Chapter 1: What is NIMO?

Learning Objectives

• Understand the challenges of traditional materials exploration
• Learn what closed-loop automation means
• Get an overview of NIMO's features and capabilities
• Successfully install NIMO on your system

1.1 The Challenge of Materials Discovery

Materials science research faces a fundamental challenge: the search space is enormous. Consider developing a new alloy with 5 elements. If each element can have 100 different composition levels, the total number of possible combinations is:

$$N = 100^5 = 10^{10} \text{ (10 billion combinations)}$$

Even if each experiment takes just 1 hour, testing all combinations would require over 1 million years! This is known as the combinatorial explosion problem.

Traditional Approaches and Their Limitations

1.2 What is Closed-Loop Automation?

Closed-loop automation addresses these challenges by creating an intelligent feedback cycle between AI algorithms and experimental systems. The key insight is: instead of testing everything, let AI guide us to the most promising candidates.

This approach offers several advantages:

• Efficiency: AI learns from each experiment to propose better candidates
• Speed: Robots can run experiments 24/7 without fatigue
• Objectivity: Decisions based on data, not human bias
• Reproducibility: Automated processes are consistent and documented

1.3 Introducing NIMO

NIMO (formerly NIMS-OS: NIMS Orchestration System) is a Python library developed by the National Institute for Materials Science (NIMS) in Japan. It provides a complete framework for implementing closed-loop materials exploration.

NIMO's Modular Architecture

1.4 Installing NIMO

NIMO can be installed easily via pip:

# Install NIMO from PyPI (recommended)
pip install nimo

# Verify installation
python -c "import nimo; print('NIMO installed successfully!')"

For development or custom modifications, you can install from source:

# Clone the repository
git clone https://github.com/NIMS-DA/nimo

# Navigate to the directory
cd nimo

# Install in development mode
pip install -e .

1.5 Your First NIMO Code

Let's verify that NIMO is working by examining its available modules:

import nimo

# Check available selection methods
print("NIMO provides the following optimization methods:")
methods = ["RE", "PHYSBO", "BLOX", "PDC", "SLESA", "PTR", "BOMP", "ES", "COMBI", "RSVM"]
for method in methods:
    print(f"  - {method}")

# Check visualization tools
print("\nVisualization tools:")
print("  - nimo.visualization.plot_history")
print("  - nimo.visualization.plot_distribution")
print("  - nimo.visualization.plot_phase_diagram")

Output:

NIMO provides the following optimization methods:
  - RE
  - PHYSBO
  - BLOX
  - PDC
  - SLESA
  - PTR
  - BOMP
  - ES
  - COMBI
  - RSVM

Visualization tools:
  - nimo.visualization.plot_history
  - nimo.visualization.plot_distribution
  - nimo.visualization.plot_phase_diagram

1.6 A Simple Example: Random Selection

Let's see NIMO in action with a simple random selection:

import nimo
import pandas as pd

# Create a simple candidates file
# Format: descriptor columns + objective column (NaN for untested)
candidates_data = {
    'x1': [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0],
    'x2': [1.0, 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1],
    'objective': [float('nan')] * 10  # All untested
}
df = pd.DataFrame(candidates_data)
df.to_csv('candidates.csv', index=False)

# Use NIMO to randomly select 3 candidates
nimo.selection(
    method="RE",           # Random Exploration
    input_file="candidates.csv",
    output_file="proposals.csv",
    num_objectives=1,
    num_proposals=3,
    re_seed=42             # For reproducibility
)

# Check the proposals
proposals = pd.read_csv('proposals.csv')
print("Selected candidates for experiments:")
print(proposals)

Output:

Selected candidates for experiments:
    x1   x2  objective
0  0.7  0.4        NaN
1  0.2  0.9        NaN
2  0.5  0.6        NaN

Exercises

import nimo
import physbo
import numpy as np
print(f"NIMO ready! NumPy version: {np.__version__}")

Summary

• Materials discovery faces the combinatorial explosion problem
• Closed-loop automation combines AI algorithms with robotic experiments
• NIMO is a modular Python library for automated materials exploration
• NIMO provides 11 AI algorithms, 3 robot interfaces, and visualization tools
• Installation is simple via pip install nimo