1.Faculty of Intelligent Manufacturing, Wuyi University, Jiangmen 529020, China 2.School of Materials Science and Engineering, South China University of Technology, Guangzhou 510640, China
Abstract:The martensitic transformation between the high-temperature face-centered cubic (FCC) phase and the low-temperature body-centered cubic (BCC) phase in iron-based alloys has been studied for years, which plays a critical role in controlling microstructures and hence properties of the alloys. Generally, the BCC structure martensitic phase forms from the FCC parent phase, involving a collective motions of atoms over a distance less than the interatomic distance in the vicinity of the interphase boundary. Thus the structure of interphase boundary separating the FCC and BCC phases is the key characteristics to quantitatively understanding the mechanism and kinetics of martensitic transformation. Due to the difficulty in observing the atomic motions taking place at a velocity as high as the speed of sound, the experimental investigation on the migration of FCC/BCC interphase boundary during the transformation is as yet limited. Noteworthily, molecular dynamics (MD) simulation has been applied to studying the martensitic transformation, in particular for investigating the mobility of the FCC/BCC interphase boundary in iron. However, in most of the MD studies the atomistically planar interfaces of {111}FCC // {110}BCC are considered as the initial configuration of the interphase boundary between FCC and BCC phases, which is in contradiction to the high-resolution TEM observations. In fact, the FCC/BCC interphase boundary, which is known as the macroscopic habit plane, is a semi-coherent interface consisting of several steps and terrace planes on an atomic scale. In the present work, the atomic configuration of a terrace-step FCC/BCC interphase boundary of iron is built in terms of the topological model. The MD simulation is conducted to clarify the mechanism of interphase boundary migration in the FCC-to-BCC transformation. The results show that the FCC/BCC boundary migrates along its normal at the expense of FCC phase as a result of the lateral motions of the transformation dislocations. Meanwhile, the interphase boundary maintains the stable terrace-step structure during the transformation. Further examinations reveal that the transformation dislocations move steadily at a velocity as high as (2.8 ± 0.2) × 103 m/s, affecting the migration of the interphase boundary with a constant velocity of about (4.4 ± 0.3) × 102 m/s. The effective migration velocity of FCC/BCC interface exhibits dynamic properties consistent with the characteristic features commonly observed in a displacive martensitic transformation. Additionally, the motion of transformation dislocations gives rise to the macroscopic shape strain composed of a shear component $ {\varGamma _{{\rm{yz}}}} = 0.349$ parallel to the boundary and a dilatation $ {\varGamma _{{\rm{zz}}}} = 0.053$ normal to the boundary in the MD simulation, which is close to the crystallographic calculations by the topological model. Keywords:transformation dislocation/ interphase boundary migration/ molecular dynamics simulation/ topological model of martensitic transformation
全文HTML
--> --> -->
2.FCC/BCC相界面台阶结构初始模型的构建及模拟条件在分子动力学模拟中, 势函数能否正确描述模拟对象的物理性质对于模拟结果的合理性至关重要, 而获得与Fe势函数的选择密切相关的稳定FCC/BCC相界面结构作为模拟的初始模型亦非常关键. Engin等[28]曾根据不同类型的势函数分别计算了Fe中FCC和BCC相的自由能, 计算结果显示Finnis-Sinclair型势函数[29]所描述的FCC相自由能存在一个局部最小值, 表明FCC相在FCC-BCC复相体系中处于亚稳态, 它在转变为BCC相时需要克服一定的能垒. 更重要的是, 该能垒有利于在FCC-BCC复相体系中获得结构稳定的FCC/BCC相界面, 从而为后续界面迁移模拟提供合理的界面初始模型[21-26]. 因此, 本文采用Fe的Finnis-Sinclair型势函数作为初始参量, 其对应FCC和BCC相的晶格常数分别为${a_{{\rm{FCC}}}} = 0.3668\;{\rm{ nm}}$和${a_{{\rm{BCC}}}} = 0.2996\;{\rm{ nm}}$[29]. 由于FCC和BCC晶体结构和晶格常数存在差异, 导致两相在界面上通常存在晶格错配. 根据Finnis-Sinclair型势函数对应的FCC和BCC相晶格常数, 计算可得两相晶格在y轴方向(即$[7\;\overline{10}\;7]_{\rm FCC}$//${[\overline 5 70]_{{\rm{BCC}}}}$)上的错配度δ较小(< 0.1%), 即在该方向上两相能够较好地匹配. 因而本模拟中相界面的晶格错配主要存在于x轴方向(即${[10\overline 1 ]_{{\rm{FCC}}}}$//${[001]_{{\rm{BCC}}}}$)上. 图2(a)是两相在x轴方向上的晶格匹配状态图, 其中Fe原子根据中心对称参数(centrosymmetry parameter, CSP)着色, 该原子参数为大于0的数值, 表征原子近邻晶格的无序程度, 可用于描述晶体材料中的位错缺陷结构. 对于占据理想晶格点阵位置的原子, 其CSP值为0, 而位于晶格缺陷如位错附近的原子, 由于晶格畸变而导致该原子CSP值为一较大数值. 从图2(a)中可见, 自然状态下FCC和BCC相在x轴方向上存在明显的晶格错配, 这些错配可用在FCC/BCC相界面沿x轴方向分布并具有一定间距的失配位错阵列(misfit dislocation array)表示, 显然该相界面在自然状态下为非共格界面. 图 2 自然状态(a)和约束共格状态(b)下FCC和BCC相晶格点阵在x轴方向上的匹配状态图 Figure2. The atomic configuration of interphase boundary between FCC and BCC crystals in (a) natural state and (b) constraint coherent state (in which the atoms are colored by their centrosymmetry parameters).
图6是约束共格复相体系FCC→BCC晶体结构转变过程模拟结果及宏观尺度相界面的位置-时间关系曲线, 其中相界面位置以界面原子z坐标的统计平均值进行标定. 在相变过程中, FCC/BCC相界面沿其界面法线方向稳定迁移至25 ps时转变完成, 且在迁移过程中与两相${(575)_{{\rm{FCC}}}}$//${(750)_{{\rm{BCC}}}}$初始界面始终保持平行, 并在宏观尺度下呈现相对平直的界面结构特征[5,6], 在此期间未观察到有新生BCC相在FCC晶体内部的形核. 从相界面位置-时间关系曲线可见, 相界面沿法线方向的迁移距离与时间呈单调线性变化关系, 且相界面具有接近恒定的迁移速度, 并可通过线性拟合相界面位置-时间关系曲线计算获取. 为了更好地反映本文所采用相界面模型的可靠性并评价模拟结果的数值偏差或不确定度, 我们在同一FCC/BCC双相系统初始结构模型下采用10组具有不同模拟盒子尺寸和原子总数量的模拟参数分别进行计算, 结果显示FCC/BCC相界面的迁移速度是(4.4 ± 0.3) × 102 m/s. 此外, 相变过程中复相体系产生一个平行于相界面的剪切应变εyz以及一个垂直于相界面的法向应变εzz, 其数值分别为0.349和0.053, 这是非扩散切变型马氏体相变的典型特征. 图 6 FCC→BCC相变中约束共格台阶型相界面位置随时间变化关系曲线及晶体结构转变过程模拟结果 Figure6. Curves of the constraint coherent FCC/BCC boundary location versus time within 20 ps and snapshots of the crystal structure evolution process by MD simulation.
在另一方面, 可以用FCC和BCC晶体的${(111)_{{\rm{FCC}}}}$和${(110)_{{\rm{BCC}}}}$密排面构建非台阶型FCC/BCC相界面初始模型, 其界面结构如图7所示, 图中Fe原子根据其势能值进行着色. 该相界面为半共格界面, 两相间的晶格错配通过界面内两组失配位错${{{b}}_1} = \dfrac12{[1\overline 1 0]_{{\rm{FCC}}}}// \Big(\dfrac12{[\overline 1 11]_{{\rm{BCC}}}}\Big)$和${{{b}}_2} = \dfrac12 {[10\overline 1 ]_{{\rm{FCC}}}}$$\Big(//\dfrac12{[1\overline 1 1]_{{\rm{BCC}}}}\Big)$松弛以降低界面的应变能. 图 7 FCC和BCC晶体非台阶型${(111)_{{\rm{FCC}}}}$//${(110)_{{\rm{BCC}}}}$相界面初始结构 Figure7. View on the FCC/BCC boundary of ${(111)_{{\rm{FCC}}}}$//${(110)_{{\rm{BCC}}}}$ interface (in which the atoms are colored in terms of their potential energy).
图8是非台阶型${(111)_{{\rm{FCC}}}}$//${(110)_{{\rm{BCC}}}}$相界面迁移及晶体结构演变过程模拟结果, 同上, 晶体结构根据原子配位数进行着色. 图8中新生BCC相通过在初始BCC晶体${(110)_{{\rm{BCC}}}}$密排面上外延形核并生长, 由于失配位错的Burgers矢量b1和b2平行于相界面, 因此界面的迁移需要借助失配位错在热激活作用下沿界面法线方向作非保守攀移, 导致相界面在迁移过程中呈现出粗糙界面, 与实验观察中马氏体相界面具有相对平直的形貌特征不符. 图 8${(111)_{{\rm{FCC}}}}$//${(110)_{{\rm{BCC}}}}$非台阶型相界面迁移及两相晶体结构演变过程模拟结果 (a) 0 ps; (b) 15 ps; (c) 30 ps Figure8. Snapshots of the evolution of the local structure and propagation of the ${(111)_{{\rm{FCC}}}}$//${(110)_{{\rm{BCC}}}}$ boundary at different times: (a) 0 ps; (b) 15 ps; (c) 30 ps (in which the atoms are colored by their coordinate number where green: FCC, red: BCC and yellow: phase boundary).
23.3.马氏体相界面迁移微观机制 -->
3.3.马氏体相界面迁移微观机制
如图9所示, 将约束共格复相体系在10.0 ps时台阶面上的原子构型与10.2 ps时的原子构型重叠获得原子位移前后状态图, 对Fe原子使用CNA方法标识晶体结构并着色, 绿色和蓝色分别代表FCC和BCC相, 白色为未知结构. 图中红色箭头表示原子位移方向, 其长度表征位移量. 相变过程中FCC相在x方向上的界面原子列A1沿台阶面${[1\overline 2 1]_{{\rm{FCC}}}}$方向整体切变式迁移并占据BCC相的A2位置, 即FCC晶格沿${[1\overline 2 1]_{{\rm{FCC}}}}$方向产生剪切形变, 同时FCC相的$\dfrac12 {[101]_{{\rm{FCC}}}}$转变为BCC相的${[010]_{{\rm{BCC}}}}$晶格矢量, 实现FCC晶格向BCC晶格的结构转变. 图 9 台阶面上原子在10.0和10.2 ps 时的位移前后状态图 Figure9. Atomic displacements on the terrace plane of (111)FCC//(110)BCC (in which the transformed configuration at 10.2 ps is superposed over the configuration at 10.0 ps).
图10是约束共格复相体系在10.0和10.2 ps时FCC/BCC相界面的微观结构及界面近邻原子位移状态图, 显然CNA方法能辨识相界面近邻两侧的FCC和BCC晶体结构, 并可清晰展示出相界面的台阶形貌特征. 根据马氏体相变拓扑模型理论, 图中相界面的台阶前沿存在一组相变位错结构阵列, 如图10(a)中白色原子所示, 其位错线沿x方向. 计算可得相变位错的Burgers矢量bD垂直于台阶面的分量${{b}}_ \bot ^{\rm{D}}$为8.095 × 10–4, 远小于平行于台阶面的分量${{b}}_{//}^{\rm{D}}$ = -0.727, 且它对台阶面上的共格应变不具有松弛作用. 实际上, 相变位错平行于台阶面的Burgers矢量分量${{b}}_{//}^{\rm{D}}$为FCC晶体中1/2 Shockley不全位错, 即$\dfrac12 {[1\overline 2 1]_{{\rm{FCC}}}}\Big(//\dfrac16 {[\overline 1 01]_{{\rm{BCC}}}} \Big)$, 它垂直于位错线即x轴方向, 具有刃型位错结构特征. 该相变位错阵列随相变的进行而沿台阶面${[1\overline 2 1]_{{\rm{FCC}}}}$//${[1\overline 1 0]_{{\rm{BCC}}}}$方向滑移, 计算可得相变位错的平均迁移速度高达(2.8 ± 0.2) × 103 m·s–1. 从原子位移状态图10(c)可见, 相变位错滑移过程中相变位错应力场引起界面前沿FCC相原子在${(111)_{{\rm{FCC}}}}$密排面上沿${[1\overline 2 1]_{{\rm{FCC}}}}$方向作整体切变式迁移, 在界面处Fe原子的堆垛结构由FCC晶体${(111)_{{\rm{FCC}}}}$密排堆垛转变为BCC晶体${(110)_{{\rm{BCC}}}}$的堆垛结构, 即界面处FCC相${(111)_{{\rm{FCC}}}}$晶面切变为BCC相${(110)_{{\rm{BCC}}}}$晶面, 相变位错滑过区域内的FCC相转变为BCC相. 综上, 相变位错的保守型滑移对FCC向BCC晶体结构转变起到关键作用, 马氏体相变的原子机制是相变位错沿台阶面的横向滑移. 图 10 约束共格FCC/BCC相界面在 (a) 10.0 ps和 (b) 10.2 ps时的微观结构以及相界面近邻原子位移状态图(c) Figure10. Configuration of the step-like constraint coherent interface between FCC and BCC crystals at (a) 10.0 ps and (b) 10.2 ps evaluated by common neighbor analysis method; and (c) the atomic displacements near the transforming boundary.
另一方面, 相变位错除了作横向保守型滑移导致相变进行之外, 同时还起到松弛两相晶体内共格应变的作用. 图11是1.0 ps时FCC和BCC两相以及界面附近的应变分布图, 其中应变值根据各原子相对于截断半径3.2 ?内近邻原子的位移计算获得[33]. 两相晶格在台阶面上处于约束共格状态, 台阶面上FCC晶体一侧存在拉应变, 而BCC晶体为压应变. 虽然共格应变施加于两相晶体, 但在相变开始后模拟体系逐渐驰豫至平衡状态, 两相晶体内的共格应变场得到松弛, 并只局限于台阶面近邻处, 而远离相界面的晶体内部并不存在显著的长程应变. 值得一提的是, 有HRTEM实验报道在ZrO2陶瓷材料的马氏体惯习面上观察到类似的共格应变特征[34], 表明马氏体相界面共格应变为短程应变场, 其在两相晶体内的长程作用通过间隔分布的相变位错阵列得到有效松弛. 图 11 FCC和BCC相在1.0 ps时约束共格台阶面上的应变分布图 Figure11. Distribution of the strain field along the constraint coherent terrace plane calculated by relative displacements of atoms’ neighbors within a given cutoff radius of 3.2 ?.
23.4.FCC→BCC马氏体相变晶体学特征参量计算 -->
3.4.FCC→BCC马氏体相变晶体学特征参量计算
如图12所示, 在台阶面上建立坐标系(${x_{{\rm{TP}}}}$, ${y_{{\rm{TP}}}}$, ${z_{{\rm{TP}}}}$)以确定台阶面内的共格应变. 根据FCC和BCC满足N-W位向关系, 可知台阶面上存在${x_{{\rm{TP}}}}$和${y_{{\rm{TP}}}}$方向的共格应变, 其中, 具有特定间距的相变位错阵列可松弛${y_{{\rm{TP}}}}$方向的共格应变, 而对于${x_{{\rm{TP}}}}$方向的共格应变则往往需要另一组晶体缺陷阵列加以松弛, 但是, 在本文的拓扑模型计算中将抑制该方向上的共格应变, 从而能够得到具有一组相变位错阵列的相界面结构特征参数, 进而与MD模拟结果进行对比分析. 根据本文采用的Finnis-Sinclair势函数, 计算可得台阶面上沿${y_{{\rm{TP}}}}$方向的共格应变${\varepsilon _{yy}} = 0.05{\rm{85}}$. 根据(2)式得到相变位错Burgers矢量在台阶面坐标系下的表达式为${{{b}}^{\rm{D}}} \!= \![0{{, - 0}}{{.727474, 8}}{{.095331}} \!\times \!{{1}}{0^{{{ - 4}}}}{{]}}$, 其中, ${{t}}_{{\rm{Martensite}}}\! = $$ {[010]_{{\rm{BCC}}}}$, 且其位错线方向ξD平行于x轴. 图 12 拓扑模型中台阶型FCC/BCC相界面上的相变位错结构示意图 Figure12. Schematic illustration of the transformation dislocation configuration at the stepped FCC/BCC boundary in the topological model.
表1FCC → BCC相变晶体学特征参量的分子动力学模拟结果与拓扑模型计算值的比较 Table1.Comparison of the FCC → BCC transformation crystallographic characteristics obtained by MD simulation and topological model.