中文摘要：

制备了bcc结构Al-Fe-Mn和fcc结构Al-Cu-Ni三元铝合金单相固态半无限长和有限长（或薄膜型）扩散偶,通过EPMA测得相应成分-距离曲线,并采用实用型数值回归法计算了相应合金体系1273 K时随成分变化的互扩散系数,所得结果均满足热力学稳定条件.基于所测得互扩散系数及Fick第二定律,重现了实验测定的成分-距离曲线,证实了所获得互扩散系数的可靠性.进一步分析结果表明,该实用型数值回归法不仅可以高效准确地计算单相三元扩散偶随成分变化的互扩散系数,还能很好地解决传统方法不能处理的情况,如扩散通道上无交点的扩散偶和有限长（或薄膜型）扩散偶等.

英文摘要：

Owing to excellent mechanical properties, Al alloys are widely used in aerospace, automotive and civil industry. In order to optimize the properties and performance of the currently used Al alloys and/or even de- sign novel Al alloys, the quantitative description of the microstructure during alloys preparation is the key. In re- cent years, the phase-field simulation coupling with the CALPHAD thermodynamic and atomic mobility databases has become an effective way to quantitatively simulate the microstructure evolution. So far, the accurate thermody- namic database for Al alloys has been established. However, it is not the case for atomic mobility database for Al al- loys. The major obstacle lies in the lack of reliable diffusion coefficients in ternary and higher-order Al alloys, and thus there is an urgent need to remedy this situation. In this work, several semi-infinite and finite （thin film） single- phase solid-state diffusion couples in bcc Al-Fe-Mn and fcc Al-Cu-Ni alloys were first prepared. The concentration profiles for all the diffusion couples were then measured by means of EPMA. After that, the pragmatic numerical inverse method, which has been recently developed for high-throughput determination of the interdiffusion coeffi-cients in ternary system and validated in several systems, was employed to compute the composition-dependent in- terdiffusivities in the corresponding systems at 1273 K. In order to eliminate the possibility that different interdiffu- sivities at the same composition would be obtained from different sets of diffusion couples, only one set of adjust- able parameters was used for one system. All the obtained interdiffusivities satisfy the thermodynamic constrains. On the basis of the determined interdiffusivities as well as Fick＇s second law, all the experimental concentration profiles were reproduced nicely via numerical simulation, which verifies the reliability of the determined interdiffu- sivities. The further analysis indicates that the pragmatic numerical inverse method