Watching crystals grow: In situ TEM Observation of Microstructure Evolution in HfO2 Based Memristive Devices

Abstract number
European Microscopy Congress 2020
Corresponding Email
[email protected]
PST.6 - In-situ and in-operando microscopy
Robert Eilhardt (1), Alexander Zintler (1), Oscar Recalde (1), Déspina Nasiou (1), Dr. Stefan Petzold (2), Prof. Dr. Lambert Alff (2), Prof. Dr. Leopoldo Molina-Luna (1)
1. AEM, Technical University of Darmstadt
2. ATFT, Technical University of Darmstadt


Hafnium oxide

In situ heating

Resistive switching

Abstract text

Summary: During heating in the TEM, local structural changes (crystallization and grain growth) were investigated in situ by Scanning Precession Electron Diffraction (SPED) (4D-STEM) using a MerlinEM Medipix3 (Quantum Detectors) direct electron detector.

Introduction: Resistive random access memory (RRAM) devices are composed of a dielectric layer sandwiched between two metal layers. Using HfO2 as the insulator material is appealing because of the back‑end‑of-line processes compatibility in the current semiconductor fabrication process and good chemical and thermal stability. [1-2] In filament type resistive switching, grain boundaries in the dielectric layer provide predefined breakdown paths for filament formation, as shown by conductive atomic force microscopy (C-AFM). [3] Careful grain boundary engineering with grain boundaries threading through the dielectric is a promising way to significantly lower the electrical stress required for filament formation, thus achieving electroforming free RRAM devices. [4] To give insight into the grain growth mechanism and the development of a grain boundary we performed a heating experiment on a RRAM device with amorphous hafnia inside a TEM while in situ monitoring structural changes with SPED.

Methods/Materials: Stack combinations of Au/Pt/HfO2/TiN/Al2O3 with amorphous hafnia were grown. A cross sectional lamella was prepared by using a Focused Ion Beam (FIB) dual beam system and then transferred onto a Micro Electrical Mechanical System (MEMS) based heating chip. During in situ heating, laterally resolved structural information was acquired by SPED (4D STEM) using a MerlinEM Medipix3 (Quantum Detectors) direct electron detector. Structural changes during in situ heating were further compared to an ex situ annealed stack combination to correlate the very local structural changes observed in the TEM to our macroscopic results.

Results and Discussion: The heating experiment in the TEM revealed, that grain growth in the hafnia layer starts at ~150 °C and the developed grain boundaries thread the complete dielectric layer. The observed microstructural changes inside the TEM coincide with structural changes of the ex situ annealed stack. Furthermore, crystallization of the dielectric layer lowered the electrical stress required for filament formation.

Conclusion: The development of grain boundaries in the hafnia layer was investigated in situ by monitoring grain growth during heating inside a TEM which yields a unique insight into the grain growth mechanism. The ex situ annealed stack complemented structural changes on a very local scale and showed improved electroforming voltages. The results of this study serve as a basis for a direct structure-property correlation in this system. 


[1] Pan, F., et al., Recent progress in resistive random access memories: Materials, switching mechanisms, and performance. Mat. Sci. a. Eng.: R: Reports, 2014. 83: p. 1-59.

[2] Miranda, E.A., et al., Model for the Resistive Switching Effect in HfO2 MIM Structures Based on the Transmission Properties of Narrow Constrictions. IEEE Elec. Dev. Let., 2010. 31(6): p. 609-611.

[3] M. Lanza et al., ‘Grain boundaries as preferential sites for resistive switching in the HfO2 resistive random access memory structures’, Appl. Phys. Lett., vol. 100, no. 12, p. 123508, Mar. 2012.

[4] S. Petzold, et al., Forming-Free Grain Boundary Engineered Hafnium Oxide Resistive

 Random Access Memory Devices, Advanced Electronic Materials. 5 (2019) 1900484.


The authors acknowledge funding from DFG grant MO 3010/3-1 and the European Research Council (ERC) “Horizon 2020” Program under Grant No. 805359-FOXON.