Related Experiment Video
Updated: May 1, 2026
12:35
Atomically Traceable Nanostructure Fabrication
Published on: July 17, 2015
8.3K
SPACESHIP: Autonomous Mapping of Hardware-Dependent Synthesizable Space in Solution-Phase Gold Nanomaterials
Nayeon Kim1,2, Hyuk Jun Yoo1, Daeho Kim1,3
1Computational Science Research Center, Korea Institute of Science and Technology, Seoul 02792, Republic of Korea.
Journal of the American Chemical Society
|April 29, 2026
Summary
SPACESHIP, an AI-robotics framework, accelerates materials discovery by adaptively exploring chemical spaces without predefined limits. This system overcomes reproducibility issues by dynamically adjusting to hardware-specific synthesis conditions.
Area of Science:
- Materials Science
- Artificial Intelligence
- Robotics
Background:
- Autonomous laboratories promise accelerated materials discovery but are limited by predefined experimental boundaries.
- Existing constraints, based on human intuition or literature, risk excluding feasible synthesis regions.
- Synthesizable conditions can vary with hardware and environmental factors, impacting reproducibility.
Purpose of the Study:
- To present SPACESHIP, an AI framework integrated with automated hardware for adaptive chemical space exploration.
- To overcome literature- or expert-derived feasibility constraints in materials synthesis.
- To address the reproducibility gap in science by adapting to system-specific synthesizable boundaries.
Main Methods:
- SPACESHIP employs AI-based prediction, robotic synthesis, and real-time characterization for iterative model updates.
- It utilizes probabilistic models with an 'Autopilot' acquisition strategy to refine synthesizable regions.
- The system dynamically switches between models, incorporating both successful and failed experiments.
Main Results:
- The AI-robotics system achieved 90% accuracy in gold nanoparticle (NP) and nanorod (NR) synthesis within 23 experiments, significantly less than the 512 for ground truth.
- Distinct growth regimes were uncovered across optical property classes.
- Synthesizable regions were expanded by 8x for NPs and 4x for NRs beyond literature maps.
Conclusions:
- SPACESHIP enables adaptive exploration of chemical spaces, free from fixed constraints.
- The framework successfully adapts to hardware-specific conditions, enhancing the discovery of novel materials.
- By merging AI and autonomous experimentation, SPACESHIP addresses scientific reproducibility challenges.
