Ramanathan Sh. (Ed.) Thin Film Metal-Oxides: Fundamentals and Applications in Electronics and Energy

Подождите немного. Документ загружается.

128 J. Lu et al.

64. Lee S, Fursina A, Mayo JT, Yavuz CT, Colvin VL, Soﬁn RGS, Shvets IV, Natelson D (2008)

Electrically driven phase transition in magnetite nanostructures. Nat Mater 7:130–133

65. West KG, Lu JW, Yu J, Kirkwood D, Chen W, Pei YH, Claassen J, Wolf SA (2008) Growth

and characterization of vanadium dioxide thin ﬁlms prepared by reactive-biased target ion

beam deposition. J Vac Sci Technol A 26:133–139

66. Hu G, Suzuki Y (2002) Negative spin polarization of Fe

in magnetite/manganite-based

junctions. Phys Rev Lett 89:276601

67. Alldredge LMB, Chopdekar RV, Nelson-Cheeseman BB, Suzuki Y (2006) Complex oxide-

based magnetic tunnel junctions with nonmagnetic insulating barrier layers. J Appl Phys

99:08K303

68. Alldredge LMB, Chopdekar RV, Nelson-Cheeseman BB, Suzuki Y (2006) Spin-polarized

conduction in oxide magnetic tunnel junctions with magnetic and nonmagnetic insulating

barrier layers. Appl Phys Lett 89:182504

69. Ohno H, Chiba D, Matsukura F, Omiya T, Abe E, Dietl T, Ohno Y, Ohtani K (2000) Electric-

ﬁeld control of ferromagnetism. Nature 408:944–946

70. DiVincenzo DP, Bacon D, Kempe J, Burkard G, Whaley KB (2000) Universal quantum com-

putation with the exchange interaction. Nature 408:339–342

71. Ohya S, Ohno K, Tanaka M (2007) Magneto-optical and magnetotransport properties of heav-

ily Mn-doped GaMnAs. Appl Phys Lett 90:112503

72. Mack S, Myers RC, Heron JT, Gossard AC, Awschaloma DD (2008) Stoichiometric growth

of high Curie temperature heavily alloyed GaMnAs. Appl Phys Lett 92:192502

73. Dietl T, Ohno H, Matsukura F, Cibert J, Ferrand D (2000) Zener model description of ferro-

magnetism in zinc-blende magnetic semiconductors. Science 287:1019–1022

74. Matsumoto Y, Murakami M, Shono T, Hasegawa T, Fukumura T, Kawasaki M, Ahmet P,

Chikyow T, Koshihara S, Koinuma H (2001) Room-temperature ferromagnetism in transpar-

ent transition metal-doped titanium dioxide. Science 291:854–856

75. Chambers SA (2006) Ferromagnetism in doped thin-ﬁlm oxide and nitride semiconductors

and dielectrics. Surf Sci Rep 61:345–381

76. Janisch R, Gopal P, Spaldin NA (2005) Transition metal-doped TiO

and ZnO – present status

of the ﬁeld. J Phys Condens Matter 17:R657–R689

77. Abraham DW, Frank MM, Guha S (2005) Absence of magnetism in hafnium oxide ﬁlms.

Appl Phys Lett 87:252502

78. Shinde SR, Ogale SB, Higgins JS, Zheng H, Millis AJ, Kulkarni VN, Ramesh R, Greene

RL, Venkatesan T (2004) Co-occurrence of superparamagnetism and anomalous Hall effect

in highly reduced cobalt-doped rutile TiO

2-delta

ﬁlms. Phys Rev Lett 92:166601

79. Ando K (2006) Seeking room-temperature ferromagnetic semiconductors. Science 312:

1883–1885

80. Zhao T, Shinde SR, Ogale SB, Zheng H, Venkatesan T, Ramesh R, Das Sarma S (2005)

Electric ﬁeld effect in diluted magnetic insulator anatase Co W TiO

. Phys Rev Lett 94:126601

81. Spaldin NA, Fiebig M (2005) The renaissance of magnetoelectric multiferroics. Science

309:391–392

82. Eerenstein W, Mathur ND, Scott JF (2006) Multiferroic and magnetoelectric materials. Nature

442:759–765

83. Ramesh R, Spaldin NA (2007) Multiferroics: progress and prospects in thin ﬁlms. Nat Mater

6:21–29

84. Cheong SW, Mostovoy M (2007) Multiferroics: a magnetic twist for ferroelectricity. Nat

Mater 6:13–20

85. Ederer C, Spaldin NA (2006) Recent progress in ﬁrst-principles studies of magnetoelectric

multiferroics. Curr Opin Solid State Mater Sci 9:128–139

86. Wang J, Scholl A, Zheng H, Ogale SB, Viehland D, Schlom DG, Spaldin NA, Rabe KM,

Wuttig M, Mohaddes L, Neaton J, Waghmare U, Zhao T, Ramesh R (2005) Response to

comment on “Epitaxial BiFeO

multiferroic thin ﬁlm heterostructures”. Science 307:1103959

87. Haeni JH, Irvin P, Chang W, Uecker R, Reiche P, Li YL, Choudhury S, Tian W, Hawley ME,

Craigo B, Tagantsev AK, Pan XQ, Streiffer SK, Chen LQ, Kirchoefer SW, Levy J, Schlom

DG (2004) Room-temperature ferroelectricity in strained SrTiO

. Nature 430:758–761

3 Novel Magnetic Oxide Thin Films 129

88. Baettig P, Ederer C, Spaldin NA (2005) First principles study of the multiferroics

BiFeO

; Bi

FeCrO

, and BiCrO

: structure, polarization, and magnetic ordering tempera-

ture. Phys Rev B 72:214105

89. Ederer C, Spaldin NA (2005) Weak ferromagnetism and magnetoelectric coupling in bismuth

ferrite. Phys Rev B 71:060401

90. Eerenstein W, Morrison FD, Dho J, Blamire MG, Scott JF, Mathur ND (2005) Comment on

“Epitaxial BiFeO

multiferroic thin ﬁlm heterostructures”. Science 307:1203

91. Zhao T, Scholl A, Zavaliche F, Lee K, Barry M, Doran A, Cruz MP, Chu YH, Ederer C,

Spaldin NA, Das RR, Kim DM, Baek SH, Eom CB, Ramesh R (2006) Electrical control

of antiferromagnetic domains in multiferroic BiFeO

ﬁlms at room temperature. Nat Mater

5:823–829

92. Nogues J, Schuller IK (1999) Exchange bias. J Magn Magn Mater 192:203–232

93. Bea H, Bibes M, Cheriﬁ S, Nolting F, Warot-Fonrose B, Fusil S, Herranz G, Deranlot C,

Jacquet E, Bouzehouane K, Barthelemy A (2006) Tunnel magnetoresistance and robust room

temperature exchange bias with multiferroic BiFeO

epitaxial thin ﬁlms. Appl Phys Lett

89:242114

94. Dho JH, Qi XD, Kim H, MacManus-Driscoll JL, Blamire MG (2006) Large electric polariza-

tion and exchange bias in multiferroic BiFeO

. Adv Mater 18:1445

95. Martin LW, Chu YH, Zhan Q, Ramesh R, Han SJ, Wang SX, Warusawithana M,

Schlom DG (2007) Room temperature exchange bias and spin valves based on

BiFeO

=SrRuO

=SrTiO

=Si (001) heterostructures. Appl Phys Lett 91:172513

96. Martin LW, Chu YH, Holcomb MB, Huijben M, Yu P, Han SJ, Lee D, Wang SX, Ramesh R

(2008) Nanoscale control of exchange bias with BiFeO

thin ﬁlms. Nano Lett 8:2050–2055

97. Chu YH, Martin LW, Holcomb MB, Gajek M, Han SJ, He Q, Balke N, Yang CH, Lee D,

Hu W, Zhan Q, Yang PL, Fraile-Rodriguez A, Scholl A, Wang SX, Ramesh R (2008) Electric-

ﬁeld control of local ferromagnetism using a magnetoelectric multiferroic. Nat Mater 7:

678–678

98. Nan CW, Bichurin MI, Dong SX, Viehland D, Srinivasan G (2008) Multiferroic magnetoelec-

tric composites: historical perspective, status, and future directions. J Appl Phys 103:022511

99. Zheng H, Wang J, Loﬂand SE, Ma Z, Mohaddes-Ardabili L, Zhao T, Salamanca-Riba L,

Shinde SR, Ogale SB, Bai F, Viehland D, Jia Y, Schlom DG, Wuttig M, Roytburd A,

Ramesh R (2004) Multiferroic BaTiO

-CoFe

nanostructures. Science 303:661–663

100. Zavaliche F, Zheng H, Mohaddes-Ardabili L, Yang SY, Zhan Q, Shafer P, Reilly E,

Chopdekar R, Jia Y, Wright P, Schlom DG, Suzuki Y, Ramesh R (2005) Electric ﬁeld-induced

magnetization switching in epitaxial columnar nanostructures. Nano Lett 5:1793–1796

101. Zavaliche F, Zhao T, Zheng H, Straub F, Cruz MP, Yang PL, Hao D, Ramesh R (2007) Elec-

trically assisted magnetic recording in multiferroic nanostructures. Nano Lett 7:1586–1590

102. Brinkman A, Huijben M, van Zalk M, Huijben J, Zeitler U, Maan JC, van der Wiel WG,

Rijnders G, Blank DHA, Hilgenkamp H (2007) Magnetic effects at the interface between

non-magnetic oxides. Nat Mater 6:493–496

Chapter 4

Bipolar Resistive Switching in Oxides

for Memory Applications

Rainer Bruchhaus and Rainer Waser

Abstract Resistance change Random Access Memory (RRAM) devices in which

at least two resistance states are used are a top candidate for future nonvolatile

data storage. Simple Metal-Insulator-Metal (MIM) structures form the memory ele-

ment which can be easily incorporated in large arrays. In particular, in the so-called

Valence Change Memories (VCM) the drift of anions, typically oxygen, is consid-

ered as the key step to explain the bistable resistive switching behavior. A ﬁrst-order

classiﬁcation of the observed material changes is related to the geometrical location.

In “ﬁlamentary” type switching the formation and rupture of a thin ﬁlament is re-

sponsible for the resistance change. In the “distributed” systems the switching can

be traced back to modiﬁcations at interfaces. Oxygen ion migration into thin tun-

nel oxides in high electric ﬁelds and Schottky barrier engineering with metals and

complex perovskites are two mechanisms under discussion for the distributed sys-

tems. In the ﬁlamentary type of switching fast oxygen ion transport along extended

defects is demonstrated to be the key step for the formation of the conducting ﬁl-

aments. The bistable resistance characteristics with switching induced by voltage

pulses is a promising approach for future nonvolatile memory technologies. Excel-

lent scaling behavior to sizes below 20 nm has been demonstrated.

4.1 Introduction and Motivation

As the standard high density mass memory technologies like Dynamic Random Ac-

cess Memory devices (DRAM) and Flash are approaching their scaling limits it has

become increasingly difﬁcult and expensive to progress into the next technology

nodes and density generations. In order to overcome these limitations alterna-

tive principles for data storage are under investigation. Among these “emerging”

memory technologies the so-called Resistance change Random Access Memory

(RRAM) devices in which at least two resistance states in suitable materials are

used to store the memory bits are looking very promising.

R. Bruchhaus (



)andR.Waser

Institute of Solid State Research, Forschungszentrum Juelich, 52425 Juelich, Germany

e-mail: r.bruchhaus@fz-juelich.de

S. Ramanathan (ed.), Thin Film Metal-Oxides: Fundamentals and Applications

in Electronics and Energy, DOI 10.1007/978-1-4419-0664-9

 Springer Science+Business Media, LLC 2010

131

132 R. Bruchhaus and R. Waser

This new type of memory device uses materials preferably embedded into a

parallel plate like capacitor structure with metal electrodes which alleviate the

electrical stimulation to form different resistance states. This simple metal-insulator-

metal (MIM) memory element structure facilitates the incorporation into large

arrays of memory cells including a suitable select device like a transistor or a diode.

A resistance change effect after suitable electrical stimulation has been found in

many material systems. In the Phase Change RAM (PCRAM) the resistance differ-

ence between the amorphous and the crystalline state of ternary metal alloys like

GeSbTe is utilized [1]. In the Conductive Bridge RAM (CBRAM), also called Elec-

trochemical Metallization Memory (ECM) [2] or Programmable Metallization Cells

(PCM) [3], the electrochemical formation and rupture of very thin metal ﬁlaments

of Ag or Cu in an ion-conducting material like GeSe or GeS is used to form dif-

ferent resistance states [4]. Another class of RRAM, which will be in the focus

of this chapter, is based on the drift of anions, typically oxygen ions (often the

oxygen vacancy motion is considered), in transition metal oxides (TMO) and a sub-

sequent valence change in the cation sublattice. We name this class Valence Change

Memories (VCM). The classiﬁcation of the nanoionics-based memories into cation

migration and anion migration types is described in more detail in [5].

A ﬁrst report about resistance switching in MIM devices in which the isolator

“I” is an oxide dates back in 1962 in which Hickmott reported a hysteretic resis-

tance change in Al=Al

=Al MIM structures after application of voltage pulses

[6]. Later, similar effects were observed in NiO [7]andSiO[8]. This early period

of research was based mainly on relatively thick ﬁlms which in some cases needed

the application of high voltages in the range of several hundred volts was compre-

hensively reviewed earlier [9–11].

In this earlier work, the ﬁlms were switched using the so-called unipolar or non-

polar mode, i.e., the switching event does not depend on the polarity. Switching

from the high resistive OFF state to the low resistive ON state was achieved by the

same polarity (either positive or negative) (Fig. 4.1a).

Current [µA to mA]

Voltage [few volts]

Current [µA to mA]

Voltage [few volts]

Unipolar

Bipolar

SET

RESET

OFF

states

processes

SET

RESET

OFF

Fig. 4.1 Comparison of unipolar (a) and bipolar (b) switching characteristics. For the bipolar

switching SET and RESET processes occur at different polarity. The states ON and OFF received

after the switching processes SET and RESET are indicated in gray. CC is a current compliance

used to limit the current preventing damage to the memory element

4 Bipolar Resistive Switching in Oxides for Memory Applications 133

A memory element in the high resistive “OFF” state is biased either with positive

or negative polarity until it switches into the low resistive “ON” state. This process

is frequently called as the “SET” process. During the set process the current is lim-

ited by a compliance unit of the measurement unit to avoid damage to the sample

by excessive current. The reverse operation is called the “RESET” process in which

the memory element transforms from the ON state back to the OFF state. In gen-

eral, this process is driven by the higher current and the reset voltage is found to

be lower than the set voltage, i.e., V

set

> V

reset

,andI

reset

> I

set

. In the bipolar

switching characteristics the SET and RESET process are observed at different po-

larities (Fig. 4.1b). Inherently, the MIM structure must include some asymmetry for

the bipolar switching characteristics. Asymmetry may include different electrode

materials or a graded composition of the isolator or the voltage polarity during the

initial electroforming step. To our knowledge, the ﬁrst report of bipolar switching,

however not explicitly mentioned, is included in the paper of Hiatt and Hickmott

[12]. In this paper, the bistable switching of Nb

ﬁlms with asymmetric Nb and

Bi electrode structure has been described.

Increased interest in new or emerging memory technologies in which a variety

of hysteretic effects in different materials is investigated for data storage lead to re-

newed interest in switching effects in oxides. This period started in the late 1990s

and now the focus is on the preparation of memory elements in thin ﬁlm form with

switching voltages in the range of the supply voltages of semiconductor memory

circuits to achieve compatibility and integration with CMOS processing. An exam-

ple is the paper of Beck [13] in which thin SrZrO

ﬁlms were reproducibly switched

at ˙0:5 V. Careful adjustment of the applied voltage lead to multiple levels of re-

sistance states opening the ﬁeld for potential multibit memory devices. Another

example is the observation of reversible resistance change effects on epitaxially

grown Pr

0:7

0:3

MnO

(PCMO) ﬁlms by Liu et al. [14] in 2000.

In the transition metal oxides of interest for the use in RRAM oxygen vacan-

cies frequently have a higher mobility than the cations. Under bias voltage oxygen

vacancies as a positively charged species are attracted by the cathode. Most of the

electrodes are not permeable for oxygen and as a result an oxygen-deﬁcient region is

formed in the vicinity of the cathode. As an accommodation process to compensate

for the oxygen deﬁciency electrons emitted from the cathode get trapped. For TiO

and the titanates, for example, this results in a reduction of the Ti-ion according to



C Ti

! Ti

.4n/C

The increasingly reduced valence state typically changes the oxide into a metalli-

cally conducting phase, for Ti-oxide typically for n>1:5in TiO

22=n

.Thisregion

expands into the bulk of the material and forms a “virtual cathode,” which moves

toward the anode [15]. Finally this reduction will lead to a conduction path between

the cathode and the anode. At the anode an oxidation reaction is observed accord-

ing to

! V

C 2e



;

134 R. Bruchhaus and R. Waser

which may result in the evolution of oxygen or the oxidation of the electrode

material. In single crystals, the progress of the chemically reduced region can be di-

rectly observed and it may take several hours at room temperature to form the

conducting ﬁlament. In thin ﬁlms it will occur on a much shorter timescale and

voltages in the range of 1–5 V may be sufﬁcient.

In the following paragraphs, bipolar switching effects in transition metal oxides

will be reviewed. Special attention will be drawn to a classiﬁcation of the different

materials and electrodes related to the underlying switching mechanisms. It will be

shown that the mobility of oxygen ions or oxygen vacancies and their local distribu-

tion on the nanometer scale plays the key role for the understanding of the observed

bipolar switching effects in transition metal oxides. However, it will also become

clear that many observations and results have not been completely understood and

that substantially more research needs to be done. In this exciting ﬁeld more explo-

rative work is deﬁnitely rewarding.

4.2 Classiﬁcation of Bipolar Switching Effects

4.2.1 Geometrical Localization of the Switching Event

The geometrical location of the material property change responsible for the switch-

ing event can serve as a ﬁrst-order classiﬁcation of the bipolar switching ef-

fects. It is either a locally conﬁned ﬁlamentary effect or an interface-related effect

distributed homogeneously over the complete metal electrode area of the MIM

structure (Fig. 4.2).

Filaments in the MIM structure can be arranged vertically or horizontally with

respect to the sample surface (Fig. 4.3).

Fig. 4.2 First-order classiﬁcation of bipolar switching effects. (a) Filamentary conduction path,

(b) homogeneously distributed interface effect (after [16])

4 Bipolar Resistive Switching in Oxides for Memory Applications 135

Fig. 4.3 Schematic image of the ﬁlamentary conduction in MIM structures. The red-colored ﬁl-

ament indicates the lowest resistance path responsible for the ON state resistance. (a) Vertical

arrangement between two metal electrodes M (b) horizontal arrangement between two metal elec-

trodes M (after [16])

The ﬁlaments are formed during the ﬁrst set process which is very often called

the “forming process.” Switching between the ON and the OFF state is achieved

by ﬁlament rupture and restoration. In the horizontal arrangement (Fig. 4.3b), the

ﬁlament can be observed by an optical microscope if the ﬁlament size is sufﬁciently

large and the ﬁlament is formed in the near surface region of the sample [17]. Ini-

tially, numerous ﬁlaments start to form. However, due to current channeling and

electric ﬁeld distribution effects it is expected that one ﬁlament will form a low

resistance path between the two electrodes. Due to the limitations with respect of

patterning of the metal electrodes the distance between the electrodes is usually

much larger for the horizontal arrangement compared to the vertical arrangement.

Thus switching times are expected to be much longer and every surface modiﬁcation

by a possible contamination or reaction with the environment can lead to signiﬁcant

changes in the switching behavior. The vertical arrangement is very close to the ap-

plication in the memory devices. However, due to the conﬁnement of the ﬁlament

down to the size of several nanometers it is very hard to identify the ﬁlaments in this

arrangement using microscopy.

4.2.2 Memory Window Considerations

The window between the OFF and the ON state is an important parameter for the

scaling potential of the resistance change memory device technology. This window

can be characterized by the R

off

ratio of the OFF and the ON state. For the

resistance modulation of a homogeneously distributed interface effect the R

off

ratio does not depend on the area, i.e., R

off

is constant when memory elements

with different sizes are fabricated. For a ﬁlamentary conduction in the ON state and

a homogenous conduction in the OFF state the R

off

ratio covers a huge range

as will be illustrated with a simple example.

136 R. Bruchhaus and R. Waser

film

filament

50nm

OFF ON

SET

RESET

Fig. 4.4 Model to calculate R

off

ratios for ﬁlamentary switching (see text)

Let us assume the ﬁlament formed by a set process in a MIM structure is a fairly

good conductor with a resistivity of 1  10

4

cm and only one ﬁlament is formed.

Further, it is assumed that the RESET process completely removes the ﬁlament

and the virgin isolator is completely recovered. The insulator has a resistivity of

110

cm and a thickness of 50 nm. The length a (area a

) is varied, as well as the

ﬁlament cross-section. Figure 4.4 gives a schematic representation of the memory

element in the OFF and the ON state for this simple one ﬁlament model [18].

The R

off

ratio is increasing with decreasing memory element size and be-

comes >10

for sizes smaller than 100  100 nm

(Fig. 4.5). This is a size which

is expected to be at the higher end for fully integrated structures in future high den-

sity memory device applications. On the other hand, many laboratories engaged in

research of resistance switching effects do not use sophisticated lithography tools

to prepare samples with the target sizes for future technologies. They rely on pat-

terning of top electrodes by contact lithography or so-called metal mask or shadow

mask experiments in which the top electrode material is deposited through openings

in a metal foil in intimate contact with the insulator surface. Typical sizes for this

approach are 30  30 m

to 100  100 m

, i.e., sizes found on the higher end of

thescaleofFig.4.5. Figure 4.5 clearly demonstrates that for these larger structures

with sizes larger than 10  10 m

much lower R

off

ratios are observed due to

the parallel resistance of the insulator. Thus, for the comparison of R

off

ratios of

different materials the memory element size must be given. In addition, the ﬁlament

size itself has a considerable impact. In the range given in Fig. 4.5 a difference of

more than one order of magnitude is observed. To minimize the range of the resis-

tance of the ON state a well-controlled switching process with reproducible ﬁlament

size for the ﬁlamentary type of switching process is needed.

4 Bipolar Resistive Switching in Oxides for Memory Applications 137

100 x 100 nm

10 x 10 μm

Ronf50nm1E6R2_grp

off

Area [nm

]

Fil. 2x2 nm

Fil. 4x4 nm

Fil. 6x6 nm

Fil. 8x8 nm

Fil. 10x10 nm

Fig. 4.5 Calculated R

off

ratio for ﬁlamentary switching. Memory element area is varied be-

tween 400 nm

and 10

, the ﬁlament size between 2  2 nm

to 10  10 nm

.TheR

off

ratio varies over more than 8 orders of magnitude

4.2.3 Observed Area Dependencies of the Resistive

Switching Effect

Experimentally two approaches can be used to understand the ﬁlamentary or in-

terface type effect. A measurement series switching current versus electrode size

reveals a ﬂat behavior for the ﬁlamentary type of conduction as the current is solely

ﬂowing through the ﬁlament at least for the ON state (Fig. 4.6a). For the interface-

related behavior the current is proportional to the electrode size for the OFF and the

ON state (Fig. 4.6b).

A second method to test the switching mechanism requires a destructive test.

It is to cut the sample into two pieces and remeasure the parts separately. For the

case that one ﬁlament is formed for one part a current related to the OFF state will

be observed and the sample part which includes the ﬁlament will remain in the ON

state. For a sample exhibiting the interface related effect the currents will be reduced

proportional to the electrode sizes of the different parts.

138 R. Bruchhaus and R. Waser

10 100 1000

TO_RRAM_(b).grp

Resistance [Ohm]

Area [

]

R_initial

R_reset (OFF)

R_set (ON)

0,1 1 10 100

R_set (ON)

R_reset (OFF)

R_initial

Resistance [Ohm]

Area [µm

]

Fig. 4.6 (a) Area dependence of the resistance of NiO in the initial state prior to forming and the

SET and RESET process indicating a ﬁlamental type of conduction for both states [19]. (b)Area

dependence of the resistance for a purely interface-related switching process [20]

4.3 Distributed Bipolar Switching Effects in Transition

Metal Oxides

4.3.1 Oxide Dual Layer Memory Element

Bipolar resistance switching with area scaling for both the ON and the OFF state is

observed in a double layer of a complex metal oxide and a tunnel barrier between

Pt electrodes [20](Fig.4.7).

A sequence of a conductive metal oxide ﬁlm with a thickness of 25 nm and

a thin tunnel oxide (2–3 nm thick) and a Pt top electrode is deposited by sput-

tering onto a Pt bottom electrode. Substrate temperatures during deposition range