Importance of feature engineering and database selection in a machine learning model: A case study on carbon crystal structures
Graz University of Technology（格拉茨工业大学）；Know-Center GmbH, Research Center for Data-Driven Business & Big Data Analytics（知识中心有限公司（数据驱动业务与大数据分析研究中心））；Università di Roma La Sapienza（罗马第一大学）
Franz M. Rohrhofer, Santanu Saha, Simone Di Cataldo, Bernhard C. Geiger, Wolfgang von der Linden, Lilia Boeri
Drive towards improved performance of machine learning models has led to the creation of complex features representing a database of condensed matter systems. The complex features, however, do not offer an intuitive explanation on which physical attributes do improve the performance. The effect of the database on the performance of the trained model is often neglected. In this work we seek to understand in depth the effect that the choice of features and the properties of the database have on a machine learning application. In our experiments, we consider the complex phase space of carbon as a test case, for which we use a set of simple, human understandable and cheaply computable features for the aim of predicting the total energy of the crystal structure. Our study shows that (i) the performance of the machine learning model varies depending on the set of features and the database, (ii) is not transferable to every structure in the phase space and (iii) depends on how well structures are represented in the database.
Since hydrogen fuel involves the highest energy density among all fuels, production of this gas through the solar water splitting approach has been suggested as a green remedy for greenhouse environmental issues due to extensive consumption of fossil fuels. Low dimensional materials possessing a large surface-to-volume ratio can be a promising candidate to be used for the photocatalytic approach. Here, we used extensive first principles calculations to investigate the application of newly fabricated members of two dimensional carbon nitrides including tg-C3N4, hg-C3N4, C2N, and C3N for water s…
A recently predicted two dimensional (2D) carbon allotrope, tetrahex-carbon consisting of tetragonal and hexagonal rings, draws research interests due to its unique mechanical and electronic properties. Tetrahex-C shows ultrahigh strength, negative Poisson ratio, a direct band gap and high carrier mobility. In this work, we employ first-principles density-functional theory calculations to explore the directional dependence of electronic properties such as carrier effective mass and mobility in tetrahex-C. Tetrahex-C demonstrates strong anisotropicity in effective mass of charge carrier and the…
Bottom-up prepared carbon nanostructures appear as promising platforms for future carbon-based nanoelectronics, due to their atomically precise and versatile structure. An important breakthrough is the recent preparation of nanoporous graphene (NPG) as an ordered covalent array of graphene nanoribbons (GNRs). Within NPG, the GNRs may be thought of as 1D electronic nanochannels through which electrons preferentially move, highlighting NPG's potential for carbon nanocircuitry. However, the π-conjugated bonds bridging the GNRs give rise to electronic cross-talk between the individual 1D channels,…
Graphdiyne, atomically-thin 2D carbon nanostructure based on sp-sp2 hybridization, is an appealing system potentially showing outstanding mechanical and optoelectronic properties. Surface-catalyzed coupling of halogenated sp-carbon-based molecular precursors represents a promising bottom-up strategy to fabricate extended 2D carbon systems with engineered structure on metallic substrates. Here, we investigate the atomic-scale structure and electronic and vibrational properties of an extended graphdiyne-like sp-sp2 carbon nanonetwork grown on Au(111) by means of on-surface synthesis. The formati…
Based on structure prediction method, the machine learning method is used instead of the density function theory (DFT) method to predict the material properties, thereby accelerating the material search process. In this paper, we established a data set of carbon materials by high-throughput calculation with available carbon structures obtained from the Samara Carbon Allotrope Database. We then trained an ML model that specifically predicts the elastic modulus (bulk modulus, shear modulus, and the Young's modulus) and confirmed that the accuracy is better than that of AFLOW-ML in predicting the…
Topology as a Design Variable for Multiproperty Engineering in Synthesized 4-5-6-8 Carbon Nanoribbons
State University of Campinas（坎皮纳斯州立大学）；Federal University of Campina Grande（坎皮纳格兰德联邦大学）；Fluminense Federal University（弗鲁米嫩塞联邦大学）；University of Campinas (UNICAMP)（坎皮纳斯大学）
Modification of graphene to open a robust gap in its electronic spectrum is essential for its use in field effect transistors and photochemistry applications. Inspired by recent experimental success in the preparation of homogeneous alloys of graphene and boron nitride (BN), we consider here engineering the electronic structure and bandgap of C$_{2x}$B$_{1-x}$N$_{1-x}$ alloys via both compositional and configurational modification. We start from the BN end-member, which already has a large bandgap, and then show that (a) the bandgap can in principle be reduced to about 2 eV with moderate subst…
A Spectral Neighbor Analysis (SNAP) machine learning interatomic potential (MLIP) has been developed for simulations of carbon at extreme pressures (up to 5 TPa) and temperatures (up to 20,000 K). This was achieved using a large database of experimentally relevant quantum molecular dynamics (QMD) data, training the SNAP potential using a robust machine learning methodology, and performing extensive validation against QMD and experimental data. The resultant carbon MLIP demonstrates unprecedented accuracy and transferability in predicting the carbon phase diagram, melting curves of crystalline …
We present an accurate machine learning (ML) model for atomistic simulations of carbon, constructed using the Gaussian approximation potential (GAP) methodology. The potential, named GAP-20, describes the properties of the bulk crystalline and amorphous phases, crystal surfaces and defect structures with an accuracy approaching that of direct ab initio simulation, but at a significantly reduced cost. We combine structural databases for amorphous carbon and graphene, which we extend substantially by adding suitable configurations, for example, for defects in graphene and other nanostructures. T…
We study the structural and mechanical properties of nanoporous (NP) carbon materials by extensive atomistic machine-learning (ML) driven molecular dynamics (MD) simulations. To this end, we retrain a ML Gaussian approximation potential (GAP) for carbon by recalculating the a-C structural database of Deringer and Csányi [Phys. Rev. B 2017, 95, 094203] adding van der Waals interactions. Our GAP enables a notable speedup and improves the accuracy of energy and force predictions. We use the GAP to thoroughly study the atomistic structure and pore-size distribution in computational NP carbon sampl…
We have theoretically investigated the thermoelectric properties of impurity-doped one-dimensional semiconductors, focusing on nitrogen-substituted (N-substituted) carbon nanotubes (CNTs), using the Kubo formula combined with a self-consistent $t$-matrix approximation. N-substituted CNTs exhibit extremely high thermoelectric power factor ($PF$) values originating from a characteristic of one-dimensional materials where decrease in the carrier density increase both the electrical conductivity and the Seebeck coefficient in the low-N regime. The chemical potential dependence of the $PF$ values o…
Wake-up effect is still an obstacle in the commercialization of hafnia-based ferroelectric thin films. In this work, we investigate the effect of defects, controlled by ozone dosage, on the field cycling behavior of the atomic layer deposited Hf0.5Zr0.5O2 (HZO) films. A nearly wake-up free device was achieved after reduction of carbon contamination and oxygen defects by increasing the ozone dosage. The sample which was grown at 30 sec ozone pulse duration shows about 98% of the woken-up Pr at the pristine state while those grown below 5 sec ozone pulse time show a pinched hysteresis loop, unde…
By means of ab initio calculations we investigate the possibility of existence of a boron nitride (BN) porous two-dimensional nanosheet which is geometrically similar to the carbon allotrope known as biphenylene carbon. The proposed structure, which we called Inorganic Graphenylene (IGP), is formed spontaneously after selective dehydrogenation of the porous Boron Nitride (BN) structure proposed by Ding et al. We study the structural and electronic properties of both porous BN and IGP and it is shown that, by selective substitution of B and N atoms with carbon atoms in these structures, the ban…
Mechanical properties of steels are significantly enhanced by retained austenite. Particularly, it has been shown that a recently developed heat-treatment technique called Quenching and Partitioning (Q\&P) stabilises austenite effectively. In the present work, the phase-field approach is adopted to simulate the phase transformation and carbon diffusion, which respectively accompanies the quenching and partitioning process of the polycrystalline Fe-C system. By incorporating the chemical driving-force from the CALPHAD database, the elastic phase-field model, which recovers the sharp-interface s…
The development of efficient photoelectrodes via defect engineering of wide-band gap metal oxides has been the prime focus for many years. Specifically, the effect of carbon defects in wide-band gap metal oxides on their performance in photoelectrochemical (PEC) applications raised numerous controversies and still elusive. Herein, the effect of various carbon defects in m-ZrO2 was investigated using the density functional theory to probe the thermodynamic, electronic, and optical properties of the defective structures against pristine m-ZrO2. The defect formation energies revealed that elevati…
