HfO2 and ITO Resistive Random-Access Memory
(2020) FYSM30 20201Synchrotron Radiation Research
Department of Physics
- Abstract
- The purpose of this work is of evaluating the choice of HfO2 and ITO as the dielectric and the top electrode in high performance resistive random-access memory (RRAM), respectively. The study is twofold, as it quantifies performance according to standard figures of merit for this technology, as well as provides insight into the physics of current conduction for this material choice. Results are achieved thanks to the probing of structures processed by the Nano Electronics group in Lund, followed by direct observations on the produced I−V characteristics and more involved data analysis. Specifically, different models according to different mechanisms of current conduction in a dielectric are fitted to the measured data, and physical parameter... (More)
- The purpose of this work is of evaluating the choice of HfO2 and ITO as the dielectric and the top electrode in high performance resistive random-access memory (RRAM), respectively. The study is twofold, as it quantifies performance according to standard figures of merit for this technology, as well as provides insight into the physics of current conduction for this material choice. Results are achieved thanks to the probing of structures processed by the Nano Electronics group in Lund, followed by direct observations on the produced I−V characteristics and more involved data analysis. Specifically, different models according to different mechanisms of current conduction in a dielectric are fitted to the measured data, and physical parameter extraction is used as a mean to evaluate goodness of fit. It is discovered that several modes that were thought to potentially describe conduction in RRAM can instead not be the case, for confusing (if not impossible) parameters are extrapolated for them. An important result is thus that no matter how great a fit may be numerically, one cannot simply claim that a given mechanism of conduction is a good physical interpretation of a studied system without corroborating its fit with relevant parameter extraction. It is concluded that the only reasonable and available interpretation of measured devices is that of Ohmic type conduction, followed by space-charge (at higher voltages) in the high resistance state of RRAM, and solely Ohmic type in the low resistance state. The latter relates to not observing conduction through a barrier, which was expected as HfO2 and ITO RRAM has previously been found to be modulated by a self-compliance effect. Based on this interpretation, a computer model is developed in Python, and its output is put to the test by comparison with data from a different sample (i.e. not related to the observations that lead to the creation of the model itself). The universality of the model is found to be satisfying, hinting at the achieved interpretation being adequate. Concerning raw performance of the studied devices, resistive switching at low and steady voltages is observed, which relates to envisioning low power operation for HfO2 and ITO structures. (Less)
- Popular Abstract
- The computer memory technology that we currently rely on is either fast, when it comes to writing and reading programmed states, yet volatile (i.e. it loses its programmed state if power is not being supplied to it), or non-volatile yet slow in terms of write operations and facing issues with further scaling (reduction in size). It should then be obvious that finding a replacement for the above is an important task, as volatility is an undesired trait that implies power consumption, and slow and non-scalable devices are inevitably going to become a bottleneck for new, high performance systems. Multiple new technologies have been proposed, yet a common opinion seems to be that resistive random-access memory (RRAM) is the most promising... (More)
- The computer memory technology that we currently rely on is either fast, when it comes to writing and reading programmed states, yet volatile (i.e. it loses its programmed state if power is not being supplied to it), or non-volatile yet slow in terms of write operations and facing issues with further scaling (reduction in size). It should then be obvious that finding a replacement for the above is an important task, as volatility is an undesired trait that implies power consumption, and slow and non-scalable devices are inevitably going to become a bottleneck for new, high performance systems. Multiple new technologies have been proposed, yet a common opinion seems to be that resistive random-access memory (RRAM) is the most promising alternative for the future of computer memory.
To comprehend the benefits of RRAM a good approach is to first understand how it works, and only then draw conclusions about its implications for the current state of memory applications. Imagine the task of having to make a stream of water pass through a patch of compact soil. The latter, in its pristine state, will impose a high resistance to water flow, as no paths are available in it for water to cross it. However, if a high enough water inlet pressure is applied, the soil will eventually reshape and a conductive path will arise. Subsequently, one could use a shovel to pat on the soil and cause one end of the patch to become compact again, and after that apply a high pressure to reopen the path for facilitated water flow. That is, one can switch between high and low water conductivity by applying appropriate stress on the soil. If we now think of the soil as an insulator, of the water as electrons, and of soil manipulation as the application of electrical bias, we suddenly begin to understand RRAM basics.
The reshaping of soil resembles quite well the principles of operation of RRAM, and understanding how these principles are put to practice becomes easier once a clear picture of an RRAM cell is had in mind. The latter may be visualised as a sandwich-like structure, consisting of an electrode-insulator-electrode stack. Given an appropriate initial electrical bias, a conductive filament arises in the insulator, and the latter’s pristine high electrical resistance is changed to a low resistance; this is known as a FORM operation. The tip of the filament can then be ruptured to reinstate high resistance (RESET), and regrown to yield low resistance (SET). High and low resistance are the “0s” and “1s” of RRAM, and achieving them makes the programming of cells possible.
The switching between resistance states has been shown to correlate with the drift of oxygen ions and of vacancies that are left behind by moving oxygen. The response of these entities to electric fields in RRAM, which arise when different biases are applied, is very fast, and allows for writing speeds down to no more than a billionth of a second! Moreover, programmed states are permanent, and do not require power to be retained, and the simplicity of cell design makes scalability a straightforward task for this technology. Further benefits may be achieved with appropriate material selection. In fact, it has been shown that using hafnium oxide as the insulator and indium-tin-oxide as an active electrode results in the possibility to operate RRAM units at biases down to a fraction of a volt. This, together with current limitation, opens the door for the realisation of low power-consumption devices! All in all, a study dedicated to characterising RRAM units featuring promising materials seems like a perfect starting point for deepening our knowledge about systems that may make the necessary difference in the electronics of the world of tomorrow. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9018656
- author
- Åstrand, Mattias LU
- supervisor
- organization
- course
- FYSM30 20201
- year
- 2020
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- RRAM, Resistive switching, HfO2, ITO
- language
- English
- id
- 9018656
- date added to LUP
- 2020-06-18 08:46:16
- date last changed
- 2020-06-18 08:46:16
@misc{9018656,
abstract = {{The purpose of this work is of evaluating the choice of HfO2 and ITO as the dielectric and the top electrode in high performance resistive random-access memory (RRAM), respectively. The study is twofold, as it quantifies performance according to standard figures of merit for this technology, as well as provides insight into the physics of current conduction for this material choice. Results are achieved thanks to the probing of structures processed by the Nano Electronics group in Lund, followed by direct observations on the produced I−V characteristics and more involved data analysis. Specifically, different models according to different mechanisms of current conduction in a dielectric are fitted to the measured data, and physical parameter extraction is used as a mean to evaluate goodness of fit. It is discovered that several modes that were thought to potentially describe conduction in RRAM can instead not be the case, for confusing (if not impossible) parameters are extrapolated for them. An important result is thus that no matter how great a fit may be numerically, one cannot simply claim that a given mechanism of conduction is a good physical interpretation of a studied system without corroborating its fit with relevant parameter extraction. It is concluded that the only reasonable and available interpretation of measured devices is that of Ohmic type conduction, followed by space-charge (at higher voltages) in the high resistance state of RRAM, and solely Ohmic type in the low resistance state. The latter relates to not observing conduction through a barrier, which was expected as HfO2 and ITO RRAM has previously been found to be modulated by a self-compliance effect. Based on this interpretation, a computer model is developed in Python, and its output is put to the test by comparison with data from a different sample (i.e. not related to the observations that lead to the creation of the model itself). The universality of the model is found to be satisfying, hinting at the achieved interpretation being adequate. Concerning raw performance of the studied devices, resistive switching at low and steady voltages is observed, which relates to envisioning low power operation for HfO2 and ITO structures.}},
author = {{Åstrand, Mattias}},
language = {{eng}},
note = {{Student Paper}},
title = {{HfO2 and ITO Resistive Random-Access Memory}},
year = {{2020}},
}