Workflow Architecture ===================== .. container:: architecture-diagram .. code-block:: text Input Molecule (XYZ) │ ▼ ┌────────────────────────────┐ │ Node 1 (Lower level) │ │ Geometry Opt. │ └────────────────────────────┘ │ optimized_xyz ▼ ┌────────────────────────────┐ │ Node 2 (Higher level) │ │ Geom Opt + Energy Calc. │ └────────────────────────────┘ │ ▼ Final Results .. note:: The *lower level* method can be **xTB** or **Hybrid-ML**, while the *higher level* typically refers to a **DFT** level of theory. Node Configuration ================== Input Parameters (Node 1) -------------------------- See :ref:`geometry_optimization` for detailed input parameters for Node 1. It depends on the selected module. Input Parameters (Node 2) -------------------------- See :ref:`geometry_optimization` for detailed input parameters for Node 2. It depends on the selected module. Key Benefits ============ .. container:: benefits-grid 🚀 **Computational Efficiency** Initial lower level of theory optimization provides a good starting geometry without the computational cost of full DFT optimization. This allows for faster convergence in the subsequent higher level calculations. .. container:: benefits-grid 🎯 **High Accuracy** DFT calculations on pre-optimized structures ensure accurate energetic and electronic properties. .. container:: benefits-grid ⚖️ **Resource Balance** Combines speed and accuracy by using appropriate methods at each stage. .. container:: benefits-grid 🔄 **Seamless Integration** Automated data flow between nodes and better user interface eliminates manual intervention.