ArticlePDF Available

Intrinsic resistive switching and memory effects in silicon oxide


Abstract and Figures

Resistive switching behaviors are described in silicon oxide (SiOx ) systems employing vertical E/SiOx /E (E denotes the electrode) structures. The switching is largely independent of the electrode material and attributed to the intrinsic properties of SiOx . Based on the recent experimental observation (Yao et al. in Nano Lett. 10:4105, 2010) of a silicon filament embedded in the SiOx matrix, we further discuss the switching mechanism in light of the measured electrical phenomena. The set voltages are largely SiOx -thickness independent, consistent with the mechanistic picture of point switching in the silicon filament. The multi-state switching and shifts in the set voltages with respect to the reset voltages are consistent with an electrochemical redox process (Si ↔ SiOy ) at the switching site.
Content may be subject to copyright.
Appl Phys A (2011) 102: 835–839
DOI 10.1007/s00339-011-6267-6
Intrinsic resistive switching and memory effects in silicon oxide
Jun Yao ·Lin Zhong ·Douglas Natelson ·
James M. Tour
Received: 2 November 2010 / Accepted: 22 December 2010 / Published online: 26 January 2011
© Springer-Verlag 2011
Abstract Resistive switching behaviors are described in sil-
icon oxide (SiO
) systems employing vertical E/SiO
denotes the electrode) structures. The switching is largely
independent of the electrode material and attributed to the
intrinsic properties of SiO
. Based on the recent experimen-
tal observation (Yao et al. in Nano Lett. 10:4105, 2010)of
a silicon filament embedded in the SiO
matrix, we further
discuss the switching mechanism in light of the measured
electrical phenomena. The set voltages are largely SiO
thickness independent, consistent with the mechanistic pic-
ture of point switching in the silicon filament. The multi-
state switching and shifts in the set voltages with respect to
the reset voltages are consistent with an electrochemical re-
dox process (Si SiO
) at the switching site.
J. Yao
Applied Physics Program Through the Department
of Bioengineering, Rice University, 6100 Main St., Houston,
TX 77005, USA
L. Zhong (
) · D. Natelson
Department of Electrical and Computer Engineering, Rice
University, 6100 Main St., Houston, TX 77005, USA
L. Zhong · J.M. Tour
Department of Computer Science, Rice University, 6100 Main
St., Houston, TX 77005, USA
D. Natelson (
Department of Physics and Astronomy, Rice University, 6100
Main St., Houston, TX 77005, USA
J.M. Tour (
Department of Chemistry, Rice University, 6100 Main St.,
Houston, TX 77005, USA
1 Introduction
Resistance-change materials have been extensively studied
as potential candidates for future nonvolatile memory [1].
A wide range of materials have been studied to show re-
sistive switching behaviors [2], including metal oxides [3].
While the switching mechanism is still largely unclear and
may vary in different systems, metal elements are usu-
ally considered as the active and indispensable compo-
nents constituting the conducting pathways and hence the
switching phenomena. For example, metal-filament forma-
tion and ionic-state change are two of the commonly pro-
posed causes [3]. For this reason, some metal-oxide systems
even rely on further metal doping for the formation of resis-
tive switching [3, 4]. This focus on metal elements may have
contributed to the comparatively little attention paid to the
most prevalent oxide, SiO
, as a candidate for resistance-
change memory. While some programmable metallization
memory cells have involved SiO
, the memory effect is ex-
trinsic to SiO
, as the oxide merely serves as the solid elec-
trolyte glass matrix for the mobile metal ions [5].
In fact, the idea of extrinsic switching in SiO
by metal
injection from the electrodes has been largely unchanged
since the first observation of resistance-change phenomena
in silicon monoxide (SiO) [6], and was frequently revived in
metallic nanogap systems residing on SiO
substrates [7].
The lack of consideration of potential intrinsic electrical
phenomena from SiO
itself further lead to proposals of
electromechanical/electrochemical switching from the elec-
trodes in various nanogap systems on SiO
substrates [8, 9].
Recently, however, we revealed electrode-independent re-
sistive switching behaviors in various E–SiO
–E systems,
with the electrode materials ranging from metals to semi-
conductors, amorphous carbon and even carbon nanotubes
(CNTs) [10]. This depicted a picture of intrinsic switching
836 J. Yao et al.
Fig. 1 Electrical characterizations in E/SiO
/E devices. (a) Sche-
matic of the structure of E/SiO
/E vertical cell and its electrical
characterization setup. (b) IV curves in a formed TiW/SiO
(120 nm/40 nm/120 nm) device. The numbers beside the curves in-
dicate the voltage sweep order and the black arrows indicate the
sweep directions. Inset shows 80 consecutive IV curves from a
/polySi (70 nm/40 nm/70 nm) vertical device. (c)Mem-
ory cycles in a TiW/SiO
/TiW device using +5V,+ 12 V, and + 1V
as set, reset, and read voltages, respectively. The set and reset currents
are not shown here
solely from SiO
. Following this implication, we demon-
strated resistive switching in a metal-free SiO
system us-
ing polycrystalline silicon (polySi) as the electrode ma-
terial [11]. Furthermore, by examining a localized CNT
–CNT switching system using transmission electron
microscopy, we visualized the conduction path—an aligned
pathway of silicon nanocrystals—embedded in the amor-
phous SiO
matrix, providing direct evidence for and in-
sights in the switching process intrinsic to SiO
[11]. This
intrinsic switching renders SiO
an attractive material for
constructing resistance-change memory, as SiO
is one of
the most frequently used materials in the semiconductor in-
dustry, and hence enables fully CMOS-compatible process-
ing. Here we further discuss the resistive switching behav-
iors in SiO
and their implications to the mechanism.
2 Experimental
The tested memory cells were typical capacitor-like struc-
tures, each with the thin SiO
layer sandwiched between
the top and bottom electrodes (see schematic in Fig. 1a).
The diameters of the cells were 50 µm for an easy probe-tip
landing. TiW and polySi were used as electrode materials to
form TiW/SiO
/TiW and polySi/SiO
/polySi (x 1.9–2)
sandwiched structures. The TiW layers (120 nm thick) were
grown by physical vapor deposition method. The polySi lay-
ers (70 nm thick) were grown by plasma enhanced chemical
vapor deposition (PECVD) method, followed by low-energy
boron implantation to achieve a resistivity 0.005 cm.
The SiO
layers were also grown by the PECVD method.
Standard photolithography and lift-off methods were used
to define a layer (40 nm thick) of circular chromium pat-
terns (50 µm in diameters) as the sacrificial mask. Reactive-
ion etching (SF
for polySi etching and CHF
for both
TiW and SiO
etching) was then used to etch away the
top-electrode layer and the SiO
layer, leaving the regions
underneath the circular chromium patterns protected. The
chromium layer was then dissolved in a chromium etchant
(CEP-200). Electrical characterizations were performed at
room temperature in a vacuum (10
Torr) probe station
connected to an Agilent 4155C semiconductor parameter
3 Results and discussion
Figure 1b shows the typical current-voltage (IV) curves in a
formed TiW/SiO
/TiW (40 nm SiO
) device, demonstrat-
ing unipolar switching behaviors. Unlike some other unipo-
lar systems [1, 2], here the set voltage has a lower magnitude
than the reset voltage; starting from a high-resistance (OFF)
state, a sudden conductance increase (the set process) begins
at 3.5 V. A less abrupt conductance decrease in the low-
resistance (ON) state begins at a higher voltage (5.5V),
defining the reset region. Nonvolatile memory states be-
tween ON and OFF can be written to the device by volt-
age pulses of corresponding magnitudes in the set and reset
regions (see the memory cycle in Fig. 1c). Pulse width as
narrow as 100 ns is achievable [11].
It should be noted that while the metal elements in the
electrodes are susceptible to potential metal-migration re-
lated switching, the electrical behavior here is attributed
solely to the SiO
layer. Programmable metallization mem-
ory using SiO
as solid electrolyte features bipolar switch-
ing behavior [5], as opposed to the unipolar behavior here.
This can be further confirmed by replacing the electrode
material with nonmetals. The same switching behaviors
Intrinsic resistive switching and memory effects in silicon oxide 837
were observed in a metal-free polySi/SiO
/polySi (40 nm
) system (see inset in Fig. 1b) and other planar carbon-
electrode SiO
systems [10]. The visualization of a sili-
con nanocrystalline pathway (silicon filament) embedded
in the SiO
matrix revealed the intrinsic cause of resistive
switching in SiO
: the oxide feeds the formation of the sil-
icon filament as well as providing the solid supporting ma-
trix [11]. Like in many metal-oxide based resistive switch-
ing systems [3], a forming process is needed to transform
the pristine non-conducting state into the switching state in
. From the atomic structural point of view, the forming
process is to strip away some oxygen atoms and enrich the
silicon atoms locally to form silicon filament. The forming
process here involves voltage sweeps to a high value to in-
duce an initial soft breakdown locally in the SiO
layer [11].
In particular, as the etched vertical SiO
edge is more de-
fective or more susceptible to initial soft breakdown, the
switching was revealed to be localized at the vertical SiO
edge [11, 12]. Thermal annealing (600°C for 5 min) in vac-
uum (100 mTorr) was found to facilitate the electroforming
process as it induced an easier soft breakdown, and hence
was adopted in these vertical structures prior to electrical
While the visualization of the silicon filament provides
substantial insights into the intrinsic switching in SiO
the dynamics of the switching process is not yet revealed.
Conductance modulation by trapped charges in quasi-one-
dimensional semiconductors can cause memory effects in
a two-terminal configuration [13, 14]. Nevertheless, the
charging and discharging processes usually require differ-
ent voltage polarities, and hence result in bipolar switching
behavior [14], as opposed to the unipolar one observed here.
Memory states were recorded after the devices sat in an am-
bient environment for three months (Fig. 2). The good air
stability further indicates that the switching is unlikely to
be charge-based. Consequently, a voltage-driven structural
change in the silicon filament is more likely to be the cause.
In this regard, we varied the thickness of the SiO
to vary the length of the silicon filament, to study the effect
on the electrical behaviors. Vertical devices with SiO
thickness ranging from 10 nm to 200 nm were fabricated and
electroformed to switching states. The threshold set voltage
, defined as the voltage value at which the set process
begins) trend with respect to SiO
thickness is plotted in
Fig. 3a. It shows largely thickness-independent behaviors,
with the V
concentrated between 3.5 and 4 V. This weak
-thickness dependence or filament-length dependence
supports the mechanistic picture of point switching in the
filament (see illustration in Fig. 3b): for the large OFF resis-
tance, the set voltage drops mainly across the broken point
in the filament regardless of the filament length. The V
also seems to indicate an intrinsic threshold voltage value
at which some silicon conversion/transfer process begins to
Fig. 2 Memory state retention in polySi/SiO
/polySi devices. (a)ON
states from 22 individual devices after 3 months in ambient environ-
ment. (b) OFF states from 22 individual devices after 3 months in am-
bient environment. Inset shows reading cycles for different memory
states, with each symbol indicating each different memory state. The
memory states were recorded by voltage pulses at a rate of 1 read per
sec. All the memory states were read at +1V
Fig. 3 (a) V
from devices with different SiO
thicknesses. Data
points at 40 nm, 80 nm, 160 nm and 200 nm were collected from
/TiW structures, while data points at 10 nm and 20 nm were
collected from polySi/SiO
/polySi structures. The reset voltages were
at +12 V (+10 V for the 10 nm devices to avoid hard breakdown in
the SiO
layer). (b) Schematics of ON and OFF states resulting from a
point switching in a silicon filament embedded in the SiO
838 J. Yao et al.
Fig. 4 (a) IV curves of the set processes in a same TiW/SiO
(40 nm SiO
) device. Each set IV curve was obtained after a reset
process, with different colors indicating different reset voltage values.
The black, magenta, blue,andgreen colors correspond to set IV curves
after +8V,+10 V, +12 V, and +14 V reset operations, respectively.
(b) Schematics of the valence states of silicon (or oxygen concentra-
tions) at the switching site in the filament with respect to different OFF
conduction states
heal and reconnect the broken point. It is thus likely that the
set process correlates to certain electrochemical processes
which require specific electrical potentials. As the only ele-
ments in the system are oxygen and silicon, we propose that
an electrochemical reduction process of SiO
Si is in-
volved in the set process to reconnect the silicon filament.
The reset process, on the other hand, is through an oxidiza-
tion process of Si SiO
mainly driven by current-induced
local heating, given its voltage-polarity independence and
higher voltage magnitude.
The above-proposed electrochemical redox process can
also account for the multi-level states and corresponding V
shifts observed in the SiO
devices. In the same device, a
lower magnitude in the reset voltage usually results in an in-
creased OFF conductance and it also slightly shifts the V
to a lower magnitude (see Fig. 4a). The conductance mod-
ulation in the OFF states by different reset values enables
the possibility of multi-bit memory [12, 15]. Meanwhile,
the device-size independence in the ON states [11] indicates
that this conductance modulation is through single or few
filaments, not a collective effect from a large number of fila-
ments. In other words, it is the structural/compositional vari-
ation at the switching point in the filament that governs the
conductance modulation. In the mechanistic picture that a
reset process is through Si SiO
, it is expected that the
final valence state of the Si
in SiO
will affect the con-
duction. More generally, a higher valence state in Si
richer oxygen concentration) is likely to result in a more in-
sulating state (see illustration in Fig. 4b). A higher magni-
tude in the reset voltage should result in a more complete
oxidization process and hence lower conductance. On the
other hand, for the electrochemical process of SiO
involved in a set process, the lower the valence state in the
the less energy or electrical potential (applied volt-
age) is required for the reduction. Therefore, a decrease in
the magnitude of the reset voltage results in the shift of V
toward a lower magnitude as well. Note that valence states
of Si
, and Si
were indeed observed and
identified in the SiO
, and the binding energy increases as
the valence state increases [16]. The stability of the various
valence states of silicon is also consistent with the robust
nonvolatile property and stability of the multi-level states
(see inset in Fig. 2b).
4 Other implications
Despite the large device sizes studied here (for the purpose
of easy testing), the observed small size of the silicon fil-
ament (d 5nm)[11] indicates promising device scaling.
Memory switching in confined SiO
volume was demon-
strated in vertical structures [11] or in planar structures us-
ing CNTs as electrodes [10]. Switching at dimensions below
50 nm was demonstrated in SiO
nanowires using graphitic
carbon as effective electrodes, although the switching was
not well understood at the time [8].
Based on the filament formation mechanism, the switch-
ing is largely independent of the SiO
stoichiometry. In
other words, similar switching can be induced in a variety
of SiO
(e.g., 1 x 2). Generally, a surface area is more
defective and more easily induced into the switching state.
This intrinsic conduction and switching from the soft break-
down in SiO
is no exception) may even call some
molecular electronics and nanoelectronics studies into ques-
tion with respect to the actual cause for the observed electri-
cal phenomena [17], as SiO
is so frequently used in various
electronic devices as a nominal insulator. A detailed study
and discussion of this aspect may be found in [17].
5 Conclusion
In summary, we discussed the resistive switching behav-
iors intrinsic to SiO
. The SiO
-thickness independence or
filament-length independence in the switching voltage is
consistent with the mechanistic picture of point switching in
the silicon filament. The multi-state switching and V
with respect to the reset voltage is further explainable by an
electrochemical redox process of Si SiO
at the switch-
ing site.
Intrinsic resistive switching and memory effects in silicon oxide 839
Acknowledgements D.N. acknowledges the support of the David
and Lucille Packard Foundation. L.Z. acknowledges support from the
Texas Instruments Leadership University Fund and National Science
Foundation Award No. 0720825. J.M.T. acknowledges support from
the Army Research Office (W911NF-08-C-0133), the Office of Naval
Research (N00039-10-0056) and the Air Force Office of Scientific Re-
search (FA9550-10-C-0098) through SBIR and STTR programs ad-
ministrated by PrivaTran, LLC.
1. R. Waser, M. Aono, Nat. Mater. 6, 833 (2007)
2. R. Waser, R. Dittmann, G. Staikov, K. Szot, Adv. Mater. 21, 2632
3. A. Sawa, Mater. Today 11, 28 (2008)
4. M. Janousch, G.I. Meijer, U. Staub, B. Belley, S.F. Karg, B.P. An-
dreasson, Adv. Mater. 19, 2232 (2007)
5. C. Schindler, M. Weides, M.N. Kozicki, R. Waser, Appl. Phys.
Lett. 92, 122910 (2008)
6. J.G. Simmon, R.R. Verderber, Proc. R. Soc. Lond. Ser. A, Math.
Phys. Sci. 301, 77 (1967)
7. S. Furuta, T. Takahashi, Y. Naitoh, M. Horikawa, T. Shimizu, M.
Ono, Jpn. J. Appl. Phys. 47, 1806 (2008)
8. Y. Li, A. Sinitskii, J.M. Tour, Nat. Mater. 7, 966 (2008)
9. B. Standley, W. Bao, H. Zhang, J. Bruck, C. Lau, M. Bockrath,
Nano Lett. 8, 3345 (2008)
10. J. Yao, L. Zhong, Z. Zhang, T. He, Z. Jin, P.J. Wheeler, D. Natel-
son, J.M. Tour, Small 5, 2910 (2009)
11. J. Yao, Z. Sun, L. Zhong, D. Natelson, J.M. Tour, Nano Lett. 10,
4105 (2010)
12. J. Yao, L. Zhong, D. Natelson, J.M. Tour, Appl. Phys. Lett. 93,
253101 (2008)
13. X. Duan, Y. Huang, C.M. Lieber, Nano Lett. 2, 487 (2002)
14. J. Yao, Z. Jin, L. Zhong, D. Natelson, J.M. Tour, ACS Nano 3,
4122 (2009)
15. A. Beck, J.G. Bednorz, Ch. Gerber, C. Rossel, D. Widmer, Appl.
Phys. Lett. 77, 139 (2000)
16. F. Rochet, Ch. Poncey, G. Dufour, H. Roulet, C. Guillot, F. Sirotti,
J. Non-Cryst. Solids 216, 148 (1997)
17. J. Yao, L. Zhong, D. Natelson, J.M. Tour, J. Am. Chem. Soc.
(2010). doi:10.1021/ja108277r
... Most of the RS studies in Si-ncs are based on SiO x monolayers [3][4][5][6][7][8]. However, control of the size, distribution and separation of Si-ncs within SiO x monolayer structures remains a complicated issue. ...
... Nevertheless, the same trend is observed for different devices measured. The nonlinear electrical behavior and multilevel current states have been also reported in the RS behavior of SiO x monolayers [8,33]. Nevertheless, these devices exhibit unipolar RS where a lower RESET voltage produce an increased OFF conductance but the same current at the LRS [8]. ...
... The nonlinear electrical behavior and multilevel current states have been also reported in the RS behavior of SiO x monolayers [8,33]. Nevertheless, these devices exhibit unipolar RS where a lower RESET voltage produce an increased OFF conductance but the same current at the LRS [8]. Our devices show bipolar RS and a lower set voltage results in a decrease of the ON conductance and a lower RESET voltage, as observed in Fig. 6. ...
We report on the bipolar resistive switching behavior of a SiO2/Si multilayer (ML) structure measured without current compliance for non-volatile memory applications. The SiO2/Si ML structure was thermally annealed to induce the formation of Si nanocrystals, which act as conductive nodes. It was found that the SET/RESET voltages and the ON/OFF resistance ratio (107-102) depend on the current-voltage measurement cycles. Moreover, the low resistance state from the initial cycle can be recovered by increasing the stop voltage in a subsequent cycle. The space charge limited conduction (SCLC) is the mechanism responsible for the charge transport in the high resistance state (HRS) during the first measurement cycle, while the Ohmic conduction predominates in the low resistance state (LRS). This suggest the formation of a conductive path within the SiO2/Si ML, which breaks apart and becomes reconstructed upon switching between the LRS to HRS and vice versa. Performing a sequence of switching cycles, one can observe that the SCLC becomes predominant for the both LRS and HRS, suggesting the partial formation and breakdown of several different conductive paths.
In this work we use conductive atomic force microscopy (cAFM) to study the charge injection process from a nanoscale tip to a single isolated bilayer 2D MoS 2 flake. The MoS 2 is exfoliated and bonded to ultra-thin SiO 2 /Si substrate. Local current–voltage (IV) measurements conducted by cAFM provides insight in charge trapping/de-trapping mechanisms at the MoS 2 /SiO 2 interface. The MoS 2 nano-flake provides an adjustable potential barrier for embedded trap sites where the charge is injected from AFM tip is confined at the interface. A window of (ΔV <1.8 V) is obtain at a reading current of 2 nA between two consecutive IV sweeps. This is a sufficient window to differentiate between the two states indicating memory behavior. Furthermore, the physics behind the charge entrapment and its contribution to the tunneling mechanisms is discussed.
With the explosion of data in the information universe and the approaching of fundamental limits in silicon-based flash memories, the exploration of new device architectures and alternative materials is necessary for next-generation memory technology. Accordingly, emerging two-dimensional (2D) material-based memristive devices have attracted increasing attention due to their unique properties and great potential in flexible and wearable devices, and even neuromorphic computing systems. Herein, we provide an overview of the recent progress on memristive devices based on 2D materials beyond graphene. The device structures and choice of active materials and electrodes materials are summarized for various types of 2D material-based memristive devices. Following the discussion and classification on the device performances and mechanisms, the challenges and perspectives on future research based on 2D materials are also presented.
Electron emission (EE) from electron sources based on island metal films (IMFs) was mainly attributed to field emission or thermionic emission from IMFs. Here, we propose a new mechanism of EE from IMF-based sources, namely, EE from horizontal tunneling junctions formed in the substrate. The devices with and without IMFs fabricated on silicon oxide substrates are found to exhibit similar EE properties, while the island-metal-film-based devices fabricated on Si <sub xmlns:mml="" xmlns:xlink="">3</sub> N <sub xmlns:mml="" xmlns:xlink="">4</sub> /Si substrate show no EE. The comparative results indicate that EE originates from the underlying silicon oxide substrate that was ignored in previous mechanisms, but not from IMFs. EE from the devices is thought to be generated from horizontal tunneling junctions in electroformed silicon oxide substrate due to the rupture of conducting filaments. Even though metal island films are not the origin of EE, they can greatly decrease the forming voltage of the devices. The insights into the emission mechanism are helpful for optimizing the electron source performances.
On‐chip electron sources with the advantages of high emission current and density, high emission efficiency, low working voltage, and easy fabrication are highly desired for scaling down free electron‐based devices and systems, especially for realizing those on a chip, but remain challenging. Here, such an on‐chip electron source is reported simply based on electroformed silicon oxide between concentric graphene films on silicon oxide substrate. It is demonstrated that electron emission from an electron emitter can be driven by a low voltage about 11 V, and a high emission efficiency of 33.6%. An on‐chip electron source with 36 × 36 emitter array in an area of 594 × 594 µm² exhibits an emission current up to 1 mA at 38 V working voltage, corresponding to a high emission density of 283 mA cm⁻². Electron emission from the sources is thought to be generated from horizontal tunneling diodes formed in electroformed silicon oxide. Combined advantages of high emission current and density, high emission efficiency, low working voltage, and easy fabrication make this on‐chip electron sources promising in realizing miniature and on‐chip free electron‐based devices and systems.
Full-text available
The pervasiveness of information technologies is generating an impressive amount of data, which need to be accessed very quickly. Nonvolatile memories (NVMs) are making inroads into high‐capacity storage to replace hard disk drives, fuelling the expansion of the global storage memory market. As silicon‐based flash memories are approaching their fundamental limit, vertical stacking of multiple memory cell layers, innovative device concepts, and novel materials are being investigated. In this context, emerging 2D materials, such as graphene, transition metal dichalcogenides, and black phosphorous, offer a host of physical and chemical properties, which could both improve existing memory technologies and enable the next generation of low‐cost, flexible, and wearable storage devices. Herein, an overview of graphene and related 2D materials (GRMs) in different types of NVM cells is provided, including resistive random‐access, flash, magnetic and phase‐change memories. The physical and chemical mechanisms underlying the switching of GRM‐based memory devices studied in the last decade are discussed. Although at this stage most of the proof‐of‐concept devices investigated do not compete with state‐of‐the‐art devices, a number of promising technological advancements have emerged. Here, the most relevant material properties and device structures are analyzed, emphasizing opportunities and challenges toward the realization of practical NVM devices.
Magnetic tunnel junction (MTJ), the core of the magnetic random access memory (MRAM), have attracted intensive interest for low-power storage and information processing technologies. This thesis focuses on the discussion of the influence of edge effect for the magnetic proprieties of nanodevices. Two typical magnetic nanostructures are involved: MTJ nanopillar under a special encapsulation process, and magnetic nanodots with the same structure as the free layer of MTJ stack.First, we develop the full fabrication process for both MTJ nanopillar and magnetic nanodots, which is compatible with standard CMOS technology.Then, a novel heterogeneous memristive device composed of an MTJ nanopillars surrounded by resistive silicon switches is investigated by transport measurement and structural characterization. The potential application as a logic-in-memory device with memory encryption function is discussed.Finally, the magnetic nanodots with perpendicular magnetic anisotropy (PMA) is investigated using Kerr microscope. A physical model of Laplace pressure applied on a DW motion at the edges of nanodots is responsible for explaining the unexpected shifting of switching field distribution (SFD) for nanodots with varied sizes.
Full-text available
The SiOx-based resistive random access memory (RRAM) is reported as the gas sensor with the spontaneous response i.e. memory window closure as air-vacuum switches, meanwhile enlarged the switching voltages in the vacuum. The HfOx/SiOx stacked structure is proposed with a superior resilience with response to ambient changes through inserting an interface structure, which attributed to the gain of the source of high defect density and Hf-encapsulation of the filamentary region as compared to single layer SiOx structures. Besides, the intrinsic nonlinear behavior is demonstrated by integrating a low dielectric layer, which suppresses the sneak path issues for large-scale SiOx-based RRAM crossbar array implementation in near future. © The Author(s) 2018. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY,, which permits unrestricted reuse of the work in any medium, provided the original work is properly cited.
Electrically driven on‐chip electron sources that do not need to be heated are long pursued, but their realization remains challenging. Here, it is shown that a nanogap formed by two electrodes on a silicon oxide substrate functions as an electron‐emitting nanodiode after the silicon oxide in the nanogap is electrically switched to a high‐resistance conducting state. A nanodiode based on graphene electrodes can be turned on by a voltage of ≈7 V in ≈100 ns and show an emission current of up to several microamperes, corresponding to an emission density of ≈10⁶ A cm⁻² and emission efficiency as high as 16.6%. We attribute the electron emission to be generated from a metal–insulator–metal tunneling diode on the substrate surface formed by the rupture of conducting filaments in silicon oxide. An array of 100 nanodiodes exhibits a global emission density of 5 A cm⁻² and stable emission with negligible current degradation over tens of hours under modest vacuum. The combined advantages of a low operating voltage, fast temporal response, high emission density and efficiency, convenient fabrication and integration, and stable emission in modest vacuum make silicon oxide electron‐emitting nanodiodes a promising on‐chip electron sources.
Full-text available
Oblique deposition was used to fabricate two metal electrodes separated by a gap of less than 10 nm on a SiO2 substrate. By sweeping voltage between these electrodes, a negative resistance change of several digits was observed in vacuum. In this work, electrodes made of Au, Pd, Pt, and Ta were fabricated, and their electric properties were measured in vacuum. The negative resistance was observed for all of the four metals. The result of the measurements clearly shows the correlation between the voltage at the minimum resistance and the melting point. Also, the calculated temperature rise shows a correlation with the melting point. These facts support the effect that the thermal change of the electrode metal has a considerable effect on the electric properties of the nanogap switch (NGS).
Full-text available
Vertical structures of SiO2 sandwiched between a top tungsten electrode and conducting nonmetallic substrate were fabricated by dry and wet etching methods. Both structures exhibit similar voltage-controlled memory behaviors, in which short voltage pulses (1 mu s) can switch the devices between high- and low-impedance states. Through the comparison of current-voltage characteristics in structures made by different methods, filamentary conduction at the etched oxide edges is most consistent with the results, providing insights into similar behaviors in metal/SiO/metal systems. High ON/OFF ratios of over 10(4) were demonstrated.
Full-text available
Rapid advances in information technology rely on high-speed and large-capacity nonvolatile memories. A number of alternatives to contemporary Flash memory have been extensively studied to obtain a more powerful and functional nonvolatile memory. We review the current status of one of the alternatives, resistance random access memory (ReRAM), which uses a resistive switching phenomenon found in transition metal oxides. A ReRAM memory cell is a capacitor-like structure composed of insulating or semiconducting transition metal oxides that exhibits reversible resistive switching on applying voltage pulses. Recent advances in the understanding of the driving mechanism are described in light of experimental results involving memory cells composed of perovskite manganites and titanates.
It has been observed that when thin (200 to 3000 {circ} { A}) film insulators have been formed by the electrolytic introduction of gold ions from one of the electrodes they can draw appreciable currents. They then show temperature-independent conductivity, voltage-controlled negative resistance and reversible voltage and thermal-voltage memory effects. It is postulated that the injected ions introduce a broad band of localized impurity levels within the normally forbidden band of the insulator. The electrons are assumed to move through the insulator by tunnelling between adjacent sites within the impurity band; it is also assumed that the electrons can, under certain conditions, be trapped within the impurity band. A model based on these ideas accounts in a self-consistent manner for all the experimental observations, and calculations of current-voltage characteristics based on the model are in fact in agreement with them.
Thin oxide films with perovskite or related structures and with transition metal doping show a reproducible switching in the leakage current with a memory effect. Positive or negative voltage pulses can switch the resistance of the oxide films between a low- and a high-impedance state in times shorter than 100 ns. The ratio between these two states is typically about 20 but can exceed six orders of magnitude. Once a low-impedance state has been achieved it persists without a power connection for months, demonstrating the feasibility of nonvolatile memory elements. Even multiple levels can be addressed to store two bits in such a simple capacitor-like structure. © 2000 American Institute of Physics.
Bistable nanoscale switches have been assembled using nanowires and redox active molecules as building blocks. The nanodevices consist of nanowire field-effect transistors (NW-FET) functionalized with redox active molecules, where the redox species can store charges and thereby maintain the NW-FETs in either a logic on or off state with high or low channel conductance, respectively. Single NW and crossed NW-FET devices have been assembled and shown to switch reversibly as two- and three-terminal devices. Individual NW devices with on/off ratios exceeding 104 and retention times in excess of 20 min, and integrated arrays in which devices are independently addressable, have been assembled. The characteristics of the molecule-gated NW-FET devices suggest that they could serve as key elements in a range of nanoelectronics applications, including nonvolatile memory and programmable logic array.
Resistive switching in Ir / Si O <sub>2</sub>/ Cu memory cells was investigated. The proposed switching mechanism is the formation and dissolution of a Cu filament. Under positive bias, Cu cations migrate through Si O <sub>2</sub> and are reduced at the counterelectrode forming a filament. The filament is dissolved under reverse bias. The write current can be reduced down to 10 pA which is four orders of magnitude below published values and shows the potential of extremely low power-consuming memory cells. Furthermore, a comparison of the charge flow in the high resistance state and the energy for writing is given for write currents between 25 pA and 10 nA .
Synchrotron radiation X-ray photoemission spectroscopy (SRXPS), used with an optimized surface and interface sensitivity, is a unique tool to determine the chemical composition and spatial extension of the suboxide layer present at the Si/SiO2 interface. The bonding at ‘thermal interfaces’ appear to be essentially dependent on the Si crystal orientation. For Si(001)/SiO2 a detailed photon energy-dependent and angle-dependent. SRXPS study (in the 120 ÷ 175 eV photon energy range) has been performed. Previous results on higher-oxidation states cross-section resonances around hv = 130 eV are re-examined. The various oxide states do not present photoelectron diffraction peaks, when polar scans are performed in the (110) azimuthal plane. On the other hand, there is evidence that substrate silicon atoms close to the interface retain orientational order. Limits in the vertical distribution of the suboxides are given. The compatibility of our experimental findings with models of the current literature — in particular the so-called ‘dimerized interface’ model — is examined.
Various classes of resistive random access memory (RRAM) classes involving chemical effects related to redox processes in the MIM cell are presented. The electrochemical metallization (ECM) cells systems involve Ag and Cu as electrochemically active metals and phase separated amorphous selenides and sulfides and various oxides, acting as solid electrolytes. The polarity of the bipolar switching cycle is determined by many factors such as the work function and the oxygen affinity of the electrode metals. The SET and RESET switching of an individual filament at the surface of a SrTiO3 single crystal shows that the resistance of the filament can be switched between an ONand OFFstate by a voltage applied to the tip of the LC-AFM. The dissolution of cylindric metal filaments studied by electrothermal simulation is described by a thermally activated process of the filament surface.
Silicon oxide (SiO(x)) has been widely used in many electronic systems as a supportive and insulating medium. Here, we demonstrate various electrical phenomena such as resistive switching and related nonlinear conduction, current hysteresis, and negative differential resistance intrinsic to a thin layer of SiO(x). These behaviors can largely mimic numerous electrical phenomena observed in molecules and other nanomaterials, suggesting that substantial caution should be paid when studying conduction in electronic systems with SiO(x) as a component. The actual electrical phenomena can be the result of conduction from SiO(x) at a post soft-breakdown state and not the presumed molecular or nanomaterial component. These electrical properties and the underlying mechanisms are discussed in detail.