Vertical hetero-integration of 2D-materials-based integrated circuits for artificial intelligence

The monolithic 3D integration of 2D material-based electronics realizing artificial intelligence (AI) processing is demonstrated. Six layers transistor and memristor layers are vertically integrated into a 3D nanosystem with unprecedented integration density and multifunctionality.
Vertical hetero-integration of 2D-materials-based integrated circuits for artificial intelligence
Like

The vision of stacking different semiconductor layers [1, 2] permits novel three-dimensional (3D) integration and van der Waals integration scenarios [3, 4] to create vertically integrated electronic and optoelectronic systems addressing current system-on-a-chip limitations [5-7]. To this end, monolithic 3D (M3D) integration showcase efficient connection of chips with the absence of rigid wafers. Recent advents on atomically thin two-dimensional materials [8-11] offers excellent candidates for electronic devices with extremely low stiffness, thickness, and decent performances [12].

A recently published article, led by MIT, Washington University in St. Louis, Yonsei University, and international collaborators, reported the  successful monolithic 3D integration of stacked 2D material-based electronic circuits demonstrating  artificial intelligence (AI) hardware functionality. As illustrated in Fig. 1, all layers can be monolithically integrated into the 3D heterostructure: an AI computing layer consists of a 2D material-based memristor array (top) and 2D material-based transistor array (middle) to construct an integrated 2D-based AI processors (bottom). "The fully monolithic-3D-integrated AI system significantly reduces processing time, voltage drops, latency, and footprint due to its densely packed AI processing layers with dense interlayer connectivity". The successful realization of the M3D system synergize the benefits of 2D materials for advanced hetero-integrated electronics, paving the way for unrivaled multifunctional computing hardware architecture with ultimate parallelism.

F1
Figure 1. Schematic  illustration of ultimate edge computing system based on monolithic 3D (M3D) integrated 2D material-based electronics.  

Article information

Ji-Hoon Kang, Heechang Shin, Ki Seok Kim, Min-Kyu Song, et al. Monolithic 3D integration of 2D materials-based electronics towards ultimate edge computing solutions, Nature Materials (2023).
DOI: 10.1038/s41563-023-01704-z

Link:  https://www.nature.com/articles/s41563-023-01704-z

References

[1] S.-H. Bae, et al. Nature Materials, 18,  550–560 (2019).
[2] H. S. Kum, et al. Nature, 578, 75-81 (2020).

[3]  Y. Liu, et al. Nature, 567, 323–333 (2019). https://www.nature.com/articles/s41586-019-1013-x

[4] Y. Meng, et al. Nature Reviews Materials, 8, 498–517 (2023). https://www.nature.com/articles/s41578-023-00558-w

[5] C. Choi, et al. Nature Electronics, 5, 386–393 (2022).

[6] J. Shin, et al. Nature, 614, 81–87 (2023). https://www.nature.com/articles/s41586-022-05612-1

[7] Y. Meng, et al. Light: Science & Applications, 10, 235 (2021)

[8] W. Kong, et al. Nature Nanotechnology, 14, 927–938 (2019).

[9] K. S. Kim, et al. Nature, 614, 88–94 (2023)

[10] H. Zhong, et al. Nanoscale Horizons,  8, 1345-1365 (2023)

[11] H. Kim, et al. Nature Nanotechnology, 17 , 1054-1059 (2022). https://www.nature.com/articles/s41565-022-01200-6

[12] T. He, et al. Nature Photonics (2023). https://www.nature.com/articles/s41566-023-01309-7

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in