DETAILED RESEARCH NOTES

MEMRISTOR
The Missing Circuit Element

Based on Primary Source:

Leon O. Chua, "Memristor — The Missing Circuit Element"
IEEE Transactions on Circuit Theory, Vol. CT-18, No. 5, September 1971

Subject Detail

Topic Memristor — Theory, Math & Applications

Document Type Study-Based Research Notes (College Level)

Source Paper Chua (1971), IEEE Trans. Circuit Theory

Coverage Theory → Mathematics → Properties → Applications

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 1 — BACKGROUND & MOTIVATION

1.1 What is the Memristor?

The word memristor is a portmanteau of memory + resistor. It was introduced as a theoretical construct by
Leon O. Chua in 1971 to fill a logical gap in circuit theory. It is a two-terminal passive circuit element whose
resistance at any instant depends on the entire history of current (or voltage) that has ever passed through it.

Simple analogy: Imagine a pipe whose diameter changes permanently based on how much water has
flowed through it in the past. The pipe 'remembers' its flow history. That is what a memristor does —
electrically.

1.2 The Symmetry Argument — Why a 4th Element Must Exist

📄 Primary Source:
Chua, Section II: "Memristor — The Fourth Basic Circuit Element"

The three classical circuit elements are defined by pairwise relationships among four fundamental circuit
variables:

Variable Pair Circuit Element Constitutive Relation

Voltage (v) ↔ Current (i) Resistor (R) v = R · i (Ohm's Law)

Charge (q) ↔ Voltage (v) Capacitor (C) q=C·v

Flux-linkage (φ) ↔ Current Inductor (L) φ=L·i

(i)

Flux-linkage (φ) ↔ Charge Memristor (M) dφ = M(q) · dq ← MISSING!

(q)

The four variables v, i, q, φ can form C(4,2) = 6 pairwise combinations. Two of these are already defined by
physics:
● dq/dt = i (charge is the time-integral of current)
● dφ/dt = v (flux-linkage is the time-integral of voltage)
The remaining three combinations give R, L, C. The sixth combination — between φ and q — was undefined.
Chua argued that from symmetry and completeness, a fourth element corresponding to this relationship
MUST exist.

KEY INSIGHT
The memristor's existence was predicted purely from mathematical symmetry — before any physical
device was ever built. This is similar to how Mendeleeff predicted undiscovered chemical elements from
the periodic table's structure.

Page 2 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

1.3 Definitions of the Four Circuit Variables

Variable Symbol Physical Meaning Mathematical Definition

Current i(t) Flow of charge per unit time i = dq/dt

Voltage v(t) Energy per unit charge v = dφ/dt

Charge q(t) Accumulated flow of current q(t) = ∫ i(τ) dτ from −∞ to t

Flux-linkage φ(t) Accumulated voltage over time φ(t) = ∫ v(τ) dτ from −∞ to t

EXAM TRAP
Flux-linkage φ used here is NOT the same as magnetic flux in a solenoid. Here it is the time-integral of
voltage across the two terminals of a device — a purely mathematical quantity used to characterise the
element.

Page 3 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 2 — CONSTITUTIVE RELATIONS & MATHEMATICS

2.1 The Memristor's Defining Equation

📄 Primary Source:
Chua (1971), Equations (1)–(4), pp. 511

A memristor is characterised by a relation of the form g(φ, q) = 0. There are two dual cases:

Case A: Charge-Controlled Memristor

The flux-linkage φ is expressed as a single-valued function of charge q:
v(t) = M(q(t)) · i(t) ... (1)
where M(q) is the incremental memristance:
M(q) ≡ dφ(q) / dq ... (2)

Case B: Flux-Controlled Memristor

The charge q is expressed as a single-valued function of flux-linkage φ:
i(t) = W(φ(t)) · v(t) ... (3)
where W(φ) is the incremental memductance (conductance):
W(φ) ≡ dq(φ) / dφ ... (4)

Quantity Symbol Unit Analogy

Memristance M(q) Ohm (Ω) Like resistance, but depends on charge

history

Memductance W(φ) Siemens (S) Like conductance, but depends on flux

history

2.2 Why M(q) is NOT a Constant

In an ordinary resistor, R is a fixed constant. In a memristor, M(q) is a function of q(t), which is itself the time-
integral of all past currents. Therefore:

● At any single instant t₀, the memristor looks like a resistor with resistance M(q(t₀))
● But M(q(t₀)) depends on the entire current history from t = −∞ to t₀
● This is why it is called a memory resistor: past inputs determine present resistance

From the paper: "...while the memristor behaves like an ordinary resistor at a given instant of time t₀, its
resistance (conductance) depends on the complete past history of the memristor current (voltage)."

Page 4 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

2.3 The φ-q Curve: Visual Fingerprint of a Memristor

Just as a resistor is characterised by its v-i (voltage vs current) curve, a memristor is characterised by its φ-q
(flux-linkage vs charge) curve. Key observations from Chua's experimental results:

● The φ-q curve is SMOOTH and monotonically increasing for a passive memristor
● The slope at any point on the φ-q curve gives the memristance M(q) at that charge value
● A straight-line φ-q curve gives a linear (constant) memristance → reduces to an ordinary resistor
● Chua synthesised memristors with piecewise-linear φ-q curves using active circuits (M-R mutators)

IMPORTANT
The φ-q curve is traced in the φ-q plane, NOT the v-i plane. Chua designed a 'φ-q curve tracer' (Fig. 3 of
the paper) specifically for this purpose. A standard oscilloscope cannot directly show this curve.

2.4 The Pinched Hysteresis Loop (v-i Characteristic)

When we drive a memristor with a sinusoidal AC source and plot v(t) vs i(t) on an oscilloscope, we get a
characteristic 'pinched hysteresis loop'. This is the most recognisable experimental signature of a memristor.

Properties of the Pinched Hysteresis Loop:

● Always passes through the origin (0,0) — this is the 'pinch'
● Has a figure-eight or lemniscate shape
● The loop area decreases as input frequency increases
● At very high frequencies: loop collapses to a straight line through the origin (memristor becomes a
linear resistor)
● At very low frequencies: loop is wide, showing maximum memory effect

Physical reason for frequency dependence: at high frequencies, charge q(t) oscillates rapidly and averages
near zero, so M(q) barely changes — the device appears static (resistor-like). At low frequencies, significant
charge accumulates per cycle, causing large swings in M(q).

EXAM QUESTION
Q: How can you distinguish a memristor from a nonlinear capacitor or inductor? A: The v-i hysteresis loop
of a memristor always passes through the origin. Capacitor or inductor loops do NOT pass through the
origin.

2.5 Special Case: Linear Memristor

If the φ-q curve is a straight line, then M(q) = R = constant, and the memristor reduces to a linear time-
invariant resistor. Chua notes that introducing a 'linear memristor' in circuit theory adds no new information.
The interesting behaviour only arises with nonlinear φ-q curves.

Page 5 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 3 — CIRCUIT-THEORETIC PROPERTIES

📄 Primary Source:
Chua (1971), Section III: "Circuit-Theoretic Properties of Memristors", pp. 511–514

3.1 Theorem 1: Passivity Criterion

Statement:
A memristor characterised by a differentiable charge-controlled φ-q curve is passive if, and only if, its
incremental memristance M(q) ≥ 0 for all q.

Proof Sketch:
Instantaneous power dissipated by the memristor is:
p(t) = v(t) · i(t) = M(q(t)) · [i(t)]² ... (5)
If M(q) ≥ 0 for all q, then p(t) ≥ 0 always → power is never negative → device is passive (no internal energy
source). Conversely, if M(q) < 0 at some point, negative power is possible → active device.

Physical implication:
● A physically realisable (passive) memristor in device form must have a monotonically increasing φ-q
curve
● Only memristors with M(q) ≥ 0 everywhere can exist as stand-alone devices without an internal
power supply
● Memristors with M(q) < 0 in some region require active circuits (with power supplies) to realise them

3.2 Theorem 2: Closure Theorem

Statement:
Any one-port circuit containing ONLY memristors is equivalent to a single memristor.

Intuition:
Apply KCL and KVL to the network, integrate over time → all equations become relations between flux-
linkages and charges of individual branches → the composite port relationship is also a φ-q relation → it is a
memristor.

ANALOGY
Just as a network of only resistors is equivalent to a single resistor (via Thevenin/Norton), a network of
only memristors is equivalent to a single memristor. The memristor is 'closed' under the operation of
interconnection.

3.3 Theorem 3: Existence & Uniqueness

Statement:

Page 6 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

Any network containing only memristors with positive incremental memristances has one, and only one,
solution.

Proof follows because the governing equations of an all-memristor network are mathematically identical in
form to those of an all-nonlinear-resistor network, and uniqueness is already established for the latter.

3.4 Theorem 4: Principle of Stationary Action

Analogous to Fermat's principle or Hamilton's principle in mechanics, the solution of a memristor network
corresponds to a stationary point of a scalar 'action' function A(q).

Definitions:
● Action A(q) for a charge-controlled memristor: A(q) ≡ ∫₀ᵂ φ(q) dq
● Coaction Â(φ) for a flux-controlled memristor: Â(φ) ≡ ∫₀φ q(φ) dφ

These are the memristor analogues of 'magnetic energy' (for inductors) and 'electric energy' (for capacitors).
Using these potential functions, one can derive Manley–Rowe-like power formulas for memristors operating
as frequency converters.

3.5 Theorem 5: Order of Complexity

For a general network containing resistors (b_L inductors, b_C capacitors, b_M memristors), independent
sources, the order of complexity m (number of state variables) is:
m = (b_L + b_C + b_M) − n_M − n_CE − n_LM − n̂ _M − n̂ _LJ −
n̂ _CM

Key point: each memristor adds exactly one state variable (its charge or flux) unless it is in a loop with other
memristors or cut-set with other memristors, which reduces the count. This is directly analogous to how each
inductor/capacitor adds one state variable to an RLC network.

Page 7 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 4 — ELECTROMAGNETIC INTERPRETATION

📄 Primary Source:
Chua (1971), Section IV: "An Electromagnetic Interpretation of Memristor Characterisation", pp. 514–515

4.1 Circuit Theory as Quasi-Static Electromagnetics

Circuit theory is a special (quasi-static) limiting case of Maxwell's full electromagnetic field equations. The
three classical circuit elements map to EM systems as follows:

Element EM System Dominant Field

Resistor Instantaneous (memoryless) E₀ ↔ H₀ relationship Both zero-order fields dominant

Inductor First-order magnetic field negligible Zero-order E, first-order H

Capacitor First-order electric field negligible Zero-order H, first-order E

Memristor BOTH first-order fields are non-negligible Both first-order E₁ and H₁ significant

4.2 Quasi-Static Expansion of Maxwell's Equations

Define a 'family time' τ = αt where α is a small time-rate parameter. Maxwell's equations expanded as a
power series in α give nth-order Maxwell's equations for n = 0, 1, 2, ...

Zero-Order (n=0):
curl E₀ = 0 and curl H₀ = J₀

First-Order (n=1):
curl E₁ = − ∂B₀/∂τ and curl H₁ = J₁ + ∂D₀/∂τ

4.3 The Memristor Device Requirements

For a device to behave as a memristor, it must satisfy both:
● Both zero-order fields are negligible → E ≈ E₁, H ≈ H₁, D ≈ D₁, B ≈ B₁
● The device material must be nonlinear, i.e., J₁ = g(E₁), B₁ = B(H₁), D₁ = D(E₁)

These two conditions together imply that D₁ is instantaneously (memorylessly) related to B₁:
D₁ = g(B₁) ... (32)

Since the surface integral of D₁ is proportional to charge q(t), and the surface integral of B₁ is proportional to
flux-linkage φ(t), equation (32) is the EM representation of the memristor's constitutive relation.

KEY CONCLUSION
A physical memristor is essentially an AC device. Applying DC current would cause q(t) → ∞ and φ(t) →
∞, making the device non-physical. This is consistent with the pinched hysteresis loop collapsing only

Page 8 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

near the origin for AC signals.

Page 9 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 5 — LABORATORY REALISATION (ACTIVE

CIRCUITS)
📄 Primary Source:
Chua (1971), Section II, pp. 508–510: Mutators and Active Circuit Realisations

5.1 The Problem in 1971

Chua proved theoretically that the memristor must exist, but no physical passive device had been found. He
demonstrated that any prescribed φ-q curve can be SYNTHESISED using active circuits called mutators
connected to nonlinear resistors.

5.2 Mutators — What Are They?

A mutator is a two-port active network that transforms the v-i characteristic of a nonlinear element at Port 2
into a different characteristic type at Port 1. Three types are relevant:

Mutator Type Transform Port 1 Element Port 2 Element

M-R Mutator (Type 1) (q,φ) → (i_R, v_R) Memristor Nonlinear Resistor

M-L Mutator (q,φ) → (φ_L, i_L) Memristor Nonlinear Inductor

M-C Mutator (q,φ) → (q_C, v_C) Memristor Nonlinear Capacitor

5.3 How M-R Mutator Works

● Type-1 M-R mutator: transforms the v_R − i_R curve of the resistor into the φ-q curve of a memristor
● Type-2 M-R mutator: transforms the i_R − v_R curve (x and y axes swapped) into the φ-q curve
● These mutators are realisable as two-port active networks using operational amplifiers and
controlled sources

Chua built practical Type-1 M-R mutator circuits using 2N4236 transistors and OMC-V operational amplifiers
(Fig. 2 of the paper). He verified that a memristor IS realised at Port 1 when a nonlinear resistor is connected
to Port 2.

5.4 The φ-q Curve Tracer

To verify and display the φ-q curve of a synthesised memristor, Chua designed a dedicated φ-q curve tracer
(Fig. 3 of the paper) using integrator circuits:
● Horizontal axis: v_x(t) = k_x ∫ v(τ) dτ = k_x · φ(t) → connected to oscilloscope X-input
● Vertical axis: v_y(t) = k_y ∫ i(τ) dτ = k_y · q(t) → connected to oscilloscope Y-input
The oscilloscope then directly displays the φ-q curve.

Page 10 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

5.5 Experimental Observations from the Paper

Chua synthesised three memristors with different φ-q curves and measured their oscilloscope waveforms
(Fig. 4 and 5 of the paper):
● The v(t) and i(t) waveforms appeared 'peculiar' even with smooth φ-q curves
● The waveforms were tested with both 63 Hz sinusoidal and triangular input signals
● Results confirmed the unique signal-processing behaviour predicted theoretically

Page 11 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 6 — APPLICATIONS (FROM CHUA'S PAPER)

📄 Primary Source:
Chua (1971), Section V: "Some Novel Applications of Memristors", pp. 515–519

6.1 Application A: Device Modelling

Chua demonstrated that memristors can model the behaviour of two newly-discovered but poorly-
understood devices of the 1970s:

Example 1: Amorphous 'Ovonic' Threshold Switch

● An ovonic switch is a two-terminal amorphous glass device that switches from high to low resistance
after a time delay T_d when a voltage pulse is applied
● Chua showed the memristor circuit (Fig. 6) with a piecewise-linear φ-q curve accurately models this
behaviour
● The state equation is: dq/dt = v_s(t) / [R₁ + R₂ + M(q)]
● The time delay T_d = [Φ₀ + (R₁ + R₂)Q₀] / E — decreases as voltage amplitude E increases
● The memristor output waveforms (Fig. 6d, g, h) showed striking resemblance to published ovonic
switch waveforms

Example 2: Electrolytic E-Cell (Coulomb Cell)

● An E-Cell is an electrochemical timing device: silver ions plate from cathode to anode during timing
interval, then the cell switches from low to high resistance
● Its behaviour is modelled precisely by a memristor with a specific φ-q curve (Fig. 7b)
● The model accurately reproduces the output waveform and timing delay T_d

RESEARCH NOTE
Both these 1971 device models are precursors to modern resistive switching memory (ReRAM). The
ovonic switch is the direct ancestor of phase-change memory (PCM).

6.2 Application B: Signal Processing — Staircase Waveform

Generator
Memristors can generate staircase waveforms (used in sampling oscilloscopes, curve tracers). The method:

● Design a memristor with a multi-segment piecewise-linear φ-q curve (N segments)

● Drive the circuit with a square-wave input
● Each segment of the φ-q curve corresponds to one step of the staircase output
● A 9-segment φ-q curve generates a 10-step staircase waveform (Fig. 9 of the paper)

The nonlinear resistor required for this is built using back-to-back Zener diode ladder networks connected to
an M-R mutator.

Page 12 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

6.3 Application C: Frequency Conversion

From the Manley-Rowe-like formula derived in Section III (Theorem 4), memristors can be used in:
● Memristor upconverter: gain proportional to (frequency ratio)² is achievable
● Memristor harmonic generator: conversion efficiencies approaching 100% are theoretically possible
● Memristor mixer: significant conversion loss is expected (disadvantage for this application)

Page 13 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 7 — MODERN CONTEXT (POST-1971)

7.1 Physical Realisation — HP Labs (2008)

For 37 years, the memristor remained an active-circuit simulation. In 2008, Strukov et al. at HP Labs identified
that TiO₂ thin-film switches — which had been studied for years — are physical memristors:

● Device: TiO₂ thin film (~5 nm) between two Platinum electrodes
● One sub-region is TiO₂ (insulating), other is TiO₂₋ₓ (oxygen-vacancy doped, conducting)
● Applying voltage moves the dopant boundary → changes effective resistance
● This is exactly Chua's 1971 constitutive relation, physically realised

CITATION
D.B. Strukov, G.S. Snider, D.R. Stewart, R.S. Williams, "The missing memristor found," Nature, 453, pp.
80–83, 2008.

7.2 Generalisation: Memristive Systems (Chua & Kang, 1976)

In 1976, Chua and Kang extended the concept to memristive systems — devices with multiple internal state
variables. The general equations are:
V = R(w, I) · I
dw/dt = f(w, I)
where w is a vector of state variables. Most real-world resistive switching devices (with oxide films, phase-
change materials, ferroelectrics) are better described as memristive systems rather than ideal memristors.

7.3 Modern Applications Overview

Application How Memristor Helps Example

Non-Volatile Memory LRS/HRS = binary 1/0; retain state with no Panasonic HfO₂ ReRAM, Intel
(ReRAM) power Optane

Neuromorphic Computing Conductance ≈ synaptic weight; STDP IBM PCM chip, Intel Loihi
learning

In-Memory Computing Matrix-vector multiply inside memory Avoids von Neumann bottleneck
crossbar

Stochastic Computing / PUFs Device variability as cryptographic Hardware security keys

fingerprint

Analogue Signal Processing Implements filters and transforms in FIR/IIR filter crossbars
hardware

Page 14 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 8 — EQUATIONS SUMMARY SHEET

Quick reference for all key equations from Chua (1971) and the HP model:

From Chua (1971)

Equation Expression Meaning

Eq. (1) v(t) = M(q(t)) · i(t) Voltage-current relation for charge-controlled

memristor

Eq. (2) M(q) ≡ dφ(q)/dq Definition of memristance

Eq. (3) i(t) = W(φ(t)) · v(t) Current-voltage relation for flux-controlled memristor

Eq. (4) W(φ) ≡ dq(φ)/dφ Definition of memductance

Eq. (5) p(t) = M(q(t)) · [i(t)]² Instantaneous power dissipation

Theorem 1 M(q) ≥ 0 ⇔ passive Passivity condition

From HP TiO₂ Model (Strukov et al., 2008)

Equation Expression Meaning

Memristance M(w) = Rᴏᴇ (w/D) + Rᴏᶠᶠ (1 − w/D) Total device resistance as function of
state

State Eq. dw/dt = (μᵥ · Rᴏᴇ / D) · I(t) Rate of change of dopant boundary
position

Parameters μᵥ = ion mobility, D = film thickness Physical device parameters

Page 15 | Memristor Research Notes — For Academic Study Only

Research Notes: Memristor — Leon O. Chua (1971) | College Study Reference

UNIT 9 — CRITICAL ANALYSIS & RESEARCH POINTERS

9.1 Strengths of Chua's 1971 Paper

● Mathematically rigorous and complete — five theorems with proofs
● Made a concrete, falsifiable prediction: a passive two-terminal device with a monotonically
increasing φ-q curve must exist in nature
● Experimentally demonstrated realisability using active circuits
● Derived both circuit-theoretic and electromagnetic justifications

9.2 Limitations / Open Questions from the 1971 Paper

● No passive physical memristor existed in 1971 — all realisations required active circuits
● The linear drift model (HP 2008) predicts ions can exit device boundaries (w < 0 or w > D), which is
unphysical — requires nonlinear window functions
● Ideal memristor vs. memristive systems: most real devices are better described as memristive
systems with multiple state variables, not ideal single-variable memristors

9.3 Open Research Problems (for your study)

● Is every resistive switching device a memristor, or must it satisfy additional conditions?
● Can a true passive memristor (with single state variable) exist, or are all physical realisations
memristive systems?
● How does thermal noise and stochastic switching affect the memristive constitutive relation?
● Can memristors implement universal computing — is Material Implication (IMP) logic truly Turing-
complete?

9.4 Recommended Reading Sequence for Research

Ste Paper / Resource Why Read It
p

1 Chua (1971) — in hand Primary source; all theory originates here

2 Chua & Kang (1976), Proc. IEEE Generalises to memristive systems

3 Strukov et al. (2008), Nature Physical realisation; HP model equations

4 Joglekar & Wolf (2009), Eur. J. Phys. Window functions; linear model limitations

5 Biolek et al. (2009), Radioengineering SPICE simulation models

6 Burr et al. (2017), Advances in Physics: X Neuromorphic applications review

End of Detailed Research Notes — Based on Chua (1971) IEEE Trans. Circuit Theory

Page 16 | Memristor Research Notes — For Academic Study Only