Machine learning-assisted ultrafast flash sintering of high-performance and flexible silver–selenide thermoelectric devices

Abstract

The first machine learning-assisted ultrafast flash sintering of flexible silver–selenide TE devices. Bayesian optimization of flash sintering variables led to a PF of 2205 μW m−1 K−2 and a zT of 1.1 at room temperature realized with a sintering time less than 1.0 second.

Advances in Ag2Se-based thermoelectrics from materials to applications

“…Furthermore, the sintering treatment can effectively reduce the n of 1.14 × 10 19 cm −3 of the unsintered film to 7.3 × 10 18 cm −3 of the film sintered under optimum variables. 160 The reduced n leads to a high S of −161.7 μV K −1 and a moderate σ of 840 S cm −1 at room temperature. Moreover, the bending test result suggested that the decrement after 1000 bending cycles with a bending radius of 5 mm was less than 9%, indicating the excellent robustness of the sintered film.…”

“…258,259 A data-driven algorithm such as Bayesian optimization (BO) was utilized in a high-throughput experiment to accelerate the discovery of the optimum variables of the fabrication process of Ag 2 Se films. 160…”

“…The workflow begins with the intense pulsed light sintering of Ag 1.9 Se films using original variables such as voltage, the number of pulses, pulse duration, and pulse delay. 160 Then, the thermoelectric performance of the as-fabricated films is measured. The measured S 2 σ and employed original sintering variables are used as the input data for BO to predict the next set of variables to test.…”

“…The measured S 2 σ and employed original sintering variables are used as the input data for BO to predict the next set of variables to test. 160…”

“…Moreover, the bending test result suggested that the decrement after 1000 bending cycles with a bending radius of 5 mm was less than 9%, indicating the excellent robustness of the sintered film. 160…”