國 立 交 通 大 學
電子工程學系 電子研究所碩士班
碩 士 論 文
新穎材料 DyMn
2O
5在電阻式記憶體上的機制研究與應用
The Researching Mechanism and Application of the Novel
Material DyMn2O5 (DMO) in Resistive Random Access
Memory
研 究 生:黃偉立
指導教授:施敏
院士
指導教授:
張鼎張 博士
新穎材料 DyMn
2O
5在電阻式記憶體上的機制研究與應用
The Researching Mechanism and Application of The Novel Material
DyMn2O5 (DMO) in Resistive Random Access Memory
研 究 生:黃偉立 Student:Wei-Li Huang
指導教授:施敏 院士 Advisor:Prof. S. M. Sze
張鼎張 博士 Prof. Ting-Chang Chang
國 立 交 通 大 學
電子工程學系 電子研究所
碩 士 論 文
A Thesis
Submitted to Department of Electronics Engineering and Institute of Electronics
College of Electrical and Computer Engineering National Chiao Tung University
in partial Fulfillment of the Requirements for the Degree of
Master in
Electronics Engineering July 2011
Hsinchu, Taiwan, Republic of China
I
新穎材料 DyMn
2O
5在電阻式記憶體上的機制研究
與應用
研 究 生:黃偉立 指導教授:施敏 院士
張鼎張 博士
國立交通大學電子工程學系電子研究所碩士班
摘要
電阻式記憶體具備了可微縮化,低功率的消耗,快速操作特性以及穩定的容 忍度的這些特點,使得它能夠成為下一世代的記憶體結構。在這篇論文研究中, 我們將進一步的去研究這種金屬-介電質-金屬 Pt/DyMn2O5 (DMO)/TiN 的電阻 式記憶體的結構。 在這篇論文裡,這類新型材料 DMO 的特性和電性特性的是藉由 XPS,TEM 等 材料分析系統和安捷倫 B1500 電性測量系統所分析得到。在第二章描述中,可經 由 X-射線光電子能譜(XPS)觀察到一些氧空缺存在於原始 DMO 的薄膜層中。我 們的元件也可藉由變溫的電性量測方式測得低電阻狀態(LRS)是較屬於金屬的 特性而在高電阻狀態下則是偏向半導體類特性機制。我們也可經由容忍度測試以 及開關特性的穩定度於高溫狀態下,來探討此類的電阻式記憶體的可靠度,從它 的容忍度可達到十萬次的操作下,一樣可以維持穩定的開關特性,而它的開關特 性也可以在高溫(850 C)環境下,穩定維持到 104 秒的。 以 DyMn2O5為基底下的電阻式記憶體,其電流電壓特性有出現一個特殊的現 象,稱為負微分電阻(NDR)現象。事實上,NDR 的現象是由於氧離子斷鍵,飄II 移和聚集所造成的現象。這種 DyMn2O5組成的電阻式記憶體可以產生出雙重電阻 轉態機制行為也可以經由電性量測方式觀察到。事實上雙極電阻開關的雙重轉態 的特性是由於金屬絲和介面的轉態的機制所共存在的一種現象,然而我們也可經 由不同的操作電壓來產生出這兩種轉態機制的出現。甚至它可以經由不同的厚度 元件而產生單一主導的轉態機制。 在最後一章節裡,我們針對它的容忍度作進一步的研究發現,氧離子與氧空 缺結合時間與熱的效應,也是影響整個容忍度的一項重要因素。因此,我們採用 不同的脈衝週期條件,來提升容忍度的效應,甚至可讓我們的元件達到 107 的操 作次數的測試,這類物理的研究而能夠優化我們元件特性。
III
The Researching Mechanism and Application of
The Novel Material DyMn2O5 (DMO) in Resistive
Random Access Memory
Student:Wei-Li Huang
Advisor:Prof. S. M. Sze
Prof.
Ting-Chang Chang
Department of Electronics Engineering and Institute of Electronics
National Chiao Tung University
Abstract
The resistive switching random access memories (RRAMs) possess some advantages of scalability, low power consumption, fast operating time and stable endurance. The RRAM with these advantages has high potential for next generation memory applications. The switching mechanism and electrical characteristics of Pt/DyMn2O5 (DMO)/TiN RRAM devices are investigated by material analysis and
electrical measurement system. In chapter1, some oxygen vacancies are observed in the pristine DMO film through x-ray photoelectron spectroscopy (XPS). The low resistive state (LRS) of RRAM is found metal-like and the high resistive state is semiconductor-like properties by electric measurement at different temperatures. The endurance of the RRAM can achieve 105 times and its retention time can also achieve 104 seconds under high temperature (850C) thermal stress.
IV
For the DyMn2O5-based RRAM, the current-voltage characteristics possess a
special phenomenon which is called Negative Differential Resistance (NDR). In fact, the NDR phenomenon is due to the breaking bond, migration and accumulation of oxygen ions. The dual resistive switching behaviors are also observed in the device structure. The dual bipolar resistance switching behaviors of filament-type and homogenous-type can coexist in the devices by applying appropriate sweep voltages. It can be found that the thicker DMO films possess only homogenous-type mechanism.
In last chapter, we can find that the recovery time of oxygen and oxygen vacancy is an important factor for the endurance of the RRAM devices. Therefore, we apply different pulse conditions to enhance the endurance and the device can achieve 107 times endurance tests at optimized condition.
V
Acknowledgement
時間過得飛快,才覺得考上研究沒多久,到現在也經過兩年了,在新竹修課然 後再到高雄做實驗,我所經歷,我所得到的不僅僅只是課業上的理論,也讓我體 會到很多待人處事的道理,我十分感謝施敏老師以及張鼎張老師,總是不厭其煩 的在 group meeting 中教導相關的知識,使我在半導體相關的領域可以懂得更多, 看得更廣,以及在我研究上的指導與論文的修改,使得當我在困難以及不知所措 時可以得到解決。 在我做實驗的時候往往能夠得到學長姊的幫忙,使得研究不會毫無目標的亂 做一通,因此我非常感謝的是所有實驗室的學長姐,尤其是蔡侑廷學長,在他帶 我做實驗的過程中,我學到很多思考的方式,做實驗的態度,知識成長以及數據 分析,還有一點是他都很有耐心執導我,在做實驗過程中很感謝他對我的鼓勵以 及建議,也很感謝一起走過這段日子的同學們:柔妙、岳恆、奕介、志誠、儀憲、 凱弘、冠任、菀琳、雅琪、承瑋、國孝、祐松以及慶恩等,總是可以在我實驗碰 到瓶頸的時候陪我聊天運動以及加油打氣,以及碩一的學弟妹們:昌蓓、明諺、 君昱、峻豪、華茂、哲丘以及天健等,和你們嘴砲讓可以讓我有紓壓的功效,看 你們在實驗室專心的準備期中期末考,讓我看到實驗室未來之光,還有真的很感 謝遠在新竹的慶恩Q博,在我們來高雄做實驗的期間,總是很辛苦地幫我打點好 學校一些大大小小的事情,讓我不用來回奔波,以及新竹的學弟謝謝你們幫我們 簽名。 最後要感謝的是我親愛的家人,在爸爸媽媽姐姐的鼓勵下,是我最大的精神 支柱,也是因為你們讓我有繼續升學的衝勁,謝謝你們給我好的讀書環境,讓我 無憂慮的專研學業。因為有你們的幫助,得以讓我完成此本論文,感謝~~再感謝 大家。 黃偉立 謹識 交通大學 2011 年VI
Contents
Chinese Abstract
………..………..IAbstract
……….…IIIAcknowledgement
……….……….…VContents
………...….…VIFigure Captions
……….…IXTable captions
………...……...…XIVChapter 1 Introduction
……….11-1 The Evolution of Memory ... 1
1-2 The Novel Memory Device... 2
1-2.1 FeRAM ... 2
1-2.2 PCRAM ... 3
1-2.3 MRAM... 4
1-2.4 RRAM ... 5
1-2.5 The Switching Mechanism of RRAM ... 7
1-2.6 The Conduction Mechanism of RRAM ... 11
Chapter 2 Basic material characteristics and electrical properties of
DyMn2O5
……….222-1 Process Flow... 22
2-1.1 Substrate Prepare ... 22
VII
2-2 Materials Analysis of Pt / DMO / TiN (STD)………..23
2-2.1 X-ray Photoelectron Spectroscopy (XPS) ... 24
2-2.2 Fourier Transform Infrared Spectroscopy (FTIR) ... 24
2-2.3 Transmission Electron Microscopy (TEM) ... 24
2-3 Electric Characteristic Measure……….25
2-3.1 Forming Process ...25
2-3.2 Current-Voltage after Forming Process ...27
2-3.3 The Carrier Conduction Mechanism Fitting………...….……….27
2-3.4 Temperature effect of LRS and HRS Characteristic…….………...………..28
2-3.5 Reliability ...28
2-3.5-1 Retention………...28
2-3.5-2 Endurance………...………...29
Chapter3 Coexistence of Filament and Homogenous Resistive
Switching for DyMn
2O
5film
……….……….……….403-1 Electric Characterisitic Measure before Forming……….…40
3-1.1 Temperature Effect ...40
3-1.2 Constant Voltage Sampling Effect ...41
3-1.3 Thinckness Effect ...42
3-1.4 Mechanism of the Phenomenon of NDR before Forming ...42
3-2 Electric Characterisitic Measure after Forming……….…45
3-2.1 Normal Operation...45
3-2.2 Temperature Effect for sub-RRAM ...46
VIII
3-2.4 Normal Switching Mechanism after Forming Process (coclusion) ...47
3-2.4-1 Original - RRAM...48
3-2.4-2 sub- RRAM...48
3-3 Forming-Free Switching Mechanism……….…..50
3-3.1 Operating Current-Voltage before Forminf ...51
3-3.2 Pure Interface-type Operated Mode ...52
3-4 The thin DMO+Nitrogen (DMON) Layer Effect...54
3-5 Conclude Coexistence of Interface and Filament Mechanism...………56
Chapter 4 Investigation of Improving Endurance Performance by
Using Fast Measurement systems
………..……….734-1 The Observation of Current by Using Pulse Current-Voltage ……….73
4-2 Pulse Cycle Effect ……….74
4-2.1 Less Pulse Cycle Effect ...74
4-2.2 Many Pulse Cycles Effect-Part One ...75
4-2.3 Many Pulse Cycles Effect-Part Two……...76
4-3 300ns and 900ns for The Reset Time Effect………79
4-4 The Result of Best Endurance………..………79
Chapter 5 Conclusion
………..………..………...90Reference
………..…92IX
Figure Captions
Fig-1.2.1-1.The ABO3 structure………. 15
Fig-1.2.1-2. The curve is P-V hysteresis……….…….15
Fig-1.2.2-1.The main memory data storage part is in the middle layer of GeSbTe (GST) thin film………...………..16
Fig-1.2.2-2. The main memory data storage part is in the middle layer of GeSbTe (GST) thin film………...………..16
Fig-1.2.2-3. The heating time and transitional process. [8]……….16
Fig-1.2.3-1. The MRAM structure……….……….17
Fig-1.2.3-2. The “1” state and “0” state of MRAM……….17
Fig-1.2.4-1. The forming process…………..………..………18
Fig-1.2.4-2. The reproducible behavior of RRAM……..………..………..…....18
Fig-1.2.4-3. The rupture and formation process of filament [34]…………...……….19
Fig-1.2.4-3. The unipolar current voltage curve, the left-side is log scale and the right-side is linear scale. [25] [26]……...19
Fig-1.2.4-4. The oxygen ions migrate to the electrode and the oxygen vacancies remain in the film………...20
Fig-1.2.4-5. The bipolar switching behavior. The left-side is log scale and the right-side is linear scale[34]……….………...…..20
Fig-1.2.4-6. The interface model [34] [37]………..21
Fig-2.1.1. The substrate cross-section………..31
Fig-2.1.2. The device structure of sample1………..………...31
Fig-2.1.3. The cross-section of the sample2……….………...32
X
Fig-2.2.2. The O 1s spectrum
[40]
……….…...33Fig-2.2-3. The spectrum of Far-FTIR for STD DMO film………..…....33
Fig-2.2-4. The mid-FTIR ………34
Fig-2.2-5.The MIM device cross-section……….………...……34
Fig-2.3-1. The probes contact with pads…….………...….35
Fig-2.3-2 .The forming of STD device
……….
…..35Fig-2.3-3.Typical current-voltage characteristics of STD device
.
…….…………...36Fig-2.3-4. The carrier conduction mechanism are at a piece of I-V curve...36
Fig-2.3-5. Two different characteristics, one is metal-like at LRS and the other one is semiconductor-like at HRS.……..………....37
Fig-2.3-5.The retention of our device and kept the thermal stress at 85 0C…………37
Fig-2.3-6. The endurance characteristics between HRS and LRS of STD device…...38
Fig-2.3-7(a).
The condition of reverse bias pulse
……….……….……….38Fig-2.3-7(b). The condition of reverse bias pulse……….…………...38
Fig-2.3-7(c).
on/off pulse cycles
………..……….…38Fig-2.3-8.The on / off ratio is about 100 resistance, the HRS fluctuation is acute than LRS
.
………..………39Fig-3.1-1. The NDR voltage depend on temperature………..……59
Fig-3.1-2.
The voltage of NDR depends on temperature
. ……….……...59Fig-3.1-3. Normal relation between current and time………..60
Fig-3.1-4(a). The
current time curve are at 303k
………...……….…...…..60Fig-3.1-4(b). The
current time curve are at 323k
………...……….…...…..60Fig-3.1-4(c). The
current time curve are at 343k
………...……….…...…..60XI
Fig-3.1-5.
These relations of current-time in the same plot and then compares
them with each other………61
Fig-3.1-6. The forming processes of four different thicknesses into the same plot…61 Fig-3.1-7. The NDR voltage and makes a plot recording the distribution of NDR
voltage with different thickness……….62 Fig-3.1-8 (a). The first step of oxygen ions model in pristine state……….62 Fig-3.1-8 (b)
. The second step of oxygen ions model and the bonds are broken
and migrate by applying voltage………..62
Fig-3.1-9 (a). The oxygen ions will accumulate near TiN interfacial and form an oxygen-rich region……….63 Fig-3.1-9 (b). The oxygen ions drift into TiN electrode by applying forming voltage
and the filament is formed in the bulk, finally……….63 Fig-3.1-10 (a). The flat band is at original state………...…63
Fig-3.1-10 (b).The step of breaking bond and migration of the oxygen ions…...63 Fig-3.1-10 (c). The oxygen ions accumulate near the interface and bend the band
upward………..63 Fig-3.1-11. The larger cell size device has larger leakage than the small cell size…..64 Fig-3.2-1(a). Original current was higher than later current, it is like reset
Phenomenon……….64 Fig-3.2-1 (b). The voltage was swept from 0V to -1.8V and then swept from -1.8V to 0V……….64 Fig-3.2-2 (a). The device is operated between SHRS and HRS for 100 cycles……...65 Fig-3.2-2 (b). The DC endurance HRS/ SHRS ratio was about 1.5 orders…………..65
Fig-3.2-3 (a). The SHRS is different between at 800K and at RT by applying the same
Condition………....65
Fig-3.2-3 (b). The Vstop condition was changed from 0.7V to 0.9V at 800K…………65
XII
Fig-3.2-4. The conduction mechanism of carrier by fitting…………..………...66
Fig-3.2-5(a). The status of forming CP………66
Fig-3.2-5(b). The reset process. ………...………66
Fig-3.2-5 (c). The NDR phenomenon………...………….…..67
Fig-3.2-5 (d). The set process……….………..67
Fig-3.2-6 (a). The migration of oxygen ion……….…….67
Fig-3.2-6 (b). The accumulation of oxygen ion………...67
Fig-3.2-6 (c). The oxygen ions migrate to bulk and the oxygen-rich layer is vanished when we apply reverse bias……….………...67
Fig-3.2-7. SHRS and HRS depended on cell size but LRS was independent to cell Size………..………..68
Fig-3.3-1(a). The swept loop from 0V→5V→0V ………..…68
Fig-3.3-1(b). The swept loop from 0V→ -5V →0V………68
Fig-3.3-1 (c) The state is not at on- state………..68
Fig-3.3-2 (a). The sub-RRAM the operating condition is at forward bias of 0.6V and at about reverse bias of -1.8V.………..………..69
Fig-3.3-2 (b). The operating mode before forming, and we discover Vstop was about 5V at forward bias………..69
Fig-3.3-3 (a). The properties also possessed the NDR phenomenon and the NDR voltage was at about 2V………70
Fig-3.3-3 (b). The state is transition from ON state to OFF state………...….70
Fig-3.3-3 (c) the current was different from OFF state………...….70
Fig-3.3-3 (d) indicates the state transition from OFF state to ON state……….
..70
Fig-3.3-4.The phenomenon of size effect………71
XIII
Fig-3.3-5(b).The oxygen went back to bulk ……….……..71
Fig-3.3-5(c). The Schottky barrier is bend the band upward………...………71
Fig-3.3-5 (d). The Schottky barrier is bend the band upward and downward……….71
Fig-3.4-1. The huge NDR phenomenon characteristic at forming process…………..72
Fig-3.4.2(a) The DMON possess the huge NDR than the DMO……….72
Fig-3.4.2(b) The sub-RRAM HRS and SHRS ratio for DMON and DMO………….72
Fig-4.1-1. The current-voltage of pulse………...81
Fig-4.1-2. The current-voltage of pulse of set process……….81
Fig-4.1-3(a). The three kinds condition of voltage pulse……….82
Fig-4.1-3(b).
The current value is detected by voltage pulse
………...…82Fig-4.1-3(c).
The state is read under low bias………...
………...…82Fig-4.1-4. The resistance increases with temperature increasing………...…..83
Fig-4.2-1(a). The
endurance
reset pulse width is 300 ns………..…….….83Fig-4.2-1(b). The endurance of reset pulse width is 900ns………..………83
Fig-4.2-2. The endurance test under many pulse cycles………..84
Fig-4.2-3. The statistics of different pulse cycles effect………..85
Fig-4.2-4. The reset process of oxygen ions………....85
Fig-4.2-5. The oxygen vacancies still exist in the bulk………...86
Fig-4.2-6. The width between reset and set is called “a” and the width between set and “next reset” is called “b”……….……….86
Fig-4.2-7. Each pair of parameters possesses some different effects………...87
Fig-4.2-8. The width between reset and set is extended………..87
Fig-4.2-9. The on/off ratio is enlarged with the a-region lengthening gradually…....88
Fig-4.3-1.The endurance of different reset width in many pulse cycles………..88
Fig-4.3-2. The different width phenomenon affects the state of device………...89
XIV
Table Cpations
Table 4-1. The endurance test condition of giving less pulse cycles………83 Table 4-2. The endurance test condition of giving many pulse cycles……….84
1
Chapter 1
Introduction
1-1 The Evolution of Memory
In recently, Dynamic RAM (DRAM), Static RAM (SRAM) and Flash Memory have been widely used in our live. These memories have existed the development of numerous electronic systems especially for computer, communication and consumer applications (3C). Generally, these memories can be divided into two different types, one is the volatile-type memory and the other one is the non-volatile-type memory (NVM). Volatile memories cannot store any data without power. In other words, the information disappears once the system power is turned off, such as DRAM and SRAM. And the other type NVM is able to store the data (information) without power, such as the Flash memory. The NVMs have become more and more important because NVMs have widely been used to store information in the portable electronic system such as MP3, Cell Phone etc. Examples of NVMs include Read-Only-Memory (ROM), EEPROM (electrically-erasable programmable read-only memory), and the flash memory. The flash memory has been more and more important since 1990. The first programmable non-volatile memory is the floating-gate MOSFET, invented by D. Kahng and S. M. Sze in 1967 [1]. But the floating gate MOSFET possesses some disadvantages such as scaling limitation, high operating voltage, and long write/program speed. We have to improve the floating gate device or create some novel NVMs structure. A limiting of the floating gate structure is the MNOS (metal-nitride-oxide-semiconductor) structure which is formed when the floating gate
2
thickness approach zero. First, The MNOS has been evolved to form SONOS and TANOS.
Second, some novel NVMs structures are invented one by one. Generally, they can be divided into four species: FeRAM (Ferroelectric RAM), MRAM (Magnetic RAM), PCRAM (Phase Change Memory) and RRAM (Resistive RAM). In these novel-RAM we have to select a type of devices need to possess some advantages such as fast access time, scalability, low power consumption and high endurance/retention. The RRAM is a promising candidate for the next-generation device to be next-generation memory because it has many outstanding attributed. For example, its switching time can be less than 10 ns, its cell structure can less than 8F², it has simple structure, and its density can be increased high. However, RRAM still has many issues to the resolves of many characteristic to be improved.
1-2 The Novel Memory Devices
In this part, we would discuss the four kinds of novel memory structures. And then we can compare them with each other.
1-2.1 Ferroelectric Random Access Memory (FeRAM)
Ferroelectric random access memory (FeRAM) has been widely researched because it can perform as a non-volatile memory. [2][3][4] It can be fabricated by using ferroelectric material to construct a FeRAM structure. The ferroelectric atomic arrangement has a crystal structure type of ABO3. Fig-1-2.1-1 shows the ABO3
structure. The ferroelectric interior atomic arrangement structure would be changed if we apply an external field to the FeRAM device. In fact, the B-type atom can be
3
driven upward or downward in the interior arrangement. Besides, the electric dipole is induced because of the position of ion modulation by applying external field. For example, the B-type atom would be drove to upper position if we apply an upward field to the ferroelectric layer. On the contrary, the B-type atom would be drove to lower position if we apply a downward field. The atomic displacement cannot be changed and the atomic arrangement is still a stable state when we remove external field. We can understand the phenomenon from the polarization-voltage curve. The Fig-1.2.1-2 clearly shows the curve is P-V hysteresis. The electric dipole cannot disappear after removing the operating voltage that phenomenon called remnant polarization (Pr). The hysteresis curve is following counter clockwise curve. And the remnant polarization direction is able to change by changing different external field. The signals of “1” and “0” could be distinguished from the direction of remnant polarization, i.e., Pr and -Pr.[5]
1-2.2 Phase Change Random Access memory (PCRAM)
Phase-change memory (PCM, PRAM and Ovonic Unified Memory) is studied as a candidate for next generation non-volatile memory. The concept of phase change RAM was proposed by R.G. Neale, D.L. Nelson and G.E. Moore as early as in 1970s.
[6] [7] And the depositing materials are known as chalcogenide and the materials after absorption thermal energy can transfer two different types, one is amorphous and the other one is poly-crystal. This material can achieve the memory storage ability because its characteristic possesses two different resistive characteristics by thermal effect changing the structure of material. The state would not be changed with time once we don’t heat it again by external thermal stress so it can store the information for a long time. Therefore, PCRAM is also able to belong to non-volatile memory.
4
The Fig-1.2.2-1 and Fig-1.2.2-2 clearly depicts the memory data storage situation is in the middle layer of GeSbTe (GST) thin film. It just only needs a simple two-terminal point metal-GST-metal structure so we also only need to control the locally current flow the thin film. And achieve two different kinds of resistance. The Fig-1.2.2-3 shows the heating time and transitional process. [8]
1-2.3 Magnetic Random Access Memory (MRAM)
The Magnetic RAM (MRAM) was discovered in 1972. It is construct a prototype and display by Honeywell in 1992
.
Magnetic random access memory (MRAM) is not similar to conventional RAM technologies; it is that information/data is stored by magnetic storage elements rather than by electric charges or different atomic structure. In fact, the mechanism of program/erase is the electric spin orientation is changed in magnetic material layer by the electric current pass the top and bottom both conductive metal layers to induce the magnetic field. And then the magnetic field can change the spin orientation of middle layer which is called tunnelingmagneto-resistance (TMR) or Giant-Magneto-resistance (GMR) cell. We can see Fig-1.2.3-1 shows the MRAM structure. The barrier layer is deposited by Al2O3. [9]
[10] The top conductive metal layer is as the bit line, and the bottom conduction metal layer is as the word line. The MRAM cell will affect the polarization direction shift in free layer if we applied a pulse current to bit line because the bit line current induced magnetic field. On the contrary, we apply a pulse current in word line and then the induced magnetic will change the polarization direction of free layer completely. Therefore, the polarization of two ferroelectric layers will be forward arranged. On the other hand, we can define the lower state as “0” state when the magnetic resistive is low. In other words, the polarization of two ferroelectric layers is
5
reverse arranged and the magnetic resistive is higher, so we can define it as “1” state.
Fig-1.2.3-2 shows the “1” state and “0” state of MRAM.
1-2.4
Resistive Random Access Memory (RRAM)
The resistive random access memory (RRAM) is a type of RAM by controlling voltage or current to make the device inducing different resistance, i.e., it can be at high or low resistive state by external field. RRAM is a simple metal-insulator-metal (MIM) structure and its operating mode is easily to achieve the transitional process. The device can be operated after we apply the breakdown voltage of device which is called forming voltage and the process we call forming process. We have to set compliance current in order to avoid hard-breakdown and make the device produce the phenomenon of soft-breakdown. Fig-1.2.4-1 shows the forming process. And then we can apply bias to make the state of device from high resistive state (HRS) to low resistive state (LRS) and the process we called reset. On the contrary, we can also apply bias to make it from HRS to LRS which is called set. Therefore, it can produce reproducible behavior through set and reset process, i.e., we can turn the device on or turn it off. Fig-1.2.4-2 depicts we mention above. The ON state and OFF state can be as “1” and”0” so it is also a novel NVM. However, the RRAM device possesses some advantages than other novel RAM. The operating speed of RRAM can is less than 10ns. (Reference), the structure and cell size are simple and small, the operating voltage is low and hence it can be used in low power consumption. [12] [13] [14] [15] Finally, the RRAM structure still has many problems to solve them. We will focus on the issue of RRAM device. Therefore, we will divide into three different parts to discuss the materials, switching mechanisms of RRAM, and conductive mechanism of carriers.
6
1-2.4-1
The materials of RRAM
There are some kinds of insulators can be used in RRAM. They are the perovskite-type, the binary transition metal oxide and so on. The two species materials we mentioned above are common RRAM materials.
1-2.4-2 Perovskite-Type
The structure of Perovskite-type is ABO3. The type A atom in the structure is a
cation and sits at cubic corner position (0,0,0) with the larger radius and the type B is also a cation and sits at body center position (1/2, 1/2, 1/2) with the smaller radius. The type O means the oxygen atom and it sits at face center position (1/2, 1/2, 0). In the common unit cell, type A occupies at the 8 corner positions, type B occupies at the body center position and O occupies at the 6 face center positions, like the Fig-1.2.1-1 . There are some common Perovskite structures such as PrCaMnO3
(PCMO) [16] [17], SrZrO3 (SZO) [18], and SrTiO3 (STO) [19].
1-2.4-3 Binary Transition Metal Oxide-Type
There are some famous transition metal oxide materials such as TiO2, NiO and so on; they have been applied for thin film research for many years. S. Seo and coworker reported reproducible resistance switch of the NiO thin films deposited on Pt/Ti/SiO2/Si substrate in2004. [20] There are still other type transition metals oxides have been widely used to match CMOS devices. Thus they can also compatible with modern CMOS process. However, this material group of binary oxides has simpler element components. It is easier to control the proportion of metal and oxygen
7
elements.
1-2.5 The Switching Mechanism of RRAM
In this part, we will introduce some different switching mechanisms of RRAM. Presently, the RRAM switching mechanism is not yet clear. But, there are many ideas and papers to offer some reasonable interpretations. Generally, the switching mechanism can be divided into two parts: one is bulk control type and the other one is interface control type. In bulk control type, the switching mechanism is relative to the insulator layer, i.e., the material characteristic is changed in the bulk. And the interface control type switching is relative to interfacial layer which is between the electrode and insulator layer. The transition of bulk is belonged to filamentary or conduction paths (CPs) model and the interface transition phenomenon is belonged to modified Schottky barrier model. We will discuss these mechanisms respectively.
1-2.5-1 Bulk type mechanism
1-2.5-1.1 The Filament type
First, we have to apply forming voltage to produce the breakdown phenomenon of device and the current value will suddenly increase which is called forming process. Next, we apply a bias to make the current reducing which is called reset process. Subsequently, we apply a bias to the device again and the current increasing sharply is like forming process we called set process. The switching phenomenon cause the filament is formed in the insulator layer. And the filament can also be ruptured and formed again and again so that the device possesses switching behavior. [21] [22] [23] Fig-1.2.4-3 illustrates what we mean. In addition, the device has to compliance the
8
current value (<10mA) to avoid the device becoming hard breakdown at the forming if we produce reproducible switching behavior. There are some reasons can explain the phenomenon of formation and rupture the filament result from thermal effect (Joel heating effect), reactive-oxidation (redox) processes by anion migration to the interface between the metal electrode and the insulator layer. And then we will discuss them respectively.
1-2.5-1.2 Thermal Effect (Joule Heating Effect)
This typical resistive switching behavior base on thermal effect shows in the unipolar characteristics. The unipolar characteristic operating mode is independent polarization. On the other hand, the device state can be changed the state from LHS to HRS or from HRS to LRS by applying forward bias. On the contrary, we can also apply reverse bias to transit the device state from LHS to HRS or from HRS to LRS. [24] Fig-1.2.4-3 depicts the unipolar current voltage curve. [25] [26] The device can turn on and turn off because the filament ruptures and forms when we apply a voltage to the device. At first, we apply the forward bias to make the device building the
filament and the current increase suddenly to higher current. The device can transit from HRS to LRS because the filament is formed. Subsequently, we observe the device can transit from LRS to HRS if we apply the smaller forward bias but we don’t compliance the current in this operating. Because we don’t set the current limit, the crowd of electrons will induce heat and melt the filament at the local area in the bulk. The phenomenon is called electron migration. On the contrary, we can produce the same phenomenon under the reverse bias. In addition, there is another similar to electron migration effect such that the structure rearrange because current flow the film induce heating. Most of the current will easily pass the local area, so the
9
temperature is very high at local area. Therefore, the higher temperature at the local area can easily form or rupture the filament. In other words, it is a type of Joule heating effect and makes the filament rupture and formation because of electron migration or structure rearrangement by heat. [27] [28]
1-2.5-1.3 Redox Processes induced by Anion Migration
Another filament type is redox processes induced by
Anion
migration.
The mechanism is the bond between oxygen and other metal will be broken by external electric field when we apply the forward bias on the bottom electrode at first.T
he oxygen will become the oxygen ion in the bulk and then the oxygen ion possesses free migrated ability. The oxygen ion will follow external field and move to the cathode. The oxygen vacancies can also provide spaces and the electron hop the vacancies through external field. Finally, the oxygen vacancies connect each other, and form a conduct path when we apply a larger bias. [25] [29] Fig-1.2.4-4 shows the oxygen ions migrate to the electrode and the oxygen vacancies remain in the bulk.
Next, the device can transit from LRS to HRS if we apply a reverse bias. We have to use an opposite polarized, i.e., forward bias, to change the state from HRS to LRS. In other words, the set and reset process depend on polarization. It means that we apply forward bias to achieve set process and the device has to be applied reverse bias to achieve reset process. The operating mode is bipolar switching behavior. Fig-1.2.4-5 depicts the bipolar switching behavior. M. Fujimoto and coworker [30] observed that resistive switch is dependent on the operated polarity by using Pt/TiO2/TiN/Pt devicesin 2006
.
C. Yoshida and coworker [31] also found the bipolar resistive switching with the same structure of devices in 2007. In fact, the phenomenon is a type of redox10
reaction because a bond between other atoms and oxygen transit to oxygen ions and then migrate to TiN electrode. First, the oxygen ions will react with TiN if we apply a forward bias on TiN electrode. Next, the oxygen ions will go back to bulk if we apply a reverse bias on TiN. The role of bottom electrode TiN is as an oxygen reservoir. [32] [33] It is a reaction and oxidation process. [30]
1-2.5-2 Interface Type Mechanisms (Modified Schottky Barrier Model
)
The insulator contact with the metal electrodes will produce some interfacial characteristics in the metal-insulator-metal (MIM) structure. We have to deposit different metal electrode in MIM structure if the device perform the interface-type model. The device can possess the interface effect due to the work function are different between these two electrode such as Pt and TiN. Many papers research the interfacial problems by different working function in MIM system. [34]-[37] Generally, the interface contact which is between metal and insulator condition can be divided into two types, one is Ohmic contact and the other one is Schottky contact. The Schottky contact model plays a very important role In RRAM conductive mechanism models. A. Sawa and coworker earliest brought up the mechanism of Schottky barrier model. [17] [34] The device structure is Ti/PCMO/SRO and PCMO is p-type semiconductor. The Schottky contact occurs in the interface of Ti/ PCMO. We are able to know one important that Ti material possess a very strong attraction for oxygen ions in many papers researching. [32] [33] Ti electrode will catch many oxygen ions at the interface between Ti and PCMO if we apply a bias to make the bond between metal and oxygen breaking caused the interface. Therefore, there are many oxygen vacancies generation in the bulk. Oxygen vacancies perform the positive charge; they will bend the energy band. We will use oxygen vacancies
11
(oxygen ions) to think about the model. First, for p-type PCMO, a large amount of oxygen vacancies migrate to the interface between SRO/PCMO when the Ti electrode is gave a positive bias (reverse field). In other words, it means the oxygen ions can drift to the Ti/PCMO interface by external forward field. And then the energy band is bent downward because the oxygen vacancies accumulate in the interface between SRO/PCMO. [17] The step is called reset process. Fig-1.2.4-6 depicts we mentioned above. [34] Second, the oxygen vacancies will migrate back to bulk when we apply a negative bias (forward) to the Ti electrode. On the other hand, the oxygen ions can drift back to the bulk through external reverse field. And then the energy band is bent downward because the oxygen vacancies not accumulate in the interface between SRO/PCMO and go back to the bulk. The step is called set process. It is shown in the Fig-1.2.4-6. They think that the more or less of charged ions can change the Schottky barrier height or width and the reason will cause the resistance to become high or low.
1-2.6 The Conduction Mechanism of RRAM
Transition metal oxides can be divided into three different materials; one is insulators, second is semiconductors and third is metals oxide materials. Therefore, the carrier conductive characteristic may be different from each other due to their different chemistry, structure and physics. Generally, the most materials which were discussed in RRAM application belong to insulators, semiconductors and metal oxides according to the constitution and stoichiometry and the carriers conductive mechanisms mostly involved are Ohmic conduction, Schottky emission, Frenkel-Poole emission and Tunneling emission. Following we mentioned above, we will introduce these type carrier transport. [38]
12
1-2.6-1 Ohmic Conduction
Ohmic conduction transportation can divide into two different circumstances; first-type is at OFF state, second-type is at ON state. First, we introduce the first-type Ohmic transportation which takes place while the injected carrier density is less than the thermally-generated carrier density. In other words, the most carriers are from the valence band of material excited to conduction band by thermal effect so the carriers can drift to the other side from one side along the direction of external field. Generally, it can be applied to low electric bias to drive the carrier which are thermally generated transportation are dominate in Ohmic conduction. Second, the filament-type mechanism which is like metallization process (conduction paths) so the carriers can transport through filaments. And then the carriers’ transportation is also an Ohmic mechanism because of metal (conduction paths). The current-voltage curve property in Ohm’s law is the current passing through a resistor is proportional to the voltage drop cross the two points. Therefore, the temperature effect is also relation to Ohmic conduction behavior because the electron and phonon scattering effects. So the conductivity increases with increasing temperature for conduction in semiconductor while with decreasing temperature for metal conduction. The following is the expression for Ohmic conduction:
exp ac i E J kT
13
1-2.6-2Tunneling carrier transportation
There is a conduction mechanism in high electric field it is tunneling effect. And it also often appears in insulator current conduction. The primary theory of tunneling conduction effect is quantum mechanical theory that the electrons can pass through any kinds of barrier. Tunneling emission possess a strong relationship with the apply voltage and less relationship with the external temperature so it appears in high voltage. Subsequently, tunneling emission can be divided into two types: direct tunneling effect (DT) and Fowler-Nordheim tunneling effect (F-N T). Direct tunneling is relation to the thickness of film. And Fowler-Nordheim tunneling is not relation to thickness of barrier.So we introduce the tunneling emission formula:
* 3 2 2
4 2
(
)
exp
3
B i im q
J
E
q E
or
2exp
b
J
V
V
Where the Ei is electric field, the ψB is barrier height and m
*
is effective mass. [38]
1-2.6-3 Schottky Barrier Emission
In the tunneling effect, we emphasize the field effect and affect the barrier for device but there is another type emission is the electron hop through the metal/insulator interface barrier height by the thermal effect. It is called Schottky barrier emission. In the type emission, the thermal effect is larger than field effect. The Schottky emission possesses a strong relationship with temperature and the image charge lowering the barrier height effect caused the electrons can hop through the barrier height easily. The barrier height is determined by the band of set between the work-function of material and metal work function which contact. And the interface trap, defect, carrier density, process condition can make the barrier height changing.
14
The electrons can easy jump the lower barrier height, and the electrons are difficult to jump to the higher barrier height. Following is the formula of Schottky emission.
** 2 4 exp B i i q qE J A T kT
or
2exp
B
q
J
T
a V
kT
Where
A
** is effective Richardson constant,
i
r 0 ,
r is dielectric constant, and
B= barrier height. [38]15
Fig-1.2.1-1 shows the ABO
3structure.
Fig-1.2.1-2 shows the curve is P-V hysteresis
(Pr = Remnant Polarization, Ps = Saturation Polarization,
Vc = Positive Coercive Voltage).
16
Fig-1.2.2-1 Fig-1.2.2-2
Fig-1.2.2-1 and Fig-1.2.2-2
clearly depicts the main memory data storage
part is in the middle layer of GeSbTe (GST) thin film.
17
Fig-1.2.3-1 shows the MRAM structure
18
Fig-1.2.4-1 shows the forming process.
Fig-1.2.4-2 shows the reproducible behavior of RRAM and the current is
large we call LRS or ON state and the current is small we call HRS or
OFF state.
19
Fig-1.2.4-3 depicts there is no any filament in the film at initial state. 1.) The
device produces filaments in the film after forming process. 2.) These filaments
rupture after reset process by applying bias. 3.) The filaments formed again
after set process.
[34]Fig-1.2.4-3 depicts the unipolar current voltage curve, the left-side is log
20
Fig-1.2.4-4 shows the oxygen ions migrate to the electrode and the
oxygen vacancies remain in the film.
Fig-1.2.4-5 depicts the bipolar switching behavior. The left-side is log
21
Fig-1.2.4-6 depicts, first, the oxygen vacancies accumulate at the interface of
SRO/PCMO after reverse bias so it is off- state for p-type PCMO. Second, the
oxygen vacancies go back to the bulk after forward bias so it is on-state for
p-type PCMO.
[34] [37]22
Chapter 2
Basic material characteristics and electrical
properties of DyMn
2
O
5
In this chapter, we will introduce the process flow, materials analysis and basic properties of our device.
2-1 Process Flow
2-1.1
Substrate prepare
In this work, the structure of TiN/Ti/SiO2/p+-Si was used as substrate. Subsequently, low temperature Oxide (LTO) is deposited on TiN film. But this structure was divided into two type morphology on TiN film which is bottom electrode. One was covered by photo-resist is called bottom electrode. The other one was not covered by photo-resist is called via hole which can be formed device area. And then the substrate will be used to deposit other film. Fig-2.1.1 depicts the substrate cross-section.
2-1.2
Sample fabrication
2-1.2-1 Sample1 Pt / DMO (10nm) / TiN
First, DyMn2O5 (DMO) material is used as the insulator of smaple1, and then the
film of DMO by radio- frequency (RF) magnetron sputter is deposited on the substrate. The magnetron sputter was carried out in an ambient of argon (30 sccm)
23
with power of 120 W and the working pressure of 5 mtorr. The thickness of DMO film is about 12nm. Afterwards Pt was the top electrode of sample1, and the thickness of Pt film is about 100 nm was deposited by magnetron sputter. Finally, the sample 1was lifted off by acetone. Fig-2.1.2 shows the device structure of sample1. The sample1 will be used as the standard (STD) Metal-Oxide-Metal (MIM) structure device for RRAM researching. There are also many different kinds of the via hole size in the sample1. The via hole sizes are 64 (um2
) , 16(um2 ), 4(um2 ), 1(um2 ), and 0.64(um2 ).
2-1.2-2 Sample2 Pt / DMO (10nm) / DMO+NH3 (2.4nm) / TiN
The smaple2 used the same substrate to fabricate a contrastive device. It is a double layer structure. The substrate was putted into magnetron sputter system. First, magnetron sputter system was operated in an ambient of argon (30 sccm) and NH3 (25 sccm) with power of 120 W and the working pressure of 10 mtorr. The switch gear of NH3 pipeline was shut after 100 seconds, and only argon switch was opened until the thickness of DMO and Pt film were completed. Lift-off the sample2 was the ultimate step. Fig-2.1.3 shows the cross-section of the sample2.
2-2 Materials Analysis of Pt / DMO / TiN (STD)
For this topic, we would introduce our materials characteristic by X-ray Photoelectron Spectroscopy (XPS), Transmission Electron Microscopy (TEM), and Fourier Transform Infrared Spectroscopy (FTIR). We prepared many pieces of dummy wafer which has cleaned, and then we deposited the DMP film on different pieces of dummy wafer for analyzing the characteristic of materials and film
24
thickness.
2-2.1
X-ray Photoelectron Spectroscopy (XPS)
To confirm the composition of the DMO thin film, we have performed XPS analysis of DMO. Fig-2.2.1 shows the XPS Dy 4d core-level photoemission spectrum, which consists of two main peaks for the Dy-Mn and Dy-O bonds, respectively. [39] The O 1s spectrum is shown in Fig-2.2.2, and contains the two peaks of O-Mn and O-Dy. [40] From the XPS signal area ratio of three elements with sensitivity correction, the quantified composition ratio of the film is DyMn2.4O4.6. These results
indicate that some oxygen vacancies exist in the pristine resistive switching film.
2-2.2
Fourier Transform Infrared Spectroscopy (FTIR)
First, we can observe the spectrum of far-FTIR for our DMO film from Fig-2.2-3. It shows the Dy-O bonds are near 458 cm-1 and 550 cm-1 and the Mn-O bonds are near 610 cm-1 and 688 cm-1. [41] [42] Second, we can see the chemical bonds in mid-FTIR. The mid-FTIR detects the longer frequency band range so there is some overlap wave peak. Therefore, we can observe the Mn-H-O bond is near 650 cm-1. Fig-2.2-4 depicts the mid-FTIR. [41] [42]
2-2.3
Transmission Electron Microscopy (TEM)
We can direct analyze the material characteristic and thickness of STD by TEM. Fig-2.2-5 not only shows the STD device cross-section but also manifests the crystallinity of film. The Pt film shows a blacker region at the top, DMO film is under Pt film, and TiN is under the DMO film at the bottom. We got some other information of film. For instance, the thickness of DMO is about 12 nm and DMO was not
25
crystallization.
2-3 Electric Characteristic Measure
In this section, we measure the STD by Agilent B1500A and Agilent 4156C semiconductor characterization analyzer. The analyzer is controlled by the computer. First, we placed the STD device on the thermal chuck, and used probes contact with the pad of sample 1. Fig-2.3-1shows the probes contact with pads. There are two pads in each device of STD, so one of pads is linked top electrode (TE) and the other one is linked bottom electrode (BE). Second, the computer is inputted parameters controlled the analyzer. We biased the bottom electrode and grounded the top electrode is like Fig-2.3-1. Here, the forward bias is defined by the current flow from TiN electrode to Pt electrode we called (by using a positive voltage to TiN), and the current flow from Pt to TiN is reverse bias (by using a negative voltage to TiN). Subsequently, we will observe three parts in the STD device, forming process, after forming process, and reliability.
2-3.1
Forming Process
For the STD device, the state of device is still an insulator at initial state before forming process. On the other hand, we bias the BE under small voltage and the current shows low value. However, the device is applied a forming voltage and the current will increase sharply. The forming process has mentioned above former chapter1 so we don’t explain them again. Here, we will see our device during forming process. Fig-2.3-2 depicts the forming process of our STD device; we will divide into three regions during forming process. In region1, the voltage range is from 0V to 3V.
26
In region2, the voltage range is from 3V to 6V and the voltage from 6V to 8V is region3.
The region1 shows the state is high resistance and it is still original state because the electrons can’t pass the insulator. Therefore, the current value is low under small bias in pristine film. The region2 illustrated two steps, one is a negative differential resistance (NDR) phenomenon from 3V to 4V and the other one is the current trend toward smooth situation when the voltage is from 4V to 6V. The NDR phenomenon means the current decreases with the voltage increasing. In region3, the device can achieve the breakdown condition near 7V and the current can increase sharply. We set the compliance current which is about 1mA to avoid hard-breakdown when the current increases sharply. The device had been changed its material structure after forming process.
Here, there are two different kinds of breakdown, one is called hard- breakdown and the other one is called soft-breakdown. The hard-breakdown is formed due to the insulator was produced a strong CPs which were consist of many defects, and the device is normally ON state. [28] [43] [44] And then it is difficult to recover a defect-rich conduction path by using of voltage when hard-breakdown is reached. It is different between soft-breakdown and hard-breakdown because soft-breakdown is reversible behavior. It mean that the device state can be changed by applying voltage from ON to OFF state. The device is able to recover the defects because the appropriate amount of bonds between oxygen and metal are broken. Soft-breakdown process is happened if we set compliance current condition. For hard-breakdown, it would be difficult to attract the ionic oxygen back by external bias, if the phenomenon is happen. [28] [43] [44]
27
2-3.2
Current-Voltage after Forming Process
Fig-2.3-3 shows typical current-voltage characteristics of STD device. We can divide into 6 steps in the Fig-2.3-3. After positive forming process, the state was at LRS. This is step1 when the device is given a negative bias from 0V to about -1 V and the state of device is still LRS. It is a little different step1 between step2 until the voltage is over -1V and the -1V is like a critical value which make device change the state. Step2 is swept from -1V to -1.8V. The swept range is from -1.8V to 0V is the step3. We discovered the sample1 shows two characteristic current curves, one is LRS and the other one is HRS. The bias is once over critical value (-1V), the state changed from LRS to HRS. When bias was increased from 0V to about +1V, this step we called step4. In step4, the current-voltage (I-V) curve possess NDR phenomenon near about 0.7V but the resistive switching behavior was still HRS. The current value increased rapidly when the voltage is over about 1V, and then the current achieved our set compliance current. The compliance current is 10-2A . We sweep the voltage range from about 1V to 2V in the step5. The step6 swept voltage range is from 2V to 0V. The state is HRS in step4, but the state is changed from HRS to LRS through the step5.
2-3.3
The Carrier Conduction mechanism fitting
From Fig-2.3-4 we can discover the carrier conduction mechanism at a piece of I-V curve. First, we discuss the LRS carrier transportation. The carrier transportation mechanism is Ohmic conduction at LRS. Second, the carrier at HRS is also ohmic mechanism under small bias. Subsequently, the voltage range from -0.7V to -1.3V is Schottky barrier conduction mechanism and we also discover the voltage range from -1.5V to -1.8V is thermionic field emission.
28
2-3.4
Temperature effect of LRS and HRS characteristic
In order to verify the STD device transportation behavior we investigate the LRS and HRS with temperature effect. From Fig-2.3-5 depicts two different characteristics, one is metal-like at LRS and the other one is semiconductor-like at HRS. First, we observe HRS, the resistance state decrease with temperature increasing. It means the HRS is semiconductor conduction because the carriers have higher mobility under high temperature than under lower temperature. Second, the state increase with temperature increase at the LRS. It means the carriers have low mobility under higher temperature than under lower temperature.
2-3.5
Reliability
Here, we will discuss the reliability test and the reliability divide into two indices, one is retention and the other one is endurance test.
2-3.5-1 Retention
First, we observe the retention test. Retention is one of important reliability test for NVM device. The device is required retention time longer than 10 years for universal NVM. On the other hand, it has to maintain the state over 10 years if the device is on or off state. We test method is a kind of thermal stress so the retention time must to keep at thermal stress at 850C. We would apply a smaller voltage from 0V to 0.2V read our device state. At 1s, 30s, 100s, 300s, 1000s, 3000s, 5000s, 7000s, 9000s, 10000s, the device would measure the current applying 0.2V in order to inspect the state is changed or not.
Fig-2.3-5 shows the retention of our device and kept the thermal stress at 85 C. On- Off ratio was still about 100 resistance ratio at 850C. The state was stable over 10000
29
seconds. It indicated the state of our device was not changed by thermal stress at 85 C. It is a good appearance for our STD device.
2-3.5-2 Endurance
The endurance is another important reliability test, and it is an indicator of reliability. For our device can divide into two different kinds of endurance, one is DC endurance and the other one is AC endurance. First, we chose DC endurance as simple analysis. We applied DC sweep to operate our device. And then the device is changed the state from HRS to LRS by applying forward bias when the state was HRS. Next, the device sweep reverse bias made the state from LRS to HRS. That we mentioned above, it is a sweep cycle (HRS→LRS, LRS→HRS) and the LRS and HRS can be read by 0.2V. Subsequently, Fig-2.3-6 depicts the endurance characteristics between HRS and LRS of STD device. The switch behavior of this device (on / off ratio) was about 100 resistance ratio at 0.2 V, and the resistance state was gradually stabilized over 100 times of cycles. From DC endurance, it is a good device for our sample.
Contemporary Non-Volatile Memory (NVM) shows resistive cycle times could achieve between 103 and 107 [45], and our DC measurement only achieve 100 cycle times. It is not enough to compare to modern NVM, so we have to use AC pulse measurement test application for our device. The AC measurement is different DC. I-V curve took on a full state (from HRS to LRS, or From LRS to HRS), and I -V curve is successive by DC measurement for our device.
First, the device state is at LRS and we apply a reverse bias pulse to change its state from LRS to HRS. Fig-2.3-7(a) shows the condition of reverse bias pulse. Next, we bias a smaller voltage to read its state. Second, we give a forward bias pulse to make HRS switching to LRS, and read it by 0.2V. Fig-2.3-7(b) depicts the condition of
30
reverse bias pulse. We applied thousands of forward and reverse bias pulse into the device at a time in order to reduce the time what we gave a pulse make it changed the state and read its state. It is like Fig-2.3-7(c). And then we gave a smaller bias to read its state again. Repeat the process for hundred times. It could obtain endurance of 105 times. The y-axis indicates resistance, the x-axis indicates switching times. Fig-2.3-8 shows the on / off ratio is about 10 resistance, the HRS fluctuation is acute than LRS. There were some error points in the AC endurance.
31
Fig-2.1.1 depicts the substrate cross-section.
32
Fig-2.1.3 shows the cross-section of the sample2.
Dy 4d
Binding Energy (eV)
148
150
152
154
156
158
160
Inten
sity
Experiment data Fitting resultDy-O
(155.8 eV)
Dy-Mn
(153.5 eV)
Fig-2.2.1 shows the XPS Dy 4d core-level photoemission spectrum, which
33
O 1s
Binding Energy (eV)
526
528
530
532
534
536
Inten
sity
Experiment data Fitting resultO-Mn
(530 eV)
O-Dy
(531.9 eV)
The O 1s spectrum is shown in Fig-2.2.2, and contains the two peaks of
O-Mn and O-Dy. [40]
34
Fig-2.2-4 depicts the mid-FTIR.
35
Fig-2.3-1
shows the probes contact with pads.
36
Fig-2.3-3 shows typical current-voltage characteristics of STD device.
Fig-2.3-4 we can discover the carrier conduction mechanism at a piece of
37
Fig-2.3-5 depicts two different characteristics, one is metal-like at LRS
and the other one is semiconductor-like at HRS.
38
Fig-2.3-6 depicts the endurance characteristics between HRS and LRS of STD device.
The switch behavior of this device (on / off ratio) was about 100 resistance ratio at 0.2 V, and the resistance state was gradually stabilized over 100 times of cycles.
Fig-2.3-7(a) shows the condition of reverse bias pulse and Fig-2.3-7(b) depicts
39
Fig-2.3-8 shows the on / off ratio is about 100 resistance, the HRS
40
Chapter 3
Coexistence of Filament and Homogenous
Resistive Switching for DyMn
2
O
5
film
Resistive switching memories can divide into some mechanisms such as filament-type, interface-type and other types. However, our device I-V curve possesses a strange NDR phenomenon. Subsequently, we will investigate the NDR before forming, after forming and forming-free. Therefore, we discover the device possess filament and homogenous resistive switching behavior.
3-1 Electric Characteristic Measure before Forming
In the chapter3, we have observed a strange phenomenon from forming process. It is a little different from normal forming process because it exist the negative differential resistance (NDR) phenomenon nearly 3V. We will divide into three parts to discuss the NDR, one is temperature effect, another is constant voltage stress and the other is thickness effect.
3-1.1 Temperature Effect
We observed the forming process by using the STD sample in the experiment. First, we select same cell size devices which are on the STD device. And then we sweep the forming process condition at different temperature, i.e., 303k, 323k, 343k, 363k, 383k and 403k. We could observe a trend that the voltage of NDR depends on temperature. On the other hand, the voltage of NDR decreased with increasing temperature, i.e.,
41
303k is at about 4.2V, 343k is at about 2.7V, and 403k is at about 2V and so on. Fig-3.1-1 depicts what we mentioned above. We select the occurrence voltage of NDR and make a plot recording the distribution of NDR voltage. Fig-3.1-2 clearly shows the voltage of NDR depends on temperature and it shows the NDR voltage decrease with temperature increasing.
3-1.2 Constant Voltage Sampling Effect
In this experiment, we apply a constant voltage to stress the device which is a pristine film and then observe the current of device with time. First, the device is stressed through fixed 2V under different temperature, i.e., 303k, 323k, 343k and 363k. The current-time curve possess two kinds of curves, first is the current gradually increase with time increasing and second is the current possess smooth with time increasing. Fig-3.1-3 shows normal relation between current and time. Subsequently, we applied a fixed voltage 2V to the device. Fig-3.1-4(a) and Fig-3.1-4(b) shows current time curve at 303k and at 323k, respectively. We can observe the different between experiment and we think phenomenon. Under 303k and at 323k, the current-time possesses two trends, first is the current increase with time increasing and second is current decrease with time increase. Finally, the current achieve saturated current value. Fig-3.1-4(c) and Fig-3.1-4(d) depicts the current time curve at 343k and 363k, respectively. Fig-3.1-4(c) and Fig-3.1-4(d) are a little different from temperature under 303k and at 323k. Because there is no a step which is the current increase with time increasing under 343k and 363k. Therefore, we put these relations of current-time in the same plot and then compare them with each other. From Fig-3.1-5 we can discover the rate of current decrease to saturated state depend on the temperature, i.e., the current decay rate at higher temperature is faster than at