Oxygen mobility and ferroelectric switching in scaled hafnia synapses

On conventional computers, the performance of AI models is limited by the data transfer between the memory and the processor. Compute-in-Memory architectures offer a new paradigm: Vector-Matrix Multiplications may be performed by a voltage drop through a matrix of programmable resistances, the “synaptic weights”. Ferroelectric materials are excellent candidates for their realization: in a two- or three-terminals geometry and in combination with a semiconducting oxide,^[1–3] the conductance is programmed by controlling the configuration of the ferroelectric domains.
Single crystal perovskite oxide ferroelectrics such as BiFeO₃ exhibit a sharp coercive field distribution and are excellent candidates for unsupervised learning schemes.^[4] We will see that they are also suitable for selector-less operations in crossbar arrays.^[5] In recent years, interest in ferroelectric hafnium oxide, which can be deposited using atomic layer deposition, has grown significantly. The unique fluorite unit cell allows for the stabilization of ferroelectricity below 3 nm,^[6] facilitating the scaling of synaptic weights. The relatively low crystallization temperature of polycrystalline hafnium oxide / zirconium oxide superlattices (HZO-SL) allows the Back-End-Of-Line (BEOL) integration of functional devices to CMOS transistors.^[7,8]
The mechanisms governing the resistive switching in WO_x / HZO-SL (5 nm) bilayers are discussed. The effect of the programming pulse duration and amplitude on the polarization switching are investigated, from milliseconds to nanoseconds timescales. Devices of different sizes and shapes are measured down to 500 nm in dimension. For an octagonal device size of 1 micrometer, an On/Off ratio as high as 8 is obtained for 20 ns pulses, a 4-fold improvement compared to 40 um devices. A gradual RESET is observed as every timescale, however we found that the SET becomes sharper as the pulse duration decreases. The results suggest the exchange of oxygen between the WO_x and the hafnia at timescales longer than microseconds.
[1] L. Bégon-Lours, M. Halter, F. M. Puglisi, L. Benatti, D. F. Falcone, Y. Popoff, D. D. Pineda, M. Sousa, B. J. Offrein, Advanced Electronic Materials 2022, 2101395.
[2] M. Halter, L. Bégon-Lours, V. Bragaglia, M. Sousa, B. J. Offrein, S. Abel, M. Luisier, J. Fompeyrine, ACS Appl. Mater. Interfaces 2020, 12, 17725.
[3] M. Halter, L. Bégon-Lours, M. Sousa, Y. Popoff, U. Drechsler, B. J. Offrein, COMMUNICATIONS MATERIALS 2023, 4, DOI 10.1038/s43246-023-00342-x.
[4] S. Boyn, J. Grollier, G. Lecerf, B. Xu, N. Locatelli, S. Fusil, S. Girod, C. Carrétéro, K. Garcia, S. Xavier, J. Tomas, L. Bellaiche, M. Bibes, A. Barthélémy, S. Saïghi, V. Garcia, Nature Communications 2017, 8, 14736.
[5] M. Halter, E. Morabito, A. Olziersky, C. Carrétéro, A. Chanthbouala, D. F. Falcone, B. J. Offrein, L. Bégon-Lours, Journal of Materials Research 2023, 38, 4335.
[6] L. Bégon-Lours, M. Halter, M. Sousa, Y. Popoff, D. D. Pineda, D. F. Falcone, Z. Yu, S. Reidt, L. Benatti, F. M. Puglisi, B. Offrein, Neuromorph. Comput. Eng. 2022, 2, DOI 10.1088/2634-4386/ac5b2d.
[7] L. Bégon-Lours, S. Slesazeck, D. F. Falcone, V. Havel, R. Hamming-Green, M. M. Fernandez, E. Morabito, T. Mikolajick, B. J. Offrein, Advanced Electronic Materials 2024, n/a, 2300649.
[8] R. Hamming-Green, M. S. Ram, D. F. Falcone, B. Noheda, B. J. Offrein, L. Bégon-Lours, in 2024 8th IEEE Electron Devices Technology & Manufacturing Conference (EDTM), IEEE, Bangalore, India, 2024, pp. 1–3.