Abstract
Developing high dielectric permittivity material is vital to satisfy the ongoing demands for the miniaturization of electronic and energy storage devices. Recent investigations uncover the role of a thermodynamical tricritical phenomenon on enhancing the dielectric response. However, such a tricritical point always locates in an extremely narrow composition region, which makes it time-consuming for exhaustive experimental searching of the optimal dielectric permittivity in a given material system. In the present paper, we employ an accelerated discovery strategy to seek the largest dielectric permittivity in Ba(Ti1-x%Hfx%)O-3 ceramic material by using an iterative method between computational machine learning and the experimental synthesis and property measurement. The optimal composition is found to be x = 11 with the highest permittivity of epsilon(r) = 4.5 x 10(4) after 4 loops of iteration involving 6 compositions, which shows higher efficiency compared with conventional experimental searching. Further thermal analysis study suggests that such a permittivity-maximum location on the phase diagram is indeed a tricritical point. Moreover, the microstructure investigation by TEM observation indicates that the tricritical point shows a mottled morphology consisting of numerous nanodomains with multiple phases coexisting, and a phenomenological thermodynamic model based on the experimental result implies that the tricriticality is responsible for the enhanced dielectric permittivity.