FEASIBILITY AND APPLICATION OF A CuxO-BASED MEMRISTOR FOR
SENSING OXYGEN AND OTHER GASES
By
Chinwe Pamela Nyenke

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Electrical Engineering - Doctor of Philosophy
2017

ABSTRACT
FEASIBILITY AND APPLICATION OF A CuxO-BASED MEMRISTOR FOR
SENSING OXYGEN AND OTHER GASES
By
Chinwe Pamela Nyenke
This dissertation introduces the design, fabrication, and application of a
copper-oxide-based memristor for the passive sensing of oxygen and other gases.
The device design was as follows: Deposition of copper (Cu) bottom electrodes,
(oxygen) vacancy-rich copper oxide (CuxO) switching layers, and tungsten (W) top
electrodes in a crossbar array structure. The CuxO layer was deposited via reactive
sputtering of a Cu target with an argon-oxygen (Ar/O2) mixture. A portion of this
layer was extended from each array cell to be exposed for sensing. Memristive
devices of different switching layer thicknesses were initially explored for
irreversible sensing of oxygen in ambient air. Results of this first experiment
demonstrated an increase in resistance states upon prolonged exposure to
ambient air. For the second experiment, memristive devices were fabricated with
sub-micron holes that were etched into the W top electrode to better reveal the
switching layer surface. The devices were also subjected to ambient oxygen at 180
deg C to induce passive sensing in minutes. Resistance results were consistent
with the first experiment but also revealed a dependence on the surface area of
the exposed oxide. Finally, memristive devices were investigated in a third
experiment for reversible sensing of an oxidizing gas and reducing gas at room

temperature. This time, changes were not only observed in resistance but also in
hysteresis (current versus voltage) depending on the type of gas introduced.
Overall, this work demonstrates a step towards the use of the memristor as a gas
sensor, which we have named “memsensors”, by taking advantage of the device’s
ability to memorize (or record) historical information.

Dedicated to my parents, sisters, husband, and son.

iv

ACKNOWLEDGMENTS

Firstly, I would like to thank my advisor – Dr. Lixin Dong – for his guidance
and support throughout these years. Without his intellectual probing and ideas, this
work would not be what it is. He challenged me to go down the path of considering
a new application for memristors that had never before been explored.
Additionally, I would like to thank Dr. Percy Pierre for his longtime
mentorship, encouragement, and motivation. His belief in me helped to instill the
perseverance I needed to complete this research and dissertation. I would also
like to acknowledge his support of my work, including through the Alfred P. Sloan
Foundation M.PhD Program and GAANN Fellowship Program.
I would also like to thank my dissertation committee – Dr. Ning Xi, Dr. Tim
Hogan, Dr. Wen Li, and Dr. Junghoon Yeom – for their assistance and motivation
in completing this project. They helped me to ponder and explore possibilities I had
not initially realized.
I would like to acknowledge Dr. Baokang Bi and Dr. Per Askeland for
information and equipment training related to fabrication of the devices for this
project. Additionally, I would like to acknowledge Dr. Zheng Fan and Amberley
Huang for critical assistance in data collection.

v

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................. viii
LIST OF FIGURES ............................................................................................... ix
CHAPTER 1 INTRODUCTION.............................................................................. 1
1.1 MEMRISTOR – WHAT IS IT? ...................................................................... 1
1.2 BACKGROUND AND MOTIVATION ........................................................... 3
1.3 STATEMENT OF THE PROBLEM .............................................................. 4
1.4 PURPOSE OF THE STUDY ........................................................................ 5
1.5 RESEARCH QUESTIONS ........................................................................... 5
1.6 DISSERTATION OUTLINE .......................................................................... 6
REFERENCES ...................................................................................................... 8
CHAPTER 2 LITERATURE REVIEW ................................................................. 12
2.1 METAL-OXIDE MEMRISTOR: SWITCHING LAYER DESIGN ................. 12
2.2 MEMRISTOR RESEARCH IN ELECTRONIC APPLICATIONS ................ 16
2.3 PASSIVE GAS SENSING WITH METAL OXIDES .................................... 18
2.4 MEMRISTORS ADVANCES IN SENSING ................................................ 21
2.5 SUMMARY................................................................................................. 24
REFERENCES .................................................................................................... 26
CHAPTER 3 MEMRISTOR DESIGN AND FABRICATION ................................ 34
3.1 INTRODUCTION ....................................................................................... 34
3.2 PROTOTYPE MEMRISTOR DESIGN ....................................................... 34
3.2.1 Design Overview ................................................................................. 34
3.2.2 Fabrication Process ............................................................................ 36
3.2.2 Maskplate Design ................................................................................ 39
3.3 IMPROVED DEVICE - MEMRISTOR WITH SUB-MICRON HOLES......... 41
3.3.1 Design Overview ................................................................................. 41
3.3.2 Fabrication Process ............................................................................ 42
3.3.3 Maskplate Design ................................................................................ 43
3.3.4 Improving Sensitivity with Holes .......................................................... 48
3.3.5 Preventing Contact Oxidation ............................................................. 57
3.4 SUMMARY................................................................................................. 58
APPENDICES ..................................................................................................... 59
APPENDIX A: SPUTTERING PARAMETERS ................................................ 60
APPENDIX B: RECIPES ................................................................................. 61
REFERENCES .................................................................................................... 62

vi

CHAPTER 4 DETECTION OF AMBIENT OXYGEN USING W/CuxO/Cu
MEMRISTORS OF DIFFERENT THICKNESSES .............................................. 64
4.1 INTRODUCTION ....................................................................................... 64
4.2 EXPERIMENTAL SETUP .......................................................................... 64
4.3 DISCUSSION OF RESISTIVE SWITCHING BEHAVIOR.......................... 66
4.3.1 Formation of Conductive Filaments .................................................... 66
4.3.2 Characterizing Memristive Behavior ................................................... 67
4.3.3 Identifying the Conduction Mechanism ............................................... 69
4.4 DISCUSSION OF PASSIVE SENSING RESULTS ................................... 71
4.4.1 Increased Oxygen Composition with Time .......................................... 71
4.4.2 Progression of Resistance with Time under Exposure ....................... 73
4.5 SUMMARY................................................................................................. 76
REFERENCES .................................................................................................... 77
CHAPTER 5 INVESTIGATION OF W/CuxO/Cu MEMRISTORS WITH SUBMICRON HOLES FOR PASSIVE SENSING OF OXYGEN ................................ 80
5.1 INTRODUCTION ....................................................................................... 80
5.2 EXPERIMENTAL SETUP .......................................................................... 80
5.3 DISCUSSION OF RESULTS ..................................................................... 82
5.3.1 Characterizing Memristive Behavior ................................................... 82
5.3.2 Sensing of Ambient Oxygen under Heat ............................................. 86
5.3.3 Sensing Mechanism – Adsorption and Diffusion ................................. 91
5.6 SUMMARY................................................................................................. 95
REFERENCES .................................................................................................... 96
CHAPTER 6 EFFECT OF NITROGEN DIOXIDE AND AMMONIA ON
RESISTIVE SWITCHING .................................................................................... 99
6.1 INTRODUCTION ....................................................................................... 99
6.2 EXPERIMENTAL SETUP .......................................................................... 99
6.3 DISCUSSION OF RESULTS ................................................................... 100
6.3.1 Evaluation of NO2 Detection ............................................................. 100
6.3.2 Evaluation of NH3 Detection ............................................................. 103
6.4.3 Mechanism of Gas Detection ............................................................ 106
6.6 SUMMARY............................................................................................... 110
REFERENCES .................................................................................................. 111
CHAPTER 7 CONCLUSIONS AND FUTURE WORK ...................................... 114
7.1 SUMMARY OF CONCLUSIONS ............................................................. 114
7.2 FUTURE RESEARCH ............................................................................. 115
7.2.1 Further Understanding of Reversible Sensing Mechanism ............... 115
7.2.2 Different Switching Layer Composition and Design .......................... 116
7.2.3 Influence of Memristor Area on Sensing ........................................... 116
REFERENCES .................................................................................................. 117

vii

LIST OF TABLES

Table 1. EDS Quantitative Results for CuxO (35 nm). ......................................... 72
Table 2. EDS Quantitative Results after 22 Days. .............................................. 72
Table 3. EDS Quantitative Results for CuxO (11 nm). ......................................... 72
Table 4. EDS Quantitative Results after 22 Days. .............................................. 72
Table 5. Progression of low resistance states with time ...................................... 75
Table 6. Percentage change in low resistance states with time .......................... 75
Table 7. EDS Results - Oxygen Composition for Non-Mesh ............................... 87
Table 8. EDS Results - Oxygen Composition for Nano-Mesh ............................. 89
Table 9. EDS Results - Oxygen Composition for Micro-Mesh ............................ 91
Table 10. Nitrogen dioxide detection of device #1 (negative bias) .................... 103
Table 11. Nitrogen dioxide detection of device #1 (positive bias) ..................... 103
Table 12. Ammonia detection of device #2 (negative bias) ............................... 104
Table 13. Ammonia detection of device #2 (positive bias) ................................ 104

viii

LIST OF FIGURES

Figure 1. Semiconductor film with a high dopant region (ROFF) and low dopant
region (RON). ................................................................................................... 1
Figure 2. Formation of conductive filaments comprising of a) titanium ions only
and b) oxygen vacancies and ions with titanium ions. Reprinted from H. Z.
Zhang, D. S. Ang, K. S. Yew, and X. P. Wang, “Enhanced stability of
complementary resistance switching in the TiN/HfOx/TiN resistive randomaccess memory device via interface engineering,” Applied Physics Letters,
vol. 108, art. no. 083505, Feb 2016, with the permission of AIP Publishing.13
Figure 3. [LEFT] When a sufficient negative bias is applied, the ZnWOx layer
acts as an oxygen ion reservoir to form the conductive filament. [RIGHT]
When a sufficient positive is bias, the layer becomes an oxygen supplier
leading to the rupture of the conductive filament. Reprinted with permission
from S.-M. Lin, J.-S. Huang, W.-C. Chang, T.-C. Hou, H.-W. Huang, C.-H.
Huang, S.- J. Lin and Y.-L. Chueh, “Single-step formation of ZnO/ZnWOx
bilayer structure via interfacial engineering for high performance and low
energy consumption resistive memory with controllable high resistance
states,” ACS Applied Materials & Interfaces, vol. 5, no. 16, pp. 7831-7837,
Aug. 2013. Copyright 2013 American Chemical Society.............................. 14
Figure 4. Reprinted with permission from S. Jo, T. Chang, I. Ebong, B. Bhadviya,
P. Mazumder, and W. Lu, "Nanoscale Memristor Device as Synapse in
Neuromorphic Systems," Nano Letters, vol. 10, pp. 1297-1301, Apr 2010.
Copyright 2010 American Chemical Society. ............................................... 17
Figure 5. Reproduced from L. G. Wang, W. Zhang, Y. Chen, Y. Q. Cao, A. D. Li,
and D. Wu, “Synaptic plasticity and learning behaviors mimicked in single
inorganic synapses of Pt/HfOx/ZnOx/TiN memristive system,” Nanoscale
Research Letters, vol 12, art. no. 65, Jan. 2017. Publisher: Springer. ......... 17
Figure 6. Reprinted from Materials Chemistry and Physics, vol. 141, M. Ahmad,
A. Sadek, J. Ou, M.H. Yaacob, et al., “Facile synthesis of nanostructured
WO3 thin films and their characterization for ethanol sensing,” pp. 912-919,
Copyright 2013, with permission from Elsevier. ........................................... 19
Figure 7. Reprinted from Sensors and Actuators B-Chemical, vol. 202, V. Hien, J.
You, K. Jo, S. Kim, et al., “H2S-sensing properties of Cu2O submicron-sized

ix

rods and trees synthesized by radio-frequency magnetron sputtering,” pp.
330-338, Copyright 2014, with permission from Elsevier. ............................ 20
Figure 8. Reprinted from Sensors and Actuators B-Chemical, vol. 171, S.
Carrara, D. Sacchetto, M. Doucey, C. Baj-Rossi, G. De Micheli, and Y.
Leblebici, "Memristive-biosensors: A new detection method by using
nanofabricated memristors,” pp. 449-457, Copyright 2012, with permission
from Elsevier. ............................................................................................... 21
Figure 9. X. Wang, Y. Chen, Y. Gu, and H. Li, "Spintronic Memristor Temperature
Sensor," IEEE Electron Device Letters, vol. 31, pp. 20-22, JAN 2010. Š 2010
IEEE. ............................................................................................................ 24
Figure 10. Initial memristor design. During testing of the memristor, the bottom
electrode is grounded. .................................................................................. 35
Figure 11. SEM of prototype – crossbar array memristor There are a total of five
bottom electrodes and one top electrode. The copper oxide switching layer
(which is not easily seen here) sits between the copper bottom electrode and
top tungsten electrode with regions exposed for sensing. This design results
in five 10-micron-x-15-micron memristors. ................................................... 36
Figure 12. Fabrication process. (a) The fabrication begins on a clean silicon die.
(b) PECVD of silicon dioxide is performed to provide an insulating layer
between the memristor and silicon surface. (c) Photoresist is spun and
baked onto the die followed by photolithography. (d) Cu and CuxO are
deposited and lift-off reveals a bottom electrode and switching layer,
respectively. (e-f) The procedures are repeated for the W top electrode. .... 38
Figure 13. Maskplate design for layering of copper electrodes and contacts. .... 39
Figure 14. Maskplate design for layering of tungsten electrode and contact. ..... 40
Figure 15. SEM of memristor array. .................................................................... 41
Figure 16. Holes in tungsten layer. ...................................................................... 42
Figure 17. Maskplate design for photolithography. This is the complete design. 44
Figure 18. Maskplate design for photolithography. This is the first layer consisting
of the bottom electrodes, bottom contacts, and markers. ............................ 45
Figure 19. Closeup of the first layer. ................................................................... 46

x

Figure 20. Maskplate design for photolithography. This is the first layer consisting
of the top electrodes, top contacts, and markers. ........................................ 47
Figure 21. Closeup of the second layer. .............................................................. 48
Figure 22. Initial NPGS design with two sets of 4x4 500-nm holes (in two
quadrants) and one set of larger holes......................................................... 50
Figure 23. Failed etching attempt. Note that much of the tungsten layer has been
etched through for numerous memristors. ................................................... 51
Figure 24. Closeup of failed etching attempt. ...................................................... 52
Figure 25. Improved attempt. .............................................................................. 53
Figure 26. Closeup SEM of 700nm-diameter meshes ("micro meshes"). ........... 54
Figure 27. One of the micro meshes. .................................................................. 54
Figure 28. Closeup of 300-nm holes ("nano meshes"). ....................................... 55
Figure 29. One of the nano meshes. ................................................................... 56
Figure 30. Comparison of (a) usuable nano mesh and over-etched nano mesh. 56
Figure 31. Closeup of final quadrant. The holes on the right were not used for
testing but for etching guidance. The remaining non-etched devices were
used as a control group (“non meshes”). ..................................................... 57
Figure 32. Schematic of memristor with etched tungsten layer. .......................... 58
Figure 33. Measurement setup. .......................................................................... 65
Figure 34. Current-voltage characteristics of the W/CuxO(11nm)/Cu memristor. 67
Figure 35. Current-voltage characteristics of the W/CuxO(35nm)/Cu memristor. 68
Figure 36. Close-up of the W/CuxO(35nm)/Cu memristor current-voltage
characteristics between -1 and 1 V. ............................................................. 69
Figure 37. Log (I) vs. log (V) plot of memristor in the positive bias region. ......... 70

xi

Figure 38. Best fit lines on log-log plots of memristor in the positive bias region.
..................................................................................................................... 71
Figure 39. Plot of resistance versus applied voltage before and after exposure to
ambient oxygen. ........................................................................................... 73
Figure 40. Closeup of low resistance state (LRS). .............................................. 74
Figure 41. Comparison of LRS states of W/CuxO(35nm)/Cu device. ................. 74
Figure 42. Measurement setup. .......................................................................... 81
Figure 43. Current-voltage characteristics of nano-mesh before testing. ............ 82
Figure 44. Current-voltage characteristics of nano-mesh before testing. ............ 83
Figure 45. Current-voltage characteristics of nano-mesh before testing. ............ 83
Figure 46. Current-voltage characteristics of micro-mesh before testing. ........... 84
Figure 47. Current-voltage characteristics of micro-mesh before testing. ........... 84
Figure 48. Current-voltage characteristics of non-mesh before testing. .............. 85
Figure 49. Current-voltage characteristics of micro-mesh before testing. ........... 85
Figure 50. Current-voltage characteristics of non-mesh before testing. .............. 86
Figure 51. Current-voltage plots for two non mesh devices in (a) and (b) and for
two nano mesh devices in (c) and (d). The black data points are after
substrate has been subjected to ambient air at 180 deg C for 25 minutes
while the red is after an additional five minutes............................................ 88
Figure 52. Current-voltage plots for a micro mesh. The black data points are
before testing while the blue is after substrate has been subjected to
ambient air at 180 deg C for a total of 30 minutes. The plot in (a) is an
overlay of both sets of data while (b) is a closeup of the latter set. .............. 89
Figure 53. Current-voltage plots for a second micro mesh. The black data points
are after substrate has been subjected to ambient air at 180 deg C for 25
minutes while the blue is after an additional five minutes. The plot in (a) is an
overlay of both sets of data while (b) is a closeup of the latter set. .............. 90

xii

Figure 54. Schematic of adsorption and vacancy filling. (a) Oxygen molecules
from the ambient air result in adsorbed oxygen ions and holes near the
CuxO surface. (b) Over time, the adsorbed ions fill oxygen vacancies within
the layer. (c) The memristor now has higher resistance states due to fewer
vacancies. .................................................................................................... 93
Figure 55. Schematic of (a) nano mesh device and (b) micro mesh device. The
micro mesh design experiences more adsorbed oxygen ions. .................... 94
Figure 56. IV characteristics before NO2 exposure. .......................................... 101
Figure 57. Current-voltage plots for device #1 after 20 ppm NO2 exposure. Plots
(a-f) demonstrate gradual recovery of the device at five-minute intervals. The
final plot in (f) is 30 minutes after exposure................................................ 102
Figure 58. Current-voltage plots for device #1. The three plots are an overlay of
two voltages sweeps (a) prior to exposure to 100ppm NH3, (b) after 50 and
55 minutes of exposure, and (c) after 30 and 35 minutes of recovery. ...... 105
Figure 59. Depiction of gas detection mechanism. ........................................... 109

xiii

CHAPTER 1 INTRODUCTION

1.1 MEMRISTOR – WHAT IS IT?
More than forty-five years ago, there were only three fundamental twoterminal circuit elements known to relate charge or current (dq = idt) to flux or
voltage (dj = vdt). These elements were the resistor (dv = Rdi), capacitor
(dq=Cdv), and inductor (dj=Ldi). In 1971, professor Leon Chua mathematically
recognized a fourth fundamental circuit element to relate charge q and flux j via
the equation dj=Mdq [1]. He named this element the memory resistor, or
memristor.
It was not until 2008 that a feasible physical model of the memristor was
presented by Strukov and others [2] in Nature. The researchers described a
semiconductor device divided into two regions – one with a high concentration of

w

RON

ROFF

D
Figure 1. Semiconductor film with a high dopant region (ROFF) and low dopant
region (RON).

1

dopants and the other with a low concentration of dopants. By consequence, the
region with high dopant concentration has high electrical conductivity translating
to low resistance RON. Alternatively, the region with low dopant concentration has
low electrical conductivity and thus, large resistance ROFF. An illustration of this
device is provided in Figure 1.
When an external bias is applied across the device, the charged dopants
will then drift, thus moving the boundary between the two regions. Assuming
Ohmic conduction and linear ionic drift in a uniform field [2], we then obtain the
following:
𝑣 𝑡 = 𝑅%&

𝑤 𝑡
𝑤 𝑡
+ 𝑅%** 1 −
𝐷
𝐷

𝑖 𝑡 																																																									(1)

𝑑𝑤(𝑡)
𝑅%&
= 𝜇3
𝑖 𝑡 																																																																																																							(2)
𝑑𝑡
𝐷
where Âľv is the electrical mobility, w(t) is the width of the high dopant region over
time t, D is the thickness of the semiconductor film, and i(t) is the current. If we
integrate equation (2) to solve for w(t), we get:

𝑤 𝑡 = 	 𝜇3

𝑅%&
𝑞 𝑡 																																																																																																									(3)
𝐷

Substituting equation (3) into equation (1) then gives us the voltage in terms of
charge q(t) and i(t):

2

𝑣 𝑡 = 	 𝜇3

𝑅%& 7
𝜇3 𝑅%&
𝑞
𝑡
+
𝑅
1
−
𝑞 𝑡
%**
𝐷7
𝐷7

𝑖 𝑡 																																												(4)

Thus, the term on the right side of equation (4) in front of the term i(t) is just the
memristance of the device, which is resistance that is dependent on the drift of
charged dopants. For the case where RON << ROFF, the memristance simplifies to
the following:

𝑀 𝑞 = 	 𝑅%** 1 −

𝜇3 𝑅%&
𝑞(𝑡) 																																																																																					(5)
𝐷7

1.2 BACKGROUND AND MOTIVATION
In metal-oxide based memristors, it is believed that resistive switching
depends on the migration of oxygen vacancies [3-6] or oxygen ions [7,8] between
electrodes. Based on the external bias applied, these devices can switch between
low and high resistance states that are hundreds to thousands of ohms and tens
to thousands of kilo-ohms, respectively. The understanding of this phenomenon
combined

with

the

increasing

limits

of

complementary

metal–oxide

semiconductors [3,5,9] have led to a surge in memristor research for memory
applications, particularly for non-volatile memory (NVM) devices, or resistive
random-access memories (RRAMs) [9,10]. Considerable attention has also been
given to the memristor for logic applications, including as synapses in
neuromorphic systems [10–12]. Over the past several years, improvements in

3

memristive switching behavior have also been reported, including increased
speeds [5,13] and endurances [10] for particular metal oxide combinations and
stoichiometric ratios [5,14].
Excluding memristor research, metal oxides have been heavily studied for
gas and chemical sensing [15–19] because of inherent defective sites, which allow
for the adsorption of oxygen and other chemical species. For example, in 2014, V.
Hien and others developed Cu2O nanorods and nano-trees to irreversibly sense
hydrogen sulfide (H2S) [18]. The sensing mechanism of their samples involved a
phase transformation of Cu2O to copper sulfide (CuS). In 2013, M. Ahmad and
other researchers fabricated tungsten oxide (WO3) thin films for ethanol (C2H6O)
sensing. They concluded that the reaction between the sensor's WO3 surface and
the ethanol molecules was attributed to the oxygen ion species adsorbed by the
metal oxide while in air [19].
This ability of metal oxides to adsorb oxygen ions begs the question of how
can we exploit metal-oxide based memristance for gas or chemical sensing.
(This section is reproduced from publication [20].)

1.3 STATEMENT OF THE PROBLEM
In this study, metal-oxide based memristors are fabricated and investigated
for feasibility and application in passive sensing of oxygen and other gases. The

4

traditional memristor is influenced by voltage only and typically researched for
electronic applications in memory and logic. This study, however, examines the
influence of gas and voltage on the memristor, thus applying the device with the
added function of gas sensing.
A nano-sized metal-oxide memristor is fabricated with a portion of the
switching layer exposed. Sensing is determined as it relates to adsorption of
oxygen and other gases to the exposed metal oxide.

1.4 PURPOSE OF THE STUDY
The purpose of this project is to expand memristor research, especially that
of metal oxides, into the area of gas sensing. Another purpose of this project is to
increase the understanding of environmental effects, such as tuning, on
memristors. Finally, there is the potential application of such passive sensors as
sensing alternatives for devices which typically have no batteries integrated due to
their small sizes including micro/nanorobots, such as catalytically-driven microjets
[21] or artificial bacteria flagella [22] designed for medical applications.

1.5 RESEARCH QUESTIONS
In particular, the following questions will be addressed:

5

•

Can the memristance of a metal-oxide based memristor be exploited for
sensing oxygen and other gases?

•

If so, what is the nature of sensing in this memristor?
o What is happening at the physical level?
o How do the surface layer physics (and bulk layer physics) come into
play?

•

How will heat affect sensing?

•

How will design of the device affect sensing?

•

Finally, how do the memristive effects from exposure to oxygen compare to
those from exposure to other gases.

1.6 DISSERTATION OUTLINE
The rest of this dissertation is outlined as follows:
Chapter 2 will give a review of relevant literature on memristor research,
including the design of metal-oxide based devices, applications of the memristor
in electronics, and advances in sensing. Chapter 3 will discuss the fabrication of
the memristor as well as choices in approach and design. Chapter 4 will cover the
first study – passive sensing of ambient oxygen at room temperature with a
prototype device. Chapter 5 will cover an improved “memsensor” design and

6

passive sensing of oxygen with heat. Chapter 6 will discuss the reversible
detection of an oxidizing gas and a reducing gas with the improved “memsensor”.
Chapter 7 will sumamrize the dissertation with conclusions and future work
concerning each study.

7

REFERENCES

8

REFERENCES

1. L. O. Chua, “Memristor - the missing circuit element,” IEEE Transactions
Circuit Theory, vol. 18, pp. 507–519, 1971.
2. D. Strukov, G. Snider, D. Stewart, and R. Williams, "The missing memristor
found," Nature, vol. 453, pp. 80-83, May 2008.
3. R. Waser and M. Aono, “Nanoionics-based resistive switching memories,”
Nature Materials, vol. 6, pp. 833−840, Nov. 2007.
4. Y. Yang, S. Choi, and W. Lu, “Oxide heterostructure resistive memory,” Nano
Letters, vol. 13, pp. 2908-2915, May 2013.
5. J. J. Yang, M. D. Pickett, X. Li, D. A. A. Ohlberg, D. R. Stewart, and R. S.
Williams. “Memristive switching mechanism for metal/oxide/metal nanodevices,”
Nature Nanotechnology, vol. 3, pp. 429-433, Jun. 2008.
6. J.P. Strachan, et al. “Direct identification of the conducting channels in a
functioning memristive device,” Advanced Materials, vol. 22, pp. 3573–3577,
2010.
7. T.-M. Tsai, K.-C. Chang, T.-C. Chang, R. Zhang, T. Wang, C.-H. Pan, and K.H. Chen, “Resistive switching mechanism of oxygen-rich indium tin oxide
resistance random access memory,” IEEE Electron Device Letters, vol. 37, no. 4,
pp. 408-411, Apr. 2016.
8. Y.-J. Huang, S.-C. Chao, D.-H. Lien, C.-Y. Wen, J.-H. He, and S.-C. Lee,
“Dual-functional memory and threshold resistive switching based on the push-pull
mechanism of oxygen ions,” Scientific Reports |, vol. 6, art. no. 23945, Apr. 2016.
9. D. Sacchetto, G. De Micheli, and Y. Leblebici, "Multiterminal memristive
nanowire devices for logic and memory applications: a review," Proceedings of
the IEEE, vol. 100, pp. 2008-2020, Jun. 2012.
10. D. Jeong, R. Thomas, R. Katiyar, J. Scott, H. Kohlstedt, A. Petraru, and C.
Hwang, “Emerging memories: Resistive switching mechanisms and current
status,” Reports on Progress in Physics, vol. 75, art. no. 076502, Jul. 2012.
11. S. Jo, T. Chang, I. Ebong, B. Bhadviya, P. Mazumder, and W. Lu,
"Nanoscale memristor device as synapse in neuromorphic systems," Nano
Letters, vol. 10, pp. 1297-1301, Apr. 2010.

9

12. T. Chang, Y. Yang, and W. Lu, "Building neuromorphic circuits with
memristive devices," IEEE Circuits and Systems Magazine, vol. 13, pp. 56-73,
2013.
13. J. Tedesco, L. Stephey, M. Hernandez-Mora, C. Richter, and N. GergelHackett, "Switching mechanisms in flexible solution-processed TiO2 memristors,"
Nanotechnology, vol. 23, art. no. 305206, Aug. 2012.
14. P. R. F. Rocha, A. Kiazadeh, D. M. De Leeuw, S. C. J. Meskers, F. Verbakel,
D. M. Taylor, and H. L. Gomes. “The role of internal structure in the anomalous
switching dynamics,” Journal of Applied Physics, vol. 113, no. 13, art. no.
134504, Jul. 2013.
15. G. F. Fine, L. M. Cavanagh, A. Afonja, and R. Binions, “Metal oxide
semiconductor gas sensors in environmental monitoring,” Sensors, vol. 10, pp.
5469-5502, Jun. 2010.
16. R. Ionescu, A. Hoel, C.G. Granqvist, E. Llobet, P. Heszler, “Low-level
detection of ethanol and H2S with temperature-modulated WO3 nanoparticle gas
sensors,” Sensors and Actuators B, vol. 104, pp. 132-139, 2005.
17. Z. Wen, L. Tian-mo, and L. De-jun, “Formaldehyde gas sensing property and
mechanism of TiO2–Ag nanocomposite,” Physica B: Condensed Matter, vol. 405,
pp. 4235-4239, Oct. 2010.
18. V. Hien, J.-L.You, K.-M. Jo, S.-Y. Kim, J.-H. Lee, et al., "H2S-sensing
properties of Cu2O submicron-sized rods and trees synthesized by radio
frequency magnetron sputtering," Sensors and Actuators B, vol. 202, pp. 330338, Oct. 2014.
19. M. Ahmad, A. Sadek, J. Ou, M.H. Yaacob, et al., “Facile synthesis of
nanostructured WO3 thin films and their characterization for ethanol sensing,"
Materials Chemistry and Physics, vol. 141, pp. 912-919, Jun. 2013.
20. C. Nyenke and L. Dong, “Sensing ambient oxygen using a W/CuxO/Cu
memristor,” IEEE 10th International Conference on Nano/Micro Engineered and
Molecular Systems, pp. 254-258, Apr. 2015. Š 2015 IEEE.
21. V. Magdanz, M. Guix, and O. G. Schmidt, “Tubular micromotors: from
microjets to spermbots,” Robotics and Biomimetics, vol. 1, no. 11, pp. 1-10, Oct.
2014.

10

22. L. Zhang, J. J. Abbott, L. Dong, B. E. Kratochvil, D. Bell, and B. J. Nelson,
“Artificial bacterial flagella: Fabrication and magnetic control,” Applied Physics
Letters, vol. 94, art. no. 064107, Feb. 2009.

11

CHAPTER 2 LITERATURE REVIEW

2.1 METAL-OXIDE MEMRISTOR: SWITCHING LAYER DESIGN
While the simplest design of a metal-oxide based memristor consists of a
transition metal oxide sandwiched between two metal electrodes, consideration for
improved endurance and performance have propelled design of the device into the
use of nonstoichiometric switching layers [1]-[3]. This is because the chemistry of
such layers allows for increased oxygen vacancies [4]-[6] – or increased oxygen
ions [7,8] – giving rise to more efficiently formed conductive filaments. To date,
some of the more studied nonstoichiometric switching layers include copper-rich
oxide (CuxO) [9]-[13], oxygen-deficient titanium oxide (TiO2-x) [14]-[16], oxygendeficient silicon oxide (SiOx) [6,17-20], oxygen-rich tungsten oxide (WOx) [21]-[24],
and oxygen-rich zinc oxide (ZnOx) [25]-[27].
Another area of memristor design improvement has been the sandwiching
of the metal oxide between an inactive metal electrode and an active metal
electrode. With this particular arrangement, the active electrode can contribute
metal ions to the formation of conductive filaments in the switching layer. Such
electrodes have been constructed from metals like silver (Ag) [28]-[30] and copper
(Cu) [31]-[33] that will react with the metal-oxide layer and emit metal ions. On the
other hand, the inactive electrode is a metal, such as platinum (Pt) [28,33], gold
(Au) [30], and tungsten (W) [30-31] that does not chemically react with the
switching layer. For example, work by Yan and his colleagues [28] in a 2010

12

publication incorporated active electrodes fabricated from silver (Ag). The
researchers were able to attribute resistive switching to Ag+ ions contributed from
the electrode. Additionally, recent work by Zhang and his colleagues and others in
2016 suggested formation of titanium (Ti) conductive filaments from the
incorporation of a TiN electrode in their device [34]. An illustration of this process
is shown in Figure 2.
In recent years, bilayer structuring of the metal-oxide switching layer has
been used to further enhance memristor performance and endurance. Work by
Lin, Huang et al. [35] has demonstrated that the addition of ZnO1-x nanorod arrays
on a zinc oxide (ZnO) thin film reduces the operation voltage for a Pt/ZnO1-x
/ZnO/Pt memristor by acting as a reservoir for supplemental oxygen vacancies.

Figure 2. Formation of conductive filaments comprising of a) titanium ions
only and b) oxygen vacancies and ions with titanium ions. Reprinted from
H. Z. Zhang, D. S. Ang, K. S. Yew, and X. P. Wang, “Enhanced stability of
complementary resistance switching in the TiN/HfOx/TiN resistive randomaccess memory device via interface engineering,” Applied Physics Letters,
vol. 108, art. no. 083505, Feb 2016, with the permission of AIP Publishing.

13

This particular bilayer structure also exhibited hydrophobic behavior, which
allowed for good switching endurance and the prevention of short circuits in a wet
environment. Fabrication of the device involved deposition via sputtering (for the
ZnO thin film) and a chemical solution process (for the ZnO1-x nanorod array).
In a later work, Huang, Lin et al. further explored bilayer structures to
improve the performance of traditional ZnO-based memory systems. This more

Figure 3. [LEFT] When a sufficient negative bias is applied, the ZnWOx layer
acts as an oxygen ion reservoir to form the conductive filament. [RIGHT]
When a sufficient positive is bias, the layer becomes an oxygen supplier
leading to the rupture of the conductive filament. Reprinted with permission
from S.-M. Lin, J.-S. Huang, W.-C. Chang, T.-C. Hou, H.-W. Huang, C.-H.
Huang, S.- J. Lin and Y.-L. Chueh, “Single-step formation of ZnO/ZnWOx
bilayer structure via interfacial engineering for high performance and low
energy consumption resistive memory with controllable high resistance
states,” ACS Applied Materials & Interfaces, vol. 5, no. 16, pp. 7831-7837,
Aug. 2013. Copyright 2013 American Chemical Society.

14

recent work [36] demonstrated that the inclusion of a thin ZnWOx layer between a
zinc oxide (ZnO) film and bottom tungsten (W) electrode can improve switching
endurance and enable less power consumption by reducing the compliance
current. Other additional benefits of the ZnO/ZnWOx bilayer structure include
enhancements in switching uniformity and increases in high resistance states. In
this study, the ZnO/ZnWOx bilayer structure was spontaneously formed during a
one-step sputtering process of ZnO onto the W contact. The switching process of
the device is illustrated in Figure 3.
In addition to the aforementioned ZnO-based devices, other materials have
been used to implement bilayer structures for memristors. For example, work by
Lee et al. [37] has also demonstrated improvements in switching and power
consumption but via a TaO-based structure. A TaO2−x base layer was deposited
via reactive sputtering followed by a TaO2O5-x layer that was formed using oxygen
plasma to control thickness and provide medium resistivity. The overall design
allowed for device stability and a reduction in the low-resistance-state current.
Similarly, work by Pi and others [38] showed a reduction in total switching energy
consumption via the addition of a thin yttrium (Y) layer onto a yttrium oxide (Y2O3)
layer. Many other arrangements of bilayer structures have been demonstrated
over the years [39, 40].

15

2.2 MEMRISTOR RESEARCH IN ELECTRONIC APPLICATIONS
Most memristor research has focused on applications in electronics,
particularly in the areas of memory and logic [41, 42]. One intrigue is the
memristor’s potential to become an option for non-volatile memory [41] due to its
ability to remember past information even after power has ceased. Additionally,
since the resistance of these devices can switch between very small values (e.g.,
hundreds to thousands of ohms) and extremely large values (e.g., tens to
thousands of kilo-ohms), such low and high states can represent 0s and 1s.
Additionally, memristors are an appealing replacement for transistors due to
density scaling [42] because while the transistor has two dimensions – gate width
and length – the memristor has only one critical dimension – switching layer
thickness [42, 43]. (Because the conductive filaments are few and localized rather
than homogenous throughout the switching layer [44, 45], the area of the device
is not a critical dimension.)
The more recent attention towards the memristor’s application in logic has
increased research in neuromorphic systems [46-50]. For example, Jo and others
demonstrated the use of nanoscale silicon-based memristors, as shown in Figure
4, to model synapses supporting important neural functions, such as spike timing
dependent plasticity [46]. In a more recent paper by Wang and his colleagues, the
plasticity and learning behaviors of synapses are mimicked in HfOx/ZnOx-based
memristors [47], as illustrated in Figure 5. The attraction to memristors in

16

neuromorphic research is largely due to “the local adaptive learning and large
connectivity that can be obtained in such devices” [48].

Figure 4. Reprinted with permission from S. Jo, T. Chang, I. Ebong, B.
Bhadviya, P. Mazumder, and W. Lu, "Nanoscale Memristor Device as
Synapse in Neuromorphic Systems," Nano Letters, vol. 10, pp. 12971301, Apr 2010. Copyright 2010 American Chemical Society.

Figure 5. Reproduced from L. G. Wang, W. Zhang, Y. Chen, Y. Q. Cao, A. D.
Li, and D. Wu, “Synaptic plasticity and learning behaviors mimicked in single
inorganic synapses of Pt/HfOx/ZnOx/TiN memristive system,” Nanoscale
Research Letters, vol 12, art. no. 65, Jan. 2017. Publisher: Springer.

17

2.3 PASSIVE GAS SENSING WITH METAL OXIDES
Most metal oxides are suitable for gas sensing because of inherent
defective sites, which allow for the adsorption of oxygen from the ambient air.
These defective sites are typically the result of vacancies [51] generated in the
fabrication process. In the case of cuprous oxide (Cu2O) and cupric oxide (CuO),
for example, such sites are the result of Cu vacancies [52]. Chemical bonding
between adsorbed oxygen and Cu results in the presence of holes (O2 ↔ O2-+ h),
which leads to an accumulation layer near the grain surface. Certain gases can
then react with the ionically-adsorbed oxygen (O2-/O-) thus changing the resistance
of the material.
In 2013, M. Ahmad and his colleagues fabricated and investigated simple
tungsten oxide (WO3) thin films for reproducible ethanol (C2H6O) sensing [53]. The
films were most sensitive to a concentration of 100ppm of ethanol at 300 to 400
degrees Celsius and exhibited a reversible decrease in resistance. This change in
resistance was believed to be attributed to oxygen ions which were adsorbed by
the metal oxide surface from the ambient air. Upon exposure to ethanol, these
adsorbed oxygen ions react with the ethanol molecules leading to the desorption
of surface oxygen ions as demonstrated in the following reactions [53]:
C
𝐶𝐻= 𝐶𝐻7 𝑂𝐻(?@A) + 6𝑂(?@A)
→ 2𝐶𝑂7(E) + 3𝐻7 𝑂(3?F) + 6𝑒 C 																																				(6)

7C
𝐶𝐻= 𝐶𝐻7 𝑂𝐻(?@A) + 6𝑂(?@A)
→ 2𝐶𝑂7(E) + 3𝐻7 𝑂(3?F) + 12𝑒 C 																																	(7)

18

This chemical process, which is depicted in Figure 6, translates to a temporary
increase in the WO3 conductivity. When the ethanol source is cut off, the electrons
and conductivity gradually decrease.
Similarly, in 2014, Hien and his colleagues investigated passive sensing of
hydrogen sulfide (H2S) via fabricated cuprous oxide (Cu2O) structures [52]. The
sensing mechanism of their samples involved an irreversible phase transformation
from Cu2O to CuS [52] through an interaction of H2S with Cu2O (at high
temperatures) as follows:
2𝐶𝑢7 𝑂 + 𝑂7 → 4𝐶𝑢𝑂																																																																																																										(8)
𝐶𝑢𝑂 + 𝐻7 𝑆 → 𝐶𝑢𝑆 + 𝐻7 𝑂																																																																																																		(9)

Figure 6. Reprinted from Materials Chemistry and Physics, vol. 141, M.
Ahmad, A. Sadek, J. Ou, M.H. Yaacob, et al., “Facile synthesis of
nanostructured WO3 thin films and their characterization for ethanol
sensing,” pp. 912-919, Copyright 2013, with permission from Elsevier.

19

Additionally, there was a removal of adsorbed oxygen ions (where n = 1, 2) on the
oxide layer via the following reactions:
MC
2𝐻7 𝑆(E) + 3𝑂7(?@A)
→ 2𝐻7 𝑂(E) + 2𝑆𝑂7(E) + 3𝑛𝑒 C 																																																			(10)

MC
2𝑆𝑂7(E) + 𝑂7(?@A)
→ 2𝑆𝑂= + 𝑛𝑒 C 																																																																																		(11)

More importantly, Hien and his colleagues compared the sensitivities
between Cu2O thin films, Cu2O nano-rods, and Cu2O nano-trees, with the nanotrees outperforming the other structures. This research underlined the importance
of structural considerations – particularly, surface area – when fabricating gas
sensors.

Figure 7. Reprinted from Sensors and Actuators B-Chemical, vol. 202,
V. Hien, J. You, K. Jo, S. Kim, et al., “H2S-sensing properties of Cu2O
submicron-sized rods and trees synthesized by radio-frequency
magnetron sputtering,” pp. 330-338, Copyright 2014, with permission
from Elsevier.

20

2.4 MEMRISTORS ADVANCES IN SENSING
Though the main focus of memristor research continues to be in electronics,
a few of the more recent literature have exploited memristive effects for sensing.
For example, in 2012, Carrara et al demonstrated bio-sensing based on
memristive signals obtained on silicon freestanding nanowires (SiNW) [54].
(Silicon nanowires already manifest a memory effect due to charge trapping
occurring inside the structures [55].) As shown in Figure 8, the silicon channels
were functionalized by covalent attachment of rabbit polyclonal antibodies with
GPTS (glycidoxypropyltrimethoxysilane), which contributed positive charges. This
contribution resulted in a virtual “all-around bio-gate” around the channels, thus
affecting the hysteresis. The functionalized SiNWs were then exposed to antigens,

Figure 8. Reprinted from Sensors and Actuators B-Chemical, vol. 171, S.
Carrara, D. Sacchetto, M. Doucey, C. Baj-Rossi, G. De Micheli, and Y.
Leblebici, "Memristive-biosensors: A new detection method by using
nanofabricated memristors,” pp. 449-457, Copyright 2012, with permission
from Elsevier.

21

which further impacted the hysteresis. Thus, the authors of this study
demonstrated that both probe antibodies and target antigens affect the memristive
behavior of SiNWs.
Following that research, the same authors reported also using SiNWs for
the detection of different concentrations of H+, and thus pH detection in 2014 [55].
H+ ions from a solution go to form OH- groups with the native oxide on the
nanowires causing an increase in conductance and hysteresis of the wires. The
observed change in the electrical properties with pH in dry conditions is related to
the formation of a wet film at the nanowire surface. Ions from the initial solution
were free to move in the final water thin film at the sensing interface with
consequent polarization of the NW surface. Carrara and his colleagues [55]
believed that the detection of hydrogen ions after the device had dried was due to
adsorption of water molecules from the surrounding environment to the native
oxide of the nanowire. According to the Gouy-Chapman-Stern model, “the only ion
species that approaches the surface of the insulator is the H+ ions, since these
ions are small enough and poorly hydrated by water molecules, unlike the ions of
other dissolved species, such as NaCl [55]”.
In 2012, Ungureanu and his colleagues demonstrated sensing via a
memristor in which the resistance was not only controlled by means of voltage
pulses, but also by light [56]. The device allowed for dual functions: 1) data
encoding by using a certain light irradiance during the writing process and 2)
capability to work as a light sensor in the ultraviolet (UV) to infrared (IR) range. To

22

fabricate this memristor, an Al2O3 film was deposited on a SiO2/Si substrate,
followed by the fabrication of Pd top contacts. As displayed in Figure 9, lightexposed Al2O3/SiO2/Si rings were fabricated around the circular Pd contacts, thus
allowing for light to reach the optically active Si after passing through the
transparent oxide layers. Ungureanu interpreted the change in the remnant current
as a modulation of the trapped electrons in the Al2O3 layer, where the resistivity
increases upon the introduction of electrons from the Si layer. The photogenerated electrons are injected in the Al2O3 layer only if the electric field is high
enough to overcome this energy barrier [56]. Consequently, the observed behavior
of the memristor was attributed to the Al2O3 layer and the Al2O3/SiO2 interface.
Many other researchers have since looked at using light as a second control
mechanism for the memristor [57, 58, 59, 60], inlcluding most recently, Wei Wang
and his colleagues with a MoS2-based memristor in 2016 [61].
Lastly, in 2010, Xiaobin Wang and others proposed a spintronic memristor
acting as a temperature sensor whereby the resistance dependence was achieved
through

a

combination

of

magnetoresistance

and

spin-torque-induced

magnetization motion [62]. As shown in Figure 10, the resistance of the memristor
depends on the domain wall position, which is moved through current-induced spin
torque excitation. The domain wall velocity at finite temperature depends on spintorque excitation strength and thermal fluctuation magnitude [62, 63]. (As
temperature increases, domain wall velocity increases.) For temperature sensing,
a biasing voltage pulse with constant magnitude is applied to the device, and the

23

resistance before and after the voltage pulse is measured. This resistance
difference is calibrated to sense temperature.

Figure 9. X. Wang, Y. Chen, Y. Gu, and H. Li, "Spintronic Memristor Temperature
Sensor," IEEE Electron Device Letters, vol. 31, pp. 20-22, JAN 2010. Š 2010
IEEE.

2.5 SUMMARY
In summary, this chapter was a review of literature regarding memristor
research over the past decade. Much of this research involves a movement from
the basic metal-insulator-metal (MIM) design of the device to a bilayer structure of
the metal-oxide switching thin film for improved performance and endurance.
These enhancements are essential for memristor applications in electronics,
including for memory and logic. Though much memristor research continues to be

24

in the area of electronics, more recent literature shows an extension to applications
in sensing, especially light sensing. However, there remains to be an investigation
of metal-oxide based memristors for passive gas sensing, despite the inherent
ability of oxides to respond to such stimuli. Thus, this dissertation, to our
knowledge, is the first exploration of memristor research into this area.

25

REFERENCES

26

REFERENCES

1. J. J. Yang, M. D. Pickett, X. Li, D. A. A. Ohlberg, D. R. Stewart, and R. S.
Williams. “Memristive switching mechanism for metal/oxide/metal nanodevices,”
Nature Nanotechnology, vol. 3, pp. 429-433, Jun. 2008.
2. P. R. F. Rocha, A. Kiazadeh, D. M. De Leeuw, S. C. J. Meskers, F. Verbakel,
D. M. Taylor, and H. L. Gomes. “The role of internal structure in the anomalous
switching dynamics,” Journal of Applied Physics, vol. 113, no. 13, art no. 134504,
Jul. 2013.
3. M. Ismail, C.-Y. Huang, D. Panda, C.-J. Hung, T.-L. Tsai, J.-H. Jieng, C.-A.
Lin, U. Chand, A. Rana, E. Ahmed, I. Talib, M. Nadeem, and T.-Y. Tseng,
“Forming-free bipolar resistive switching in nonstoichiometric ceria films,”
Nanoscale Research Letters, vol. 9, no. 1, art. no. 45, Jan. 2014.
4. R. Waser and M. Aono, “Nanoionics-based resistive switching memories,”
Nature Materials, vol. 6, pp. 833-840, Nov. 2007.
5. Y. Yang, S. Choi, and W. Lu, “Oxide heterostructure resistive memory,” Nano
Letters, vol. 13, pp. 2908-2915, May 2013.
6. A. Mehonic, S. Cueff, M. Wojdak, S. Hudziak, O. Jambois, C. Labbe´, B.
Garrido, R. Rizk, and A. J. Kenyon. “Resistive switching in silicon suboxide films,”
Journal of Applied Physics, vol. 111, art. no. 074507, Feb. 2012.
7. T.-M. Tsai, K.-C. Chang, T.-C. Chang, R. Zhang, T. Wang, C.-H. Pan, and K.H. Chen, “Resistive switching mechanism of oxygen-rich indium tin oxide
resistance random access memory,” IEEE Electron Device Letters, vol. 37, no. 4,
pp. 408-411, Apr. 2016.
8. Y.-J. Huang, S.-C. Chao, D.-H. Lien, C.-Y. Wen, J.-H. He, and S.-C. Lee,
“Dual-functional memory and threshold resistive switching based on the push-pull
mechanism of oxygen ions,” Scientific Reports |, vol. 6, art. no. 23945, Apr. 2016.
9. H. B. Lv, M. Yin, Y. L. Song, X. F. Fu, L. Tang, P. Zhou, C. H. Zhao, T. A.
Tang, B. A. Chen, and Y. Y. Lin. “Forming process investigation of CuxO memory
films,” IEEE Electron Device Letters, vol. 29, no. 1, pp. 47-49, Jan. 2008.
10. N. Mcdonald, S. Bishop, B. Briggs, J. V. Nostrand, and N. Cady, “Influence of
the plasma oxidation power on the switching properties of Al/CuxO/Cu memristive
devices,” Solid-State Electronics, vol. 78, pp. 46–50, Dec. 2012.

27

11. P. Zhou, H. B. Lv, M. Yin, L. Tang, Y. L. Song, T. A. Tang, Y. Y. Lin, A. Bao,
A. Wu, S. Cai, H. Wu, C. Liang, and M. H. Chi, “Performance improvement of
CuOx with gradual oxygen concentration for nonvolatile memory application,”
Journal of Vacuum Science & Technology B, vol. 26, pp. 1030-1032, Apr. 2008.
12. L. L. Wei, D. S. Shang, J. R. Sun, S. B. Lee, Z. G. Sun, and B. G. Shen,
“Gradual electroforming and memristive switching in Pt/CuOx/Si/Pt systems,”
Nanotechnology, vol. 24, no. 32, art. no. 325202, Aug. 2013.
13. K.-D. Liang, C.-H. Huang, C.-C. Lai, J.-S. Huang, H.-W. Tsai, Y.-C. Wang, Y.C. Shih, M.-T. Chang, S.-C. Lo, and Y.-L. Chueh, “Single CuOx Nanowire
Memristor: Forming-Free Resistive Switching Behavior,” ACS Applied Materials
& Interfaces, vol. 6, no. 19, pp. 16537-16544, Oct. 2014.
14. P. Bousoulas, I. Michelakaki, and D. Tsoukalas, “Influence of oxygen content
of room temperature TiO2-x deposited films for enhanced resistive switching
memory performance,” Journal of Applied Physics, vol. 115, no. 3, art. no.
034516, Jan. 2014.
15. K. Tsunoda, Y. Fukuzumi, J. R. Jameson, Z. Wang, P. B. Griffin, and Y.
Nishi, “Bipolar resistive switching in polycrystalline TiO2 films,” Applied Physics
Letters, vol. 90, no. 11, art. no. 113501, Dec. 2007.
16. E. Goren, M. Ungureanu, R. Zazpe, M. Rozenberg, L. E. Hueso, P. Stoliar, Y.
Tsur, and F. Casanova, “Resistive switching phenomena in TiOx nanoparticle
layers for memory applications,” Applied Physics Letters, vol. 105, no. 14, art. no.
143506, Jun. 2014.
17. R. Huang, L. Zhang, D. Gao, Y. Pan, S. Qin, P. Tang, Y. Cai, and Y. Wang,
“Resistive switching of silicon-rich-oxide featuring high compatibility with CMOS
technology for 3D stackable and embedded applications,” Applied Physics A, vol.
102, no. 4, pp. 927–931, Mar. 2011.
18. Y.-F. Chang, B. Fowler, Y.-C. Chen, Y.-T. Chen, Y. Wang, F. Xue, F. Zhou,
and J. C. Lee, “Intrinsic SiOx-based unipolar resistive switching memory. II.
Thermal effects on charge transport and characterization of multilevel
programing,” Journal of Applied Physics, vol. 116, no. 4, art. no. 043709, Jul.
2014.
19. Y.-F. Chang, B. Fowler, Y.-C. Chen, Y.-T. Chen, Y. Wang, F. Xue, F. Zhou,
and J. C. Lee, “Intrinsic SiOx-based unipolar resistive switching memory. I. Oxide
stoichiometry effects on reversible switching and program window

28

optimization,” Journal of Applied Physics, vol. 116, no. 4, art. no. 043708, Jul.
2014.
20. C.-Y. Liu, Y.-Y. Tsai, and C.-H. Lai, “Effect of plasma treatment of resistive
layer on a Cu/SiOx/Pt memory device,” Journal of Vacuum Science & Technology
A: Vacuum, Surfaces, and Films, vol. 32, no. 2, art. no. 02B111, Mar. 2014.
21. C. Ho, E. K. Lai, M. D. Lee, C. L. Pan, Y. D. Yao, K. Y. Hsieh, R. Liu, and C.
Y. Lu, “A Highly Reliable Self-Aligned Graded Oxide WOx Resistance Memory:
Conduction Mechanisms and Reliability,” 2007 IEEE Symposium on VLSI
Technology, 2007.
22. D. S. Shang, L. Shi, J. R. Sun, B. G. Shen, F. Zhuge, R. W. Li, and Y. G.
Zhao, “Improvement of reproducible resistance switching in polycrystalline
tungsten oxide films by in situ oxygen annealing,” Applied Physics Letters, vol.
96, no. 7, art. no. 072103, Feb. 2010.
23. K. P. Biju, X. Liu, M. Siddik, S. Kim, J. Shin, I. Kim, A. Ignatiev, H. Huang,
“Resistive switching characteristics and mechanism of thermally grown WOx thin
films,” Journal of Applied Phyics, vol. 110, art. no. 064505, Sep. 2011.
24. R. Yang, K. Terabe, T. Tsuruoka, T. Hasegawa, and M. Aono, “Oxygen
migration process in the interfaces during bipolar resistance switching behavior
of WO3-x-based nanoionics devices,” Applied Physics Letters, vol. 100, no. 23,
art. no. 231603, Apr. 2012.
25. G. Chen, C. Song, C. Chen, S. Gao, F. Zeng, and F. Pan, “Resistive
Switching and Magnetic Modulation in Cobalt-Doped ZnO,” Advanced Materials,
vol. 24, no. 26, pp. 3515–3520, Aug. 2012.
26. X. Zhu, W. Su, Y. Liu, B. Hu, L. Pan, W. Lu, J. Zhang, and R.-W. Li,
“Observation of Conductance Quantization in Oxide-Based Resistive Switching
Memory,” Advanced Materials, vol. 24, no. 29, pp. 3941–3946, 2012.
27. J.-J. Ke, Z.-J. Liu, C.-F. Kang, S.-J. Lin, and J.-H. He, “Surface effect on
resistive switching behaviors of ZnO,” Applied Physics Letters, vol. 99, no. 19,
art. no. 192106, Jul. 2011.
28. X.B. Yan, K. Li, J. Yin, Y. D. Xia, H. X. Guo, L. Chen, and Z. G. Liua, “The
resistive switching mechanism of Ag/SrTiO3/Pt memory cells,” Electrochemical
and Solid State Letters, vol. 13, pp. H87-H89, Jan. 2010.

29

29. Y. Li, S. Long, M. Zhang, Q. Liu, L. Shao, S. Zhang, Y. Wang, Q. Zuo, S. Liu,
and M. Liu, “Resistive Switching Properties of Au/ZrO2/Ag Structure for LowVoltage Nonvolatile Memory Applications,” IEEE Electron Device Letters, vol. 31,
pp. 117-119, Feb. 2010.
30. N. Ge, M-X Zhang, L. Zhang, J. Joshua Yang*, Z. Li, and R S. Williams,
“Electrode-material dependent switching in TaOx memristors”, Semiconductor
Science and Technology, vol. 29, no. 10, art. no.104003, Oct. 2014.
31. C. Schindler, S. C. P. Thermadam, R. Waser, and M. N. Kozicki, “Bipolar and
unipolar resistive switching in Cu-doped SiO2,” IEEE Transactions Electron
Devices, vol. 54, pp. 2762-2768, Sep 2007.
32. T. Sakamoto, K. Lister, N. Banno, T. Hasegawa, K. Terabe, and M. Aono,
“Electronic transport in Ta2O5 resistive switch,” Applied Physics Letters, vol. 91,
art. no. 092110, Aug 2007.
33. Y. Li, S. Long, Q. Liu, Q. Wang, M. Zhang, H. Lv, L. Shao, Y. Wang, S.
Zhang, Q. Zuo, S. Liu, and M. Liu, “Nonvolatile multilevel memory effect in
Cu/WO3/Pt device structures,” Physica Status Solidi (RRL) – Rapid Research
Letters, vol. 4, no. 5-6, pp. 124-126, Apr 2010.
34. H. Z. Zhang, D. S. Ang, K. S. Yew, and X. P. Wang, “Enhanced stability of
complementary resistance switching in the TiN/HfOx/TiN resistive random-access
memory device via interface engineering,” Applied Physics Letters, vol. 108, art.
no. 083505, Feb 2016.
35. C.H. Huang, J. S. Huang, S. M. Lin, W. Y. Chang, J. H. He, and Y. L. Chueh,
“ZnO1-x nanorod arrays/ZnO thin film bilayer structure: from homojunction diode
and high-performance memristor to complementary 1D1R application,” ACS
Nano, vol. 6, no. 9, pp. 8407-8414, Aug. 2012.
36. S.-M. Lin, J.-S. Huang, W.-C. Chang, T.-C. Hou,. H.-W. Huang, C.-H. Huang,
S.- J. Lin and Y.-L. Chueh, “Single-step formation of ZnO/ZnWOx bilayer
structure via interfacial engineering for high performance and low energy
consumption resistive memory with controllable high resistance states,” ACS
Applied Materials & Interfaces, vol. 5, no. 16, pp. 7831-7837, Aug. 2013.
37. M.J. Lee, C. B. Lee, D. Lee, S. R. Lee, M. Chang, J. H. Hur, Y. B. Kim, C. J.
Kim, D. H. Seo, S. Seo, U. I. Chung, I. K. Yoo, and K. Kim, “A fast, highendurance and scalable non-volatile memory device made from asymmetric
TaO2O5-x /TaO2−x bilayer structures,” Nature Materials, vol. 10, no. 8, pp. 625-630,
Jul. 2011.

30

38. C Pi, Y. Ren, Z. Q. Liu, and W. K. Chima, “Unipolar memristive switching in
yttrium oxide and RESET current reduction using a yttrium interlayer,”
Electrochemical and Solid-State Letters, vol. 15, no. 3, pp. G5-G7, Dec. 2011.
39. S. Gao, F. Zeng, M. Wang, G. Wang, C. Song and F. Pan, “Tuning the
switching behavior of binary oxide-based resistive memory devices by inserting
an ultra-thin chemically active metal nanolayer: a case study on the Ta2O5–Ta
system,” Physical Chemistry Chemical Physics, vol. 17, no. 19, pp. 12849-12856,
2015.
40. A. Prakash, S. Maikap, C. S. Lai, T. C. Tien, W. S. Chen, H. Y. Lee, F. T.
Chen, M.-J. Kao, and M.-J. Tsai, “Bipolar resistive switching memory using
bilayer TaOx/WOx films,” Solid-State Electronics, vol. 77, pp. 35-40, Nov. 2012.
41. A. Siemon, T. Breuer, N. Aslam, S. Ferch, W. Kim, J. van den Hurk, V. Rana,
S. Hoffmann-Eifert, R. Waser, S. Menzel, E. Linn, “Realization of boolean logic
functionality using redox-based memristive devices,” Advanced Functional
Materials, vol. 25, pp. 6414-6423, Oct. 2015.
42. D.B. Strukov and H. Kohlstedt, “Resistive switching phenomena in thin films:
Materials, devices, and applications,” MRS Bulletin, vol. 37, pp. 108-114, Feb.
2012.
43. J. Hou, S.S. Nonnenmann, W. Qin, and D.A. Bonnell, “Size dependence of
resistive switching at nanoscale metal-oxide interfaces", Advanced Functional
Materials, vol. 24, no. 26, pp. 4113-4118, Jul. 2014.
44. R. Dittmann, R. Muenstermann, I.K. D. Park, T. Menke, J. Mayer, A.
Besmehn, F. Kronast, C. Schneider, and R. Waser, “Scaling potential of local
redox processes in memristive SrTiO3 thin-film devices,” Proceedings of the IEEE
100, vol. 6, pp. 1979-1990, Jun. 2012.
45. Y. Yang, P. Gao, S. Gaba, T. Chang, X. Pan, and W. Lu, “Observation of
conducting filament growth in nanoscale resistive memories,” Nature
Communications, vol. 3, art. no. 732, Mar. 2012.
46. S. Jo, T. Chang, I. Ebong, B. Bhadviya, P. Mazumder, and W. Lu,
"Nanoscale memristor device as synapse in neuromorphic systems," Nano
Letters, vol. 10, pp. 1297-1301, Apr. 2010.
47. L. G. Wang, W. Zhang, Y. Chen, Y. Q. Cao, A. D. Li, and D. Wu, “Synaptic
plasticity and learning behaviors mimicked in single inorganic synapses of

31

Pt/HfOx/ZnOx/TiN memristive system,” Nanoscale Research Letters, vol 12, art.
no. 65, Jan. 2017.
48. T. Chang, Y. Yang, and W. Lu, "Building neuromorphic circuits with
memristive devices," IEEE Circuits and Systems Magazine, vol. 13, pp. 56-73,
2013.
49. K. Cantley, A. Subramaniam, H. Stiegler, R. Chapman, and E. Vogel, "Neural
learning circuits utilizing nano-crystalline silicon transistors and memristors,"
IEEE Transactions on Neural Networks and Learning Systems, vol. 23, pp. 565573, Apr. 2012.
50. T. Chang, S-H Jo, K-H Kim, P. Sheridan, S. Gaba, and W. Lu, “Synaptic
behaviors and modeling of a metal oxide memristive device,” Applied Physics A,
vol. 102, no. 4, pp. 851-855, Mar. 2011.
51. N. Barsan and U. Weimar, “Conduction model of metal oxide gas sensors,”
Journal of Electroceramics, vol. 7, no. 3, pp. 143–167, Dec. 2001.
52. V. Hien, J. You, K. Jo, S. Kim, J. Lee, J. Kim, et al., "H2S-sensing properties
of Cu2O submicron-sized rods and trees synthesized by radio-frequency
magnetron sputtering,” Sensors and Actuators B-Chemical, vol. 202, pp. 330338, Oct. 2014.
53. M. Ahmad, A. Sadek, J. Ou, M.H. Yaacob, et al., “Facile synthesis of
nanostructured WO3 thin films and their characterization for ethanol sensing,"
Materials Chemistry and Physics, vol. 141, pp. 912-919, Jun. 2013.
54. S. Carrara, D. Sacchetto, M. Doucey, C. Baj-Rossi, G. De Micheli, and Y.
Leblebici, "Memristive-biosensors: A new detection method by using
nanofabricated memristors,” Sensors and Actuators B-Chemical, vol. 171, pp.
449-457, Sep. 2012.
55. F. Puppo, M. Di Ventra, G. De Micheli, and S. Carrara, "Memristive sensors
for pH measure in dry conditions," Surface Science, vol. 624, pp. 76-79, Jun.
2014.
56. M. Ungureanu, R. Zazpe, F. Golmar, P. Stoliar, R. Llopis, F. Casanova, et al.,
"A light-controlled resistive switching memory," Advanced Materials, vol. 24, pp.
2496-2500, May 2012.

32

57. B. Sun, X. Li, D. Liang, and P. Chen, “Effect of visible-light illumination on
resistive switching characteristics in Ag/Ce2W3O12/FTO devices,” Chemical
Physics Letters, vol. 643, pp. 66-70, Jan. 2016.
58. A. Chiolerio, I. Roppolo, V. Cauda, M. Crepaldi, S. Bocchini, K. Bejtka, A.
Verna, and C. F. Pirri, “Ultraviolet mem-sensors: Flexible anisotropic composites
featuring giant photocurrent enhancement,” Nano Research, vol. 8, no. 6, pp.
1956-1963, Jun. 2015.
59. B. Sun, J. Wu, X. Jia, F. Lou, and P. Chen, “Preparation and light-controlled
resistive switching memory behavior of CuCr2O4,” Journal of Sol-Gel Science and
Technology, vol. 75, no. 3, pp. 664-669, May 2015.
60. B. Sun, X. J. Jia, J. H. Wu, and P. Chen, “Photo-electron double regulated
resistive switching memory behaviors of Ag/CuWO4/FTO device,” Solid State
Communications, vol. 223, pp. 1-5, Dec. 2015.
61. W. Wang, G. N. Panin, X. Fu, L. Zhang, P. Ilanchezhiyan, V. O. Pelenovich,
D. Fu and T. W. Kang, “MoS2 memristor with photoresistive switching,” Scientific
Reports, vol. 6, pp. 1-10, Nov. 2016.
62. X. Wang, Y. Chen, Y. Gu, and H. Li, "Spintronic memristor temperature
sensor," IEEE Electron Device Letters, vol. 31, pp. 20-22, Jan. 2010.
63. H. Mahmoudi, V. Sverdlov, and S. Selberherr, "Influence of geometry on the
memristive behavior of the domain wall spintronic memristors and its applications
for measurement," Journal of Superconductivity and Novel Magnetism, vol. 26,
pp. 1745-1748, May 2013.

33

CHAPTER 3 MEMRISTOR DESIGN AND FABRICATION

3.1 INTRODUCTION
The metal-insulator-metal (MIM) memristor design was chosen for ease and
repeatability in fabrication for this project. A nonstoichiometric metal oxide (copper
oxide) was sandwiched between an inactive metal electrode (tungsten) and an
active electrode (copper). Further design and fabrication details are covered in this
chapter.

3.2 PROTOTYPE MEMRISTOR DESIGN
3.2.1 Design Overview
The prototype was designed using a crossbar array structure, as
demonstrated in Fig. 11, to increase device yield per substrate. Five 15-Îźm copper
(Cu) bottom electrodes were overlaid with non-stoichiometric copper oxide (CuxO)
switching layers. They shared a 10-Îźm tungsten (W) top electrode, which resulted
in five 10-Îźm-x-15-Îźm memristors per substrate. As depicted in the schematic, a
portion of the CuxO layer is left uncovered by the W top layer to allow for sensing
via oxygen exposure. Figure 12 displays a scanning electron microscopy (SEM)
image of the fabricated device. This design overview is reproduced from
publication [1].

34

Figure 10. Initial memristor design. During testing of the memristor, the
bottom electrode is grounded.

35

Figure 11. SEM of prototype – crossbar array memristor There are a total
of five bottom electrodes and one top electrode. The copper oxide
switching layer (which is not easily seen here) sits between the copper
bottom electrode and top tungsten electrode with regions exposed for
sensing. This design results in five 10-micron-x-15-micron memristors.

3.2.2 Fabrication Process
The fabrication process for this initial memristor was as follows and is
reproduced from publication [1]:
Silicon substrates underwent plasma-enhanced chemical vapor deposition
of 500-nanometer thick silicon dioxide (SiO2) via an Oxford PECVD System
PlasmaLab 80 plus. This process was necessary to provide an insulating layer
between the to-be-deposited Cu bottom electrodes and the substrate. The

36

substrates were then subjected to photolithography for patterning of these bottom
electrodes. W/CuxO/Cu devices were prepared on the SiO2/Si substrates using DC
magnetron sputtering via a Kurt Lesker Axxis PVD System at room temperature.
A copper layer of 30 nanometers was first deposited onto the pattern using a 99.9%
pure Cu target, which was presputtered to remove any native oxide layer. The Cu
electrodes were then partially shielded, and a non-stoichiometric copper oxide
(CuxO) switching layer was deposited using the same Cu target but with an Ar/O2
mixture injected into the system. The parameters for the CuxO deposition were as
follows: power of 200 Watts, ~27 cubic centimeters per minute of argon, and ~7
cubic centimeters per minute of oxygen. After deposition of the Cu and CuxO
layers, lift-off was performed to complete patterning of the bottom electrode and
switching layer, respectively. The substrate was then prepared for patterning of the
top electrode by photolithography. A tungsten (W) layer of 30 nanometers was
sputtered using a 99.9% pure W target that had been pre-sputtered to remove any
oxide layer or surface contamination. A final metal lift-off was then performed to
complete the W top electrode. The fabrication process is illustrated in Fig. 13.

37

Figure 12. Fabrication process. (a) The fabrication begins on a clean silicon die.
(b) PECVD of silicon dioxide is performed to provide an insulating layer between
the memristor and silicon surface. (c) Photoresist is spun and baked onto the die
followed by photolithography. (d) Cu and CuxO are deposited and lift-off reveals
a bottom electrode and switching layer, respectively. (e-f) The procedures are
repeated for the W top electrode.

38

3.2.2 Maskplate Design
In order to achieve this memristor design, a maskplate was drawn using
Nanometer Pattern Generation Software (NPGS) in combination with CAD
software for construction by PhotoSciences. The maskplate was to consist of two
sections as pictured in Figures 14 and 15. Each section corresponded to a layer
of metal that would be exposed to ultraviolet light (via the mask aligner) and remain
after metal lift-off.

Figure 13. Maskplate design for layering of copper electrodes and contacts.

39

Figure 14. Maskplate design for layering of tungsten electrode and contact.

40

3.3 IMPROVED DEVICE - MEMRISTOR WITH SUB-MICRON HOLES
3.3.1 Design Overview
The device, which is shown in Fig. 16, was designed using a larger crossbar
array structure to increase memristor yield. Ten 10-um bottom electrodes were
overlaid with switching layers followed by ten 10-um top electrodes resulting in 100
memristors per substrate. Additionally, holes were etched through the tungsten top
layer to allow for sensing via oxygen exposure to the copper oxide layer, as
illustrated in Fig. 17. This design overview references publications [2] and [3].

Figure 15. SEM of memristor array.

41

Figure 16. Holes in tungsten layer.

3.3.2 Fabrication Process
Fabrication of the device consisted of the following procedure and is
reproduced from publications [2] and [3]:
Silicon substrates underwent plasma-enhanced chemical vapor deposition
of 500-nm thick SiO2 to provide an insulating layer. Next, photolithography was
performed to pattern the bottom electrodes. A 30-nm W layer was then deposited
onto the SiO2/Si substrate using a 99.9% pure W target, which had been presputtered to remove any native oxide layer or surface contamination. Next, a 30nm Cu layer was deposited — except onto the W contact pads — using a 99.9%
pure Cu target, which had also been pre-sputtered. (W contact pads explained
further in Section 3.3.5) After deposition of the W and Cu layers, lift-off was
performed to complete patterning of the bottom electrode with contacts. The
substrate was then prepared for patterning of the switching layer and top
electrodes by photolithography. A 40-nm layer of oxygen-deficient copper oxide,

42

CuxO, was then deposited using a pre-sputtered 99.9% Cu target with a mixture of
Ar and O2 gases. Next, a W layer of 30 nm was sputtered, and a final metal lift-off
was performed to complete the CuxO switching layer and W top electrode.

3.3.3 Maskplate Design
A second maskplate was constructed for this new memristor design by
PhotoSciences after sending new drawings created with Nanometer Pattern
Generation Software (NPGS) in combination with CAD software. These drawings
are included in the following figures:

43

.

Figure 17. Maskplate design for photolithography. This is the complete
design.

44

Figure 18. Maskplate design for photolithography. This is the first layer
consisting of the bottom electrodes, bottom contacts, and markers.

45

Figure 19. Closeup of the first layer.

46

Figure 20. Maskplate design for photolithography. This is the first
layer consisting of the top electrodes, top contacts, and markers.

47

Figure 21. Closeup of the second layer.

3.3.4 Improving Sensitivity with Holes
Electron beam lithography was employed in order to pattern sub-micron
holes into the tungsten layer for gas sensing. (This sub-section is reproduced from
publications [2] and [3].) The crossbar structure was visually divided into four
quadrants of twenty-five devices each. One quadrant was patterned with meshes
of 300-nm diameter holes while a second quadrant was patterned with meshes of
700-nm diameter holes. A third quadrant consisting of larger holes was used as a
visual guide. The remaining quadrant was left unpatterned for comparison.

48

After patterning, the device was submerged in a NH4OH:H2O2:H2O
(1:100:100) solution for a total of 11 seconds in order to wet etch the sub-micron
holes. The device was then thoroughly rinsed in DI water to stop etching. Figs. 2728 is an SEM image of the etched 500-nm (actual of 700-nm) holes, which will be
referred to as “micro meshes” for the remainder of this paper. Twelve testable
devices were successfully etched in this case. Figures 29-31 is an SEM image of
successfully etched 200-nm (actual of 300-nm) holes, which will be referred to as
“nano meshes” for the remainder of this paper. Most of the meshes were underetched or not etched at all, while a few were over-etched, which resulted in only
two testable devices.

.

49

Figure 22. Initial NPGS design with two sets of 4x4 500-nm holes (in two
quadrants) and one set of larger holes.

50

Figure 23. Failed etching attempt. Note that much of the tungsten layer has
been etched through for numerous memristors.

51

Figure 24. Closeup of failed etching attempt.

52

Figure 25. Improved attempt.

53

Figure 26. Closeup SEM of 700nm-diameter meshes ("micro meshes").

Figure 27. One of the micro meshes.

54

Figure 28. Closeup of 300-nm holes ("nano meshes").

55

Figure 29. One of the nano meshes.

Figure 30. Comparison of (a) usuable nano mesh and over-etched nano mesh.

56

Figure 31. Closeup of final quadrant. The holes on the right were not used
for testing but for etching guidance. The remaining non-etched devices were
used as a control group (“non meshes”).

3.3.5 Preventing Contact Oxidation
In our previous fabrication trials, only a copper layer was deposited for the
bottom electrode and contact pads. However, with continued exposure to heat
during fabrication and testing, the copper contact pads would oxidize thus
rendering the device unusable. Because W does not oxidize as easily, it was
deposited prior to the copper electrodes to create repeatedly usable contacts, as
illustrated in Fig. 33. This sub-section is reproduced from publications [2] and [3].

57

Figure 32. Schematic of memristor with etched
tungsten layer.

3.4 SUMMARY
This chapter details the memristor design and fabrication for the three
studies in this dissertation. A prototype memristor was designed and constructed
early on in the research to determine feasibility of the sensing application for the
first study. This prototype device consisted of a nonstoichiometric copper oxide
sandwiched between two metal contacts. The second memristor was designed
and fabricated later in the project with holes in the top metal contact. This design
was used for the remaining two studies in this dissertation to expose more of the
copper oxide switching layer and further investigate the sensing mechanism.

58

APPENDICES

59

APPENDIX A: SPUTTERING PARAMETERS
Sputtering Parameters for Section 4 Memristors:
W/CuxO (11 nm)/Cu device:
• Cu - Ar: ~1ccm, gas pressure: 3.47e-03 Torr, power: 50W, time: 60 sec
• CuxO - Ar: ~27ccm, O2: ~6ccm, gas pressure: 1.4e-01 Torr
power: 200W, time: 11 sec
• W - Ar: ~1ccm, gas pressure: 9e-03 Torr, power: 60W, time: 3.5 min
W/CuxO (23 nm)/Cu device:
• Cu - Ar: ~1ccm, gas pressure: 4.89e-03 Torr, power: 50W, time: 111 sec
• CuxO - Ar: ~27ccm, O2: ~5ccm, gas pressure: 1.4e-01 Torr
power: 200W, time: 22 sec
• W - Ar: ~1ccm, gas pressure: 9e-03 Torr, power: 60W, time: 3 min
W/CuxO (35 nm)/Cu device:
• Cu - Ar: ~1ccm, gas pressure: 4.89e-03 Torr, power: 50W, time: 111 sec
• CuxO - Ar: ~27ccm, O2: ~7ccm, gas pressure: 1.52e-01 Torr
power: 200W, time: 35 sec
• W - Ar: ~1ccm, gas pressure: 9e-03 Torr, power: 60W, time: 3 min
Sputtering Parameters for Section 5 Memristors:
The Ar:O2 ratio was adjusted to increase oxygen vacancies.
•
•
•
•

W - Ar: ~1ccm; gas pressure: 7.6e-3; power: 65 W; time: 3 min
Cu - Ar: ~1ccm; gas pressure: 7.7e-3; power: 85 W; time: 3 min
CuxO - Ar: ~48ccm, O2: 8ccm, gas pressure: 2.6e-1 Torr, power: 200 W,
time: 3 min
W - Ar: ~1ccm; gas pressure: 7.6e-3; power: 65 W; time: 3 min

60

APPENDIX B: RECIPES
Recipe for Photolithography of Memristors:
•
•
•
•
•
•

Apply Shipley S1318 photoresist
Spin at 3500 RPM for 1 minute
Bake at 115 deg Celsius for 1 minute
Undergo photolithography (MJB 3 Mask Aligner for 10 seconds)
Develop with Shipley Microposit MF 319 for 30 seconds
Rinse in deionized water for 1 minute

Recipe for Electron Beam Lithography of Memristors (Section 5):
•
•
•
•
•
•

Apply PMMA C2 (electron beam lithography)
Spin at 1500 RPM for 40 seconds
Bake at 180 deg Celsius for 180 seconds
Undergo electron beam lithography (JEOL 840A Scanning Electron
Microscope and Electron Beam Lithography)
Develop with 2:3 ration MIBK (methyl isobutyl ketone):IPA (isopropyl
alcohol) solution for 10 secs
Rinse in deionized water for 1 minute

Resist removal (both photoresist and electro beam resist)
•
•
•

Rinse substrate in acetone for 1 minute
Rinse in isopropanol or methanol for 1 minute
Rinse in deionized water for 1-2 minutes

61

REFERENCES

62

REFERENCES

1. C. Nyenke and L. Dong, “Sensing ambient oxygen using a W/CuxO/Cu
memristor,” IEEE 10th International Conference on Nano/Micro Engineered and
Molecular Systems, pp. 254-258, Apr. 2015. Š 2015 IEEE.
2. C. Nyenke and L. Dong, “Fabrication of a W/CuxO/Cu memristor with submicron holes for passive sensing of oxygen,” Microelectronic Engineering, vol.
164, pp. 48-52, Oct. 2016.
3. C. Nyenke and L. Dong, “Effect of NO2 and NH3 on the resistive switching
behavior of W/CuxO/Cu memristors,” Journal of Micromechanics and
Microengineering, vol. 27, no. 10, art. no. 105013, Oct. 2017. 10.1088/13616439/aa8672. Š IOP Publishing. Reproduced with permission. All rights
reserved.

63

CHAPTER 4 DETECTION OF AMBIENT OXYGEN USING
W/CuxO/Cu MEMRISTORS OF DIFFERENT THICKNESSES

4.1 INTRODUCTION
Three sets of memristors with different switching layer thicknesses – 11nm,
23nm, and 35nm – were fabricated using the methodology detailed in Section 3.2.
Each set was exposed to ambient oxygen for a period of time (days) to investigate
passive sensing. Electrical characteristics and detection of ambient oxygen were
investigated for both devices before and after the prolonged exposure to ambient
oxygen. A significant portion of this work references publication [1].

4.2 EXPERIMENTAL SETUP
The current-voltage (I-V) characteristics of the device were measured with
a Keithley 6487 Picoammeter. For the measurement setup, the copper bottom
electrode was grounded while a voltage was applied to the tungsten top electrode.
The electrode contacts were connected to the picoammeter via metal probes, as
shown in Fig. 34. Electrical measurements – voltage and current – were taken
before and after extended exposure to ambient oxygen. This data was sent from
the picoammeter to a local computer via an accompanying GPIB cable and stored
in Excel.

64

Figure 33. Measurement setup.
Atomic composition of each device layer was determined by energydispersive X-ray spectroscopy (EDS) inside a Carl Zeiss Variable Pressure
scanning electron microscope (SEM) EVO LS25. EDS analyses were performed
twice – once before the experiment and again after extended exposure to ambient
air. The thickness of each layer in the memristor was measured using a Veeco
Dektak3 surface profiler with a 0.5-nanometer vertical resolution before the
experiment.

65

4.3 DISCUSSION OF RESISTIVE SWITCHING BEHAVIOR
4.3.1 Formation of Conductive Filaments
It has been shown that formation of conductive filaments from oxygen
vacancies within the switching layer converts the metal oxide layer to a resistive
switching layer for the device to be functional as memristor [2]. Once formed,
application of a sufficient bias voltage can rupture and re-heal these conductive
filaments by repelling and attracting oxygen vacancies, respectively. This leads to
switching between a low resistance state and high resistance state, resulting in the
hysteretic nature of memristors.
In the case of the two devices that were fabricated for this study, the forming
process required voltages of -2.1 Volts, -3.8 Volts, and -4 Volts, respectively. The
dependence of the forming voltage and switching layer thickness makes sense
given the correlation between the anodizing voltage and forming voltage [3], (The
anodizing voltage, Va, is what is required to oxidize a metal layer.) Because voltage
is proportional to distance, an increase in layer thickness will translate to an
increase in Va. The correlation between Va and the forming voltage Vf means that
a thicker metal oxide layer will require a higher Vf. Thus, we observe a higher
forming voltage for the W/CuxO (35 nm)/Cu device versus the W/CuxO (11 nm)/Cu
W/CuxO (23 nm)/Cu devices. A similar relationship between the breakdown
voltage and Vf can be used to explain this thickness-dependent voltage as well [4].

66

4.3.2 Characterizing Memristive Behavior
Figs. 35 and 38 display the current-voltage (IV) characteristics for each
memristor set after two voltage sweeps. For the W/CuxO (11 nm)/Cu device in Fig.
35, we observe bipolar resistive switching between Âą2.5 V. The low resistance
state (LRS) and high resistance state (HRS) are 10.7 k-ohms and 12.4-35.0 kohms, respectively. In contrast, the W/CuxO (35 nm)/Cu device in Fig. 36 has a
larger LRS and HRS due to the thicker switching layer. Additionally, we observe
inherent complementary resistive switching (CRS) for the W/CuxO (35 nm)/Cu
device unlike for the thinner device. From Âą 1 V, the device is in HRS at ~ 400 kohms before switching to an LRS of ~ 70k-ohms at Âą 1 V. The W/CuxO (35 nm)/Cu

Figure 34. Current-voltage characteristics of the W/CuxO(11nm)/Cu memristor.

67

device maintains this LRS until it switches again to a higher and nonlinear
resistance at Âą 4 V. When the voltage approaches Âą 1 V, the device becomes linear
again with a maximum HRS of ~400 k-ohms. The memristance of this device
between Âą1 V is better observed in Figure 37.

Figure 35. Current-voltage characteristics of the W/CuxO(35nm)/Cu
memristor.

68

Figure 36. Close-up of the W/CuxO(35nm)/Cu memristor current-voltage
characteristics between -1 and 1 V.

4.3.3 Identifying the Conduction Mechanism
In order to identify the conduction mechanisms for the W/CuxO (35 nm)/Cu
device, Fig. 38 displays a double logarithmic plot for the positive bias region of Fig.
37. Here we see four segments with slopes ending and/or beginning at 1 V and 4
V. (The sweep direction is indicated by the letters a.-d.) This double logarithmic
plot is then split into four distinct graphs in Fig. 39 so as to better capture the slope
information. In Fig. 39(a), the log-log graph is linear with a slope of ~1, which
indicates Ohmic conduction (I Îą V) for voltages < 1 V. For voltages of 1 V through
4 V, the graph in Fig. 39(b) is also in agreement with Ohmic conduction. Thus, the
transition from HRS in the low-voltage region to LRS in the high-voltage region

69

corresponds with the transition from one form of Ohmic conduction to another. At
4 V, however, the device transitions from Ohmic conduction to space-chargelimited conduction (SCLC) as indicated by the slope of ~2 in Fig. 39(c). This slope
is in agreement with I Îą V2, which describes the square law dependence of SCL
current due to the dominance of hole injection [5,6]. This conduction also
corresponds with the nonlinear portion of the HRS in Fig. 18, from 4 V to 1 V. The
device then returns to Ohmic conduction at 1 V with a slope of ~1 in Fig. 39(d).
This conduction mechanism correlates with the switch to linear HRS in the lowvoltage region.

Figure 37. Log (I) vs. log (V) plot of memristor in the positive bias
region.

70

Figure 38. Best fit lines on log-log plots of memristor in the positive bias region.

4.4 DISCUSSION OF PASSIVE SENSING RESULTS
4.4.1 Increased Oxygen Composition with Time
Tables 1 and 2 show EDS results for W/CuxO (35 nm)/Cu memristor five
and eight weeks after exposure to ambient oxygen, respectively. (Because the
SEM penetration depth is greater than 0.5 microns, the oxygen content includes
that of the SiO2/Si substrate.) Comparison of both tables reveals a 0.44 % increase

71

of oxygen in those three weeks. Comparison of Tables 4 and 5 for a second
memristor, W/CuxO (23 nm)/Cu, reveals a 0.55 % increase. This observation is
consistent with adsorption of oxygen to copper atoms on the exposed copper oxide
surface in those few weeks. Additionally, there were no new elements detected
(such as nitrogen) by EDS after analysis. (NOTE: Only W/CuxO (35 nm)/Cu and
W/CuxO (23 nm)/Cu were analyzed with EDS.)

Table 1. EDS Quantitative Results for CuxO (35 nm).
Elemen Weight
tO
%
72.52
Cu
27.48

Atomic Net Int.
%
91.29 279.32
8.71
39.24

Net Int.
Error
0.02
0.11

Table 2. EDS Quantitative Results after 22 Days.
Elemen Weight
tO
%
73.63
Cu
26.37

Atomic Net Int.
%
91.73 183.03
8.27
23.97

Net Int.
Error
0.02
0.13

Table 3. EDS Quantitative Results for CuxO (23 nm).
Elemen Weight
tO
%
73.24
Cu
26.76

Atomic Net Int.
%
91.57 329.57
8.43
44.25

Net Int.
Error
0.01
0.11

Table 4. EDS Quantitative Results after 22 Days.
Elemen Weight Atomic Net Int. Net Int.
t
%
%
Error
O
74.63
92.12 159.04 0.02
Cu
25.37 7.88
19.51 0.15

72

4.4.2 Progression of Resistance with Time under Exposure
Resistances were computed from current and voltage data that were
collected before and after exposure to ambient oxygen. Fig. 40 shows plots of the
resistance versus voltage, as opposed to current versus voltage, to better illustrate
the qualitative effects on resistance states. (These plots are of the W/CuxO (11
nm)/Cu device.) Exposure to oxygen over the course of 48 days has changed the
shape of the original plot, yielding an average increase in resistance states. Fig.
41 is a plot of the low resistance state only and between +-2.1 V, which is the
setting voltage for this particular memristor. Similarly, exposure to oxygen over the

Figure 39. Plot of resistance versus applied voltage before and after exposure
to ambient oxygen.

73

Figure 40. Closeup of low resistance state (LRS).

Figure 41. Comparison of LRS states of
W/CuxO(35nm)/Cu device.
course of 23 days correlates with an increase in low resistance states for the
W/CuxO (35 nm)/Cu device in Figure 42.

74

The results of exposure to ambient oxygen as a function of time are shown
numerically in Tables 5 and 6 for the two memristors. The low resistance states
increased by 6.8 %, 2.1 %, and 0.6 % for the 11-nm, 23-nm, and 35-nm
memristors, respectively, after three days. The devices were measured again after
three to six weeks and revealed an increase of 31.3 %, 24.2 %, and 28.2 %,
respectively, from their original values. These increases in resistance states are
believed to be the result of ionically adsorbed oxygen in ambient air to the copper
oxide surface, which is then followed by the filling of oxygen vacancies in the layer
[14-15]. These results also indicate that the adsorption of oxygen was not
immediate but rather progressed over time.
Table 5. Progression of low resistance states with time
CuxO

Forming
Day 1

Resistance
Day 3
Day 23-48

thickness

voltage

11 nm

+/-2.1 V

8990 W

9600 W

11800 W (day 48)

23 nm

+/- 3.8 V

48300 W

49300 W

60000 W (day 23)

35 nm

+/- 4 V

69900 W

70300 W

89600 W (day 23)

Table 6. Percentage change in low resistance states with time
CuxO thickness

Forming voltage

Resistance
3 days 23 days 48 days

11 nm

+/-2.1 V

6.8 %

--

31.3 %

23 nm

+/- 3.8 V

2.1 %

24.2 %

--

35 nm

+/- 4 V

0.6 %

28.2 %

--

75

4.5 SUMMARY
In summary, memristors were designed, fabricated, and tested for passive
sensing of ambient oxygen at room temperature. Comparisons of resistance states
before and after exposure to ambient oxygen demonstrate increases with time (i.e.,
days). EDS profiles are consistent with the adsorption of oxygen to the copper
oxide switching layers with time. Early results also indicate an initial dependence
of sensitivity on thickness of the copper-oxide switching layer, but more
experiments are required to further investigate this observation. Future work will
also research the oxygen concentration of the switching layer with resistance
changes more quantitatively as well as the effect of temperature on sensing. In
conclusion, we have demonstrated a step towards the use of a memristor as a
passive sensor by taking advantage of the device’s ability to memorize (or record)
historical information.

76

REFERENCES

77

REFERENCES

1. C. Nyenke and L. Dong, “Sensing ambient oxygen using a W/CuxO/Cu
memristor,” IEEE 10th International Conference on Nano/Micro Engineered and
Molecular Systems, pp. 254-258, Apr. 2015. Š 2015 IEEE.
2. J. Tedesco, L. Stephey, M. Hernandez-Mora, C. Richter, and N. GergelHackett, "Switching mechanisms in flexible solution-processed TiO2 memristors,"
Nanotechnology, vol. 23, art. no. 305206, Aug. 2012.
3. F. A. Chudnovskii, L. L. Odynets, A. L. Pergament, and G. B. Stefanovich,
“Electroforming and switching in oxides of transition metals: The role of metalinsulator transition in the switching mechanism,” Journal of Solid State
Chemistry, vol. 122, pp. 95-99, 1996.
4. F. Verbakel, S. C. J. Meskers, R. A. J. Janssen, H. L. Gomes, M. CĂślle, M.
Büchel, and D. M. de Leeuw, “Reproducible resistive switching in nonvolatile
organic memories,” Applied Physics Letters, vol. 91, art. no. 192103, Nov. 2007.
5. A. Rose, “Space-Charge-Limited Currents in Solids,” Physical Review, vol. 97,
no. 6, pp. 1539-1544, Mar. 1955.
6. Y-J. Huang, S-C. Chao, D-H. Lien, C-Y. Wen, J-H. He, et al., “Dual functional
memory and threshold resistive switching based on the push-pull mechanism of
oxygen ions,” Scientific Reports, vol. 6, art. no. 23945, Apr. 2016.
7. R. Waser and M. Aono, “Nanoionics-based resistive switching memories,”
Nature Materials, vol. 6, pp. 833-840, Nov 2007.
8. Y. Yang, S. Choi, and W. Lu, “Oxide heterostructure resistive memory,”
NanoLetters, vol. 13, pp. 2908-2915, May 2013.
9. S. Kim, S. Choi, and W. Lu. “Comprehensive physical model of dynamic
resistive switching in an oxide memristor,” ACS Nano, vol. 8, no. 3, pp. 23692376, Feb. 2014.
10. H. Z. Zhang, D. S. Ang, C. J. Gu, K. S. Yew, X. P. Wang, et al., “Role of
interfacial layer on complementary resistive switching in the TiN/HfOx/TiN
resistive memory device,” Applied Physics Letters, vol. 105, art. no. 222106, Dec.
2014.

78

11. H. Z. Zhang, D. S. Ang, K. S. Yew, and X. P. Wang, “Enhanced stability of
complementary resistance switching in the TiN/HfOx/TiN resistive random-access
memory device via interface engineering,” Applied Physics Letters, vol. 108, art.
no. 083505, Feb. 2016.
12. Y. Yang, P. Sheridan, and W. Lu, “Complementary resistive switching in
tantalum oxide-based resistive memory devices,” Applied Physics Letters, vol.
100, art. no. 203112, May 2012.
13. S. Brivio, J. Frascaroli, and S. Spiga, “Role of metal-oxide interfaces in the
multiple resistance switching regimes of Pt/HfO2/TiN devices,” Applied Physics
Letters, vol. 107, art. no. 023504, Jul. 2015.
14. H. Freund, “Metal oxide surfaces: electronic structure and molecular
adsorption,” Basic Solid State Physics (Physica Status Solidi B), vol. 192, pp.
407-440, Dec. 1995.
15. A. Goswami, Thin Film Fundamentals, first ed. New Delhi: New Age
International Pvt Ltd Publishers, 1996.

79

CHAPTER 5 INVESTIGATION OF W/CuxO/Cu MEMRISTORS WITH
SUB-MICRON HOLES FOR PASSIVE SENSING OF OXYGEN

5.1 INTRODUCTION
This chapter, which references publication [1], discusses the influence of
surface area on passive sensing of oxygen. Two sets of memristors were
fabricated with sub-micron holes etched into the top W electrodes to reveal the
oxide surface. One set consisted of fifty-two holes each 300 nm in diameter. The
second set consisted of sixteen holes each 700 nm in diameter. Additionally, a
larger crossbar array (10x10) was constructed to increase memristor yield per
substrate. Each substrate was then exposed to ambient oxygen at 180 degrees
Celsius to induce passive sensing in minutes.

5.2 EXPERIMENTAL SETUP
Electrical measurements of each testable device were acquired using a
Keithley 6487 Picoammeter. The copper bottom electrode was grounded while a
bias voltage was applied to the tungsten top electrode for this setup. In order to
bring about passive oxygen sensing, the substrate was placed inside a thermal
oven and subjected to 5-25 min of surrounding air at a working temperature of 180
°C. (The devices were not tested at other temperatures or oxygen pressures since
this process was irreversible.) The electrical properties were collected at the

80

beginning and end of the experiment then analyzed for each of the devices. Atomic
composition analyses were performed, as well, by energy-dispersive X-ray
spectroscopy (EDS) inside a Carl Zeiss Variable Pressure scanning electron
microscope (SEM) EVO LS25.

Figure 42. Measurement setup.

81

5.3 DISCUSSION OF RESULTS
5.3.1 Characterizing Memristive Behavior
Figs. 44-51 show current-voltage (IV) plots for each nano mesh (Figs. 4446), micromesh (Figs. 47-49), and devices without meshes (Figs. 50-51) before
sensing. All the devices exhibit visually similar thin hystereses, which is indicative
of similar levels of memristance. After the substrate is subjected to 25 minutes of
ambient air at the working temperature of 180 deg C, each set of devices behaves
differently from the other as will now be discussed.

Figure 43. Current-voltage characteristics of nano-mesh before testing.

82

Figure 44. Current-voltage characteristics of nano-mesh before testing.

Figure 45. Current-voltage characteristics of nano-mesh before testing.

83

Figure 46. Current-voltage characteristics of micro-mesh before testing.

Figure 47. Current-voltage characteristics of micro-mesh before testing.

84

Figure 49. Current-voltage characteristics of micro-mesh before testing.

Figure 48. Current-voltage characteristics of non-mesh before testing.

85

Figure 50. Current-voltage characteristics of non-mesh before testing.

5.3.2 Sensing of Ambient Oxygen under Heat
In Fig. 53, IV plots for the nano meshes and the devices without meshes
are shown after 25 minutes in heated ambient air (black) and again after an
additional 5 minutes of exposure (red). In Fig. 53(a) and 53(b), the devices without
meshes lose their memristance after the 25-minute exposure and become strictly
resistive in behavior. Additionally, the overall resistance of the device increases
from 25-27 kilo-ohms to 89-107 kilo-ohms. Other devices without meshes
displayed similar resistive behavior at this juncture.
After another 5 minutes of exposure to ambient air at 180 deg C, the
resistances of these same devices without meshes remain virtually unchanged, as

86

shown by the overlapping plots in red. This is to be expected considering that the
diffusion of oxygen into the sides of the switching layer has a depth limit well under
the width of the CuxO. EDS analyses of these devices, which are shown in Table
6, show a 1.03%+-0.05% increase in oxygen composition after the total 30 minutes
of exposure. (The data in Table 6 are unpaired.) This slight increase in composition
is not unreasonable given that the switching layer is not completely enclosed by
the top W layer. More importantly, this measurement is used as a type of control
for comparison to the mesh devices. (The oxygen content includes that of the
SiO2/Si substrate.)
Table 7. EDS Results - Oxygen Composition for Non-Mesh
O	%	
(before)	 O	%	(after)	
30.22	
30.62	
27.25	
30.1	
30.05	
28.72	
29.62	
31.83	
AVERAGE	 AVERAGE	
=	29.285	 =	30.3175	

Error	%	
<=0.05	
<=0.05	
<=0.05	
<=0.05	
	

On the other hand, the nano meshes in 52(c) and 52(d) maintain
memristance after being subjected to 25 minutes of ambient air, thus making
memristive sensing feasible. (The devices also increase in overall resistance from
28-31 kilo ohms to 128-240 kilo-ohms.) After an additional 5 minutes of exposure
to ambient air at the operating temperature of 180 deg C, distinguishable shifts in
hystereses are observed for these same nano meshes (in red). For the case of the

87

Figure 51. Current-voltage plots for two non mesh devices in (a) and (b) and for
two nano mesh devices in (c) and (d). The black data points are after substrate
has been subjected to ambient air at 180 deg C for 25 minutes while the red is
after an additional five minutes.
device in 52(c), the high resistance state (HRS) and low resistance state (LRS)
increase up to 74.8% and 25.0%, respectively. Meanwhile, for the case of the nano
mesh in 53(d), the HRS increases by up to 72.4% while the LRS increases by up
to 22.8%. EDS analyses further show a 1.45%+-0.05% increase in oxygen
composition, which is higher than that of the control group (i.e., the devices without
meshes). (The data in Table 7 are unpaired.)

88

Table 8. EDS Results - Oxygen Composition for Nano-Mesh
O	%	
(before)	 O	%	(after)	
31.69	
32	
31.46	
32.49	
31.5	
33.29	
31.4	
34.06	
AVERAGE	 AVERAGE	
=	31.5125	
=	32.96	

Error	%	
<=0.05	
<=0.05	
<=0.05	
<=0.05	
	

For further comparison, Fig. 53 and 54 show current-voltage (IV) plots for
two micro mesh devices that were subjected to ambient air at the working
temperature of 180 deg C. Initially, the micro-mesh device measured in 53(a)
displays slight resistive switching (black). After 25 minutes of exposure to ambient
air at 180 deg C, this same micro mesh becomes capacitor-like, with a resistance
on the order of 1e10 ohms. Fig. 53(b) is a single plot of the micro mesh after
exposure for better clarity of this capacitance. In Fig. 54(a), we show a second

Figure 52. Current-voltage plots for a micro mesh. The black data points are
before testing while the blue is after substrate has been subjected to ambient
air at 180 deg C for a total of 30 minutes. The plot in (a) is an overlay of both
sets of data while (b) is a closeup of the latter set.

89

Figure 53. Current-voltage plots for a second micro mesh. The black data
points are after substrate has been subjected to ambient air at 180 deg C for
25 minutes while the blue is after an additional five minutes. The plot in (a) is
an overlay of both sets of data while (b) is a closeup of the latter set.
micro mesh which maintained slight memristance (black) after 25 minutes in
heated hair. However, after an additional five minutes of exposure, this second
micro mesh shifts from slight resistive switching (black) to capacitor-like (blue). All
other micro meshes tested on the substrate behaved similarly. It is believed that
these micro meshes develop a capacitance and high resistance after exposure to
ambient air at 180 deg C because the CuxO switching layer becomes
oversaturated with oxygen. Atomic composition results from EDS analyses show
a 1.96%+-0.05% increase in oxygen composition after exposure, which is higher
than those of both the control group and the nano meshes. (The data in Table 8
are unpaired.)

90

Table 9. EDS Results - Oxygen Composition for Micro-Mesh
O	%	
(before)	 O	%	(after)	
31.75	
33.22	
31.92	
33.29	
31.8	
33.57	
31.52	
34.73	
AVERAGE	 AVERAGE	
=	31.7475	 =	33.7025	

Error	%	
<=0.05	
<=0.05	
<=0.05	
<=0.05	
	

5.3.3 Sensing Mechanism – Adsorption and Diffusion
Thin film metal oxides can undergo adsorption and diffusion of oxygen [26] in the presence of ambient air. As demonstrated in Fig. 55(a), oxygen is first
ionically adsorbed (O2 = O2-+ h+) onto the surface from the surrounding
environment [3], [6]. We can write the change in oxygen concentration C with time
across a thin region of thickness dx at a depth x from the CuxO surface as the
difference between the flux F of oxygen per unit area entering the region and that
leaving the region [5]
𝜕𝐶
𝑑𝑥 = 𝐹 𝑥 − 𝐹 𝑥 + 𝑑𝑥 .																																																																																														(12)
𝜕𝑡
A Taylor-series expansion then allows us to approximate the last term as

𝐹 𝑥 + 𝑑𝑥 ≈ 𝐹 𝑥 +

𝜕𝐹
𝑑𝑥,																																																																																							(13)
𝜕𝑥

which allows us to rewrite the change in oxygen concentration as

91

𝜕𝐶(𝑥)
𝜕𝐹
≈ − .																																																																																																																				(14)
𝜕𝑡
𝜕𝑥
Fick’s first law states that the flux density is proportional to the concentration
gradient, where D is the diffusivity:

𝐹 ≈ −𝐷

𝜕𝐶
																																																																																																																												(15)
𝜕𝑥

Combining (14) and (15), we obtain Fick’s second law
𝜕𝐶
𝜕7𝐶
≈ 𝐷 7 ,																																																																																																																									(16)
𝜕𝑡
𝜕𝑥
which states that diffusion causes the concentration in the layer to change with
time. In our case, it is the diffusion of ionically adsorbed oxygen that causes the
oxygen concentration in the CuxO layer to increase with time. These oxygen ions
can then encounter and fill nearby vacancy sites [3, 5, 7-8], as illustrated in Fig.
55(b). It is believed that as the concentration of vacancies is reduced, fewer
conductive filament branches may be re-healed during the switching process as
modeled in Fig. 55(c). Consequently, the conductivity of the switching layer
decreases (or the resistance states increase) which agrees with our results.
As for the memristance differences between the nano meshes (300- nm
holes) and the micro meshes (700-nm holes), we look to both the number and size
of holes for each set of devices. In the case of the nanomeshes, there were 52
holes per mesh making the total area of exposed CuxO surface for each memristor

92

3.68 Îźm2. In the case of the micro meshes, there were 16 holes per mesh making
the total area of exposed CuxO surface 6.52 Îźm2. Consequently, the 700-nm
meshes expose 1.77 times (nearly twice) more Cu O surface to oxygen than the
x

300-nm meshes. Hence, more oxygen can be adsorbed and diffused into the 700nm micro mesh devices.

Figure 54. Schematic of adsorption and vacancy filling. (a) Oxygen molecules
from the ambient air result in adsorbed oxygen ions and holes near the CuxO
surface. (b) Over time, the adsorbed ions fill oxygen vacancies within the layer.
(c) The memristor now has higher resistance states due to fewer vacancies.

93

Other than oxygen content, the detailed arrangement of conductive
filaments under the tungsten electrode may have an impact on the resistance
difference seen between the micro meshes and nano meshes. Two previous
studies demonstrate that conductive filaments are few and localized rather than
homogeneous throughout the entire switching layer [9,10] after traditional
electroforming. Given the formation of a few local filaments under or near the
meshes (rather than strictly outside of the meshes), a conductive filament near a
700-nm hole would experience more exposed CuxO surface than one near a 300nm hole, as shown in Fig. 56.

Figure 55. Schematic of (a) nano mesh device and (b) micro mesh device. The
micro mesh design experiences more adsorbed oxygen ions.

94

5.6 SUMMARY
In summary, we have demonstrated the use of memristors for passive
sensing by taking advantage of the device’s ability to memorize (or record)
historical information. There was a notable increase in HRS and LRS when
memristors that were designed with 300-nm holes were exposed to ambient
oxygen for five minutes at 180 deg C. The mechanism by which this occurred is
believed to be the filling of vacancy sites within the switching layer by adsorbed
oxygen ions. EDS profiles, which show increases in oxygen composition, are
consistent with this adsorption and incorporation of oxygen into the CuxO layer.
Additionally, there is evidence to suggest that the size and number of holes is
important. In the case of devices with 700-nm holes, the memristance was
replaced by capacitance and high resistance after exposure to ambient air at 180
deg C. This observation is believed to be the result of oversaturation of oxygen
into the CuxO layer, and EDS profiles support this. There is potential application
of passive “memsensors” as sensing alternatives for devices which typically have
no batteries integrated due to their small sizes. Such devices may include
micro/nanorobots such as catalytically-driven microjets [11] or artificial bacteria
flagella [12] designed for medical applications. Additionally, this work increases
understanding of environmental effects on memristors, including tunability.

95

REFERENCES

96

REFERENCES

1. C. Nyenke and L. Dong, “Fabrication of a W/CuxO/Cu memristor with submicron holes for passive sensing of oxygen,” Microelectronic Engineering, vol.
164, pp. 48-52, Oct. 2016.
2. A. Goswami, Thin Film Fundamentals, first ed. New Delhi: New Age
International Pvt Ltd Publishers, 1996.
3. H. Freund, “Metal oxide surfaces: electronic structure and molecular
adsorption,” Basic Solid State Physics (Physica Status Solidi B), vol. 192, pp.
407-440, Dec. 1995.
4. P. R. F. Rocha, A. Kiazadeh, D. M. De Leeuw, S. C. J. Meskers, F. Verbakel,
D. M. Taylor, and H. L. Gomes, “The role of internal structure in the anomalous
switching dynamics,” Journal of Applied Physics, vol. 113, art. no. 134504, Apr.
2013.
5. R. S. Muller, T. I. Kamins, and M. Chan, Device Electronics for Integrated
Circuits. New York: Wiley & Sons, Inc., 2003, pp. 84-86.
6. G. F. Fine, L. M. Cavanagh, A. Afonja, and R. Binions, “Metal oxide semiconductor gas sensors in environmental monitoring,” Sensors, vol.10, pp. 54695502, Jun. 2010.
7. J. Tedesco, L. Stephey, M. Hernandez-Mora, C. Richter, and N. GergelHackett, "Switching mechanisms in flexible solution-processed TiO2 memristors,"
Nanotechnology, vol. 23, art. no. 305206, Aug. 2012.
8. D.H. Kwon, K.M. Kim, J.H. Jang, J.M. Jeon, M.H. Lee, G.H. Kim, X.S. Li, G.S.
Park, B. Lee, S. Han, M. Kim, and C.S. Hwang, “Atomic structure of conducting
nanofilaments in TiO2 resistive switching memory,” Nature Nanotechnology, vol.
5, no. 2, pp. 148-153, Feb. 2010.
9. R. Dittmann, R. Muenstermann, I.K. D. Park, T. Menke, J. Mayer, A. Besmehn,
F. Kronast, C. Schneider, and R. Waser, “Scaling potential of local redox
processes in memristive SrTiO3 thin-film devices,” Proceedings of the IEEE 100,
vol. 6, pp. 1979-1990, 2012.
10. Y. Yang, P. Gao, S. Gaba, T. Chang, X. Pan, and W. Lu, “Observation of
conducting filament growth in nanoscale resistive memories,” Nature
Communications, vol. 3, art. no. 732, Mar. 2012.

97

11. V. Magdanz, M. Guix, and O. G. Schmidt, “Tubular micromotors: from
microjets to spermbots,” Robotics and Biomimetics, vol. 1, no. 11, pp. 1-10, Oct.
2014.
12. L. Zhang, J. J. Abbott, L. Dong, B. E. Kratochvil, D. Bell, and B. J. Nelson,
“Artificial bacterial flagella: Fabrication and magnetic control,” Applied Physics
Letters, vol. 94, art. no. 064107, Feb. 2009.

98

CHAPTER 6 EFFECT OF NITROGEN DIOXIDE AND AMMONIA ON
RESISTIVE SWITCHING

6.1 INTRODUCTION
This final work, which references publication [1], set out to investigate the
influence of oxidizing and reducing gases on the resistive switching behavior of a
CuxO-based memristor. This capability has the potential to extend memristor
research to reversible gas sensing in the near future. While traditional resistive gas
sensors only respond with a change in resistance, the memristor can act as an
“enhanced” sensor in that it may respond with changes in resistance states, and
in turn hysteresis gap. For this study, the memristive characteristics before, during,
and after exposure to nitrogen dioxide (NO2) — an oxidizing gas — and ammonia
(NH3) — a reducing gas — is examined.

6.2 EXPERIMENTAL SETUP
Current and voltage measurements were collected with a Keithley 6487
Picoammeter and a probe setup in which the bottom Cu electrode was grounded
and a bias voltage applied to the top W electrode. The first substrate was directly
subjected to 10 then 20 ppm of NO2 through a 2-mm diameter tube connected to
a gas source in order to determine NO2 detection. A similar setup was used for the
second substrate in detecting 100 ppm of NH3. Both tests were performed at room

99

temperature, and electrical measurements were collected before, during, and after
gas exposure. These measurements were then compared and analyzed.
Additionally, sensitivity of the device to stimuli was quantified by dividing the
resistances after exposure by the resistances before exposure.

6.3 DISCUSSION OF RESULTS
6.3.1 Evaluation of NO2 Detection
Figs. 57 and 58 shows current-voltage plots for the first device – Fig. 57 is
prior to NO2 exposure, Fig. 62(a)-(d) is after exposure has ceased, and Fig. 58(e)(f) upon full recovery. Prior to gas exposure, the memristor demonstrates a
consistent pinched hysteresis in two consecutive overlapping voltage sweeps
shown in Fig. 57. After 10 minutes of exposure to 20 ppm, there is then a notable
decrease in resistance state values. Additionally, the hysteresis collapses, or
becomes thinner. After the NO2 source has been shut off, the resistance states
remain this way for an additional 5 minutes, as shown in Fig. 58(a). Then 10
minutes after shutoff, the device begins to gradually recover as observed in the
slight widening of the hysteresis depicted in Fig. 58(b). In Fig. 58(c) and (d), we
observe further widening of the hystereses coupled with increases in resistance
states while electrical measurements continue to be taken at 15 minutes, and 20
minutes, respectively, after gas exposure has ceased. It is at 25 minutes after
shutoff — shown in Fig. 62(e) — that the original current-voltage characteristics

100

appear. The complete recovery is confirmed at 30 minutes after shutoff, which is
depicted in Fig. 58(f).

Figure 56. IV characteristics before NO2 exposure.

101

Figure 57. Current-voltage plots for device #1 after 20 ppm NO2 exposure. Plots
(a-f) demonstrate gradual recovery of the device at five-minute intervals. The
final plot in (f) is 30 minutes after exposure.

102

The gas response of the device, which was calculated by the resistance gap
in ambient air divided by the resistance gap in 20 ppm NO2, is 44.5 at 3.5 V and
154.0 at -3.5 V. (These resistance gap values are shown in Tables 10 and 11.)
Table 10. Nitrogen dioxide detection of device #1 (negative bias)
NO2 (negative
bias)

Low resistance
state (at -3.5V)

High resistance Resistance gap
state (at -3.5V)
(HRS minus
LRS)

before

1.43E+05 W

4.96E+05 W

3.53E+05 W

after

9.30E+04 W

9.53E+04 W

2.30E+03 W

Table 11. Nitrogen dioxide detection of device #1 (positive bias)
NO2 (positive
bias)

Low resistance
state (at 3.5V)

High resistance Resistance gap
state (at 3.5V)

before

1.76E+05 W

2.57E+05 W

8.02E+04 W

after

9.47E+04 W

9.65E+04 W

1.80E+03 W

6.3.2 Evaluation of NH3 Detection
In Fig. 59, current-voltage plots depict an opposite trend when the second
device is exposed to 100 ppm of NH3 at room temperature. Initially there is a slight
hysteresis prior to detection, as shown in Fig. 63(a), but after 20 minutes of
exposure to NH3 the memristor begins to respond by a widening of the hysteresis
gap. Fig. 59(b) shows the device's response after 50 minutes of gas exposure,
where the hysteresis widens to its maximum. The resistance states also decrease,
as shown in Fig. 59(b) as well as in Tables III and IV. When exposure to NH3

103

ceases, electrical measurements continue to be recorded for an additional 30
minutes, at which the memristor demonstrates recovery as shown in Fig. 59(c).
The gas response of the device, which was calculated by the resistance gap in
100 ppm NH3 divided by the resistance gap in ambient air, is 43.2 (for the negative
voltage region). These values are also shown in Tables 12 and 13.

Table 12. Ammonia detection of device #2 (negative bias)
NH3 (negative
bias)

Low resistance
state (at -2.5V)

High resistance
state (at -2.5V)

Resistance gap

before

2.43E+09 W

2.46E+09 W

3.56E+07 W

after

5.15E+08 W

2.05E+09 W

1.54E+09 W

Table 13. Ammonia detection of device #2 (positive bias)
NH3 (positive
bias)

Low resistance
state (at 2.5V)

High resistance
state (at 2.5V)

Resistance
gap

before

2.43E+09 W

2.46E+09 W

3.56E+07 W

after

4.07E+08 W

3.02E+12 W

3.02E+12 W

104

Figure 58. Current-voltage plots for device #1. The three plots are an
overlay of two voltages sweeps (a) prior to exposure to 100ppm NH3,
(b) after 50 and 55 minutes of exposure, and (c) after 30 and 35
minutes of recovery.

105

6.4.3 Mechanism of Gas Detection
The mechanism by which NO2 detection occurs is related to the surface
effects of the metal oxide switching layer [2-3] and illustrated in Fig. 60. After device
fabrication, oxygen molecules from the ambient air adsorb to the CuxO surface and
extract electrons from the conduction band forming ionically adsorbed oxygen
(O2(ads)- = O2(ads) + e-) [3-5]. This results in an accumulation of holes near
the surface which is further increased in the presence of NO2, an oxidizing gas.
(NO2 takes electrons from the adsorbed oxygen ions and CuxO surface.) These
chemical reactions are as follows [3-5]:
C
C
𝑁𝑂7(E) + 𝑂7(?@A)
→ 𝑁𝑂7(?@A)
+ 𝑂7(E) 																																																																																	(17)

C
𝑁𝑂7(?@A) + 𝑒 C → 𝑁𝑂7(?@A)
																																																																																																	(18)

This increase in hole accumulation causes the p-type CuxO to increase in
conductivity (or decrease in resistance), which supports the observation made in
Fig. 4. What is not immediately clear is why the hysteresis gap also collapses
during this process. Once the NO2 source is cut off, there is an eventual recovery
of the original resistance states and hysteresis shape. This is because the
chemical process is reversed, and oxygen is readsorbed from the ambient air to
the CuxO surface [3-5].

106

Likewise, the mechanism for NH3 detection involves adsorbed oxygen ions
on the CuxO surface. Unlike NO2, NH3 is a reducing gas, thus giving up electrons
in the chemical process as demonstrated in the following reaction [6,7]:
C
𝑁𝐻=(E) + 3𝑂7(?@A)
	 → 6𝐻7 𝑂 + 2𝑁7 + 6𝑒 C 																																																																								(19)

This reaction results in the addition of electrons to the CuxO surface, which in turn
leads to a decrease in hole accumulation due to recombination. For p-type resistive
gas detectors and sensors, this process should translate to an increase in
resistance (or decrease in conductivity) [5,8]. However, the overall decrease in
resistance states for this memristor indicates that the deposited CuxO layer is
actually n-type rather than p-type. This suggestion is supported by previous
literature that points to the creation of n-type CuxO given a relatively high (>50%
PO2) [9] or low introduction of oxygen (e.g., Cu-rich) during deposition [9,10]. The
high initial resistance of this memristor (on the order of 109 ohms) versus the
resistance of the previous memristor (on the order of 105 ohms) shows evidence
of relatively high oxygen concentration in spite of both devices being the same
thicknesses.
In addition to these observations, exposure of the device to NH3 correlates
with a widening of the hysteresis gap as observed in Fig. 59. Further analysis is
necessary to determine the physical reasoning behind this phenomenon but

107

repeated tests confirm the association between NH3 exposure and increasing
distance between resistance states.

108

Figure 59. Depiction of gas detection mechanism.

109

6.6 SUMMARY
In conclusion, we have shown that we can achieve reversible gas detection
through the design of perforated memristors. In response to gas exposure, these
devices exhibit changes in low and high resistance states, which result in either a
collapse or widening of the hysteresis gap as well as a decrease in overall
conductivity. These observations suggest an association between the memristor
resistance states and exposure to an oxidizing (NO2) or reducing gas (NH3),
though further analysis is necessary to physically explain the influence on
hysteresis gap change. This work introduces the possibility of the memristor as an
“enhanced” resistive gas detector (and ultimately sensor) by providing additional
information within the shape and size of the hysteresis. This work also presents
the potential of tuning memristor characteristics after fabrication via external
stimuli, such as gases.

110

REFERENCES

111

REFERENCES

1. C. Nyenke and L. Dong, “Effect of NO2 and NH3 on the resistive switching
behavior of W/CuxO/Cu memristors,” Journal of Micromechanics and
Microengineering, vol. 27, no. 10, art. no. 105013, Oct. 2017. 10.1088/13616439/aa8672. Š IOP Publishing. Reproduced with permission. All rights
reserved.
2. N. Barsan, C. Simion, T. Heine, S. Pokhrel, and U. Weimar, “Modeling of
sensing and transduction for p-type semiconducting metal oxide based gas
sensors”, Journal of Electroceramics, vol. 25, pp. 11-19, June 2009.
3. K. Wetchakuna, T. Samerjaia, N. Tamaekonga, C. Liewhirana, C. Siriwonga,
et al., “Semiconducting metal oxides as sensors for environmentally hazardous
gases,” Sensors and Actuators B: Chemical, vol. 160, pp. 580-591, Aug. 2011.
4. G. F. Fine, L. M. Cavanagh, A. Afonja, and R. Binions, “Metal oxide
semiconductor gas sensors in environmental monitoring,” Sensors, vol. 10, pp.
5469-5502, Jun. 2010.
5. A.A. Mane, M.P. Suryawanshi, J.H. Kim and A.V. Moholkar, “Fast response of
sprayed vanadium pentoxide (V2O5) nanorods towards nitrogen dioxide (NO2)
gas detection,” Applied Surface Science, vol. 403, pp. 540–550, May 2017.
6. A. Sharma, P. Bhojane, A. K. Rana, Y. Kumar, and P. M., Shirage,
“Mesoporous nickel cobalt hydroxide/oxide as an excellent room temperature
ammonia sensor,” Scripta Materialia, vol. 128, pp. 65-68, Feb. 2017.
7. I. Jiménez, M. A. Centeno, R. Scotti, F. Morazzoni, A. Cornet, et al., “NH3
interaction with catalytically modified nano-WO3 powders for gas sensing
applications,” Journal of The Electrochemical Society, vol. 150, pp. 72-80, Feb.
2003.
8. S. Bhuvaneshwari, N. Gopalakrishnan, “Hydrothermally synthesized copper
oxide (CuO) superstructures for ammonia sensing,” Journal of Colloid and
Interface Science, vol. 480, pp. 76-84, Jul. 2016.
9. V. Figueiredo, E. Elangovan, R. Barros, J. V. Pinto, T. Busani, et al., “p-Type
CuxO films deposited at room temperature for thin-film transistors,” Journal of
Display Technology, vol. 8, pp. 41-47, Jan. 2012.

112

10. S. H. Lee, M. Shin, S. J. Yun, and J. W. Lim, “CuOx/a-Si:H heterojunction
thin-film solar cell with an n-type μc-Si:H depletion-assisting layer,” Progress in
Photovoltaics, vol. 23, pp. 1642-1648, Nov. 2015.

113

CHAPTER 7 CONCLUSIONS AND FUTURE WORK

7.1 SUMMARY OF CONCLUSIONS
In this dissertation, CuxO-based memristors were designed, fabricated, and
investigated for passive sensing of oxygen as well as reversible sensing of nitrogen
dioxide (an oxidizing gas) and ammonia (a reducing gas) in consecutive studies.
The first study set out to determine the feasibility of such memristors for passive
sensing under ambient conditions [1]. It was then concluded that resistance states
demonstrated increases with time (i.e., days) while EDS profiles were consistent
with the adsorption of oxygen to the CuxO switching layers. The second study built
upon the first study by introducing heat and an improved memristor design – which
included 1) sub-micron holes in the tungsten top electrode and 2) more devices
per substrate – for further investigation of passive sensing [2]. From this study, it
was confirmed that resistance states and atomic composition both increase with
exposure to oxygen from the air and depend on the surface area of the exposed
switching layer. The mechanism by which this detection occurred was determined
to be the diffusion of adsorbed oxygen ions and the filling of local oxygen vacancies
over time. The final study set out to demonstrate the reversible detection of an
oxidizing gas (nitrogen dioxide) and a reducing gas (ammonia) by the CuxO-based
memristors [3]. In response to gas exposure, these devices exhibited decreases
in low and high resistance states, which resulted in either a collapse (or widening)

114

of the current-voltage hysteresis gap. The mechanism by which these resistance
state changes occurred was determined to be a reversible chemical reaction
between the oxidizing (or reducing) gas and the adsorbed oxygen ions on the
exposed switching layer surface. Ultimately, this dissertation demonstrated a
potential application of the memristor as a multifunctional tool for memory and gas
detection.

7.2 FUTURE RESEARCH
7.2.1 Further Understanding of Reversible Sensing Mechanism
For future work, it is suggested that more experiments be done to build upon
the reversible sensing of oxidizing and reducing gases with memristors. Further
analysis is necessary to physically understand the influence of oxidization and
reduction on the change in current-voltage hysteresis gap size. This analysis would
require several more test trials than was explored in this dissertation in order to
gather more data. These tests could incorporate various ranges (ppm) of gas
concentrations and, perhaps, different temperatures as well.

115

7.2.2 Different Switching Layer Composition and Design
Future work may also investigate other metal oxide materials – for example
nonstoichiometric forms of zinc oxide or tungsten oxide – for the switching layers.
This is because the choice of material could be made to greatly increase the
sensing capability and switching endurance of the device. Alternatively, fabrication
of the switching layer via the bilayer approach [4, 5, 6, 7] (as discussed in the
literature review) could further improve the device performance. This approach
may require a deposition method other than reactive sputtering to better control
the composition and structure of the second layer. Additionally, x-ray photoelectron
spectroscopy (XPS) or other analytical techniques could be employed to more
accurately examine the chemical state of the switching layer as well as determine
the empirical formula.

7.2.3 Influence of Memristor Area on Sensing
Lastly, construction of thinner electrodes in a crossbar array structure – and
thus, smaller memristors in terms of area – may be investigated for influence on
sensitivity to gas or chemical stimuli. Electron beam lithography (EBL) can be used
to fabricate nanoelectrodes for the device.

116

REFERENCES

117

REFERENCES

1. C. Nyenke and L. Dong, “Sensing ambient oxygen using a W/CuxO/Cu
memristor,” IEEE 10th International Conference on Nano/Micro Engineered and
Molecular Systems, pp. 254-258, Apr. 2015.
2. C. Nyenke and L. Dong, “Fabrication of a W/CuxO/Cu memristor with submicron holes for passive sensing of oxygen,” Microelectronic Engineering, vol.
164, pp. 48-52, Oct. 2016.
3. C. Nyenke and L. Dong, “Effect of NO2 and NH3 on the resistive switching
behavior of W/CuxO/Cu memristors,” Journal of Micromechanics and
Microengineering, vol. 27, no. 10, art. no. 105013, Oct. 2017. 10.1088/13616439/aa8672.
4. C.H. Huang, J. S. Huang, S. M. Lin, W. Y. Chang, J. H. He, and Y. L. Chueh,
“ZnO1-x nanorod arrays/ZnO thin film bilayer structure: from homojunction diode
and high-performance memristor to complementary 1D1R application,” ACS
Nano, vol. 6, no. 9, pp. 8407-8414, Aug. 2012.
5. S.-M. Lin, J.-S. Huang, W.-C. Chang, T.-C. Hou,. H.-W. Huang, C.-H. Huang,
S.- J. Lin and Y.-L. Chueh, “Single-step formation of ZnO/ZnWOx bilayer
structure via interfacial engineering for high performance and low energy
consumption resistive memory with controllable high resistance states,” ACS
Applied Materials & Interfaces, vol. 5, no. 16, pp. 7831-7837, Aug. 2013.
6. M.J. Lee, C. B. Lee, D. Lee, S. R. Lee, M. Chang, J. H. Hur, Y. B. Kim, C. J.
Kim, D. H. Seo, S. Seo, U. I. Chung, I. K. Yoo, and K. Kim, “A fast, highendurance and scalable non-volatile memory device made from asymmetric
TaO2O5-x /TaO2−x bilayer structures,” Nature Materials, vol. 10, no. 8, pp. 625-630,
Jul. 2011.
7. A. Prakash, S. Maikap, C. S. Lai, T. C. Tien, W. S. Chen, H. Y. Lee, F. T.
Chen, M.-J. Kao, and M.-J. Tsai, “Bipolar resistive switching memory using
bilayer TaOx/WOx films,” Solid-State Electronics, vol. 77, pp. 35-40, Nov. 2012.

118