Welcome to Scribd. Sign in or start your free trial to enjoy unlimited e-books, audiobooks & documents.Find out more
Download
Standard view
Full view
of .
Look up keyword
Like this
1Activity
0 of .
Results for:
No results containing your search query
P. 1
Non-memristive Systems

Non-memristive Systems

Ratings: (0)|Views: 22|Likes:
Published by Blaise_Mouttet
Discusses dynamic systems models other than memristor that generate zero-crossing pinched hysteresis
Discusses dynamic systems models other than memristor that generate zero-crossing pinched hysteresis

More info:

Published by: Blaise_Mouttet on Feb 23, 2012
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

02/23/2012

pdf

text

original

 
February 20, 2012 (ver.5)
Memresistors and non-memristivezero-crossing hysteresis curves
Blaise Mouttet
 Abstract 
— It has been erroneously asserted by thecircuit theorist Leon Chua that all zero-crossingpinched hysteresis curves define memristors. Thisclaim has been used by Stan Williams of HPLabs toassert that all forms of RRAM and phase changememory are memristors. This paper demonstratesseveral examples of dynamic systems which falloutside of the constraints of memristive systems andyet also produce the same type of zero-crossinghysteresis curves claimed as a fingerprint for amemristor. This establishes that zero-crossinghysteresis serves as insufficient evidence for amemristor.
 Keywords- non-linear dynamic systems, memresistor, phase change memory, RRAM, ReRAM 
I.
 
I
NTRODUCTION
 Below is a summarized timeline of events so farpertaining to the Chua-HP “memristor”:1) In 1971 Leon Chua publishes a paper proclaimingthe
memristor 
as the missing fourth fundamentalcircuit element linking electric charge and magneticflux linkage [1]. This paper acknowledges the earlierexistence of ovonic switches (a precursor to modernphase change memory) but argues that no physicalmemristor has yet been discovered.2) In 1976 Leon Chua and Sung-Mo Kang publish apaper introducing
memristive systems
as a specialcase of state-space dynamic systems [2]. This paperdistinguishes memristive systems from thememristor by noting that:
“..there remains an even broader class of physicaldevices and systems whose characteristics resemblethose of a memristor and yet 
 cannot be realistically modeled by this element
 ,..”
3) Beginning in the late 1990’s and continuingthrough the 2000’s several companies includingMicron Technology, Sharp, Samsung, and UnitySemiconductor began experimenting with newforms of 2-terminal resistance memories based onchalcogenide and metal oxide materials. During thisperiod HPLabs was attempting to make molecularswitching devices work properly [3]. Also duringthis period researchers at Samsung invented a newtype of resistive memory based on the motion of oxygen ions in TiO
2
and other metal oxides [4].4) In 2008 researchers for HPLabs including StanWilliams claim to have found Chua’s missingmemristor and simultaneously switch their researchefforts away from molecular memory toward TiO
2
 and other metal oxide resistance switches [5]. Thephysical structure of TiO
2
discussed by the HPLabsresearchers is identical to that discussed in the patentapplication of Samsung [4].5) In 2010 I gave a presentation at the IEEEInternational Symposium on Circuits and Systems(ISCAS) noting several problems with the narrativeregarding Chua and HP’s memristor [6]. During thepresentation I noted that most real memory resistordevices could not be accurately modeled usingChua’s original 1971 definition of a memristor. Ialso noted other companies had been working onother forms of RRAM based on other devicematerials and physical operating mechanisms.6) In January of 2011 Leon Chua published a paper[7] arguing that:
“All 2-terminal non-volatile memory devicesbased on resistance switching are memristors,regardless of the device material and physicaloperating mechanisms.”
In order to justify this statement Chua pointed to the1976 definition of memristive systems rather thanthe original memristor definition therefore directlycontradicting the statement that the broader class of 
 
February 20, 2012 (ver.5)
memristive systems cannot be realistically modeledby a memristor.7) In October of 2011 Stan Williams of HPLabspublically made a claim that phase change memory,MRAM, and RRAM are all memristor technologies[8]. Thus ovonic switches which were originallydenied as a form of physical memristor by Chua in1971 are now suddenly a type of memristor.There is already evidence in the literature [9]suggesting that Chua’s analysis is flawed for failingto consider the distinction between type I (self-crossing pinched) hysteresis curves (usuallyassociated with bipolar memory) and type II (non-self-crossing pinched) hysteresis curves (usuallyassociated with unipolar memory).The goal of this paper is to further illustrate that itis possible to construct dynamic systems which falloutside of the definition of either memristor ormemristive systems and yet still produce zero-crossing pinched hysteresis curves.II.
 
E
XAMPLES OF NON
-
MEMRISTIVE DYNAMICSYSTEMS EXHIBITING ZERO
-
CROSSING HYSTERESIS
It is first noted that the definition of a memristivesystems is broadly given [2] as
(
)=
(
,
,
)
(
)

(
)

=
(
,
,
)
 
(1)
 
wherein
u(t)
= the input signal,
 y(t)
= the outputsignal,
 x
(t)
is the system state (which generally maybe a vector function),
g(
 x
 ,u,t)
is a
continuous
 function most generally dependent on the systemstate, input signal, and time, and
 f(
 x
 ,u,t)
is a
continuous
function defining the rate of change of the state dependent on the system state, input signal,and time.The following examples illustrate that zero-crossing pinched hysteresis curves may be generatedby dynamic systems falling outside of the canonicalmemristive systems framework.Example 1: 1
st
order non-linear dynamic systemwith initial condition x(0)=x
0
 
=
+2(
0
)
(
−
)

=
2
12
 
2
 
(2)
 
In this case
u
represents the input signal,
 x
is thestate variable, and
 y
is the output signal.
 A
may beinterpreted as a particular voltage or current leveland
 A
2
 /2 may be interpreted as a particular powerlevel (see example 4 for a related physical system).For u=Asin(
ω
t), d(x-x
0
)/dt = A
2
(sin
2
(
ω
t)-1/2) = -A
2
 cos(2
ω
t)/2, and x = x
0
-A
2
sin(2
ω
t)/4
ω)
(given theinitial condition x(0)=x
0
). The output equation canthen be expressed as y = Asin(
ω
t) – A
2
sin(2
ω
t)/2
ω
.The y vs. u hysteresis curve for this case isillustrated in the top figure on page 7 based on aMathcad plot for
ω
=1 and A=1.Example 2: 2
nd
order non-linear dynamic systemwith initial conditions x(0)=x
0
and dx(0)/dt = -
ωΑ
 
=
+(
0
)

(
−
)

=
2
 
(3)
 
As in the 1
st
example,
u
represents the inputsignal,
 x
is the state variable, and
 y
is the outputsignal. For u = Asin(
ω
t), x-x
0
= -Asin(
ω
t) (given theinitial conditions). The output equation can then beexpressed as y = Asin(
ω
t)-
ωΑ
2
sin(
ω
t)cos(
ω
t). The yvs. u hysteresis curve for this case is illustrated inthe middle figure on page 7 based on a Mathcad plotfor
ω
=1 and A=1. It is notable that this example maybe physically comparable to a memadmittancesystem in which ionic or oxygen vacanciesdemonstrate large inertia relative to the damping andelectrostatic forces [10,11].Example 3: Frequency independent pinchedhysteresis with x(0)= -A
=
+(
 
2
2
)

=

 
(4)
 
As in the 1
st
example,
u
represents the inputsignal,
 x
is the state variable, and
 y
is the outputsignal. For u = Asin(
ω
t), x = -Acos(
ω
t) (given the
 
February 20, 2012 (ver.5)
initial condition). The output equation can then beexpressed as y = Asin(
ω
t)-A
3
sin
2
(
ω
t)cos(
ω
t). The yvs. u hysteresis curve for this case is illustrated inthe bottom figure of page 7 based on a Mathcad plotfor
ω
=1 and A=1.In Example 1 it is evident that an increase in thesignal frequency produces decay in the hysteresiseffect (as is currently assumed to be correct frommemristive systems). In Example 2 an increase inthe signal frequency increases the hysteresis effectin contrast to memristive systems. Example 3 showsan example of frequency independent pinchedhysteresis.The first three examples prove that it is possibleto generate pinched hysteresis using dynamicsystems outside of the realm of memristive systems.The next three examples provide some cases withmore physical meaning.Example 4: Thermal memresistor
=

+
(
0
)
(
−
)

=

2
(
0
)
 
(5)
 
In this case
i
(electron current) represents theinput variable,
(temperature) is the state variable,and
v
(voltage) is the output variable.This example is based on an idealized physicalsystem in which a voltage
v
is derived based on thesum of an ohmic voltage expressed as the product of resistance
 R
and current
i
and an additional voltagebased on the product of the Seebeck coefficient
S
 and a temperature gradient
T-T 
0
. In this idealizedsystem even in cases wherein there is no appliedcurrent a temperature gradient can induce a voltage.The dynamic equation expresses the rate of variationof the temperature gradient
T-T 
0
based on thedifference between the Joule heating (
ki
2
), whichraises the temperature of the system, and thedissipated energy
α 
(T-T 
0
)
, determining how fast thethermal energy is lost to the environment.It is evident that this example has a form verysimilar to that of Example 1 with the exception thata transient component exists in the solution to thedynamic equation. However, it may be feasible todesign a system such that, given a current signali=I
0
sin
ω
t, the Joule heating in the first term exactlybalances the energy loss from the second term. Inthis case it would be possible to reproduce the sametype of hysteresis curve as in Example 1.It is also evident that this example represents avolatile memory systems but it may be converted toa non-volatile memory system such as phase changememory by including a crystallization rate equationand noting the dependence of the resistance on thecrystallization state.Example 5: Ionic memresistor
=
(
∆
,
)
(
∆
,0)
 
(
6a
)
 
∆
(
)

+
1
∆
(
)

+
(

)

∆
(
)=(
ν


)

−

(
)

 
(
6b
)
 
A full discussion of the development of thismemresistive system is given in [11]. In this case
v
a
 (voltage) is the input variable,
 x
(variation in thedepletion width) is the state variable,
i
(current) isthe output variable, and
i
(
 x
 ,v
a
)
is a functiondefining the dependence of the current on thedepletion width and the voltage. For the dynamicequation (6b) the ionic doping level is denoted by
 N 
 and the remainder of the terms may be consideredconstants.Different versions of the function
i
(
 x
 ,v
a
)
maybe formulated depending on a particular thin filminsulator or semiconductor. In general, however, fora sinusoidal voltage input v
a
=V
0
sin(
ω
t) the outputcurrent takes the form
=
(
∆
sin(
ω
t+
φ
0
),
0
sin

)
(
∆
sin(
ω
t+
φ
0
),0)
 
(
7
)
 
where
 X 
and
ϕ 
0
are constants depending on theion doping level, effective ion mass, relativepermittivity, temperature, and signal frequency (seeequations (39)-(42) of [11]). It is clear frominspection that the i vs. v curve for this system willproduce a zero-crossing hysteresis (with theexception of the case
ϕ
0
= 0 which degenerates thehysteresis).

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->