Application of the Quasi-Static Memdiode Model in Cross-Point Arrays for Large Dataset Pattern Recognition
F.L Aguirre, S. M.Pazos, F. Palumbo, J. Sune and E. Miranda
IEEE Access, vol. 8, 202174 (2020)
UAB's team investigates the use and performance of the quasi-static memdiode model (QMM) when incorporated into large cross-point arrays intended for pattern classification tasks. Following Chua’s memristive devices theory, the QMM comprises two equations, one equation for the electron transport based on the double-diode circuit with single series resistance and a second equation for the internal memory state of the device based on the so-called logistic hysteron or memory map. Ex-situ trained memdiodes with different MNIST-like databases are used to establish the synaptic weights linking the top and bottom wire networks. The role played by the memdiode electrical parameters, wire resistance and capacitance values, image pixelation, connection schemes, signal-to-noise ratio and device-to-device variability in the classification effectiveness are investigated. The confusion matrix is used to benchmark the system performance metrics. UAB's team shows that the simplicity, accuracy and robustness of the memdiode model makes it a suitable candidate for the realistic simulation of large-scale neural networks with non-idealities.
Role of Oxygen Defects in Conductive-Filament Formation in Y2O3-Based Analog RRAM Devices as Revealed by Fluctuation Spectroscopy
E. Piros, M. Lonsky, S. Petzold, A. Zintler, S.U. Sharath, T. Vogel, N. Kaiser, R. Eilhardt, L. Molina-Luna, C. Wenger, J. Müller, and L. Alff
Phys. Rev. Applied 14, 034029 (2020)
TU-Darmstadt's work, implemented within the WAKeMeUP project and which is the result of their collaboration with Dr. Martin Lonsky and Prof. Jens Müller from the Goethe University Frankfurt, has been selected as an Editors’ pick in the journal of Physical Review Applied (https://journals.aps.org/prapplied/issues/14/3).
In this publication, they take the reader on a noisy journey inside an yttrium oxide-based RRAM cell as we investigate its oxygen defect properties through low-frequency fluctuation spectroscopy. They find that the high resistive state, as well as intermediate resistive states accessed by varying the reset stop voltages, display a universal 1/fα-type noise behavior with a frequency exponent of α≈1.2 that is independent of the applied reset voltage or the device resistance. This universal behavior is attributed to the abundance of oxygen vacancies intrinsic to yttrium oxide. Additionally, they observe the striking phenomenon of systematic noise magnitude reduction in the high resistive state with dc training. They explain this effect as the stabilization of the conductive filament via the consumption of oxygen vacancies, thus reducing the number of active fluctuators in the filament vicinity.
Tailoring the Switching Dynamics in Yttrium Oxide‐Based RRAM Devices by Oxygen Engineering: From Digital to Multi‐Level Quantization toward Analog Switching
S. Petzold, E. Piros, R. Eilhardt, A. Zintler, T. Vogel, N. Kaiser, A. Radetinac, P. Komissinskiy, E. Jalaguier, E. Nolot, C. Charpin‐Nicolle, C. Wenger, L. Molina‐Luna, E. Miranda, L. Alff
Adv. Electron. Mater., Early View, 2000439 (2020)
In the framework of the WAKeMeUP project and in close collaboration with UAB and CEA-LETI, TU-Darmstadt investigated the transition from digital to gradual or analog resistive switching in yttrium oxide‐based resistive random‐access memory devices.
In their work, they demonstrate that by reducing the oxygen content of the functional layer it is possible to achieve a transition from abrupt switching to gradual characteristics. The beneficial effects of oxygen-engineering are confirmed by thorough structural and electrical analysis. They show that all operation voltages (electroforming, set, reset) can be lowered, therefore improving intra-device variability. Moreover, it is revealed that an increasing amount of homogeneously distributed oxygen vacancies in the yttria layer results in an improved accessibility of intermediate, quantized resistance states. In consequence, the set and reset processes approach an analog behavior, as required for multi-bit and neuromorphic applications. A physics-based theoretical model is proposed to account for the observed non-linear conductance quantization.
Enhanced thermal stability of yttrium oxide-based RRAM devices with inhomogeneous Schottky-barrier
E. Piros, S. Petzold, A. Zintler, N. Kaiser, T. Vogel, R. Eilhardt, C. Wenger, L. Molina-Luna, and L. Alff
Appl. Phys. Lett. 117, 013504 (2020)
Within the WAKeMeUP project, TU-Darmstadt studies the thermal stability of bipolar resistive switching in yttria-based resistive random-access memory (RRAM) devices.
By investigating the DC switching characteristics at three different operation temperatures, namely at 25 ° C, 85 °C, and 125 °C, an overall good temperature immunity is revealed for the operation voltages and the resistance levels. For both the low and high resistive state resistances, a semiconductive behavior is observed. Also, the set and reset voltages are found to scale inversely with increasing temperature. The uncommon behavior of the Schottky-barrier height being temperature-dependent (showing an increase from approximately 1.02 eV at 25 °C to approximately 1.35 eV at 125 °C, see fig.) is explained by local barrier height variance at the interface.
Back‐End‐of‐Line Compatible Low Temperature Furnace Anneal for Ferroelectric Hafnium Zirconium Oxide Formation
This paper was highlighted as a cover-page of the "Physica Status Solidi (a) - Application and Materials Sciences" Journal.
Physica status solidi (a)
Self-Organizing Neural Networks Based on OxRAM Devices under a Fully Unsupervised Training Scheme
Journal of Applied Physics 125, 234503 (2019)
Impact of roughness of TiN bottom electrode on the forming voltage of HfO2 based resistive memories
Microelectronic Engineering Volume 221, 111194, 15 January (2020)
Local crystallographic phase detection and texture mapping in ferroelectric Zr doped HfO2 films by transmission-EBSD
The local crystal phase and orientation of ferroelectric grains inside TiN/Hf0.5Zr0.5O2/TiN have been studied by the analysis of the local electron beam scattering Kikuchi patterns, recorded in transmission. Evidence was found that the ferroelectric phase of the layers is derived from an orthorhombic phase, most likely of space group Pca21. The orientation analysis reveals a strong out-of-plane texture of the polycrystalline film which is in accordance with a high remanent polarization Pr observed for P-V measurements. The results of this analysis help us to further optimize the ratio of ferroelectric grains and their orientation for many applications, e.g., in the field of emerging memory or infrared sensors.
Appl. Phys. Lett. 115, 222902 (2019); https://doi.org/10.1063/1.5129318
Forming-Free Grain Boundary Engineered Hafnium Oxide Resistive Random Access Memory Devices
Figure 1: investigated grain boundary threading the dielectric layer, and thus, interconnecting both électrodes
Stefan Petzold; Alexander Zintler; Robert Eilhardt; Nico Kaiser; Eszter Piros; Sankaramangalam Ulhas Sharath; Tobias Vogel; Márton Major; Keith Patrick McKenna; Leopoldo Molina-Luna; Lambert Alff
Advanced Electronic Materials, 5, 1900484 (2019)
Within the project of WAKeMeUP, the Advanced Thin Film Technology group (TU Darmstadt) of Prof. Alff developed a new method of directed defect engineering, namely grain boundary engineering.
With this new method, it is possible to create low energy grain boundaries of high symmetry which thread the whole dielectric layer, thus, providing a defined path for facilitated filament formation. This method renders forming free devices with a narrow distribution of forming voltages which makes it a powerful tool to overcome one of the main challenges for resistive switching devices: device-to-device variation. An exemplary transmission electron microscopy image and further analysis of the engineered grain boundaries is depicted in the Figure 1.
Comprehensive Phase-Change Memory Compact Model for Circuit Simulation (CEA-LETI, 2018):
Heavy Ion Radiation Effects on Hafnium Oxide based Resistive Random Access Memory
S. Petzold; S.U. Sharath; J. Lemke; E. Hildebrandt; C. Trautmann; L. Alff
IEEE Transactions on Nuclear Sciience, Early Access (2019)
Within the project of WAKeMeUP (https://www.ecsel.eu/projects/wakemeup), funded by BMBF and ECSEL JU, TU Darmstadt investigated the effect of ionizing radiation on resistive switching devices based on hafnium oxide.
Devices were exposed to high energetic heavy ion irradiation with Energies of 5.6 MeV per nucleus resulting in an overall Ion energy of 1.1 GeV. The effect of heavy ion irradiation on different electrical statistics of RRAM devices was investigated showing extraordinary data retention towards ionizing radiation. The written information was found to be stable up to very high total ionization doses of up to 880 Mrad. The evaluation of the data retention experiment of 177 tested devices for a heavy ion fluence of 7x1011 ions/cm2 (here about 6.3 million heavy ions hit one device) is depicted in Fig. 1. As can be seen, the devices in the high resistance states tend to become more resistive, while the devices in the low resistance states become more conducting, overall maintaining their initially written information. Therefore, RRAM devices based on hafnium oxide are a very interesting candidate for application in high-dose environments, as for example in satellites or other space applications potentially replacing FLASH.
Analysis and simulation of the multiple resistive switching modes occurring in HfOx-based resistive random access memories using memdiodes
S. Petzold, E. Miranda, S. U. Sharath, J. Muñoz-Gorriz, T. Vogel E. Piros, N. Kaiser, R. Eilhardt, A. Zintler, L. Molina-Luna, J. Suñé, and L. Alff
Journal of Applied Physics 125 (23), 234503 (2019)
Within a cooperation of Technische Universität Darmstadt and Universitat Autònoma de Barcelona in the WAKeMeUP-project, a simulation of all observed switching modes reported for hafnium oxide based resistive switching devices has been achieved.
Comparing stoichiometric vs. highly deficient hafnia devices by compact simulation, we found that a single or two memdiodes connected in anti-series can be utilized to simulate all reported switching modes. We found that all switching modes of the stoichiometric device could be simulated utilizing only one memdiode, while for the deficient device a second memdiode in opposite polarity was necessary to fit its switching modes. By correlating the different interfaces to the two memdiodes taken for compact simulation, we were able to gain additional information of the switching process, as e.g. the voltage drop at the individual interfaces.
Gradual Reset and Set Characteristics in Yttrium Oxide based Resistive Random Access Memory
S. Petzold, E. Piros, S. U. Sharath, A. Zintler, E. Hildebrandt, L. Molina-Luna, C. Wenger, and L. Alff
Semicond. Sci. Technol., Early Access (2019)
Within the project of WAKeMeUP (https://www.ecsel.eu/projects/wakemeup), funded by BMBF and ECSEL JU, we investigated the resistive switching behavior of yttrium oxide based devices.
In this study, Technische Universität Darmstadt reports for the first time the coexistence of bipolar resistive switching and unipolar resistive switching within one stack combination based on yttrium oxide. The BRS is characterized by low variability, good data retention, high DC endurance, and average on-off-ratio above 100.
The obtained concurrent gradual transitions during BRS set and reset which have been employed in pulse depression and potentiation operations (see figure) with shown plasticity over about two orders of magnitude make RRAM based on yttrium oxide an interesting alternative candidate for neuromorphic synapses. Combined with its characteristic of being forming-free and having low operation voltages, RRAM based on yttrium oxide is highlighted as a suitable material candidate for new high density nonvolatile memory technologies.