Carbonitride Metal Navigator

Descrizione del progetto

Liberare il potenziale dei carbonitruri metallici

I carbonitruri metallici, composti che combinano metalli abbondanti con carbonio e azoto, presentano notevoli proprietà meccaniche, elettroniche, catalitiche e superconduttive, aprendo la strada a potenziali applicazioni nell’ambito dell’energia pulita, della catalisi e delle tecnologie quantistiche; ciononostante, sono tuttora poco esplorati a causa dei costi e dei tempi di utilizzo dei metodi convenzionali della teoria funzionale della densità (DFT, density functional theory) intesi a esplorare sistematicamente questo vasto spazio chimico. Sostenuto dal programma di azioni Marie Skłodowska-Curie, il progetto CarMeN mira a raggiungere una precisione prossima alla DFT a una frazione del costo computazionale associato. I potenziali appresi dalla macchina e basati su dati DFT di alta qualità saranno combinati con la previsione della struttura cristallina per consentire la mappatura sistematica di composti di carbonitruri metallici stabili e metastabili dotati di proprietà desiderabili, secondo un approccio in cui i candidati previsti saranno successivamente convalidati al fine di valutarne la fattibilità sperimentale.

Obiettivo

The CarMeN (Carbonitride Metal Navigator) project aims to revolutionize the discovery of functional materials by combining machine learning (ML) with crystal structure prediction (CSP) to accelerate exploration of the vast and underexplored space of metal carbonitrides (MCNs). MCNs, composed of earth-abundant metals with carbon and nitrogen, exhibit exceptional mechanical, electronic, catalytic, and superconducting properties, making them highly relevant for sustainable energy, catalysis, quantum technologies, and advanced manufacturing. Despite this promise, only a small fraction of MCN compounds has been studied due to the limitations of conventional density functional theory (DFT) methods.

The project overcomes these bottlenecks by developing machine-learned potentials (MLPs) trained on high-quality DFT data, achieving near-DFT accuracy at orders-of-magnitude lower cost. These MLPs will enable high-throughput CSP, systematically mapping stable and metastable MCN phases while incorporating multi-objective optimization to identify compounds with technologically desirable properties. Predicted structures will be rigorously validated through phonon stability checks, property calculations (mechanical, electronic, catalytic, and energetic), and synthetic pathway analysis, ensuring experimental feasibility.

The methodology is structured across five work packages: (i) project and data management under FAIR principles, (ii) MLP development, (iii) ML-accelerated CSP, (iv) property evaluation and synthesis-informed validation, and (v) dissemination and open science practices. By uniting interdisciplinary advances in machine learning, computational chemistry, and materials science, CarMeN will establish a scalable discovery paradigm that bridges theoretical prediction and practical application. Ultimately, the project seeks to deliver sustainable, high-performance materials while advancing open, reproducible practices in computational materials research.