Predicting low-thermal-conductivity Si-Ge nanowires with a modified cluster expansion method

Kristensen, Jesper; Zabaras, Nicholas

doi:10.1103/physrevb.91.054105

Cited by 3 publications

(4 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Various models including elastic net [36], support vector regression [37], bagging (bootstrap aggregating) [38], random forest [39], gradient boosting for regression [40], artificial neural network [41], Gaussian process regression [42], clustering algorithms [43,44] While the database-screening work related with the thermal conductivity reported so far have focused mainly on the low lattice thermal conductivity crystals, we have recently performed the hierarchical screening and transfer learning screening of the high lattice thermal conductivity materials [45], which will be published soon.…”

Section: High-throughput Screeningmentioning

confidence: 99%

Materials Informatics for Heat Transfer: Recent Progresses and Perspectives

Shiomi

2019

Nanoscale and Microscale Thermophysical Engineering

View full text Add to dashboard Cite

With the advances in materials and integration of electronics and thermoelectrics, the demand for novel crystalline materials with ultimate high/low thermal conductivity is increasing. However, search for optimal thermal materials is challenge due to the tremendous degrees of freedom in the composition and structure of crystal compounds and nanostructures, and thus empirical search would be exhausting.Materials informatics, which combines the simulation/experiment with machine 2 learning, is now gaining great attention as a tool to accelerate the search of novel thermal materials. In this review, we discuss recent progress in developing materials informatics for heat transport: the exploration of crystals with high/low thermal conductivity via high-throughput screening, and nanostructure design for high/low thermal conductance using the Bayesian optimization and Monte Carlo tree search.The progresses show that the materials informatics method are useful for designing thermal functional materials. We end by addressing the remaining issues and challenges for further development.Nanostructure designing, Bayesian optimization, Monte Carlo tree search [6], ICSD [7], OQMD [8,9], and AtomWork [10], as shown in Fig. 1 (a). The thermal property of materials varies in a wide range. Taking thermal conductivity as an example, the order ranges from hundredths of Wm -1 K -1 to thousands of Wm -1 K -1 .Discovery of materials with very low or high thermal conductivity remains an experimental challenge due to high cost and time-consuming synthesis procedures.The other bottleneck is perhaps more challenging. As the length scale of materials decreases to nanoscale, heat conduction becomes more controllable through manipulating the nanostructures, as shown in Fig. 1 (b). Due to the various choices of structure parameters and coupled effects, it is difficult to quickly obtain the optimal nanostructure with desired thermal property from tremendous number of candidates.If exploring the materials or structures one by one using traditional heat transfer analysis method, it will become time-consuming and low-efficiency.The key next-generation technology to solve the above bottlenecks is materials informatics (MI): integration of material property calculations or measurements with informatics method to accelerate the material discovery and design [11][12][13]. During the past decade, MI has been successfully applied to design cathode materials of the lithium-ion battery, drugs, polymers, catalysis [14][15][16][17][18], and many others. The application of MI on thermal transport has also been gradually developed. In this review, we summarize the most recent progress of the MI application in heat transfer field. The review is organized as follows. In the first part, we summarize the recent progress of high-throughput screening for ultimate high/low lattice thermal conductivity materials. In the second part, we introduce the nanostructure

show abstract

Section: High-throughput Screeningmentioning

confidence: 99%

Materials Informatics for Heat Transfer: Recent Progresses and Perspectives

Shiomi

2019

Nanoscale and Microscale Thermophysical Engineering

View full text Add to dashboard Cite

show abstract

“…The values calculated from the DFT simulations using Eqs. (30) and (32), including the full phonon model to obtain F vib , are also shown as black dots. The uncertainty of the predictive distribution increases with the temperature, and the values calculated using DFT are within the 95% confidence interval for all the temperature range.…”

Section: Uncertainty Using the Bond Stiffness Versus The Bond Length mentioning

confidence: 99%

“…Even though some success has been achieved using ab initio structure search [20,12,21], the use of surrogate models appears as the most feasible way for the explorations of the configurational space of alloys [22,23,15]. The most important surrogate model in this field is the cluster expansion (CE) [24,25,26,27], which has been widely used in a number of different applications [22,28,23,29,15,30]. It decomposes a function f (•) of a configuration σ in contributions from different clusters of atoms in the same way as a Fourier series decomposes a periodic signal into components of different frequencies, and in the same way a Fourier series is exact if all terms are included, the CE is exact if all clusters are included.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Quantifying uncertainties in first-principles alloy thermodynamics using cluster expansions

Aldegunde¹,

Zabaras²,

Kristensen³

2016

Journal of Computational Physics

Self Cite

View full text Add to dashboard Cite

The cluster expansion is a popular surrogate model for alloy modeling to avoid costly quantum mechanical simulations. As its practical implementations require approximations, its use trades efficiency for accuracy. Furthermore, the coefficients of the model need to be determined from some known data set (training set). These two sources of error, if not quantified, decrease the confidence we can put in the results obtained from the surrogate model. This paper presents a framework for the determination of the cluster expansion coefficients using a Bayesian approach, which allows for the quantification of uncertainties in the predictions. In particular, a relevance vector machine is used to automatically select the most relevant terms of the model while retaining an analytical expression for the predictive distribution. This methodology is applied to two binary alloys, SiGe and MgLi, including the temperature dependence in their effective cluster interactions. The resulting cluster expansions are used to calculate the uncertainty in several thermodynamic quantities: ground state line, including the uncertainty in which structures are thermodynamically stable at 0 K, phase diagrams and phase transitions. The uncertainty in the ground state line is found to be of the order of meV/atom, showing that the cluster expansion is reliable to ab initio level accuracy even with limited data. We found that the uncertainty in the predicted phase transition temperature increases when including the temperature dependence of the effective cluster interactions. Also, the use of the bond stiffness versus bond length approximation to calculate temperature dependent properties from a reduced set of alloy configurations showed similar uncertainty to the approach where all training configurations are considered but at a much reduced computational cost.

show abstract