Theory and Design of Materials

The past two decades have witnessed a profound evolution in the theoretical foundations and design strategies of materials. Originating in atomistic descriptions and quantum‐mechanical methods, theory now spans multiple length and time scales to couple electronic, atomistic, mesoscale and continuum models. High‐throughput computational screening, surrogate models and machine‐learning‐based discovery have accelerated the identification of novel compounds, enlarging the pool of stable phases by orders of magnitude. Concurrently, design‐driven frameworks emphasise the material as both functional core and expressive medium, introducing methods for mapping sensorial and performative parameters, for prototyping through “tinkering,” and for integrating sustainability and circular‐economy principles. Emerging paradigms now unite material‐centric, user‐centric, contextual and case‐centered approaches, enabling designers to exploit interactive smart materials, nanomaterials, bio-grown systems and advanced composites. Together, these advances underpin a systemic, data-informed, and design-led practice that addresses pressing global challenges in energy, electronics, biomedicine and structural engineering.

Research from Nature Portfolio

Scaling deep learning for materials discovery: Large‐scale graph neural networks trained on tens of thousands of crystal structures have achieved unprecedented generalisation. These models predict formation energies, ionic conductivities and forces, uncovering over two million candidate structures below the thermodynamic convex hull. Of these, hundreds have been independently realised, and the expanded dataset now enables zero‐shot predictions of ionic transport and highly accurate interatomic potentials for dynamic simulations.

Reversible transition between polar and antipolar phases in Hf₀.₅Zr₀.₅O₂: Atomic‐resolution electron microscopy and first‐principles analysis reveal a field‐driven switch between orthorhombic polar and antipolar orthorhombic phases. Cycling beyond a critical field induces fatigue via nonpolar interfacial phases, while higher voltages can restore ferroelectricity. This mechanism informs the design of fatigue‐resistant ferroelectric thin films for nonvolatile memory and neuromorphic devices.

Highly CMOS-compatible hafnia-based ferroelectric diode: A three-dimensional, stackable ferroelectric diode using Hf₀.₅Zr₀.₅O₂ epitaxial films achieves nanosecond switching, >10⁹‐cycle endurance and intrinsic nonlinearity (>100) that obviates external selectors. Detailed STEM studies correlate polarisation with oxygen displacements, guiding the integration of hafnia ferroelectrics into advanced memory and logic circuits at sub-10 nm nodes.

Theory and Design of Materials publication trend

The graph below shows the total number of articles in theory and design of materials across all publications each year (not limited to Nature Index journals).

Technical terms

Multiscale modeling: A suite of computational methods that link descriptions of materials across atomic, mesoscopic and continuum scales to predict macroscopic properties from first principles.

Surrogate model: A machine‐learning or reduced‐order representation trained on high‐fidelity simulations to enable rapid property predictions without expensive computations.

High‐throughput screening: Automated evaluation of large libraries of candidate materials via computational or experimental workflows to identify promising compositions or structures.

Materials informatics: The application of data‐driven algorithms and databases to uncover structure–property correlations and guide materials discovery.

Design‐driven materials discovery: An approach that uses material‐centric ideation, mapping of sensorial and functional parameters, and rapid prototyping to generate novel applications and guide synthesis.

References

1. Scaling deep learning for materials discovery. Nature (2023).
2. Reversible transition between the polar and antipolar phases and its implications for wake-up and fatigue in HfO2-based ferroelectric thin film. Nature Communications (2022).
3. A highly CMOS compatible hafnia-based ferroelectric diode. Nature Communications (2020).
4. A strategy to apply machine learning to small datasets in materials science. npj Computational Materials (2018).
5. Mixing rule for calculating the effective refractive index beyond the limit of small particles.. Optics Express (2023).