More Like this

1. Fine-Tuned Language Models Generate Stable Inorganic Materials as Text

   Gruver, N; Sriram, A; Madotto, A; Wilson, AG; Zitnick, LC; Ulissi, Z (May 2024, International Conference on Learning Representations)

   We propose fine-tuning large language models for generation of stable materials. While unorthodox, fine-tuning large language models on text-encoded atomistic data is simple to implement yet reliable, with around 90% of sampled structures obeying physical constraints on atom positions and charges. Using energy above hull calculations from both learned ML potentials and gold-standard DFT calculations, we show that our strongest model (fine-tuned LLaMA-2 70B) can generate materials predicted to be metastable at about twice the rate (49% vs 28%) of CDVAE, a competing diffusion model. Because of text prompting's inherent flexibility, our models can simultaneously be used for unconditional generation of stable material, infilling of partial structures and text-conditional generation. Finally, we show that language models' ability to capture key symmetries of crystal structures improves with model scale, suggesting that the biases of pretrained LLMs are surprisingly well-suited for atomistic data. 

   more » « less
2. Secondary structures in synthetic polypeptides from N -carboxyanhydrides: design, modulation, association, and material applications

   https://doi.org/10.1039/c8cs00095f  

   Song, Ziyuan; Fu, Hailin; Wang, Ruibo; Pacheco, Lazaro A.; Wang, Xu; Lin, Yao; Cheng, Jianjun (January 2018, Chemical Society Reviews)

   Synthetic polypeptides derived from the ring-opening polymerization of N -carboxyanhydrides can spontaneously fold into stable secondary structures under specific environmental conditions. These secondary structures and their dynamic transitions play an important role in regulating the properties of polypeptides in self-assembly, catalysis, polymerization, and biomedical applications. Here, we review the current strategies to modulate the secondary structures, and highlight the conformation-specific dynamic properties of synthetic polypeptides and the corresponding materials. A number of mechanistic studies elucidating the role of secondary structures are discussed, aiming to provide insights into the new designs and applications of synthetic polypeptides. We aim for this article to bring to people's attention synthetic polymers with ordered conformations, which may exhibit association behaviors and material properties that are otherwise not found in polymers without stable secondary structures. 

   more » « less
3. Structural search for stable Mg–Ca alloys accelerated with a neural network interatomic model

   https://doi.org/10.1039/c8cp05314f  

   Ibarra-Hernández, Wilfredo; Hajinazar, Samad; Avendaño-Franco, Guillermo; Bautista-Hernández, Alejandro; Kolmogorov, Aleksey N.; Romero, Aldo H. (November 2018, Physical Chemistry Chemical Physics)

   We have combined a neural network formalism with metaheuristic structural global search algorithms to systematically screen the Mg–Ca binary system for new (meta)stable alloys. The combination of these methods allows for an efficient exploration of the potential energy surface beyond the possibility of the traditional searches based on ab initio energy evaluations. The identified pool of low-enthalpy structures was complemented with special quasirandom structures (SQS) at different stoichiometries. In addition to the only Mg–Ca phase known to form under standard synthesis conditions, C14-Mg 2 Ca, the search has uncovered several candidate materials that could be synthesized under elevated temperatures or pressures. We show that the vibrational entropy lowers the relative free energy of several phases with magnesium kagome layers: C15 and C36 Laves structures at the 2 : 1 composition and an orthorhombic oS36 structure at the 7 : 2 composition. The estimated phase transition temperatures close to the melting point leave open the possibility of synthesizing the predicted materials at high temperatures. At high pressures up to 10 GPa, two new phases at the 1 : 1 and 3 : 1 Mg : Ca stoichiometries become thermodynamically stable and should form in multi-anvil experiments. 

   more » « less
4. Totimorphic assemblies from neutrally stable units

   https://doi.org/10.1073/pnas.2107003118  

   Chaudhary, Gaurav; Ganga Prasath, S.; Soucy, Edward; Mahadevan, L. (October 2021, Proceedings of the National Academy of Sciences)

   Inspired by the quest for shape-shifting structures in a range of applications, we show how to create morphable structural materials using a neutrally stable unit cell as a building block. This unit cell is a self-stressed hinged structure with a one-parameter family of morphing motions that are all energetically equivalent. However, unlike kinematic mechanisms, the unit cell is not infinitely floppy and instead exhibits a tunable mechanical response akin to that of an ideal rigid-plastic material. Theory and simulations allow us to explore the properties of planar and spatial assemblies of neutrally stable elements, and solve the inverse problem of designing assemblies that can morph from one given shape into another. Simple experimental prototypes of these assemblies corroborate our theoretical results and show that the addition of switchable hinges allows us to create load-bearing structures. Altogether, totimorphs pave the way for structural materials whose geometry and deformation response can be controlled independently and at multiple scales. 

   more » « less
5. High‐Throughput Discovery of Novel Cubic Crystal Materials Using Deep Generative Neural Networks

   https://doi.org/10.1002/advs.202100566  

   Zhao, Yong; Al‐Fahdi, Mohammed; Hu, Ming; Siriwardane, Edirisuriya_M_D; Song, Yuqi; Nasiri, Alireza; Hu, Jianjun (August 2021, Advanced Science)

   Abstract High‐throughput screening has become one of the major strategies for the discovery of novel functional materials. However, its effectiveness is severely limited by the lack of sufficient and diverse materials in current materials repositories such as the open quantum materials database (OQMD). Recent progress in deep learning have enabled generative strategies that learn implicit chemical rules for creating hypothetical materials with new compositions and structures. However, current materials generative models have difficulty in generating structurally diverse, chemically valid, and stable materials. Here we propose CubicGAN, a generative adversarial network (GAN) based deep neural network model for large scale generative design of novel cubic materials. When trained on 375 749 ternary materials from the OQMD database, the authors show that the model is able to not only rediscover most of the currently known cubic materials but also generate hypothetical materials of new structure prototypes. A total of 506 such materials have been verified by phonon dispersion calculation. Considering the importance of cubic materials in wide applications such as solar panels, the GAN model provides a promising approach to significantly expand existing materials repositories, enabling the discovery of new functional materials via screening. The new crystal structures discovered are freely accessible atwww.carolinamatdb.org. 

   more » « less