Presentation Information
[11a-B21-6]Adsorption state study of Cu diketonates on Cu surface with Neural Network Potential
〇Yuxuan Wu1, Yichen Zou1, Noboru Sato1, Atsuhiro Tsukune1, Yukihiro Shimogaki1 (1.Univ. Tokyo)
Keywords:
Atomic layer deposition,Neural Network Potential,Cu diketonates
Atomic layer deposition (ALD) is a key thin-film deposition technique for semiconductor fabrication because of its atomic-scale thickness control and excellent conformality.Cu diketonates, like Cu(acac)2 and Cu(tmhd)2, have been used in both Cu ALD and CVD. Meanwhile, studies on the surface chemistry and adsorption mechanisms of both precursors are inadequate and dissociative adsorption mechanism are unclear. In this study, we employ an updated neural network potential (NNP) to investigate the adsorption behavior relevant to Cu ALD for two Cu dieketonates: Cu(acac)2 and Cu(tmhd)2. However, predicting these reaction mechanisms by DFT is computationally limited for large Cu diketonates since calculating dissociative adsorption pathways requires large slab models and extensive sampling. Recently, NNP calculations have emerged as a powerful and computationally efficient approach for studying ALD reaction chemistry and adsorption behavior at the atomic scale. Furthermore, this machine-learning-based method enables efficient simulations on large surface models within minutes. Using Matlantis™, we investigated the adsorption behavior of Cu(acac)2 and Cu(tmhd)2 on the Cu(111) surface. Calculations show the dissociative adsorption pathways of Cu(acac)2 and Cu(tmhd)2, respectively. Both precursors exhibited well-defined dissociative adsorption pathways that are consistent with previous DFT calculations. Although the two precursors followed similar dissociative adsorption mechanisms, quantitative differences remain in the predicted adsorption energies compared with DFT calculations.