Feature

Brains Are Made

of Memristors

Maheshwar Pd. Sah,

Hyongsuk Kim, and
Leon O. Chua

Abstract
This exposition shows that the potassium ion-channels and the
sodium ion-channels that are distributed over the entire length
brain: Image licensed by Ingram Publishing & wikimedia commons
of the axons of our neurons are in fact locally-active memris-
tors. In particular, they exhibit all of the fingerprints of memris-
tors, including the characteristic pinched hysteresis Lissajous
figures in the voltage-current plane, whose loop areas shrink
as the frequency of the periodic excitation signal increases. I. Introduction

Moreover, the pinched hysteresis loops for the potassium ion-
channel memristor, and the sodium ion-channel memristor, from
the Hodgkin-Huxley axon circuit model are unique for each
periodic excitation signal. An in-depth circuit-theoretic analysis
and characterizations of these two classic biological memris-
tors are presented via their small-signal memristive equivalent
circuits, their frequency response, and their Nyquist plots. Just
as the Hodgkin-Huxley circuit model has stood the test of time,
its constituent potassium ion-channel and sodium ion-channel
memristors are destined to be classic examples of locally-active
memristors in future textbooks on circuit theory and bio-physics.
Digital Object Identifier 10.1109/MCAS.2013.2296414
Date of publication: 20 February 2014
T
he basic computation and information processing
units in the brain are the synapses and the axons
[1]. It is well known that synapses are locally-pas-
sive non-volatile memristors [2], [3]. Such memristors
are essential building blocks for emulating memory and
learning. But learning and other higher brain functions
require that clusters of neurons be able to communicate
with each other via axons depicted schematically Fig.
1(a) and Fig. 1(b). Our main goal of this exposition is
to conduct an in-depth analysis of the classic Hodgkin-
Huxley circuit model (Fig. 1(c)) of the squid giant axons
in Fig. 1, [4] and show that it contains two locally-active

12 IEEE circuits and systems magazine 1531-636X/14/$31.00©2014IEEE first QUARTER 2014

memristors which are essential for generating action
potentials (spikes). In particular, it contains a potassium
ion-channel memristor and a sodium ion-channel memris-
tor [5] as depicted in Fig. 1(d).
As articulated in [2] and [5] the Hodgkin-Huxley cir-
cuit model in Fig. 1(c) is conceptually wrong from a cir-
cuit-theoretic perspective because the two time-varying
resistors are in fact memristors, which are time-invariant.
In the following sections, we will present an in-depth
characterization of both the potassium and the sodium
ion-channel memristors. We will show that the voltage-
current Lissajous figures due to a periodic input signal of
both potassium and sodium memristors are pinched hys-
teresis loops, whose lobe areas shrink as the frequency
of the input signal increases. These characteristics are in
fact the unique fingerprints of memristors.
Since both the potassium and the sodium ion-chan-
nel memristors in Fig. 1(d) are the fundamental building
blocks of all axons and are responsible for generating the
spikes [6] essential for learning, adaptation, intelligence,
and even consciousness, we will present a comprehen-
sive in-depth characterizations of both potassium and
sodium ion-channel memristors in the ensuing sections.
II. Characterization of Pinched Hysteresis Loops
A voltage-controlled memristor is defined by,
i = G ^ x 1, x 2, ..., x n ; v h v
dx k = f ^ x , x , ..., x ; v h, k = 1, 2, ..., n 4, (1)
k 1 2 n
dt
where G in the first equation is a piecewise continuous and
bounded1 function of ^ x 1, x 2, ..., x n ; v h, called the memduc-
tance of the memristor. The state variables (x 1, x 2, ..., x n)
in the second equation depend on the internal state of
the memristor and is defined by `n_ 1st-order differential
equations called the associated state equations.
Similarly, a current-control memristor is defined by,
v = R ^ x 1, x 2, ..., x n ; i h i
dx k f ^ x , x , ..., x ; i h, k 1, 2, ..., n 4 , (2)
= k 1 2 n =
dt
where R is a piecewise continuous and bounded func-
tion of n state variables (x 1, x 2, ..., x n), called the mem-
ristance of the memristor.
One of the unique signatures of a memristor which
distinguishes it from non-memristive devices is its asso-
ciated pinched hysteresis loop in the voltage v vs. current
1
We assume G (x 1, x 2, f, x n ; v) 1 3 and R (x 1, x 2, f, x n ; i) 1 3 for
R (x 1, x 2, f, x n ) 1 3 .
Maheshwar Pd. Sah and Hyongsuk Kim (corresponding author) are with the Division of Electronics Engineering and Intelligent Robots Research Center,
Chonbuk National University, Jeonju, Jeonbuk, 561-756, Korea (e-mails: maheshwarsah@jbnu.ac.kr and hskim@jbnu.ac.kr). Leon O. Chua is with the
Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, California 94720-1770 (e-mail: chua@eecs.berkeley.edu).
first QUARTER 2014 IEEE circuits and systems magazine 13
i plane, namely, the v-i loci always passes through origin
for any bipolar periodic input voltage v (t) (resp., input
current i (t)) waveform which assumes both positive
and negative voltages (resp., currents). In general, the
pinched hysteresis loop may intersect itself also at addi-
tional points (v, i ) ! (0, 0) .
The shape of the pinched hysteresis loop varies with
frequency f and shrinks to a single-valued function
through the origin, as the frequency tends to infinity.
This is another signature of a memristor. In case when
the memductance G (resp., memristance R) does not
depend on v (resp., i), the limiting single-valued func-
tion is a straight line (Property 6 defined in [7]).
Another fingerprint of a memristor is the dependency
of the hysteresis lobe area on the frequency f of the
periodic input signal. Beyond certain critical frequency
f *, the area of the pinched hysteresis lobe decreases
monotonically as the frequency of the periodic input
voltage v (t) (resp., current i (t)) increases.
In this section, we illustrate all these pinched hys-
teresis characteristics and fingerprints of the potassium
ion-channel memristor, and the sodium ion-channel mem-
ristor in the classic Hodgkin-Huxley circuit model of the
giant squid axon [4].
A. Examples of Pinched Hysteresis Loops
of Potassium Memristor
Chua, et. al, [5] shows the time-varying potassium con-
ductance (G K ) and the time-varying sodium conductance
(G Na) in the Hodgkin-Huxley circuit [4] are first-order
potassium memristor and second-order sodium memris-
tor, respectively. The relationship between the input volt-
age v K and the current i K in the potassium ion channel
memristor is defined by the state-dependent ohm’s law,
i K = G K (n) v K _
bb
dn = 0.01 [(v K + E K ) + 10] (1 - n) - 80.125 e K 80 K B n `
(v E )
dt ) 3
+
(v K + E K ) + 10 b
e 10 -1 ,
a
(3)
where G K (n) _ gr K n 4 is the potassium memductance
function and n is the potassium gate-activation variable.
In this paper, the parameters for E K and gr K are assumed
to be 12 mV and 36 mS/cm2 respectively, as chosen by
Hodgkin-Huxley [4].
The characteristics of the potassium ion channel
memristor are investigated by applying several bipo-
lar periodic signals with different initial states n (0) of
the potassium state variable `n_. The initial state n (0)
 v K (t), output current i K (t), gate-activating state variable
Axon n (t), and memductance G K (t) respectively. The loci (Lis-
sajous figure) of (v K (t), i K (t)) corresponding to the input
voltage v K (t), and current i K (t) is shown in Fig. 2(b).
Neuron
Note that it passes though the origin at point  and ,
(a) respectively. Moreover, observe that the memductance
G K (t) in Fig. 2(a) is always positive. The upper figure in
Dissipative (Diffusion) Couplings
Fig. 2(b) is a double-valued Lissajous figure plotted in
... i K vs. v K plane. Such a multi-valued Lissajous figure of
v (t) and i (t) which passes though the origin is called a
... pinched hysteresis loops in footnote 12, page 1850 of [8].
Hodgkin-Huxley Cells The lower figure in Fig. 2(b) shows the variation of the
(b) potassium memductance G K of the potassium ion-chan-
nel memristor with respect to the applied voltage v K (t) .
I Outside Hodgkin-Huxley Cell Consider next a potassium ion-channel memristor driven
+ by a composite waveform consisting of a sinusoidal voltage
INa IK IL
A sin (2rft) and it’s nth harmonic of amplitude B and phase
angle i; namely v K (t) = A sin (2rft) + B sin (2rNft + i),
CM E
V RNa RK RL where `N _ denotes the order of the harmonics of the sinu-
– + + soidal signal A sin (2rft) from Fig. 2(a). Fig. 3(a) shows the
+ ENa – EK – EL pinched hysteresis loop corresponding to the frequency
_
f = 100 Hz, A = B = 40, i = r/3 and N = 1, 3, 5, 8. As the
Inside value of N is varied, the input signal is modulated, result-
(c)
ing in a “multi-lobe” pinched hysteresis loci in the i K vs.
v K plane. A family of pinched loops with multiple lobes
obtained for N = 1, 3, 5, and 8. are shown in Fig. 3(b).
Observe from Fig. 3(a) and Fig. 3(b) that even though
there are multiple lobes in the i K vs. v K plane, all loops are
pinched at the origin.
I Consider a potassium ion channel memristor driven
+I by several distinct bipolar input signals, namely v K (t) =
+
ICM + INa
INa IKIK ILI A sin (2rft), v K (t) = A cos (2rft), v K (t) = A sin (2rft) +
+ + L
+ + V + G B cos (2rft) and v K (t) = a rectangular waveform. As
V VCMCM V
VNa
Na
_
GNa V
GNa VKK
_
G
GKK V_L
L
_
L
GL expected, the loci in the i K -v K plane for these input sig-
_ Na
Na _ KK
ENa
E Na EEKK ELEL nals exhibit pinched hysteresis loops at the origin as
_
_ shown in Fig. 4. Observe also, the shape of the pinched
hysteresis loops shrinks as the frequency of the input
Memristive Hodgkin-Huxley Model
signal increases. All these pinched hysteresis loops
(d) exhibit the fingerprints of a memristor.
Figure 1. (a) Schematic of a neuron and its axon. (b) One-
dimensional axon model made of resistively coupled Hodgkin-
B. Examples of Pinched Hysteresis Loops
Huxley cells. (c) Hodgkin-Huxley circuit model made of a
capacitor C M, a resistor R L, three batteries E Na, E K , and E L, a of Sodium Ion-Channel Memristor
time-varying potassium resistor R K , and a time-varying sodium The governing equations of the sodium ion channel
resistor R Na. (d) Memristive Hodgkin-Huxley axon circuit model. memristor is defined by,

is chosen such that the above 1st-order nonlinear dif-

_
ferential equation, which is driven by a periodic volt- i Na = G Na (m, h) v Na b
age v K (t) of period T = 1/f has a periodic response b
0 .1 [(v E ) 25 ] ;4e 18 E m b
* 4
dm = Na - Na + v
Na -
NaE

n (t) = n (t + T) for t $ 0. ( 1 - m ) - b
dt (v - E ) + 25
Na Na
e 10 -1 `
Let us apply first a sinusoidal voltage b
dh = ' 0.07e (v 20 1 (1 - h) -
> H
-E )
v K (t) = A sin (2rft) with amplitude A = 50 mV, and fre-
Na Na
1 h b
dt (v - E ) + 30
Na Na
b
quency f = 200 Hz to the potassium ion channel memris- e 10 +1 b
a
tor. Fig. 2(a) shows the waveform of the input voltage (4)

14 IEEE circuits and systems magazine first QUARTER 2014

 vK (t ) (mV)

60
30 2 2 iK (mA)
3 3 1 t
0 0.04
1 1 ms
–30 4 4
2
–60
140 142.5 145 147.5 150 0.014
1, 3 vK
iK (t ) (mA)
–60 –30 0 30 60 mV
–0.013
2 2
0.02 t
1 3 3 1
0 1
–0.02 4 ms 4 –0.039
4
–0.04
–0.06
140 142.5 145 147.5 150 –0.065
n(t ) n (0) = 0.45 at t = 0
GK (mS)
1 1 1 1.6
0.4 4
4 1
2
0.3 2 3 3 1.235
t
140 142.5 145 147.5 150 ms
GK (t ) (mS) 0.87

1.6 4
1 1 1
1.235 0.505 2
0.87 4 4
3
0.505 2
2 3 3 t 0.14 vK
0.14
140 142 144 146 148 150 ms –60 –30 0 30 60 mV
(a) (b)

Figure 2.  (a) Waveforms of the applied voltage v K (t) = A sin (2rft), current i K (t), state variable n (t), and memductance G K (t) of the
potassium ion-channel memristor. (b) Potassium pinched hysteresis loop in i K vs. v K plane and corresponding variation of the potas-
sium memductance G K with respect to the applied sinusoidal voltage. The simulations were performed at A = 50 mV, f = 200 Hz
and n (0) = 0.45.

where G Na (m, h) _ gr Na m 3 h is the sodium memductance
function, `m_ is the sodium gate-activation state vari-
able, and `h_ is the sodium gate inactivation state vari-
able. In this paper, the parameters for E Na and gr Na are
equal to 115 mV and 120 mS/cm2 respectively, as chosen
by Hodgkin and Huxley [4].
Let us illustrate the characteristic features of the
sodium ion channel memristor by applying various
bipolar signals with appropriate initial states m (0) and
h (0) such that the solution of the two nonlinear differ-
ential equations in (4) when driven by a periodic signal
has a periodic response, for t $ 0.
Fig. 5(a) shows the waveforms of the applied input
voltage v Na (t) = A sin (2rft), current i Na (t), m (t), h (t)
and the sodium memductance G Na (t) of the sodium
ion-channel memristor, respectively. The corresponding
Lissajous figure of i Na vs. v Na, and G Na vs. v Na are shown
in Fig. 5(b). Observe that they are “Pinched” not only
at the origin, but also at point =point . The self-
crossing of the v Na-i Na pinched hysteresis loop and the
memductance G Na hysteretic loop corresponding to
point  (resp., point ) are numerically found to be
i Na = 0.022 mA, v Na = 69.31 mV, and G Na = 0.319 mS,
respectively.
When a sodium ion-channel memristor is driven by
a composite waveform such as v Na (t) = A sin (2rft) +
B sin (2rNft + i), where `N _ denotes the order of the har-
monics of A sin (2rft) the characteristic of the pinched
loop changes with A, B, i and N. Fig. 6(a) shows the
pinched hysteresis loop corresponding to the frequency
f = 500 Hz, A = B = 50 mV, i = r/3 and N = 1, 3, 5, 8 and
Fig. 6(b) shows the corresponding pinched hysteresis

first QUARTER 2014 IEEE circuits and systems magazine 15

 iK (mA) iK (mA)
0.15 0.15

vK vK
–90 –45 0 45 90 (mV) –90 –45 0 45 90 (mV)
A = B = 40 mV A = B = 40 mV
–0.15 N = 1, θ = π /3 –0.15
N = 3, θ = π /3
f = 100 HZ f = 100 HZ
vK mV
–0.3 –0.3 vK mV
t t
ms ms
–0.45 –0.45
iK (mA) iK (mA)
0.15 0.15

vK vK
–90 –45 0 45 90 (mV) –90 –45 0 45 90 (mV)

A = B = 40 mV A = B = 40 mV
–0.15 –0.15
N = 5, θ = π /3 N = 8, θ = π /3
f = 100 HZ f = 100 HZ
vK mV vK mV
–0.3 , –0.3
t t
ms ms
–0.45 –0.45

(a)

iK (mA) iK (mA)

0.06 0.06

vK vK
–65 –32.5 0 32.5 65(mV) –65 –32.5 0 32.5 65 (mV)

–0.07 –0.07

A = 50 mV A = 50 mV
B = 10 mV B = 10 mV
N = 1, θ = π /3 N = 5, θ = π /3
f = 100 HZ f = 100 HZ
–0.2 –0.2

iK (mA) iK (mA)

0.06 0.06

vK vK
–65 –32.5 0 32.5 65(mV) –65 –32.5 0 32.5 65 (mV)

–0.07 –0.07
A = 50 mV A = 50 mV
B = 10 mV B = 10 mV
N = 10, θ = π /3 N = 20, θ = π /3
f = 100 HZ f = 100 HZ
–0.2 –0.2

(b)
Figure 3.  Pinched hysteresis loops of the potassium ion channel memristor for input voltage v K (t) = A sin (2rft) + B sin (2rNft + i),
with f = 100 Hz and i = r/3. (a) Pinched hysteresis loops for A = B = 40 mV, and N = 1, 3, 5, 8. (b) Pinched hysteresis loops for
A = 50 mV, B = 10 mV and N = 1, 5, 10, 20.

16 IEEE circuits and systems magazine first QUARTER 2014

 iK (mA) iK (mA)

0.04 0.38

vK
0.19
–60 –40 –20 0 20 40 60 (mV)
vK
–0.04 –100 –50 0 50 100 (mV)
vK = A sin(2πft) vK = A cos(2πft)
A –0.19 A
–0.08 t t
0
ms ms
–A –A
EK = 12 mV EK = 12 mV
–0.38
–0.12

Legend: Legend:
A = 50 mV A = 100 mV
f = 10 KHz, n (0) = 0.363 f = 10 KHz, n (0) = 0.524
f = 1.5 KHz, n (0) = 0.375 f = 1 KHz, n (0) = 0.51
f = 100 KHz, n (0) = 0.51 f = 500 KHz, n (0) = 0.51

(a) (b)

iK (mA) iK (mA)

1 1.2

0.5 0.6

vK vK

–260 –130 0 130 260 (mV) –90 –45 0 45 90 (mV)

vK mV
vK = A sin(2πft) +
B cos(2πft) –0.6 A
–0.5 t
t 0
0 ms
ms –A
EK = 12 mV –1.2 EK = 12 mV
–1

Legend: Legend:
A = 150 mV, B = 200 mV A = 80 mV
f = 50 KHz, n (0) = 0.474 f = 10 KHz, n (0) = 0.54
f = 5 KHz, n (0) = 0.5 f = 800 KHz, n (0) = 0.61
f = 1 KHz, n (0) = 0.58 f = 200 KHz, n (0) = 0.78

(c) (d)

Figure 4. Pinched hysteresis loops of the potassium ion channel memristor for various bipolar periodic input signals.
(a) v K (t) = A sin (2rft) . (b) v K (t) = A cos (2rft) . (c) v K (t) = A sin (2rft) + B cos (2rft) . (d) A rectangular waveform. The dotted lines
in Fig. 4(d) indicate instantaneous jump of v K and i K at t = 0, T/2, and T.

loops for f = 500 Hz, A = 50 mV, B = 15 mV, i = r/3,
and N = 1, 5, 10 and 20.
Fig. 7 shows the pinched hysteresis loops of the
sodium ion-channel memristor for v Na (t) = A sin (2rft),
v Na (t) = A cos (2rft), v Na (t) = A sin (2rft) + B cos (2rft)
and v Na (t) = a rectangular waveform for several differ-
ent frequencies. Observe that the area of the pinched
hysteresis loops shrinks with the frequency f and tend
to a straight line as predicted. All these pinched hyster-
esis loops exhibit the fingerprints of the memristor.

first QUARTER 2014 IEEE circuits and systems magazine 17

 vNa (mV)
130 4
3 5 3 5
65 2 4
2 6 6 1 t
0
1 1 ms
–65 7
7
–130
40 42 44 46 48 50
iNa (mA)

0.1 2 2 5
6 4 6 t
1
0
1 3 4 5 1 3 ms
–0.1
–0.2 7 7

–0.3
40 42 44 46 48 50
m(t ) m(0) = 1, at t = 0 INa mS
1.2 1 2 1 0.1
2 6
0.9 6 7 1 7 2 ,5
0.6 1,6 vNa
5
0.3 3 5 34 –130 –65 0 65 4 3 130 mV
t
0 4 –0.075
ms
40 42 44 46 48 50
vNa = 69.31 mV
h(t ) 7 iNa = 0.022 mA
h(0) = 0.003, at t = 0
0.07 –0.25
5 5
4 4
0.047 6 GNa mS
6 3
3 5
0.023 7
1 7 1 2
2 1 t 6
0 GNa = 0.319 mS
40 42 44 46 48 50 ms vNa = 69.31 mV
GNa (mS)
2
5
6 6 7
3.75 1 2,5
vNa
2.5
7 7 –130 –65 0 65 4 3 130 mV
1.25 1 2 3 4 1 2
5 3 4 5 1 t –1
0
40 42 44 46 48 50 ms
(a) (b)

Figure 5.  An example of a memristor exhibiting self-crossing pinched hysteresis loop in i Na vs. v Na and G Na vs. v Na plane at v Na (t) ! 0
(resp. current i Na (t) ! 0) . (a) Waveforms of the input voltage v Na (t) = A sin (2rft), current i Na (t), m (t), h (t), and the sodium mem-
ductance G Na (t) of the sodium ion-channel memristor. (b) Corresponding v Na-i Na pinched hysteresis loop and the sodium memduc-
tance G Na hysteretic loop plotted in i Na vs. v Na plane and the G Na vs. v Na plane, respectively. The simulations were performed at
A = 120 mV, f = 200 Hz, m (0) = 1, and h (0) = 0.003.

C. Computation of Lobe Area of Pinched Hysteresis
Loop Via Riemann–Stieltjes Integral
When a memristor is driven by a bipolar periodic input
voltage v (t) or current i (t), the pinched hysteresis
loop has a typical shape depending on the memristor
constitutive relation. Above a certain critical frequency
fc, the area of the hysteresis lobe in the 1st or 3rd
quadrant2 is inversely proportional to the excitation
2
Assuming each lobe is enclosed by a simple closed curve without self-
intersections except at the origin.

18 IEEE circuits and systems magazine first QUARTER 2014

 iNa (mA) iNa (mA)
0.04 0.04

0.02 0.02

vNa vNa

–100 –50 0 50 100 (mV) –100 –50 0 50 100 (mV)

A = B = 50 mV A = B = 50 mV
N = 1, θ = π /3 N = 3, θ = π /3
–0.02 –0.02 f = 500 HZ
f = 500 HZ vNa mV
vNa mV
t t
–0.04 –0.04
ms ms
iNa (mA) iNa (mA)
0.04 0.04

0.02 0.02

vNa vNa

–100 –50 0 50 100 (mV) –100 –50 0 50 100 (mV)

A = B = 50 mV A = B = 50 mV
–0.02 N = 5, θ = π /3 –0.02 N = 8, θ = π /3
f = 500 HZ f = 500 HZ
vNa mV vNa mV
t t
–0.04 –0.04
ms ms

(a)
iNa (mA) iNa (mA)
0.01 0.01

5 × 10–3 5 × 10–3

vNa vNa
–70 –35 0 35 70 (mV) –70 –35 0 35 70 (mV)
A = 50 mV A = 50 mV
–5 × 10–3 B = 15 mV –5 × 10–3
B = 15mV
N = 1, θ = π /3 N = 5, θ = π /3
–0.01 f = 500 HZ –0.01 f = 500 HZ

iNa (mA) iNa (mA)

0.01 0.01

5 × 10–3 5 × 10–3

vNa vNa
–70 –35 0 35 70 (mV) –70 –35 0 35 70 (mV)

A = 50 mV A = 50 mV
–5 × 10–3 –5 × 10–3
B = 15 mV B = 15 mV
N = 10, θ = π /3 N = 20, θ = π /3
–0.01 f = 500 HZ –0.01 f = 500 HZ

(b)

Figure 6.  Pinched hysteresis loops of the sodium ion channel memristor for input voltage v Na (t) = A sin (2rft) + B sin (2rNft + i),
with f = 500 Hz, i = r/3. (a) Pinched hysteresis loops for A = B = 50 mV, and N = 1, 3, 5, 8. (b) Pinched hysteresis loops for
A = 50 mV, B = 15 mV and N = 1, 5, 10, 20.

first QUARTER 2014 IEEE circuits and systems magazine 19

 iNa (mA) iNa (mA)

6 × 10–3
0.016
4 × 10–3 vNa = A cos(2πft)
A t
2 × 10–3 ms
–A
vNa ENa = 115 mV
–60 –40 –20 0 20 40 60 (mV) 3.5 × 10–3
vNa
–2 × 10–3 vNa = A sin(2πft)
–70 –35 0 35 70 (mV)
A
t
–4 × 10–3 ms
–A
ENa = 115 mV –9 × 10–3
–6 × 10–3

Legend: Legend:
A = 50 mV A = 60 mV
f = 10 KHz, m (0) = 0.9975, h (0) = 0.00715 f = 5 KHz, m (0) = 0.994, h (0) = 0.0011
f = 1.5 KHz, m (0) = 1, h (0) = 0.0006 f = 1 KHz, m (0) = 0.982, h (0) = 0.0011
f = 500 KHz, m (0) = 1, h (0) = 0.0004 f = 200 KHz, m (0) = 0.96, h (0) = 0.002
(a) (b)

iNa (mA) iNa (mA)

0.025 1.5 × 10–3

vNa = A sin(2πft) +
B cos(2πft)
t
0
ms 5 × 10–4
ENa = 115 mV vNa
5× 10–3 –25 –12.5 0 12.5 25
vNa (mV)
vNa mV
–75 –37.5 0 37.5 75 (mV) –5 × 10–4
A
t
0
ms
–A
–0.015 –1.5 × 10–3 ENa = 115 mV

Legend:
A = B = 50 mV
f = 8 KHz, m (0) = 0.995, h (0) = 0.0017
f = 800 KHz, m (0) = 1, h (0) = 0.0012
f = 300 KHz, m (0) = 1, h (0) = 0.0007
Legend:
A = 20 mV
f = 4 KHz, m (0) = 0.9994, h (0) = 0.00004
f = 1 KHz, m (0) = 0.997, h (0) = 0.0003
f = 100 KHz, m (0) = 0.997, h (0) = 0.0006

(c) (d)

Figure 7.  Pinched hysteresis loops of the sodium ion-channel memristor for different bipolar periodic input signals. (a) v Na (t) = A sin (2rft) .
(b) v Na (t) = A cos (2rft) . (c) v Na (t) = A sin (2rft) + B cos (2rft) . (d) v Na (t) = a rectangular waveform. The dotted lines in Fig. 7(d) cor-
respond to an instantaneous jump at t = 0, T/2, and T.

frequency f. Thus, the area of the pinched lobe gradu- Let us apply a bipolar periodic input signal to a
ally decreases with increasing frequency for f > fc . The memristor such that a pinched hysteresis loop in the
frequency dependence characteristic of the lobe area is 1st and 3rd quadrants is obtained during the posi-
another signature property of a memristor. tive (0 # t # T/2) and the negative (T/2 # t # T) signal

20 IEEE circuits and systems magazine first QUARTER 2014

amplitudes, respectively. To compute the area enclosed Similarly, the 3rd quadrant lobe area (Ar -) of the
within the pinched hysteresis loop in the 1st and pinched hysteresis loop for second half cycle T/2 # t # T
3rd quadrants, respectively, let us first consider the is given by the Riemann–Stieltjes Integral,
pinched hysteresis loop only over the bounded time
interval 0 # t # T/4 in the i-v plane as shown in the Fig. Ar - = #T/T2 i (t) dvdt(t) dt . (8)
8(a). As proposed by Chua in [2], the yellow area under
the curve of Fig. 8(a) can be computed via the following The 1st and 3rd quadrant lobe area defined by (7) and
Riemann–Stieltjes Integral [9]: (8) are illustrated in Fig. 8(d) for one cycle of the bipolar
T/4 dv (t) periodic input signal.
Ar1 = #0 i (t)
dt
dt 2 0 . (5) Observe that the advantage of using the Riemann-
Stieltjes integral, and not the conventional Riemann inte-
Similarly, the cyan area under the curve for interval gral is that we don’t need the formula describing the i-v
T/4 # t # T/2 of Fig. 8(b) can be computed as follows: curve over different time intervals and we don’t have
to break up the integration into several parts. The Rie-
T/2 dv (t)
Ar2 = #T/4 i (t)
dt
dt 1 0 . (6) mann-Stieltjes integral tracks the orientation of the loci
and add or subtract the component areas automatically,
Adding (5) and (6), we get the 1st quadrant green
lobe area shown in Fig. 8(c). It follows, that the 1st quad-
rant lobe area (Ar +) of the pinched hysteresis loop for
Lobe Area (Potassium Memristor)
the half cycle 0 # t # T/2 is given by,

dv (t)
T/4 T/2 dv (t) 2
Ar + = #0 dt
dt +
i (t) #T/4 i (t)
dt
dt 1st Quadrant Lobe Area (Ar+) f
0
T/2 dv (t) kHz
= # i (t) dt, (7)
–2
0 dt

which can be interpreted as the Riemann–Stieltjes Inte- –4 3rd Quadrant Lobe Area (Ar–)
gral of i (t) with respect to v (t) .
–6

–8

i i –10
0 1 2 3 4 5
(a)
t = T/4
Lobe Area (Sodium Memristor)
t = T/4
15
t = T/2 10 1st Quadrant Lobe Area (Ar+)
t=0 v v
5
0 0 f
(a) (b)
0
i i kHz
–5
Ar+ –10
3rd Quadrant Lobe Area (Ar–)
v –15
0
–20
Ar+ 0 2 4 6 8 10
Ar– (b)
v
0 Figure 9. Lobe area vs. frequency curve of the potassium
(c) (d)
and sodium ion channel memristors. (a) The absolute value
Figure 8.  Geometrical interpretation for computing the lobe of lobe area Ar + and Ar - of the potassium memristor
area of the pinched hysteresis loop via the Riemann-Stieltjes decreases as the frequency of the input signal increases for
integral. (a) Area of the pinched loop for period 0 # t # T/4. f > 0.1KHz in the 1st quadrant and for f > 0.03 KHz in the 3rd
(b) Area of the pinched loop for period T/4 # t # T/2. (c) Area quadrant. (b) The absolute value of the lobe area Ar + and
of the pinched loop for positive input signal over the half cycle Ar - of the sodium memristor decreases as the frequency
time interval 0 # t # T/2. (d) Area of the pinched loop for of the input signal increases for f > 1KHz in the 1st quadrant
bipolar periodic signal over the time interval 0 # t # T. and for f > 0.2 KHz in the 3rd quadrant.

first QUARTER 2014 IEEE circuits and systems magazine 21

 over the half period 0 # t # T/2, and T/2 # t # T, respec-
vK (mV) tively given only i (t) and dv (t) /dt.
60 We have used the Riemann-Stieltjes integral to derive
30 t the frequency-dependent characteristic of the lobe
0
ms area Ar + and Ar - of the potassium ion-channel and the
–30 0 1.25 2.5 3.75 5
–60 sodium ion-channel memristors. Fig. 9(a) and Fig. 9(b)
(a)
iK (mV) show the lobe area vs. frequency curve of the potassium
0.04 and the sodium ion-channel memristor, for input voltage
0.02 t v k = A sin (2rft) and v Na = A sin (2rft) with amplitude
0
–0.02 0 1.25 2.5 3.75 5 ms 50 mV and 120 mV, respectively. As shown in the Fig.
–0.04
–0.06 9(a) and Fig. 9(b), the absolute value of the lobe area
(b)
dvK decreases as the frequency of the input signal increases
65 dt for f > fc . Since the lobe area of the pinched hysteresis
32.5 loops decreases monotonically with increasing fre-
t
0 quency for f > fc, the potassium and sodium ion channel
ms
–32.5 0 1.25 2.5 3.75 5
–65 memristors exhibit the predicted signature property of
(c) a memristor.
iK dvK
2.5 dt
D. Clockwise and Counter-Clockwise Orientation
t of the Pinch Hysteresis Loop of a Memristor
0 ms
0 1.25 2.5 3.75 5 The orientation of the pinch hysteresis loop in the 1st or
–2.5 3rd quadrant of a memristor can occur in a clockwise or
(d)
T/2 dv Lobe Area over 1st counter-clockwise direction with respect to the motion
iK K dt
0 dt Half Period of the tip of a radial vector anchored at the origin. The
1.4
1st quadrant pinched hysteresis loop in Fig. 8(c) is ori-
0.7 Ar+ = 0.84
0
t ented in a clockwise direction. The actual area of the
0 1.25 2.5 3.75 5 ms lobe is computed by subtracting the area of Fig. 8(b)
–0.7
(e) (Eq. (6)) from the area of Fig. 8(a) (Eq. (5)). Since, the
T dvK
iK area of Fig. 8(a) is bigger than the area of Fig. 8(b), the
T/2 dt dt Lobe Area over 2nd
0.5 Half Period actual area of the lobe obtained in the 1st quadrant is
t
0
ms positive during first half cycle of a positive input signal.
–0.5 0 1.25 2.5 3.75 5
Ar– = –1.376 Similarly, the orientation of the pinched hysteresis loop
–1.5
(f)
in the 3rd quadrant during second half cycle for a nega-
tive input is counter-clockwise.
iK (mA) Let us illustrate the orientation of the pinch hysteresis
0.04 loop of the potassium and sodium ion channel memristors
for a sinusoidal input signal. Fig. 10(a) shows the applied
0.014 vK input voltage v K (t) = A sin (2rft) across the potassium
–60 –30 0 30 60 mV ion channel memristor with amplitude A = 50 mV, fre-
–0.013
quency f = 200 Hz, over a one cycle period 0 # t # T.
–0.039 Figs. 10(b)–(g) are the corresponding output current
i K (t), (dv K (t)) / (dt), i K (t) (dv K (t)) / (dt), 1st quadrant
–0.065
lobe area ` # i K (t) ((dv K (t)) / (dt)) dt j, 3rd quadrant lobe
T/2
(g)
area ` # i K (t) ((dv K (t)) / (dt)) dt j, and the pinched hyster-
T 0

Figure 10. An example for illustrating the orientation of the
pinched hysteresis loop in clockwise and counter-clockwise
directions for sinusoidal input voltage at A = 50 mV, f = 200 Hz
in the potassium ion channel memristor. (a) Applied input voltage
v K (t) = A sin (2rft) . (b) i K (t) . (c) (dv K (t)) /dt. (d) i K (t) (dv K (t)) /dt.
(e) 1st quadrant lobe area ` # i K (t) (dv K (t)) / (dt) dt j during first
T/2
T/2
esis loop, respectively. Observe from Fig. 10(e) and Fig.
10(f) that the 1st and 3rd quadrant lobe area during the
first half cycle 0 # t # T/2 and the second half period
T/2 # t # T are equal to Ar + = 0.84 and Ar - = - 1.376,
respectively. The areas of Fig. 10(e) and Fig. 10(f) clearly

half cycle. (f) 3rd quadrant lobe area ` # i K (t) (dv K (t)) / (dt) dt j
0 T
illustrate that, the orientation of the pinched loop in the
T/2
during second half cycle. (g) Clockwise and counter-clockwise 1st quadrant and the 3rd quadrant during the 1st and
orientation of a pinched hysteresis loop in 1st and 3rd quad-
the 2nd half cycles are clockwise and counter-clockwise,
rants, respectively.
respectively, as shown in Fig. 10(g).

22 IEEE circuits and systems magazine first QUARTER 2014

 We have also analyzed the orientation of the multi
vNa (mV)
component pinched hysteresis loop of the sodium ion- 130
channel memristor by applying a sinusoidal voltage 65
t
v Na (t) = A sin (2rft), with amplitude A = 120 mV and 0
frequency f = 200 Hz for one cycle period 0 # t # T. –65 0 1 2 3 4 5 ms
–130
The input voltage and the corresponding output cur- (a)
iNa (mA)
rent i Na (t), (dv Na (t)) /dt, i Na (t) (dv Na (t)) /dt, 1st quadrant
0.1
lobe area ` # i Na (t) (dv Na (t)) / (dt) dt j, 3rd quadrant lobe
T/2
t
0
area ` # i Na (t) (dv Na (t)) / (dt) dt j, and the pinched hys-
0
–0.075 0 1 2 3 4 5 ms
T
T/2
teresis loop, are shown in Figs. 11(a)–(g), respectively.
–0.25
Observe from Fig. 11(e) that the area of the lobe during (b)
0 # t # T/4 is varied from zero to a positive value equal dvNa

to 2.25. Therefore, the sub-loop during this period is 160 dt

80
clockwise as indicated in Fig. 11(g), which is named sub- t
0
loop 1. The area of the lobe during period T/4 # t # T/2 0 1 2 3 4 5 ms
–80
changes from a positive to a negative value and ends in –160
(c)
the negative Ar + region at Ar + = - 1.893 < 0. Thus the dv
iNa Na
area of the lobe during this period is counter-clockwise 25 dt
as indicated in Fig. 11(g) which is named sub-loop 2. The
5 t
area of the lobe during the second half cycle T/2 # t # T 0
0 1 2 3 4 5 ms
changes only in positive direction and ends at the posi- –15
tive region at Ar - = 13.099 > 0. Since, Ar - is positive, it (d)
iNadvNa dt Lobe Area over
T/2
clearly illustrates the orientation of the pinched hyster- dt
0 1st Half Period
esis loop in the 3rd quadrant is clockwise as shown in 3
2 Ar+ = –1.893
Fig. 11(g). 1 t
In summary we can say that, if the lobe area during a 0
–1 0 1 2 3 4 5 ms
half cycle of the input signal is positive (resp., negative) –2
(e)
then the orientation of the pinched loop is clockwise Lobe Area over
iNadvNa dt
T
(resp., counter-clockwise) either in the 1st quadrant, or 20 T/2 dt 2nd Half Period
the 3rd quadrant, respectively and vice versa. Table 1 15
and Table 2 show examples of the computed lobe area 10
5 Ar– = 13.099
and orientation of the pinched hysteresis loops, when 0
t
the potassium and sodium ion-channel memristors are 0 1 2 3 4 5 ms
(f)
driven by a sinusoidal voltage source with amplitude 50
0 T/4 T/2 3T/4 T/2
mV and 120 mV, respectively, for various frequencies. iNa mS
Sub-Loop 1
Table 1 and 2 clearly show the orientations of all the 0.1
At t = 0,
pinched hysteresis loops are clockwise if Ar + > 0 and T/2, T vNa
counterclockwise if Ar - < 0 and vice versa. –130 –65 0 65 130 mV
–0.075
Sub-Loop 2 At t = T/4
III. DC V-I Curves of the Potassium Ion-Channel At t = 3T/4
Memristor, Sodium Ion-Channel Memristor and –0.25
Memristive Hodgkin-Huxley Axon Circuit Model (g)
The DC V - I characteristics of the voltage-dependent
Figure 11. An example for illustrating the orientation
potassium and sodium ion channel memristors are of the pinched hysteresis loop in clockwise and coun-
obtained by solving their associated state equations ter-clockwise directions for sinusoidal input voltage at
for the equilibrium points and substituting the numeri- A = 120 mV, f = 200 Hz in a sodium ion channel memristor.
(a) Applied input voltage v Na (t) = A sin (2rft). (b) i Na (t). (c)
cally solved values into (3) and (4) for i K and i Na,
dv Na (t) /dt. (d) i Na (t) (dv Na (t)) /dt. (e) 1st quadrant lobe area
respectively. The resulting loci of all points (V K , I K ) and ` # i Na (t) (dv Na (t)) / (dt) dt j during first half cycle. (f) 3rd
T/2

quadrant lobe area ` # i Na (t) (dv Na (t)) / (dt) dt j during sec-

0
(VNa, I Na) are called the DC V K - I K and DC VNa - I Na curves T

of the potassium and the sodium ion-channel memris-
tors, respectively.
Fig. 12(a) shows the DC V K - I K curve of the potassium
ion-channel memristor when the above procedure is
T/2
ond half cycle. (g) Clockwise orientation of sub-loop 1 and
counter clockwise orientation of sub-loop 2 in the 1st quad-
rant, and clockwise orientation of the pinched loop in the 3rd
quadrant, respectively.

first QUARTER 2014 IEEE circuits and systems magazine 23

 Table 1.
Lobe area and orientation of the pinched hysteresis loops of the potassium ion-channel memristor for input voltage
v K (t) = A sin (2rft) with amplitude A = 50 mV.

f = 0 Hz f = 1 Hz f = 10 Hz

iK (mA) iK (mA) iK (mA)

2× 103 0.2 0.2

–60 –40 –20 0 20 40 60 –60 –40 –20 0 20 40 60

–150 –75 0 75 150 vK (mV) vK (mV) vK (mV)

–0.2 –0.2
–2 × 103
–0.4 –0.4

–4 × 103
–0.6 –0.6

–6 × 103 –0.8 –0.8

+=
Ar 0 Ar– = 0 Ar+= 3.004 × 10–4 Ar– = –0.689 Ar+ = 0.026 Ar– = –6.356

f = 50 Hz f = 200 Hz f = 500 Hz

iK (mA) iK (mA) iK (mA)

0.1 0.04 0.04

0.02
–60 –40 –20 0 20 40 60 vK (mV) 0.02

–60 –40 –20 0 20 40 60vK (mV)

–0.1 –60 –40 –20 0 20 40 60vK (mV)
–0.02

–0.2 –0.02
–0.04

–0.3 –0.06 –0.04

Ar+ = 1.064 Ar– = –7.019 Ar+ = 0.84 Ar– = –1.376 Ar+ = 0.369 Ar– = –0.522

f = 1 KHz f = 3 KHz f = 1 MHz

iK (mA) iK (mA) iK (mA)

0.04 0.04 0.04

0.02 0.02 0.02

–60 –40 –20 0 20 40 60 vK (mV) –60 –40 –20 0 20 40 60 vK (mV) –60 –40 –20 0 20 40 60vK (mV)

–0.02 –0.02 –0.02

–0.04 –0.04 –0.04

Ar+ = 0.186 Ar– = –0.258 Ar+ = 0.061 Ar– = –0.085 Ar+ = 1.759 × 10–4 Ar– = –9.777 × 10–5

24 IEEE circuits and systems magazine first QUARTER 2014

Table 2.
Lobe area and orientation of the pinched hysteresis loops of the sodium ion-channel memristor for input voltage
v Na (t) = A sin (2rft) with amplitude A = 120 mV.

f = 0 Hz f = 1 Hz f = 10 Hz
iNa(mA) iNa(mA) iNa(mA)
0.08 0.1 0.4

0.08
0.06 0.3
0.06
0.04 0.2
0.04
0.02 0.1
0.02

–130 –65 0 65 130v (mV) –130 –65 0 65 130 vNa(mV)

Na –130 –65 0 65 130 vNa(mV)

–0.02 –0.02 –0.1

Ar+ = 0 Ar– = 0 Ar+ = –0.426 Ar– = 3.612 × 10–4 Ar+ = –8.192 Ar– = 6.555 × 10–3

f = 50 Hz f = 200 Hz f = 500 Hz
iNa(mA) iNa(mA) iNa(mA)

0.5 0.1 0.1

0.4
–130 –65 0 65 130v (mV) –130 –65 0 65 130v (mV)
Na Na
0.3

0.2 –0.1 –0.1

0.1
–0.2 –0.2
–130 –65 0 65 130 vNa(mV)

–0.1 –0.3 –0.3

Ar+ = –16.931 Ar– = 2.471 Ar+ = –1.893 Ar– = 13.099 Ar+ = 5.358 Ar– = 6.171

f = 1 KHz f = 10 KHz f = 1 MHz

iNa(mA) iNa(mA) iNa(mA)

0.2 0.3 0.2

0.1 0.2
0.1
0.1
–130 –65 0 65 130 v (mV)
Na
–130 –65 0 65 130 vNa(mV) –130 –65 0 65 130vNa(mV)
–0.1
–0.1
–0.1
–0.2
–0.2

–0.3 –0.3 –0.2

Ar+ = 7.624 Ar– = 0.292 Ar+ = 1.711 Ar– = –1.156 Ar+ = 3.523 × 10–3 Ar– = –6.055 × 10–3

first QUARTER 2014 IEEE circuits and systems magazine 25

 IK (mA) I
1.0
+
VK INa IK IL
ICM
0 +
mV + + +
CM VNa GNa VK GK VL GL
V
– VM –
–1.0 – Na – K
ENa EK EL
–2.0 iK (mA) _
2.10–3
10–3
vK (a)
–3.0 0 mV I
–10–3
0 20 40
–4.0 +
–150 –100 –50 0 50 100 150 INa IK IL
(a) +
+ +
INa(mA) VNa GNa VK GK VL GL
V
0.1 – Na – K –
0.001 ENa EK EL
0.076 mA
0 –
–0.003
0.05 INa
–0.001 (b)
–100 –20.016 50
VNa mV Figure 13.  (a) Memristive Hodgkin-Huxley axon circuit model.
(b) DC Memristive Hodgkin-Huxley axon circuit model.
VNa
0
mV

–200 –100 0 84.215 mV 200

(b)

Figure 12.  (a) DC VK -I K curve of the potassium ion-channel I µA

150
memristor. (b) DC VNa - I Na curve of the sodium ion-channel
memristor.
100

50
V
0
applied over the range of input voltage - 150 mV # 0 mV
V K # 150 mV. The right bottom inset shows the magni- –50
fied plot of the V K - I K curve over 0 # V K # 50 mV. The
–100
positive and negative slopes of the DC V K - I K curve are –100 –50 0 50 100 150 200
printed in blue and red, respectively. Similarly, by apply- (a)
∆
ing above procedure to the sodium ion channel memris- I ′ = –I , µ A
150
tor, we obtain the DC VNa - I Na curve shown in Fig. 12(b)
over the range - 200 mV # VNa # 200 mV. The upper-left 100

inset is the enlargement of the DC VNa - I Na curve over 50

∆
- 100 mV # VNa # 50 mV. Note that the positive and nega- V ′ = –V , µ A
0
tive slopes of the DC VNa - I Na curve are printed in blue and 0
red, respectively. –50
The points (V K - I K ) on the red portion of the potas-
–100
sium DC V K - I K curve are locally-active [10] in the sense –100 –50 0 50 100 150 200
that it is possible to imbed the memristor in a passive (b)
circuit which amplifies the power of a small input sig- Figure 14. (a) V - I curve of memristive Hodgkin-Huxley
nal and produces an output signal with a larger power, axon circuit model. (b) 180˚-rotated V - I curve from (a) where
Vl and Il are corresponding voltage and current variables
at the expense of an informationless DC power sup- defined with modern reference convention.
ply, namely, the battery used to bias the potassium

26 IEEE circuits and systems magazine first QUARTER 2014

memristor at the prescribed DC equilibrium point. Simi-
larly, the points (VNa - I Na) on the red portion of the DC
VNa - I Na curve are locally active.
The memristive Hodgkin-Huxley axon circuit model
is reproduced from [5] and shown in Fig. 13(a). The DC
equivalent circuit obtained by deleting the capacitor
C M is shown in Fig. 13(b).
The DC V - I curve of the memristive Hodgkin-Huxley
axon circuit model was derived in [5] and reproduced
in Fig. 14(a). For ease of the comparison with the mod-
ern reference convention in the literature on Hodgkin-
Huxley axon circuit model, which is the reverse of those
originally adopted by Hodgkin and Huxley in their
classic 1952 paper [4], we have also plotted the corre-
sponding DC V l - I l curve in Fig. 14(b) using the modern
reference convention where V l _ - V and I l _ - I. Note
that this DC V l - I l curve resembles that of the vacuum
diode, and the pn junction, which explains and resolves
the “rectifier” anomaly that had mystified many distin-
guished physiologists including Hodgkin and Cole [11].
We remark that a “negative slope” on the DC V - I
curve of a memristor is only a sufficient condition for
local activity. An equilibrium point on a DC V - I curve
of a memristor with a positive slope can also be locally
Table 3.
Explicit Equations for L (K), R 1 (K), and R 2 (K) as functions of VK .
L (K) = 1
cn bn c n (VK ) = 4gr K n (VK ) 3 VK ,
(VK + E K + 10)e (VK + E K + 10)/10
b n (VK ) = 0.01 (n (VK ) - 1) e
10 ^e (VK + E K + 10)/10
- 1h
2
an + bn
R 1 (K) =
cn bn
0.01 (VK + E K + 10)
a n (VK ) = ,
exp " (VK + E K + 10) /10 , - 1
b n (VK ) = 0.125 exp " (VK + E K ) /80 ,,
R 2 (K) = 1
dK
d K = gr K n (VK ) 4,
a n (V K )
n (V K ) = , –
a n (VK ) + b n (VK )
E K = 12 mV, gr K = 36 mS
first QUARTER 2014 IEEE circuits and systems magazine 27
active if ReYK (f ; V K (Q)) 1 0 for the potassium memris-
tor, and ReYNa (f ; VNa (Q)) 1 0 for the sodium memristor
at some frequency f = f0 [12]. A memristor is defined
to be locally active if there exists at least one DC equi-
librium point Q where it is locally active. It follows that
both the potassium memristor and sodium memristor
are locally active.
IV. Small-Signal Equivalent Circuits
and Nyquist Plot of Ion-Channel Memristor
Small-signal equivalent circuits of ion-channel memris-
tors are fundamental concepts and tools for predicting
the nonlinear dynamics of the Hodgkin-Huxley axon.
Applying standard circuit analysis method [5], we can
derive the small-signal equivalent circuit about each
DC equilibrium point (V K , I K ) for the potassium ion-
channel memristor, and (VNa, I Na) for the sodium ion-
channel memristor. We can then derive the admittance
YK (s, Q K ) for the potassium memristor, and YNa (s, Q Na)
for the sodium memristor.
In this section, we analyze the small-signal equiva-
lent circuits and the Nyquist plot of the potassium ion-
channel memristor, the sodium-ion channel memristor
and the memristive Hodgkin-Huxley axon circuit model.
- (VK + EK +110)/10 o - 0.125n (VK )
e (VK + EK )/80 ,
^e - 1h 80
Potassium Ion-Channel
Memristor Small-Signal
Equivalent Circuit at
Equilibrium Point VK (Q)
δiK
+
L(K )
δvK
R1(K ) R2(K )
 The admittance of the small-signal equivalent circuit
of the potassium ion-channel memristor about an equi-
2 × 103
librium point Q K is given by [5]
Resistances (Ri(K )) kΩ

R2(K )
Inductance (L(K )) H

1 × 103 1
YK (s, Q K ) = + 1 . (9)
s L (K) + R 1 (K) R 2 (K)
0 We can recast (9) as a rational function of s:
L(K ) R1(K )
b 1 (K) s + b 0 (K)
–1 × 103 YK (s, Q K ) = , (10a)
a 1 (K) s + a 0 (K)

–2 × 103 where
–20 0 20
vK (mV)
(a) a 0 (K) = R 1 (K) R 2 (K), a 1 (K) = L (K) R 2 (K) (10b)
1 × 106
b 0 (K) = R 1 (K) + R 2 (K), b 1 (K) = L (K) . (10c)

The explicit formula for calculating the inductance

5 × 105
Coefficients (ai)

L (K), and resistances R 1 (K) and R 2 (K) are given in

a0 Table 3 as function of V K at the DC equilibrium point
0
QK .
a1 Substituting s = i~ = i (2rf) for the complex fre-
–5 × 105
quency s in (10a), we obtain small-signal admittance
function YK (f ; V K ) of the potassium ion-channel at the
–1 × 106
–20 –10 0 10 20 DC equilibrium point v K = V K (Q K ) . The real and imagi-
vK (mV) nary parts of YK (f ; V K ) as a function of the frequency f
(b)
is given by
1 × 104
(a 0 b 0 + a 1 b 1 ^2rf h2) (a 0 b 1 - a 1 b 0) ^2rf h
YK (f; V K ) = = G + i= G .
5× 103 (a 20 + a 21 ^2rf h2) (a 20 + a 21 ^2rf h2)
Coefficients (bi)

b0
(11a)
0

b1 Here the symbol K attached to the coefficients a i and

–5 × 103 b i is deleted to avoid clutter. Hence,

–1 × 104
–20 –10 0 10 20
vK (mV)
(c)
Figure 15.  (a) Variation of the inductance L (K), resistance
R 1 (K) and resistance R 2 (K) as a function of the potassium
ion-channel memristor at the DC equilibrium voltage VK (b)
Variation of the coefficients a 0, a 1 as a function of the potas-
sium ion-channel memristor at the DC equilibrium voltage
VK . (c) Variation of the coefficients b 0, b 1 as a function of
the potassium ion-channel memristor at the DC equilibrium
voltage VK .
A. Small-Signal Equivalent Circuit and Nyquist Plot
of the Potassium Ion-Channel Memristor
The small-signal equivalent circuit of the potassium
ion-channel memristor was derived in [5] and repro-
duced in Table 3. Observe that the values of the three
circuit elements L (K), R 1 (K) and R 2 (K) about each DC
equilibrium point at (V K (Q K ), I K (Q K )) can be calculated
via the explicit formulas given in Table 3.
a 0 b 0 + a 1 b 1 ^2rf h2
Re YK (f; V K ) = (11b)
a 20 + a 21 ^2rf h2
and
(a 0 b 1 - a 1 b 0) ^2rf h
Im YK (f; V K ) = . (11c)
a 20 + a 21 ^2rf h2
The small-signal equivalent circuit of the potassium
ion-channel memristor shown in Table 3 is composed of
one inductance and two resistances respectively. The
parameters of the small-signal equivalent circuits and
the value of the coefficients depend on v K = V K (Q K ) at
the equilibrium point Q K . The variation of the induc-
tance L (K), resistances R 1 (K) and R 2 (K), coefficients
a 0, a 1 and coefficients b 0, b 1 of the small-signal potas-
sium admittance function defined in (9) and (10) are
shown in Figs. 15(a), (b) and (c), respectively, as a
­function of the DC potassium ion-channel memristor
voltage V K .

28 IEEE circuits and systems magazine first QUARTER 2014

 VK = –12 mV VK = –16.57443 mV VK = –17.34305 mV
δiK δiK δiK

+ + +
L(K ) L(K ) L(K )
δvK 6.43 H δvK 2.302 H δvK 1.981 H

R1(K ) R2(K ) R1(K ) R2(K ) R1(K ) R2(K )

– – –
1.178 KΩ 2.727 KΩ 0.445 KΩ 1.207 KΩ 0.387 KΩ 1.066 KΩ

Re YK (f ; VK (QK)) Re YK (f ; VK (QK)) Re YK (f ; VK (QK))

1.3 3.2 3.6

1.05 2.575 2.9

mS

mS

mS
0.8 1.95 2.2

0.55 1.325 1.5

f f f
0.3 0.7 0.8
–1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 KHz

Im YK (f ; VK (QK)) Im YK (f ; VK (QK)) Im YK (f ; VK (QK))

0.5 1.2 1.4

0.25 0.6 0.7

f f
mS

mS

mS
0 0 0
KHz KHz
–0.25 –0.6 –0.7
f
–0.5 –1.2 –1.4
–1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 –1.6 –0.8 0 0.8 1.6
0.5 1.2 1.4
Im YK (f ; VK (QK))

Im YK (f ; VK (QK))

Im YK (f ; VK (QK))

f = –0.029 f = –0.031 f = –0.031

0.25 0.6 0.7
–∞

f=0 f=0 f=0

–∞

∞ –∞

0 0 0
–∞

–0.25 f = 0.029 –0.6 –0.7

f = 0.031 f = 0.031
–0.5 –1.2 –1.4
0.3 0.55 0.8 1.05 1.3 0.7 1.325 1.95 2.575 3.2 0.8 1.5 2.2 2.9 3.6
Re YK (f ; VK (QK)) Re YK (f ; VK (QK)) Re YK (f ; VK (QK))
(a) (b) (c)

Figure 16.  Small-signal equivalent circuit, admittance frequency response Re YK (f ; VK (Q K )), Im YK (f ; VK (Q K )), and the Nyquist
plot of the potassium ion-channel memristor at (a) VK = - 12 mV, (b) VK = - 16.57443 mV and (c) VK = - 17.34305 mV.

The real part Re YK (f; V K ) and imaginary part admittance YK (f; V K ) . Figs. 16(a), (b) and (c) show the
Im YK (f; V K ) obtained from the potassium admittance small-signal equivalent circuit, small-signal admittance
function YK (f; V K ) is called the small-signal admittance frequency response YK (f; V K ) = Re YK (f; V K ) + Im YK (f; V K ),
frequency response of the potassium ion-channel at and the Nyquist plot of the potassium ion-channel
equilibrium point v K = V K (Q K ) . When, the real part memristor at V K = - 12 mV, V K = - 16.57443 mV and
Re YK (f; V K ) and imaginary part Im YK (f; V K ) of the V K = - 17.34305 mV, respectively. These three volt-
admittance function of YK (f; V K ) are plotted on the ages are chosen to ­ correspond to the three equi-
horizontal and vertical axes of the Cartesian coordi- librium voltages Vm = 0, Vm = - 4.57443 mV, and
nate system with the frequency f as a parameter, the Vm = - 5.34305 mV of the DC memristive Hodgkin-Hux-
resulting plot is generally called the Nyquist plot of the ley axon circuit model.

first QUARTER 2014 IEEE circuits and systems magazine 29

 Table 4. 1 × 105
Explicit Equations for L 1 (Na), R 1 (Na), L 2 (Na), R 2 (Na)

Inductances (Li (Na)) H

and R 3 (Na) as functions of VNa . L1(Na) L2(Na)
5 × 104

L 1 (Na) = 1 c m (VNa) = 3gr Na m (VNa) 2 h (VNa) VNa, 0

cm bm

b m (VNa) = 0.1 (m (VNa) - 1) –5 × 104

(VNa - E Na + 25) e (VNa - E Na + 25)/10

e - (VNa - E Na +125)/10 o –1 × 105
10 (e (VNa - E Na + 25)/10 - 1) 2 (e - 1) –100 0 100
(VNa - E Na)/18 vNa (mV)
- 4m (VNa) e , (a)
18
1 × 103

Resistances (Ri (Na)) KΩ

L 2 (Na) = 1 C h (VNa) = gr Na m (VNa) 3 VNa,
Ch bh 500 R3 (Na)

h (VNa) e (VNa - E Na + 30)/10 0

b h (VNa) =
10 (e (VNa - E Na + 30)/10 + 1) 2
(VNa - E Na)/20 R1 (Na)
- 0.07 (h (VNa) - 1) e –500 R2 (Na)
20
–1 × 103
a m + b m a m (VNa) = 0.1 (VNa - E Na + 25) –150 –100 –50 0 50 100 150
,
R 1 (Na) =
c m +b m exp "^VNa - E Na + 25 h /10 , - 1 vNa (mV)
(b)
b m (VNa) = 4exp "^VNa - E Na h /18 ,, 0

a h (VNa) = 0.07exp "^VNa - E Na h /20 ,, a2 a0

ah + bh –2.5 × 105
R 2 (Na) =
Coefficients (ai)

ch bh b h (VNa) = 1 ,
exp "^VNa - E Na + 30 h /10 , + 1
a1
–5 × 105

R 3 (Na) = 1 d Na = gr Na m (VNa) 3 h (VNa),

d Na –7.5 × 105

a m (VNa)
m (VNa) = –1 × 106
a m (VNa) + b m (VNa) 0 20 40 60 80 100 120
vNa (mV)
a h (VNa) (c)
h (VNa) =
a h (VNa) + b h (VNa) b2
0
E Na = 115 mV, gr Na = 120 mS b0
Coefficients (bi)

b1
Sodium Ion-Channel
Memristor Small-Signal –5 × 105
Equivalent Circuit at
Equilibrium Point VNa(Q)

–1 × 106
δiNa 0 20 40 60 80 100 120
vNa (mV)
+ (d)

L1(Na) L2(Na)
Figure 17.  (a) Inductances L 1 (Na), L 2 (Na) as a function of the
δvNa sodium ion-channel memristor at the DC equilibrium voltage
VNa . (b) Resistances R 1 (Na), R 2 (Na), R 3 (Na) as a function of
R1(Na) R2(Na) R3(Na) the sodium-ion channel memristor at the DC equilibrium volt-
– age VNa . (c) Coefficients a 0, a 1, a 2, as a function of the sodium
ion-channel memristor at the DC equilibrium voltage VNa .
(d) Coefficients b 0, b 1, b 2 as a function of the sodium ion-­
channel memristor at the DC equilibrium voltage VNa .

30 IEEE circuits and systems magazine first QUARTER 2014

 VNa = 115 mV VNa = 110.42557 mV VNa = 109.65695 mV
δiNa δiNa δiNa

+ + +
L1(Na) L2(Na) L1(Na) L2(Na) L1(Na) L2(Na)
δvNa –0.549 H 118.978 H δvNa –0.216 H 23.478 H δvNa –0.189 H 18.28 H

R1(Na) R2(Na) R3(Na) R1(Na) R2(Na) R3(Na) R1(Na) R2(Na) R3(Na)

– – –
–2.317 KΩ 13.971 KΩ 94.258 KΩ –0.736 KΩ 3.02 KΩ 26.982 KΩ –0.621 KΩ 2.411 KΩ 22.314 KΩ

Re YNa(f ; VNa(QNa)) Re YNa(f ; VNa(QNa)) Re YNa(f ; VNa(QNa))

0 0 0

–0.125 –0.35 –0.4

mS

mS

mS
–0.25 –0.7 –0.8

–0.375 –1.05 –1.2

f f f
–0.5 –1.4 –1.6
–1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 KHz

Im YNa(f ; VNa(QNa)) Im YNa(f ; VNa(QNa)) Im YNa(f ; VNa(QNa))

0.3 0.7 0.9

0.15 0.35 0.45

mS

mS

mS
0 0 0

–0.15 –0.35 –0.45

f f f
–0.3 –0.7 –0.9
–1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 KHz
0.3 0.8 1.4
Im YNa(f ; VNa(QNa))

Im YNa(f ; VNa(QNa))

Im YNa(f ; VNa(QNa))

f = 0.552 0.7 f = 0.535

0.15 f = 0.678 0.4 f = 0.049

+∞
f = 0.0481
+∞
–∞ +∞

f = 0.041
0 f=0 0 f=0 0 f=0
–∞

–∞
f = –0.041 f = –0.0481 f = –0.049
–0.15 f = –0.678 –0.4 –0.7 f = –0.535
f = –0.552
–0.3 –0.8 –1.4
.5
.4
.3
.2
.1
0

.4

.9

.4

0
1

.6

.7

2
0.

0.

.1

.2

0.
–0
–0
–0
–0
–0

–1

–0

–0

–1

–0
–1

–0

Re YNa(f ; VNa(QNa)) Re YNa(f ; VNa(QNa)) Re YNa(f ; VNa(QNa))

(a) (b) (c)

Figure 18.  Small-signal equivalent circuit, ReYNa (f ; VNa (Q Na)), ImYNa (f ; VNa (Q Na)), and the Nyquist plot of the sodium ion-channel
memristor at (a) VNa = 115 mV, (b) VNa = 110.42557 mV and (c) VNa = 109.65695 mV.

B. Small-Signal Equivalent Circuit and Nyquist Plot

of the Sodium Ion-Channel Memristor δi
The small-signal equivalent circuit of the sodium +
ion-channel memristor was derived in [5] and repro-
R1(Na) L1(Na)

R2(Na) L2(Na)

duced in Table 4. Observe that the values of the five CM L(K )

circuit elements L 1 (Na), L 2 (Na), R 1 (Na), R 2 (Na) δV GL
R3(Na)

R2(K)

and R 3 (Na) about each DC equilibrium point Q Na at R1(K)

(VNa (Q Na), I Na (Q Na)) in Fig. 12(b) can be calculated via
–
the explicit formulas given in Table 4.
The expression for the admittance of the small-signal
Figure 19.  Small-signal Hodgkin-Huxley axon circuit model.
equivalent circuit of the sodium ion-channel memristor

first QUARTER 2014 IEEE circuits and systems magazine 31

 Table 5.
Formulas for calculating coefficients of Y (s ; Vm) of the Hodgkin-Huxley Axon Circuit Model.

a 0 = R 1 (Na) R 2 (Na) R 3 (Na) R 1 (K) R 2 (K)

a 1 = R 3 (Na) R 2 (K) [L (K) R 1 (Na) R 2 (Na) + R 1 (K) {R 1 (Na) L 2 (Na) + R 2 (Na) L 1 (Na)}]

a 2 = R 3 (Na) R 2 (K) [R 1 (K) L 1 (Na) L 2 (Na) + L (K) {R 1 (Na) L 2 (Na) + R 2 (Na) L 1 (Na)}]

a 3 = R 3 (Na) R 2 (K) L 1 (Na) L 2 (Na) L (K)

b 0 = R 1 (K) R 2 (K) [R 1 (Na) R 2 (Na) + R 2 (Na) R 3 (Na) + R 3 (Na) R 1 (Na)] + R 1 (Na) R 2 (Na) R 3 (Na) [R 1 (K) + R 2 (K)]
+ R 1 (Na) R 2 (Na) R 3 (Na) R 1 (K) R 2 (K) G L

b 1 = R 3 (Na)6R 1 (K)L 2 (Na) R 1 (Na) + R 2 (Na)" L (K)R 1 (Na) + R 1 (K) L 1 (Na) , + R 2 (K) " L 1 (Na) R 2 (Na) + L 2 (Na) R 1 (Na) ,@
+ R 3 (Na) R 2 (K) 66L (K) R 1 (Na) R 2 (Na) + R 1 (K) " L 1 (Na) R 2 (Na) + L 2 (Na) R 1 (Na) ,@G L + R 1 (K) R 1 (Na) R 2 (Na) C M@
+ L (K) R 2 (K) 6R 1 (Na) R 2 (Na) + R 2 (Na) R 3 (Na) + R 1 (Na) R 2 (Na)@
+ R 1 (K) R 2 (K) 6L 1 (Na) " R 2 (Na) + R 3 (Na) , + L 2 (Na) " R 1 (Na) + R 3 (Na) ,@

"^ R 1 (K) L 1 (Na) L 2 (Na) + L (K) " R 1 (Na) L 2 (Na) + R 2 (Na) L 1 (Na) ,h, G L
b 2 = R 2 (K)R 3 (Na) = G
+ " L (K) R 1 (Na) R 2 (Na) + R 1 (K) 6R 1 (Na) L 2 (Na) + R 2 (Na) L 1 (Na)@, C M
+ R 3 (Na) 6L (K) L 2 (Na) R 1 (Na) + L 1 (Na) " L (K) R 2 (Na) + L 2 (Na) 6R 1 (K) + R 2 (K)@,@
+ R 2 (K) 6L 1 (Na) L 2 (Na) R 1 (K) + L (K) " L 1 (Na) 6R 2 (Na) + R 3 (Na)@ + L 2 (Na) 6R 1 (Na) + R 3 (Na)@,@

b 3 = L 1 (Na) L 2 (Na) L (K) [R 2 (K) + R 3 (Na)]

+ R 3 (Na) R 2 (K) [L 1 (Na) L 2 (Na) L (K) G L + [R 1 (K) L 1 (Na) L 2 (Na) + L (K) {R 1 (Na) L 2 (Na) + R 2 (Na) L 1 (Na)}] C M ]

b 4 = R 3 (Na) R 2 (K) L 1 (Na) L 2 (Na) L (K) C M

from the Hodgkin-Huxley axon circuit model about an Explicit formulas for calculating L 1 (Na), R 1 (Na),
equilibrium point Q Na is given by [5] L 2 (Na), R 2 (Na) and R 3 (Na) are given in Table 4, as a
function of VNa at the DC equilibrium point Q Na .
YNa (s, Q Na) = 1 Substituting s = i~ = i (2rf) in Eq. (13a) and rearrang-
s L 1 (Na) + R 1 (Na)
ing YNa (f; VNa) into its real and imaginary parts we obtain
+ 1 + 1 . (12)
s L 2 (Na) + R 2 (Na) R 3 (Na)
(a 0 - a 2 (2rf) 2) (b 0 - b 2 (2rf) 2) + a 1 b 1 (2rf) 2
YNa (f; VNa) = = G
We can recast the admittance function (12) as a rational ^a 0 - a 2 (2rf) 2 h2 + a 21 (2rf) 2
function of s: 6(a 0 - a 2 (2rf) 2) b 1 - a 1 (b 0 - b 2 (2rf) 2)@ (2rf)
+ i= G,
^a 0 - a 2 (2rf) 2 h2 + a 21 (2rf) 2
2
b 2 (Na) s + b 1 (Na) s + b 0 (Na)
YNa (s, Q Na) = , (13a)
a 2 (Na) s 2 + a 1 (Na) s + a 0 (Na)  (14a)

where where we have deleted the symbol Na attached to the

coefficients a i and b i to avoid clutter. Hence,
a 0 (Na) = R 1 (Na) R 2 (Na) R 3 (Na)
a 1 (Na) = (L 1 (Na) R 2 (Na) + L 2 (Na) R 1 (Na)) R 3 (Na) 4 (13b) (a 0 - a 2 (2rf) 2) (b 0 - b 2 (2rf) 2) + a 1 b 1 (2rf) 2
Re YNa (f;VNa) =
a 2 (Na) = L 1 (Na) L 2 (Na) R 3 (Na) ^a 0 - a 2 (2rf) 2 h2 + a 21 (2rf) 2
(14b)
_
b 0 (Na) = R 1 (Na) R 2 (Na) + R 1 (Na) R 3 (Na) b
and
+ R 2 (Na) R 3 (Na) bb
b 1 (Na) = L 1 (Na) R 2 (Na) + L 2 (Na) R 1 (Na) ` (13c) 6(a 0 - a 2 (2rf) 2) b 1 - a 1 (b 0 - b 2 (2rf) 2)@ (2rf)
b Im YNa (f;VNa) = .
+ L 1 (Na) R 3 (Na) + L 2 (Na) R 3 (Na) b (a 0 - a 2 (2rf) 2) 2 + a 21 (2rf) 2
b 2 (Na) = L 1 (Na) L 2 (Na) b (14c)
a
32 IEEE circuits and systems magazine first QUARTER 2014
 Vm = 0
δi

+
L1(Na) L2(Na) L(K )
CM –0.549 H 118.978 H 6.43 H GT
δv 1 µF 0.677 mS

R1(Na) R2(Na) R1(K )

– –2.317 KΩ 13.971 KΩ 1.178 KΩ

GT = G2(K ) + G3(Na) + GL, where

G2(K ) = 1 1 , G3(Na) = 1 1 , GL = 1 = 0.3 mS

= =
R2(K ) 2.727 KΩ R3(Na) 94.258 KΩ RL

(a)

Vm = –4.57443 mV
δi

+
L1(Na) L2(Na) L(K )
–0.216 H 23.478 H 2.302 H GT
CM
δv 1.165 mS
1 µF
R1(Na) R2(Na) R1(K )
–0.736 KΩ 3.02 KΩ 0.445 KΩ
–

GT = G2(K ) + G3(Na) + GL, where

G2(K ) = 1 1 , G3(Na) = 1 1 , GL = 1 = 0.3 mS

= =
R2(K ) 1.207 KΩ R3(Na) 26.982 KΩ RL

(b)
Vm = –5.34305 mV
δi

+
L1(Na) L2(Na) L(K )
CM –0.189 H 18.28 H 1.981 H GT
δv 1 µF 1.283 mS

R1(Na) R2(Na) R1(K )

–0.621 KΩ 2.411 KΩ 0.387 KΩ
–

GT = G2(K ) + G3(Na) + GL, where

G2(K ) = 1 1 , G3(Na) = 1 1 , GL = 1 = 0.3 mS

= =
R2(K ) 1.066 KΩ R3(Na) 22.314 KΩ RL

(c)

Figure 20. Small-signal memristive Hodgkin-Huxley axon circuit model at (a) Vm = 0, VK = -12 mV, and VNa = 115 mV, (b) Vm =
-4.57443 mV, VK = -16.57443 mV, and VNa = 110.42557 mV, (c) Vm = -5.34305 mV, VK = -17.34305, and VNa = 109.65695 mV,
respectively.

first QUARTER 2014 IEEE circuits and systems magazine 33

 Vm = 0 Vm = –4.57443 mV Vm = –5.34305 mV

Re Y (f ; Vm (Q)) Re Y (f ; Vm (Q)) Re Y(f ; Vm (Q))

1.3 3 3

0.975 2 2

mS
mS

0.65 1 1
f f
0.325 0 0
f KHz KHz
0 –1 –1
–1.6 –0.8 0 0.8 1.6 KHz –1.6 –0.8 0 0.8 1.6 –1.6 –0.8 0 0.8 1.6

Im Y(f ; Vm (Q)) Im Y (f ; Vm (Q)) Im Y (f ; Vm (Q))

5 10.5 10.5

3 5.25 5.25
1 f f f
0
mS

0 0
–1 KHz KHz KHz
–5.25 –5.25
–3
–5 –10.5 –10.5
–1.6 –0.8 0 0.8 1.6 –1.6 –0.8 0 0.8 1.6 –1.6 –0.8 0 0.8 1.6

5 10.5 10.5
+∞

+∞

+∞
3 f = 0.0241
Im Y (f ; Vm (Q))

Im Y (f ; Vm (Q))

Im Y (f ; Vm (Q))
f = 0.455 5.25 5.25 f = 0.2417
f = 0.1516 f = 0.0883 f = 0.0238 f = 0.0933
1 f = 0.054 f = 0.022 f=0
0 f=0 0
f=0 mS 0
–1 f = –0.054 f = –0.022 f = –0.0241
f = –0.1516 f = –0.0933
f = –0.455 –5.25 f = –0.0883 f = –0.0238 –5.25
–3 f = –0.2417

–∞
–∞
∞

–5 –10.5 –10.5
0.2 0.45 0.7 0.95 1.2 –1 0 1 2 3 –1 0 1 2 3
Re Y (f ; Vm (Q)) Re Y (f ; Vm (Q)) Re Y(f ; Vm (Q))
(a) (b) (c)

Figure 21.  Small-signal Admittance Y (f ; Vm (Q)) frequency response and Nyquist plot of the Hodgkin-Huxley axon at (a) Vm = 0,
VK = -12 mV, and VNa = 115 mV, (b) Vm = - 4.57443 mV, VK = -16.57443 mV, and VNa = 110.42557 mV, (c) Vm = -5.34305 mV,
VK = -17.34305, and VNa = 109.65695 mV, respectively.

The parameters of the small-signal equivalent cir- Vm = - 4.574443 mV, and Vm = - 5.34305 mV of the DC
cuit and coefficient of the sodium ion-channel memris- memristive Hodgkin-Huxley axon circuit model.
tor defined in (12) and (13) depend on v Na = VNa (Q Na)
at the equilibrium point Q Na . Figs. 17(a)–(d) show the C. Small-Signal Equivalent Circuit and Nyquist Plot
variation of the inductances L 1 (Na), L 2 (Na), resis- of the Hodgkin-Huxley Axon Circuit Model
tances R 1 (Na), R 2 (Na) R 3 (Na), coefficients a 0, a 1, a 2 The small-signal Hodgkin-Huxley equivalent circuit
and coefficients b 0, b 1, b 2 of the small-signal sodium- about an external DC voltage V is shown in Fig. 19, where
ion channel memristor as a function of the sodium the potassium memristor and potassium Battery E K in
ion-channel memristor DC equilibrium voltage VNa, Fig. 13(a) are replaced by the small-signal equivalent
respectively. circuit of the potassium memristor about its DC equilib-
Fig. 18 shows the small-signal equivalent circuit, small- rium point V K = V - E K . Similarly, the sodium memristor
signal admittance frequency response, Re YNa (f; VNa) vs. and sodium battery E Na in Fig. 13(a) are replaced by the
f, Im YNa (f; VNa) vs. f, and the Nyquist plot, Im YNa (f; VNa) small-signal equivalent circuit of the sodium memristor
vs. Re YNa (f; VNa) of the sodium ion-channel memristor about its DC equilibrium point VNa = V + E Na .
at (a) VNa = 115 mV, (b) VNa = 110.42557 mV and (c) The admittance Y (s; Vm (Q)) at the equilibrium point
VNa = 109.65695 mV. These three voltages are chosen Q at V = Vm (Q) of the memristive Hodgkin-Huxley axon
correspond to the three DC equilibrium voltages Vm = 0, circuit model in Fig. 19 is given by [5]

34 IEEE circuits and systems magazine first QUARTER 2014

 Y (s; Vm (Q)) = G T + 1 + 1 three inductances L 1 (Na) L 2 (Na), L (K), three resis-
sL (K) + R 1 (K) sL 1 (Na) + R 1 (Na)
+ 1 + sC M ,
sL 2 (Na) + R 2 (Na)
where G T = G 2 (K) + G 3 (Na) + G L, G 2 (K) = 1/R 2 (K),
G 3 (Na) = 1/R 3 (Na), and G L = 1/R L The circuit elements
L (K), R 1 (K) and R 2 (K) from the small-signal equiva-
lent circuit of the potassium memristor and the circuit
element values L 1 (Na), L 2 (Na), R 1 (Na), R 2 (Na) and
R 3 (Na) from the small-signal equivalent circuit of the
sodium memristor are calculated from explicit formulas
given in Table 3 and Table 4, respectively.
We can recast (15) into the form of a rational function
4 3 2
Y (s; Vm) = b 4 s + b3 3 s + b2 2 s b 1 s b 0 , (16)
+ +
a3 s + a2 s + a1 s + a0
where the 9 coefficients (a 0, a 1, a 2, a 3, b 0, b 1, b 2, b 3, b 4)
depend on the circuit elements in the small-signal cir-
cuit model in Fig. 19. The explicit formula defining each
coefficient is reproduced in Table 5 from [5].
Substituting s = i~ in (16), we obtain the follow-
ing small-signal Admittance of the Hodgkin-Huxley axon
circuit model at the DC equilibrium membrane voltage
V = Vm (Q):
b 4 (i~) 4 + b 3 (i~) 3 + b 2 (i~) 2 + b 1 (i~) + b 0
Y (i~; Vm) =
a 3 (i~) 3 + a 2 (i~) 2 + a 1 (i~) + a 0
(b 0 - b 2 ~ 2 + b 4~ 4) - i (b 1 - b 3 ~ 2) ~
(17a)
= .
(a 0 - a 2 ~ 2) + i (a 1 - a 3 ~ 2) ~
Separating Eq. (17) into its real and imaginary parts, we
obtain (17b) and (17c), as shown at the bottom of the page.
The above equations are defined in terms of s = i~. If
we substitute ~ = 2rf, then the above equations can be
obtained as a function of frequency f. All graphs in this
paper are plotted as a function of frequency f.
The small-signal memristive Hodgkin-Huxley axon
circuit model in Fig. 19 is composed of a Capacitor C M ,
tances R 1 (Na) R 2 (Na), R 3 (Na), two resistances R 1 (K),
R 2 (K), and a conductance (G L) . The circuit elements
and coefficients parameters depend on the DC axon
membrane voltage Vm = Vm (Q) at the equilibrium point
Q. Figs. 20(a)–(c) show the value of the circuit elements
at equilibrium point Vm = 0, Vm = - 5.34305 mV and
Vm = - 4.57443 mV, respectively. The corresponding ad-
mittance Y (f ; Vm (Q)) = Re Y (f ; Vm (Q)) + i Im Y (f ; Vm (Q))
frequency response and Nyquist plot of the small-signal
circuit model of the memristive Hodgkin-Huxley axon
circuit model are shown in Figs. 21(a)–(c), respectively.
The first column in Fig. 21 at the DC equilibrium volt-
age Vm = 0 corresponds to the DC steady state voltage
measured across the Hodgkin-Huxley squid axon circuit
model in Fig. 13(a), when I = 0. Hence, when there is no
external stimulation signal (I = 0), the squid axon has
a normalized axon membrane voltage Vm = 0. When a
small signal current source I = I ext < 0 is applied across
the Hodgkin-Huxley circuit model3, Fig. 13(a), which
corresponds to an external excitation signal, the DC
memristive Hodgkin-Huxley axon circuit model in Fig.
13(b) is calculated to be a negative voltage Vm < 0. In
particular, when I ext = - 7.8293 nA, our calculation
gives a corresponding DC squid axon membrane volt-
age Vm = - 4.57443 mV. This corresponds to column 2
in Fig. 21 [5], [6]. At I ext = - 9.77004 nA, our calculation
gives Vm = - 5.34305 mV. This corresponds to column 3
of Fig. 21 [5], [6].
Observe that Re Y (f ; Vm (Q)) = 0 at Vm = - 4.57443 mV
in Fig. 21(b). Moreover, our calculation shows Re Y
(f ; Vm (Q)) 1 0 when Vm = - (4.57443 + DVm) mV for small
positive increments in Vm . It follows from the local activ-
ity principle that the Hodgkin Huxley axon circuit model
is at the onset of the edge of chaos at Vm = - 4.57443 mV,
a ­fundamental deep result investigated extensively in [6].
3
This external excitation signal summarizes the external stimulations
from neighboring neurons via the neuron’s dendrites.

(a 0 - a 2 ~ 2) (b 0 - b 2 ~ 2 + b 4 ~ 4) + (a 1 - a 3 ~ 2) (b 1 - b 3 ~ 2) ~ 2
Re Y (i~; Vm) =
(a 0 - a 2 ~ 2) 2 + (a 1 - a 3 ~ 2) 2 ~ 2
(a 3 b 3 - a 2 b 4 ) ~ 6 + (a 0 b 4 - a 1 b 3 + a 2 b 2 - a 3 b 1 ) ~ 4 + (a 1 b 1 - a 0 b 2 - a 2 b 0 ) ~ 2 + a 0 b 0 
= (17b)
a 23 ~ 6 + (a 22 - 2a 1 a 3) ~ 4 + (a 21 - 2a 0 a 2) ~ 2 + a 20

[(a 1 - a 3 ~ 2) (b 0 - b 2 ~ 2 + b 4 ~ 4) - (a 0 - a 2 ~ 2) (b 1 - b 3 ~ 2)] ~
Im Y (i~;Vm) = -
(a 0 - a 2 ~ 2) 2 + (a 1 - a 3 ~ 2) 2 ~ 2

a 3 b 4 ~ 7 - ( a 1 b 4 - a 2 b 3 + a 3 b 2 ) ~ 5 - (a 0 b 3 - a 1 b 2 + a 2 b 1 - a 3 b 0 ) ~ 3 + ( a 0 b 1 - a 1 b 0 ) ~ (17c)
=
a 23 ~ 6 + (a 22 - 2a 1 a 3) ~ 4 + (a 21 - 2a 0 a 2) ~ 2 + a 20

first QUARTER 2014 IEEE circuits and systems magazine 35

 Observe also that Re Y (f ; Vm (Q)) = 0 and
Im Y (f ; Vm (Q)) = 0 at Vm = - 5.34305 mV in Fig. 21(c).
This implies that the admittance function Y (s; Vm (Q)) of
the Hodgkin-Huxley axon circuit model has a zero s = 0
(or equivalently, the impedance function Z (s; Vm (Q)) has
a pole at s = 0) at Vm = - 5.34305 mV. This corresponds
to the right boundary of the edge of chaos domain, and
the onset of a sub-critical Hopf Bifurcation, which is the
mechanism which gives birth to the action potential in
the giant squid axon [6].
V. Conclusion
Since synapse and axons are the two fundamental com-
ponents of neurons responsible for computation and
information processing in the brain, and since both syn-
apses and axons can be emulated electronically by locally-
passive memristors, and locally-active memristors [12],
respectively, it follows that, from an information process-
ing, memory, and learning perspective, brains are made of
memristors. Moreover, since memristors are nano electronic
devices, the aspect ratio between the “footprints” of an elec-
tronic neuron and an electronic synapse is more than three
orders of magnitude, which is compatible with the corre-
sponding aspect ratio between a biological neuron, and a
biological synapse, it follows that memristors are the right
stuff for building brain-like machines. The race is now on for
the commercialization of low-power intelligent memristor-
based brain-like computers that can outperform digital
supercomputers on such routine tasks as facial recognition
and associative memories that our brain excels.
Acknowledgment
This work was supported in part by the US Air Force Grant
FA9550-13-1-0136 and two National Research Foundation
of Korea (NRF) grants funded by the Korea government
(No. 2013R1A1A2062282 and 2013R1A2A2A01068683).
Maheshwar Pd. Sah received the B.E.
in Electronics and Communication Engi-
neering from Pokhara University, Nepal
in 2005, M.E. and Ph.D. in Electronics
Engineering from Chonbuk National Uni-
versity, Republic of Korea in 2010 and
2013 respectively. He is currently working as a Postdoc-
toral scholar in Chonbuk National University, Republic
of Korea. His main research interests include Circuit
design, Cellular Neural Network, Analog viterbi decoder,
analysis of Memristor and Memristive System.
Hyongsuk Kim received the Ph.D. degree in Electrical
Engineering from the University of Missouri, Columbia,
in 1992. Since 1993, he has been a Professor with the
Division of Electronics Engineering, Chonbuk National
36 IEEE circuits and systems magazine first QUARTER 2014
University, Republic of Korea. From
2000 to 2002 and again from 2009 to
2010, he was with the Nonlinear Elec-
tronics Laboratory, EECS Department,
University of California, Berkeley, as a
Visiting Scholar. His current research
interests include memristors and its application to Cel-
lular Neural/Nonlinear Networks.
Leon O. Chua (Fellow, IEEE) received
the M.S. degree from the Massachusetts
Institute of Technology, Cambridge, in
1961 and the Ph.D. degree from the Uni-
versity of Illinois at Champaign-Urbana
in electrical engineering, in 1964. He
has been a Professor at the University of California
Berkeley since 1971. In 2011, he was appointed a Dis-
tinguished Professor at the Technical University of
Münich. He was awarded seven patents and 14 honorary
doctorates. When not immersed in science, he relaxes
by searching for Wagner’s leitmotifs, musing over Kan-
dinsky’s chaos, and contemplating Wittgenstein’s inner
thoughts. Prof. Chua received many awards including
the first recipient of the Gustav Kirchhoff award, the
Guggenheim Fellow award, and the European EC Marie
Curie Fellow award. He was elected a foreign member of
the Academia Europaea and of the Hungarian Academy
of the Sciences. He was elected Confrerie des Cheva-
liers du Tastevin in 2000.
References
[1] E. R. Kandel, J. H. Schwartz, and T. M. Jessell, Principles of Neural
Science, 4th ed. New York: McGraw-Hill, 2000.
[2] L. Chua. “ Memristor, Hodgkin-Huxley, and edge of chaos,” Nanotech-
nology, to be published.
[3] S. H. Jo, T. Chang, I. Ebong, B. B. Bhadviya, P. Mazumder, and W. Lu,
“Nanoscale memristor device as synapse in neuromorphic systems,”
Nano Lett. vol. 10, no. 4, pp. 1297–1301, Mar. 2010.
[4] A. L. Hodgkin and A. F. Huxley, “A quantitative description of mem-
brane current and its application to conduction and excitation in
nerve,” J. Physiol., vol. 117, no. 4, pp. 500–544, Aug. 1952.
[5] L. O. Chua, V. I. Sbitnev, and H. Kim, “Hodgkin-Huxley axon is made
of memristors,” Int. J. Bifurcation Chaos, vol. 22, no. 3, pp. 1230011(1)–
1230011(48), Mar. 2012.
[6] L. O. Chua, V. I. Sbitnev, and H. Kim, “Neurons are poised near the
edge of chaos,” Int. J. Bifurcation Chaos, vol. 22, no. 4, pp. 1250098 (1)–
1250098(49), Apr. 2012.
[7] L. O. Chua and S. M. Kang, “Memristive devices and systems,” Proc.
IEEE, vol. 64, no. 2, pp. 209–223, 1976.
[8] L. O. Chua, “Nonlinear circuit foundations for nanodevices, Part I:
The four element torus” Proc. IEEE, vol. 91, no. 11, pp. 1830–1859, Nov.
2003.
[9] R. G. Bartle, The Elements of Real Analysis, 2nd ed. New York: Wiley
1976.
[10] L. O. Chua, CNN: A Paradigm for Complexity. Singapore: World Sci-
entific, 1998.
[11] K. S. Cole, Membranes, Ions and Impulses. Berkeley: Univ. California
Press, 1972.
[12] K. Mainzer and L. Chua, Local Activity Principle. Imperial College
Press, Feb. 2013.