Int. J. Electron. Commun.

(AEÜ) 90 (2018) 130–139

Contents lists available at ScienceDirect

Int. J. Electron. Commun. (AEÜ)

journal homepage: www.elsevier.com/locate/aeue

Regular paper

A versatile window function for linear ion drift memristor model – A new T
approach
⁎
T.A. Anusudha, S.R.S. Prabaharan
School of Electronics Engineering, Vellore Institute of Technology, Chennai, Vandalur-Kelambakkam Road, Chennai 600127, Tamil Nadu, India

A R T I C LE I N FO A B S T R A C T

Keywords: The memristor is a nano-scaled resistive switching device which is widely investigated in analog and digital
Memristor applications. We report here our success in formulating a new window function as applicable to linear ion drift
Boundary effect model. Accordingly, this paper identifies the demerits of other existing window functions and the requirement of
Boundary lock a versatile window function to mimic the current-voltage characteristics of a physical memristor device. The
Pinched hysteresis loop
proposed new window function overcomes the demerits of existing window functions such as boundary effect,
Scalability
boundary lock (with respect to frequency of operation), and the scalability. The main significance of proposed
Window function
model is to facilitate the nonlinearity in linear ion drift memristor model that produces the pinched hysteresis
loop (a signature of a memristor) for any typical applied voltage within the frequency range (0.05 ≤ f < 2 Hz).
The validation of which has been verified in a memristor based op-amp circuitry. It exhibits a high gain com-
pared to other existing models and produces low power dissipation compared to CMOS based op-amp.

1. Introduction versatility of our window function and compared its performance with
respect to other existing window functions with respect to LIMM.
Memristor is a missing fourth fundamental passive circuit element
which is conceived by Leon Chua in 1971 [1]. Memristive device based 2. Linear ion drift memristor model
on TiO2 was apparently first demonstrated by Strukov et al. [2]. Con-
sidering the resistive properties of memristor, it has become an element Memristor is defined as the rate of change of flux with respect to
of surprise for various analog applications such as chaotic circuit, os- charge [1], which is represented in Eq. (1)
cillators, filters, programmable analog circuit, sensors, cellular neural
dϕ = M (q) dq (1)
networks and alike [3–5] as well as digital applications viz., fuzzy
processor [6] and non-volatile memory devices [4,5]. To throw more where M(q) is the memristance.
insight into memristors, various models have been proposed in the past The memristor consists of a bilayer of titanium-di-oxide which is
focusing on obtaining Pinched Hysteresis Loop (PHL), a typical in- sandwiched between two platinum electrodes [2,13] as shown in Fig. 1.
dicator of memristive characteristics. Hitherto, modeling of memristor The bilayer consists of two layers, one is oxygen deficient titanium-di-
is being carried out by using Linear Ion Drift Memristor Model (LIMM) oxide (TiO2−x) layer, which offers On State Resistance (RON) and an-
[2], Non-Linear Ion Drift Memristor Model [7], Simmon Tunnel Barrier other is perfect titanium-di-oxide layer (TiO2) which offers Off State
Model [8] and Team Model [9]. Resistance (ROFF).
Each of these models has its own pros and cons. In LIMM, non- In memristor, the initial state of the memristance is high, which is
linearity is very less and assumes two conditions which are (i) uniform shown in Fig. 2(a). When the positive voltage is applied across the
electric field and (ii) the average mobility of ions [13]. To mitigate this memristor, the positively charged oxygen vacancies are drifted from
problem, window functions play an important role in LIMM such as TiO2−x to TiO2.Hence, the width of the defected titanium-di-oxide in-
Jogelker [10], Biolek [11], and Prodromakis [12]. In view of various creases, simultaneously total memristance decreases and current con-
setbacks in using these revealed models with respect to LIMM, we in- ductivity increases. Therefore, the memristor switches from OFF state
troduce a new window function which overcomes the boundary lock to ON state, which is illustrated in Fig. 2(b). This process is termed as
issues pertaining to low frequency which was not thought of before the completion of growth of the conducting filaments [14]. At this stage
according to our best knowledge. Accordingly, we have proven the the polarity is reversed, that is the negative voltage is applied at the Top

⁎
Corresponding author.
E-mail addresses: [email protected], [email protected] (S.R.S. Prabaharan).

https://doi.org/10.1016/j.aeue.2018.04.020
Received 2 February 2018; Accepted 20 April 2018
1434-8411/ © 2018 Elsevier GmbH. All rights reserved.
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

TiO2-X TiO2
Pt Pt

w
Fig. 1. Schematic structure of a bilayer memristor.

Electrode (TE) with respect to the Bottom Electrode (BE), the positively
charged oxygen vacancies are repelled from TiO2 to TiO2−x, the process Fig. 3. The pinched I-V loop obtained for linear ion drift memristor model, a
is termed as rupturing of the conducting filament [14]. As a result, the characteristic signature of a bilayer memristor.
width of the defected titanium-di-oxide decreases consequently af-
fecting the total memristance to increase and the current decreases. On the other hand, if the applied voltage is large enough, then internal
Thus, the memristor switches from ON state to OFF state as illustrated state variable (x) might reach the terminal boundaries faster and fixed
in Fig. 2(c). at the boundary (+) as stated before. This phenomenon is known as
The above said behavior is mathematically described in Eq. (2) for a boundary lock.
linear ion drift memristor model,

w (t ) w (t ) ⎞ ⎞ 3. Different types of window functions

V (t ) = ⎛RON
⎜ + ROFF ⎛1− I (t )
⎟

⎝ D ⎝ D ⎠⎠ (2)
Window functions are used to retain the ions within the boundaries
where w(t) is the width of the non-stoichiometric region, V(t) is the and mimic the properties of the physical memristor device. By com-
applied voltage, I(t) is the current flow through the memristor, and D is puting the parameters of the available window functions, it is possible
the distance between the two platinum electrodes. to use memristor model in different electronic applications.
The LIMM is simulated in Cadence® software with virtuoso tool for
the applied voltage of ± 2 V at 0.5 Hz. It produces the bow-tie like loop
the so-called PHL with less nonlinearity as shown in Fig. 3. In LIMM, it 3.1. Jogelker window function
is understood that boundary effect and boundary lock are the two main
prevailing problems. When the virtual boundary reaches to the metal/ The Jogelker window function [10] is represented in Eq. (3)
oxide interface, the rate of change of width (dw/dt) is suppressed to f (x ) = 1−(2x −1)2p (3)
zero. Consequently, the memristance is fixed to either of the resistance
states. That is, it is fixed either at minimum on-state resistance or where f(x) represents the window function, p is the positive integer and
maximum off-state resistance. This effect is known as boundary effect. x is the internal state variable.

Fig. 2. Different bias conditions of a memristor (a) Before bias (b) Positive bias and (c) Negative bias.

131
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

Fig. 4. Effect of p on nonlinear I-V curves (Jogelker window function). Fig. 6. I-V characteristic for various applied voltage (U) (Jogelker window
function).
3.1.1. Effect of p
In this model, f(x) value does not scale up and scale down for the
various value of p. Hence, the nonlinear I-V curves remain the same as
shown in Fig. 4.
When p value is large enough, then inverted parabolic curve of f(x)
is changed into a rectangular curve as shown in Fig. 5. Hence, this
model provides a linkage with linear ion drift memristor model without
window function.

3.1.2. Effect of applied voltage (U)

When an applied voltage increases, the amount of charges stored in
the width of the non-stoichiometric region increases. Therefore, the
nonlinearity of the memristor model increases as shown in Fig. 6. But, if
the applied voltage is large enough, then model fails to retain the x
within the interval (0 < x < 1) as shown in Fig. 7. Therefore, this
model fails to produce the PHL.
Fig. 7. Response of x, for U = ± 5 V, p = 2, and f = 0.5 Hz (Jogelker window
function).
3.2. Biolek window function

Biolek [11] resolves the problems of Jogelker window function with where p is the positive integer, and stp(−i) is the step current flow
high nonlinearity and scalability. through the memristor device, which is represented in Eq. (5).
The Biolek window function [2] is represented in Eq. (4).
3.2.1. Effect of p
f (x ) = 1−(x −stp (−i))2p (4)
In Biolek, the size of the nonlinear I-V curve is changed for the
various value of p as shown in Fig. 8. It occurs due to f(x) value being
stp (i) = { 1 i⩾0
0 i<0 (5)
scaled up and scaled down for the various values of p, unlike Jogelker.
This model provides the better scalability because it retains the f(x)
value within the range (0 < f(x) < 1). But, if the value of p is large

Fig. 5. Response of f(x) vs x, for various values of p with the same parameters Fig. 8. PHL for the various value of p and fixed value of U = ± 3 V, and
are used in Fig. 3. f = 0.5 Hz (Biolek window function).

132
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

Fig. 9. PHL for different applied voltages and fixed value of p = 2, U = ± 2 V, and f = 0.5 Hz (a) PHL (b) Distorted Pinched Hysteresis Loop (DPHL) for U = ± 6 V
(Biolek window function).

3.3.1. Effect of p and J

The both p and J are the control parameters, which are used to
control f(x) < 1 and f(x) ≥ 1 respectively. This model retains the
nonlinearity with high scalability for the large value of p unlike other
window functions as shown in Fig. 10. Because, f(x) value lies within
the range (0 < f(x) < 1) and if the value of p is large enough, then this
model also provides a linkage with LIMM without any window func-
tions.
Nevertheless, the factor J in Prodromakis window function plays an
important role in increasing the nonlinearity as shown in Fig. 11. If the
applied voltage is large enough then factor J controls the x within the
range (0 < x < 1) and produces the large size PHL with high non-
linearity. The main function of J is scaled up and scaled down the f(x)
and produces the PHL for f(x) ≥ 1. Therefore, this model can be used in
any memristor device applications with better performances.

3.3.2. Effect of applied voltage (U) and frequency (f)

A family of I-V curves is obtained for the applied voltages
Fig. 10. Current-voltage response of the model for different value of p of ± 2 V, ± 4 V, ± 6 V, ± 8 V as shown in Fig. 12. This model retains
(Prodromakis window function). the PHL up to the voltage of ± 10 V for frequency of 0.5 Hz, and dis-
tance between the two platinum electrodes of 3 nm. Therefore, the
capacity of retaining the nonlinearity is high compared to the previous
enough, then model provides a linkage with LIMM without window
models. However, this model fails to retain the PHL for the frequency
functions.
(f) < 0.1 Hz, and applied voltages of ± 2 V due to x reaching the
boundaries much faster leading to distorted PHL. Therefore, the flex-
ibility of this model is reduced as illustrated in Fig. 19(c) for the fre-
3.2.2. Effect of applied voltage (U)
quency of 0.05 Hz.
A family of I-V curves is obtained for various applied voltages and
particular p as shown in Fig. 9(a). Here, the nonlinear dopant kinetic is
high due to introducing the step current in the window function, unlike 3.4. Jinxiang window function
Jogelker. But, this model fails to retain the PHL for high applied vol-
tages (U ≥ ± 6 V) as shown Fig. 9(b). Because, the Biolek window Jinxiang et al., window function [15] is expressed in Eq. (7) which
function does not have a control parameter to control the x within the is capable of addressing the aforementioned issues.
limited range of 0 and 1.
f (x ) = J (1−[0.25(x −stp (−i))2 + 0.75] p ) (7)

where p is the positive real number, J is the scaling parameter.

3.3. Prodromakis window function

The Prodromakis window function [12] is expressed in Eq. (6) 3.4.1. Effect of p and J
which overcame the problems of Biolek with improved nonlinearity and The Jinxiang et al., window function provides a high nonlinearity
scalability. The most significant function of this window is retaining the and scalability for the different value of p as shown in Fig. 13. It should
PHL for f(x) ≥ 1. be noted that the window function retains the f(x) value within the
range of 0 to 1. In addition, when value of p is large enough the model
f (x ) = J (1−[(x −0.5)2 + 0.75] p ) (6) provides a linkage with LIMM without a window function. On the other
hand, the J factor controls the x within the limited range of 0 to 1 and
where p and J are the control parameters. scale up/down the f(x) from its maximum value of 1.

133
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

Fig. 11. Nonlinear I-V curves for fixed U = ± 2 V, f = 0.5 Hz, and various J
(Prodromakis window function). Fig. 13. Current-voltage response of the model for different value of p (Jinxiang
et al., window function).

3.4.2. Effect of applied voltage (U) and frequency (f)

The model retains the PHL for p = 2, J = 1, f = 0.5 Hz and limited functions with high performances. The following validation ensures the
range of input voltages( ± 2 V ≤ U ≤ ± 17 V). When the input voltage success of this new proposed window function.
is (U) ≥ 18 V, the f(x) value does not remain at one. Therefore, the
model produces DPHL as shown in Fig. 14.As a result, this model fails to 3.5.1. Effect of p and J
retain the high nonlinearity. On the other hand, this model also fails to Interestingly, for any value of p, our new model incorporating the
retain the PHL for the frequency (f) of 0.05 Hz, and the applied voltage proposed window function transforms the linear ion drift model into
of ± 2 V due to x reaching the boundaries much faster leading to dis- nonlinear ion drift memristor model producing the large size PHL as
torted PHL which is also illustrated in Fig. 19(d). shown in Fig. 15. The reason being is that f(x) is controlled by the p
factor and more importantly it does not fix to any particular integer
3.5. Proposed window function constant as a Jogelker window function. Therefore, it increases the
scalability significantly.
Keep the above stated setbacks in the existing window functions; we Fig. 16 exhibits the effect of our model which produces the PHL
have proposed and defined a new window function by considering the with diminished DNDR (Dynamic Negative Differential Resistance) for
cubic parabola as given in Eq. (8). higher applied voltage (as high as ± 50 V). On the other hand, one can
also achieve a symmetric PHL by controlling the J factor. Obviously, the
f (x ) = J [1−2(x 3−x + 1) p] (8) J factor plays an important role in controlling the x within the range
from 0 to 1.
where f(x) is the window function, x is the internal state variable, p is
the positive integer, and J is the control parameter.
3.5.2. Effect of applied voltage (U)
Due to this consideration in our window function, for any value of p,
The proposed model retains the large size symmetric PHL for the
f(x) value does not fix to any integer constant and lies within the in-
voltage range ± 2 V⩽ U ⩽ ± 26 V as shown in Fig. 17. Because f(x)
terval (0 < f(x) < 1). Thus, we have successfully overcome the pro-
value lies within the range (0 < f(x) < 1) consequently, x lies within
blems of Jogelker, Biolek, Prodromakis and Jinxiang et al., window
the interval (0 < x < 1) as shown in Fig. 18. Therefore, our proposed
model is found to overcome the setbacks enunciated by the Pro-
dromakis and Jinxiang et al., window function. Even though the ap-
plied voltage is (U) > ± 26, our proposed model produces the PHL
with diminished DNDR, unlike other window functions as shown in
Fig. 16.
Generally, if the applied voltage is high, the rate at which the
changes in the width is much faster facilitating the depleted ions to
reach the boundary (metal electrode interface) causing DPHL to occur.
This problem is eradicated by our proposed window function by con-
trolling the rate of width movement yet retaining the x within the range
(0 < x < 1).

3.5.3. Effect of frequency (f)

In general, memristor models produce linear I-V curve for high
frequencies as it has direct relation to changing width with respect to
ion mobility. This limits the charge (q) accumulation in the width of the
non-stoichiometric region. Similarly, when the frequency decreases, the
charges reach the boundary quickly consequently x reaches the
Fig. 12. Current-voltage for various applied voltages (Prodromakis window boundaries with high speed as well. Therefore, the memristor model
function). produces the Distorted Pinched Hysteresis Loop (DPHL).

134
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

Fig. 17. Current-voltage response of the proposed model with the various input
voltages (U) (proposed window function).
Fig. 14. The current-voltage response of the model for input voltage
(U) ≥ ± 18 V (Jinxiang et al., window function).

Fig. 18. The state variable response of the Proposed and Prodromakis window
functions.
Fig. 15. Response of our model for the different value of p with the input
voltage U = ± 3 V, f = 0.5 Hz (proposed window function).
Interestingly, these issues are resolved by using our proposed model.
Thus, our proposed model produces the PHL within the frequency range
0.05 ≤ f < 2 Hz as illustrated in Fig. 20 unlike other window functions
as depicted in Fig. 19. It is noteworthy that the new window function
controls the width within the boundaries. On the other hand, by using
the optimal values of p and J, our model produces the PHL for f ≥ 2 Hz
as shown in Fig. 21.
The reason being is that, for any higher values of p and J our pro-
posed window function does not fix the values of f(x) to any particular
integer constant unlike other window functions known hitherto.

3.5.4. Threshold voltage controlled drift velocity (TVDV) model

Our proposed model also produces the Asymmetric Pinched
Hysteresis Loop (APHL), which is achieved by threshold voltage con-
dw (t )
trolled drift velocity of ions ( dt ) as shown in Eq. (9).
u R I (t )
⎧ v onD
· f (x ) V (t ) ⩾ Von
dw (t ) ⎪ u R I (t )
v on
= · f (x ) V (t ) ⩽ −Voff
dt ⎨ D
⎪ v on
u R I (t )
· f (x ) − Voff < V (t ) < Von
⎩ D (9)

where Von is the on-threshold voltage, Voff is the off-threshold voltage, V

Fig. 16. Symmetric PHL, for the input voltage U = ± 50 V, p = 2, and J = 0.4 (t) is the applied voltage, I(t) is the current flow through the memristor,
(proposed window function). uv is the mobility of ions, D is the distance between the two platinum

135
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

Fig. 19. The nonlinear I-V characteristics for the frequency f = 0.05 Hz, p = 2, U = ± 2 V and, J = 1 (a) Jogelker, b) Biolek, (c) Prodromakis, and (d) Jinxiang
window functions.

Fig. 21. Current-voltage response of the Proposed model for the f = 2 Hz,
p = 4, U = ± 2 V and, J = 10.
Fig. 20. The nonlinear I-V characteristic of the proposed window function for
the frequency f = 0.05 Hz, p = 2, U = ± 2 V and, J = 1.

136
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

0.0035
Model based
0.0025
Actual Device
0.0015

Current (A)
0.0005

-0.0005

-0.0015

-0.0025

Fig. 22. Asymmetric current-voltage response of the proposed memristor model

-0.0035
for x = 0.5, p = 4, J = 10, Von = 0.4 and Voff = −1.6. -5 -4 -3 -2 -1 0 1 2 3 4 5
Voltage (V)
Fig. 24. Comparison of I-V curves obtained from the fabricated Pt/TiO2−x/
TiO2/Pt memristor device [18] with our proposed model (LIMM incorporating
our new window function).

does not mimic the properties of the physical memristor device as such.
Hence, we are certain that the model requires a window function.
The Jogelker, Biolek and Prodromakis are the different types of
window functions. In these window functions, it is understood that the
nonlinearity and scalability are found to be very less. Therefore, a novel
window function is required for the linear ion drift memristor model.
In the proposed window function the nonlinearity and scalability
are very high compared to other window functions. The comparison
results are shown in Table 1.
In Jogelker and Biolek, for very small value of p, f(x) value is fixed
to one. Therefore, non-linearity and scalability are very limited. In
Fig. 23. Asymmetric current-voltage response of the proposed memristor model Prodromakis, nonlinearity and controllability is moderate. But, the
for x = 0.2, p = 4, J = 10, Von = 0.4 and Voff = −1.6. model does not produce the PHL for p = 2, J = 1, and f < 0.1 Hz. In
Jinxiang et al., resolving the boundary lock problem (higher voltage) is
electrodes. moderate and the model does not produce the PHL for p = 2, J = 1, and
Due to the existence of threshold voltages, drift velocity of ions f = 0.05 Hz. Therefore, this model fails to use in all the applications.
increases/decreases nonlinearly based on the polarity of applied vol- Hence, the linear ion drift memristor model required a versatile
tage. Correspondingly, memristance and width of the doped region window function to accommodate the observed setbacks as described
increases/decreases nonlinearly. As a result, our model produces the above.
APHL. Thus, our proposed window function overcomes all the problems
Fig. 22 illustrate the APHL which occurs during negative bias (reset with high nonlinearity as the proposed window function is defined
process). Because, off-threshold voltage decreases the width movement based on the cubic parabola and therefore, f(x) value does not fix to any
speed. As a result, memristance increases and current conductivity particular value producing the PHL for f < 0.1 Hz.
decreases slowly. Thus our model produces APHL as well.
Fig. 23 also illustrates the APHL with large sized loop for the lower
state variable and on/off threshold voltage. In general, it is known that 4.1. Application in an operational amplifier
nonlinearity is related to threshold voltage and loop size is related to
state variable. If the value of state variable is low then the ionic drift In order to validate our claim, we have used our new model in an
increases/decreases slowly based on the polarity of applied voltage in operational amplifier (op-amp) and demonstrating the amplification
turn affecting the width movement and current flow. These properties with high gain. Thus, we have used the memristor model incorporating
are responsible for obtaining large sized APHL in our memristor model. the new window function and the functionality of memristor aug-
Interestingly, our proposed model corroborates with the observed mented Op-amp has been tested successfully. Fig. 25 shows the hybrid
property of physical memristor device (Pt/TiO2−x/TiO2/Pt) fabricated op-amp circuit (memristor/CMOS gates) consisting of CMOS based
and studied in our lab [18]. The geometric parameters of the physical differential amplifier with memristor circuit topology. This circuit
memristor used in the present study for comparison is for the thickness produces a high gain with low power dissipation compared to CMOS
of bilayer is 147 nm, amplitude of the applied voltage is ± 4 V and the based operational amplifier as shown in Table 2 and its corresponding
frequency is 10 Hz as shown in Fig. 24. frequency response is illustrated in Fig. 26.
In a memristor based operational amplifier, our proposed model
4. Results and discussions produces a high gain compared to other memristor models as shown in
Table 3.
In LIMM, ionic drift is considered to be linear. Therefore, the model

137
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

Table 1
A comparison of performance factors for different window functions.
Performance factors Window functions

Jogelker Biolek Prodromakis Fuzzy [16] Sigmoidal [17] Jinxiang* et al. Proposed

Resolve Boundary effect Yes Yes Yes Yes Yes Yes Yes
Scalability No Limited Moderate Yes – Yes Yes
Control parameters No No Yes No Yes Yes Yes
Nonlinearity Limited Yes Yes Yes Yes Yes Yes
Asymmetric switching No No No Yes No No Yes
Threshold Voltage No No No Yes No No Yes
Resolve Boundary lock (Higher Voltage) Limited Limited Moderate Moderate – Moderate Yes
Resolve Boundary lock (Lower Frequency) No No No – – No Yes

– Not available.
* Jinxiang et al., window function is simulated and the results are tabulated.

Fig. 25. Memristor based operational amplifier internal circuit topology.

Table 2 5. Conclusion
Comparison of results between CMOS and memristor based operational am-
plifier. The validation of our proposed model is carried out in Cadence®
Op-amp Power dissipation (μW) GAIN (dB) software, Virtuoso 6.0 tool. Interestingly, our proposed window func-
tion increases the nonlinearity and scalability in linear ion drift mem-
CMOS 745.93 62.144 ristor model very significantly as compared to different existing models.
Hybrid (CMOS and Memristor) 263.85 66.267
Therefore, our proposed model mimics the properties of physical
memristor device as exemplified. The main significance of our proposed
model is that it retains the PHL for higher voltages as well as lower

138
T.A. Anusudha, S.R.S. Prabaharan Int. J. Electron. Commun. (AEÜ) 90 (2018) 130–139

80 Appendix A. Supplementary material

60 Supplementary data associated with this article can be found, in the

online version, at http://dx.doi.org/10.1016/j.aeue.2018.04.020.
40
Gain (dB)

References
20
[1] Chua LO. Memristor – the missing circuit element. IEEE Trans Circuit Theory
0 1971;18(5):507–19.
[2] Strukov Dmitri B, Snider Gregory S, Stewart Duncan R, Stanley Williams R. The
missing memristor found. Nature 2008;453:80–3.
-20 [3] Ayten Umut Engin, Minaei Shahram, Mehmet S. Memristor emulator circuits using
single CBTA. Int J Electron Commun (AEÜ) 2017;82:109–18.
[4] Sánchez-López C, Carrasco-Aguilar MA, Muniz-Montero. A 16 Hz–160 kHz mem-
-40 ristor emulator circuit. Int J Electron Commun (AEÜ) 2015;69:1208–19.
1 40 1600 64000 2560000 [5] Sánchez-López C, Aguila-Cuapio LE. A 860 kHz grounded memristor emulator cir-
cuit. Int J Electron Commun (AEÜ) 2017;73:23–33.
Frequency (Hz) [6] Tarkhan Mahdi, Maymandi-Nejad Mohammad. Design of a memristor based fuzzy
processor. Int J Electron Commun (AEÜ) 2018;84:331–41.
Fig. 26. Frequency response of a memristor based operational amplifier. [7] YangJ Joshua, Pickett Matthew D, Li Xuema, Ohlberg Douglas AA, Stewart Duncan
R, Stanley Williams R. Memristive switching mechanism for metal/oxide/metal
nanodevices. Nat Nanotechnol 2008;3:429–33.
[8] Pickett Matthew D, Strukov Dmitri B, Borghetti Julien L, Joshua Yang J, Snider
Table 3 Gregory S, Stewart Duncan R, Stanley Williams R. Switching dynamics in titanium
Comparison results between different window func- dioxide memristive devices. J Appl Phys 2009;106:0745081–6.
tions and proposed memristor model. [9] Kvatinsky Shahar, Friedman Eby G, Kolodny Avinoam, WeiserUri C. TEAM:
threshold adaptive memristor model. IEEE Trans Circuits Syst.-I
Window functions Gain (dB) 2013;60(1):211–21.
[10] Jogelker YN, Wolf SJ. The elusive memristor: properties of basic electrical circuits.
Jogelker 57.075 Eur J Phys 2008;30:661–75.
Biolek 7.85 [11] Biolek Z, Biolek D, Biolkova V. SPICE model of memristor with nonlinear dopant
Prodromakis 65.089 drift. Radioengineering 2009;18(2):210–4.
Jinxiang et al. 64.659 [12] Prodromakis Themistoklis, Peh Boon Pin, Papavassiliou Christos, Toumazou
Proposed 66.267 Christofer. A versatile memristor model with nonlinear dopant kinetics. IEEE Trans
Electron Devices 2011;58(9):3099–105.
[13] Williams R. How we found the missing memristor. IEEE Spectr 2008;45(12):28–35.
[14] Hu SG, Wu SY, Jia WW, Yu Q, Deng LJ, Fu YQ, et al. Review of nanostructured
resistive switching memristor and its applications. Am Sci Publ, Nanosci
Nanotechnol Lett 2014; 6(9): p. 729–757.
[15] Zha Jinxiang, Huang He, Liu Yujie. A novel window function for memristor model
frequencies and also by adjusting the fitting parameters of window with application in programming analog circuits. IEEE Trans Circuits Syst—ii:
function. Hence, our model can be used in different memristor related Express Briefs 2016;63(5):423–7.
applications with high performances. [16] Abdel-Kader Rabab Farouk, Abuelenin Sherif M. Memristor model based on fuzzy
window function. IEEE, Fuzzy Systems 2015.
[17] Georgiou PS, Yaliraki SN, Drakakis EM, Barahona M. Window functions and
Acknowledgment Sigmoidal behavior of memristive systems. Int J Circ Theor Appl
2016;44(9):1685–96.
[18] Sahoo Satyajeet, Manoravi P, Prabaharan SRS. Nano-ionic memristive crossbar
One of us (TAA) thanks VIT management, Chennai for awarding architecture- fabrication, structure, surface properties and switching characteristics.
teaching assistantship. Vacuum 2018 (Communicated).

139