運用第一原理計算探討鋁鈷鉻鎳中熵合金的相穩定度及析出行為與鈷鉻鎳機器學習模型發展

Abstract

本論文之研究目標為利用第一原理計算搭配密度泛函理論來探討鋁鈷鉻鎳中熵合金的相穩定度、析出行為與疊差能如何隨著鋁濃度變化。此外，我們也透過第一原理計算搭配密度泛函理論來發展一種可用於描述鈷鉻鎳三元系統的機器學習原子勢能模型。 在論文第一部分中，我們藉由第一原理計算，並搭配我們團隊開發的逆蒙地卡羅法，來研究鋁鈷鉻鎳隨鋁濃度變化的相穩定度。過往的研究中指出鋁鈷鉻鎳會隨著鋁濃度變化而從面心立方晶格結構逐漸轉變成體心立方晶格結構。我們使用逆蒙地卡羅法來構築不同局域排列的鋁鈷鉻鎳結構，並透過第一原理計算得到結構能量，發現鎳和鋁的介金屬相析出在熱力學上是自發性反應，且B2介金屬相的析出是鋁鈷鉻鎳從面心立方晶格結構轉變為體心立方晶格結構的關鍵因素。然而，二元鎳鋁系統實際上有兩種介金屬相，分別是L12相與B2相，且鋁濃度與鉻原子的局域分布都會決定鋁鈷鉻鎳中熵合金的析出行為。接著，我們根據前述相穩定度及鎳鋁析出行為的計算，進一步探討σ相在鋁鈷鉻鎳中熵合金的生成機制。最後，我們進行鋁鈷鉻鎳中熵合金的疊差能計算來研究鋁添加對於合金疊差能與雙晶生成能力的作用，來解釋鋁鈷鉻鎳的機械性質如何受到鎳鋁析出行為的影響。 在第二部分中，我們透過第一原理計算來開發了一組鈷鉻鎳系統的機器學習原子勢能模型，並驗證這些參數預測鈷鉻鎳系統的機械性質以及相穩定度的一致性。我們也另外發展了一種混成原子勢能模型，並比較兩種參數的整體表現。機器學習原子勢能模型可以提供高精度的預測結果，相對於第一原理計算來說可用於更大尺度的計算工具，並能應用於未來的材料開發。

The research objective of this thesis is to study the effect of Al concentration on the phase stability, precipitation behavior, and stacking fault energy (SFE) of AlxCoCrNi medium-entropy alloys (MEAs), using first-principles calculations combined with density functional theory (DFT). Besides, we develop a Spectral Neighbor Analysis Potential (SNAP) model suitable for the CoCrNi ternary system. In the first part of the thesis, we employ first-principles calculations along with the reverse Monte Carlo (RMC) algorithm to study the phase stability of AlxCoCrNi MEAs as a function of Al concentration. Previous research suggests that the phase of AlxCoCrNi MEAs transform from FCC to BCC structures with changing Al concentration. Using the RMC algorithm, we construct various local elemental arrangements of AlxCoCrNi structures and determine their structural energies through first-principles calculations. We find that the precipitation of Ni-Al intermetallic phases is thermodynamically favored, and the precipitation of the B2-NiAl intermetallic phase is a key factor in the FCC-to-BCC phase transition in AlxCoCrNi MEAs. However, there are two possible forms of Ni Al intermetallic phase precipitation: L12-Ni3Al and B2-NiAl, and both the concentration of Al and the local distribution of Cr elements affect the precipitation behavior in AlxCoCrNi MEA. Based on these calculations of phase stability and NiAl precipitation behavior, we further investigate the formation mechanism of the σ phase in AlxCoCrNi MEAs. Finally, we conduct SFE calculations in AlxCoCrNi MEAs in order to study the effects of Al addition on the twinnability, and we provide the explanation about the influence on mechanical properties of AlxCoCrNi high-entropy alloys by Ni-Al precipitation behavior. In the second part, we develop a set of SNAP machine-learning interatomic potential models for the ternary CoCrNi system using first-principles calculations, and we validate their predictions for the mechanical properties and relative energies between structures with different configuration in the CoCrNi system. We also develop and validate a hybrid MEAM+SNAP (modified embedded-atom method) model, comparing the overall performance of both sets of parameters. The SNAP model can serve as an alternative computational tool with high accuracy to for larger scale computations and can be applied in future materials development.