Advanced EIS Orazem May 2014

Advanced Electrochemical Impedance
Spectroscopy
Mark E. Orazem
Department of Chemical Engineering
University of Florida
Gainesville, Florida 32611
meo@che.ufl.edu
352-392-6207
 Mark E. Orazem, 2000-2013. All rights reserved.

Contents
• Chapter 1. Introduction
• Chapter 2. Motivation
• Chapter 3. Impedance Measurement
• Chapter 4. Representations of Impedance Data
• Chapter 5. Development of Process Models
• Chapter 6. More on Development of Process Models
• Chapter 7. Constant Phase Elements
• Chapter 8. Regression Analysis
• Chapter 9. Error Structure
• Chapter 10. Kramers-Kronig Relations
• Chapter 11. Use of Measurement Models
• Chapter 12. Conclusions
• Chapter 13. Suggested Reading
• Chapter 14. Notation
Chapter 1. Introduction page 1: 1

Spectroscopy
Chapter 1. Introduction
• How to think about impedance spectroscopy

• EIS as a generalized transfer function
• Objective of course

1992 – no logo
June 16-21, 2013

Okinawa, Japan
The Blind Men and the Elephant

John Godfrey Saxe
It was six men of Indostan

To learning much inclined,
Who went to see the Elephant
(Though all of them were blind),
That each by observation
Might satisfy his mind.
The First approached the Elephant,

And happening to fall
Against his broad and sturdy side,
At once began to bawl:
“God bless me! but the Elephant
Is very like a wall!” ...
Electrochemical Impedance
Spectroscopy
• Electrochemical technique
– steady-state
– transient
– impedance spectroscopy
• Measurement in terms of macroscopic quantities
– total current
– averaged potential
• Not a chemical spectroscopy
• Type of generalized transfer-function measurement
Interrogate Electrochemical Cell
Electrochemical
Cell
V + ∆V I + ∆I
∆V
Z (ω=
) = Z r + jZ j
∆I
M. E. Orazem and B. Tribollet, Electrochemical Impedance Spectroscopy, John Wiley & Sons,
Hoboken, New Jersey, 2008
Low Frequencies
1
I (V ) : slope =
Rp
I (V ) dI 1
=
dV Z
∆I
V
lim Z ( f ) = R p
f →0
∆V
Moderate Frequencies
I (V )
∆I
V
Z ( f ) Z=
= exp ( jφ ) Z r + jZ j
∆V
Applications of EIS
• Electrode-Electrolyte • Corrosion
Interface • Electrodeposition
– Electrical Double Layer
• Human Skin
– Diffusion Layer
• Batteries
• Reaction Mechanism
• Fuel Cells
• Electrochemical
Reactions • Materials
Characterization
Course Objectives
• Benefits and advantages of impedance spectroscopy
• Methods to improve experimental design
– Global and local impedance
• Interpretation of data
– graphical representations
– equivalent circuits
– process models
– regression
– error analysis
Contents
Comment on Notation
• IUPAC Convention: Z ' and Z '', i

• Present Work: Zr and Z j , j
Chapter 2. Motivation page 2: 1

Spectroscopy
Chapter 2. Motivation
• Comparison of measurements
– steady state
– step transients
– single-sine impedance
• In principle, step and single-sine perturbations yield same
results
• Impedance measurements have better error structure

C1 C2
Steady-State
Polarization R0
Curve R1(V) R2
0.4
0.3
Current, mA
0.2
0.1
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Potential, V
Steady-State Techniques
• Yield information on state after transient is completed

• Do not provide information on
– system time constants
– capacitance
• Influenced by
– Ohmic potential drop
– non-stationarity
– film growth
– coupled reactions
Transient Response to a Step in Potential
R 1 (V ) R2
1.5
100
} ∆V=10 mV potential input
I=
V
R0 + R1 (V ) + R2
Current / mA
Current / mA
1.0 C1 C2
10-1
R0
0.5 R1(V) R2
current response 10-2
0 2 4 6 8 10 10-5 10-4 10-3 10-2 10-1 100

(t-t0) / ms (t-t0) / s
Transient Response to a Step in Potential
R 1 (V ) R2
1.5 Short times/high frequency

} ∆V=10 mV potential input
C1 C2
Current / mA
1.0 C1 C2
R0
R0
0.5 R1(V) R2 R1(V) R2
current response long times/low frequency

0 2 4 6 8 10
(t-t0) / ms
Transient Techniques:
potential or current steps
• Decouples phenomena
– characteristic time constants
• mass transfer
• kinetics
– capacitance
• Limited by accuracy of measurements
– current
– potential
– time
• Limited by sample rate
– <~1 kHz
Sinusoidal Perturbation
V (t )= V0 + ∆V cos (ω t )
i (t ) i0 {exp(baV ) − exp( −bcV )}
=
dV
iC = C0
dt
i f = f (V )
V (t )
1.0 10 kHz
100 Hz
1 mHz
0.5
(i-i0) / max(i-i0) 0.0
-0.5
-1.0
-1.0 -0.5 0.0 0.5 1.0

(V-V0) / V0
Lissajous Representation
1.0
B
V (t ) = ∆V cos(ωt )
∆V
0.5 =I (t ) cos(ωt + φ )
Z
Y(t)/Y0
O D A
0
∆V OA
-0.5 =
|Z | =
∆I OB
OD
-1.0 sin(φ ) = −
OA
-1.0 -0.5 0 0.5 1.0
X(t)/X0
Impedance Response
-25
100 Hz
-20
-2
Zj / Ω cm
-15
-10
-5
0
0 10 20 30 40 50
Zr / Ω cm-2
Impedance Spectroscopy
• Decouples phenomena
– characteristic time constants
• mass transfer
• kinetics
– capacitance
• Gives same type of information as DC transient.
• Improves information content and frequency range by
repeated sampling.
• Takes advantage of relationship between real and
imaginary impedance to check consistency.
vs. Step-Change Transients
• Information sought is the same

• Increased sensitivity
– stochastic errors
– frequency range
– consistency check
• Better decoupling of physical phenomena
System with Large Ohmic Resistance

C
R0
R1
Circuit R0, Ω R1, Ω C, µF R0/R1
1 1,000 100 110 10
1
=
f RC = 14.5 Hz
2π R1C
M. E. Orazem, T. El Moustafid, C. Deslouis, and B. Tribollet, Journal of The Electrochemical
Society, 143 (1996), 3880-3890.
Impedance Spectrum
-60
15 Hz
-40
-Zj / Ω
-20
20
1000 1020 1040 1060 1080 1100

Zr / Ω
Error Structure
Chapter 3. Impedance Measurement page 3: 1

Spectroscopy
Chapter 3. Impedance Measurement
• Overview of techniques
– A.C. bridge
– Lissajous analysis
– phase-sensitive detection (lock-in amplifier)
– Fourier analysis
• Experimental design

Measurement Techniques
• A.C. Bridge
• Lissajous analysis
• Phase-sensitive detection (lock-in amplifier)
• Fourier analysis
– digital transfer function analyzer
– fast Fourier transform
D. Macdonald, Transient Techniques in Electrochemistry, Plenum Press, NY, 1977.

J. Ross Macdonald, editor, Impedance Spectroscopy Emphasizing Solid Materials and
Analysis, John Wiley and Sons, New York, 1987.
C. Gabrielli, Use and Applications of Electrochemical Impedance Techniques, Technical
Report, Schlumberger, Farnborough, England, 1990.
AC Bridge
Z1 Z2
• Bridge is balanced when
current at D is equal to zero
Z1Z 4 = Z 2 Z 3
D
• Time consuming
• Accurate
f ≥ 10 Hz
Z3 Z4
Generator
Lissajous Analysis
V (t ) = ∆V sin(ω t )
Current
∆V
= sin(ω t + φ )
B
I (t )
Z
A' D' O D A
Potential
∆V OA A'A
=
|Z | = =
B' ∆I OB B'B
OD D'D
sin(φ ) =− =
−
OA A'A
Phase Sensitive Detection

General Signal =A A0 sin(ω t + φ A )
∞
1 4
=
Reference Signal S ∑
π n =0 2n + 1
sin ( 2n + 1) ω t + φS 
∞
4 A01
AS = ∑
π n =0 2n + 1
sin (ω t + φ A ) sin ( 2n + 1) ω t + φS 
2π ω
ω
cos(φ A − φ S )
2 A0
2π ∫ ASdt = π
Has maximum value when
0
φ A = φS
V (t ) = V0 cos(ωt ) Fourier Analysis:

=I (t ) I 0 cos(ωt + φI ) single-frequency input
T
1
I r (ω ) = ∫ I (t ) cos(ωt )dt
T 0
T
1
I j (ω ) = − ∫ I (t ) sin(ωt )dt Vr + jV j 
T 0 Z r (ω ) = Re  
1
T  I r + jI j 
Vr (ω ) = ∫ V (t ) cos(ωt )dt
T 0 Vr + jV j 
Z j (ω ) = Im  
1
T
 I r + jI j 
V j (ω ) = − ∫ V (t ) sin(ωt )dt
T 0
Fourier Analysis:
multi-frequency input
10 10
Signal Strength
Signal Strength
0
0
Z 0
0
Time Time
-10 -10
Fast
Fourier
Transform
Z(ω)
Comparison
single-sine input multi-sine input
• Good accuracy for stationary • Good accuracy for stationary
systems systems
• Frequency intervals of ∆f/f • Frequency intervals of ∆f
– economical use of – dense sampling at high
frequencies frequency required to get
• Used for entire frequency good resolution at low
frequency
domain
• Kramers-Kronig inconsistent • Often paired with Phase-
frequencies can be deleted Sensitive-Detection (f>10 Hz)
• Correlation coefficient used to
determine whether spectrum is
inconsistent with Kramers-
Kronig relations
Measurement Techniques
• A.C. bridge
– obsolete
• Lissajous analysis
– obsolete
– useful to visualize impedance
• Phase-sensitive detection (lock-in amplifier)
– inexpensive
– accurate
– useful at high frequencies
• Fourier analysis techniques
– accurate
Experimental Considerations
• Electrode Dimension
• Frequency range
– instrument artifacts
– non-stationary behavior
– capture system response
• Linearity
– low amplitude perturbation
– depends on polarization curve for system under study
– determine experimentally
• Signal-to-noise ratio
V (t )= V0 + ∆V cos (ω t )
i (t ) = i0 exp(baV )
dV
iC = C0
dt
i f = f (V )
V (t )
Linearity
1.0 f = 1 mHz 1.0 f = 10 Hz
1 mV 1 mV
20 mV 20 mV
(I-I0)/max(I-I0)
(I-I0)/max(I-I0)
0.5 40 mV 0.5 40 mV
0.0 0.0
-0.5 -0.5
-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
(V-V0)/∆V (V-V0)/∆V
1.0 f = 100 Hz 1.0

1 mV
20 mV
(I-I0)/max(I-I0)
(I-I0)/max(I-I0)
0.5 40 mV 0.5
f = 10 kHz
1 mV
0.0 0.0 20 mV
40 mV
-0.5 -0.5
-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
(V-V0)/∆V (V-V0)/∆V
Influence of Nonlinearity on Impedance
model
1 mV
-20 20 mV
Zj / Ω cm2
40 mV
-10
0
0 10 20 30 40 50
Zr / Ω cm2
Influence of Nonlinearity on Impedance
1.0
101 model
1 mV
0.8 20 mV
|Zj| / Ω cm2
100 40 mV
model
0.6
Zr / Rt
1 mV
20 mV 10-1
0.4 40 mV
10-2
0.2
10-3
0.0
-3 -2 -1 0 1 2 3 4 5 6
10-4 -3
10 10 10 10 10 10 10 10 10 10 10 10-2 10-1 100 101 102 103 104 105 106
f / Hz f / Hz
Guideline for Linearity

b∆V ≤ 0.2
10
100(Rt-Z(0))/Rt
1
0.5
0.1
0.01
0.01 0.1 0.2 1

∆V b
Influence of Ohmic Resistance
b∆V ≤ 0.2 (1 + Re / Rt )
Re dV
iC = C0
dt
i f = f (V )
i (t ) Re η s (t )
V (t )
Influence of Ohmic Resistance

Overpotential as a Function of Frequency
Re dV
iC = C0
dt
i f = f (V )
i (t ) Re η s (t )
V (t )
Cell Design
• Use reference electrode to isolate influence of electrodes and
membranes
2-electrode 3-electrode 4-electrode
Working Working Working

Electrode Electrode Electrode
Ref 1
Ref 1
Ref 2
Counterelectrode Counterelectrode Counterelectrode

Current Distribution
Counter Electrode Lead Wires
• Seek uniform current/potential Electrolyte

Out
Electrolyte
Out
distribution
Cover
– simplify interpretation Flange
– reduce frequency dispersion

Working
Electrode
Cell Body
ID = 6 in.
L = 6 in.
Flange
Cover
Electrolyte In Electrolyte In
Working Electrode Seat

Primary Current Distribution

Modulation Technique
• Potentiostatic • Galvanostatic
– standard approach – good for nonstationary
– linearity controlled by systems
potential • corrosion
• drug delivery
– requires variable
perturbation amplitude to
 α F   maintain linearity
=in iO exp  M (V − Vcorr )  − 1
  RT 2
 
 α Fe F  1 α
 Fe F 
2

in = iO  (V − Vcorr )  +  (V − Vcorr )  + 
 RT
2
 2  RT  
I  V / Z   V  I Z  

Experimental Strategies
• Faraday cage
• Short leads
• Good wires
• Shielded wires I
• Oscilloscope WE
∆V
Ref
CE
Reduce Stochastic Noise
• Current measuring range

• Integration time/cycles
– long/short integration on some FRAs
• Delay time
• Avoid line frequency and harmonics (±5 Hz)
– 60 Hz & 120 Hz
– 50 Hz & 100 Hz
• Ignore first frequency measured (to avoid start-up
transient)
Reduce Non-Stationary Effects

• Reduce time for measurement
– shorter integration (fewer cycles)
• accept more stochastic noise to get less bias error
– fewer frequencies
• more measured frequencies yields better parameter estimates
• fewer frequencies takes less time
– avoid line frequency and harmonic (±5 Hz)
• takes a long time to measure on auto-integration
• cannot use data anyway
– select appropriate modulation technique
• decide what you want to hold constant (e.g., current or potential)
• system drift can increase measurement time on auto-integration
Reduce Instrument Bias Errors
• Faster potentiostat
• Short shielded leads
• Faraday cage
• Check results
– against electrical circuit
– against independently obtained parameters
Improve Experimental Design
• Question
– How can we isolate the role of positive electrode, negative
electrode, and separator in a battery?
• Answer
Working
Electrode
- Develop a very sophisticated process
model.
Ref 1
- Use a four-electrode configuration.
Ref 2
Counterelectrode
• Question
– How can we reduce stochastic noise in the measurement?
• Answer
- Use more cycles during integration.

- Add at least a 2-cycle delay.
- Ensure use of optimal current measuring resistor.
- Check wires and contacts.
- Avoid line frequency and harmonics.
- Use a Faraday cage.
• Question
– How can we determine the cause of variability between
experiments?
• Answer
- Use the Kramers-Kronig relations to see if data are self-

consistent.
- Change mode of modulation to variable-amplitude
galvanostatic to ensure that the base-line is not changed by
the experiment.
Facilitate Interpretation
• Question
– How can we be certain that the instrument is not corrupting
the data?
• Answer
- Use the Kramers-Kronig relations to see if data are self-

consistent.
- Build a test circuit with the same impedance response and
see if you get the correct result.
- Compare results to independently obtained data (such as
electrolyte resistance).
Experimental Considerations
• Frequency range
– instrument artifacts
– non-stationary behavior
• Linearity
– low amplitude perturbation
– depends on polarization curve for system under study
– determine experimentally
• Signal-to-noise ratio
Can We Perform Impedance on Transient

Systems?
• Timeframes for measurement

– Individual frequency
– Individual scan
– Multiple scans
EHD Experiment
-0.2
Data Set #2
-0.15 Data Set #4
Imaginary Impedance,µA/rpm
Data Set #1
-0.1
-0.05
0.05
-0.05 0 0.05 0.1 0.15 0.2
Real Impedance, µA/rpm
Time per Frequency for 200 rpm

0.01 0.1 1 10 100
1000 1000
Time per Measurement, s
100 100
10 10
Filter On
Filter Off
1 1
0.01 0.1 1 10 100

Frequency, Hz
Impedance Scans
-15
Zj / Ω cm2
-10
Time Interval
0-1183 s
-5 1183-2366 s
2366-3581 s
0
0 5 10 15 20 25 30 35 40
Zr / Ω cm2
Impedance Scans
4000
Elapsed Time / s
3000
2000
1000
0
10-2 10-1 100 101 102 103 104 105
f / Hz
Time per Frequency Measured
102
∆tf / s
5 cycles
3 cycles
101
5 seconds
10-2 10-1 100 101 102 103 104 105

f / Hz
Chapter 4. Representation of Impedance Data page 4: 1

Spectroscopy
Chapter 4. Representation of Impedance Data
• Electrical circuit components

• Methods to plot data
– standard plots
– subtract electrolyte resistance
• Constant phase elements

Impedance is a Complex Number

1000
800 = Z r + jZ i
Z
600 Z = Z e jφ
=Z Z cos (φ ) + j Z sin (φ )
-Zj / Ω
400
( Zr ) + ( Z j )
2
=
|Z| 2
Z
200
 Zj 
φ φ = arctan  
0
0 200 400 600 800 1000  Zr 
Zr / Ω
Circuit Elements
1 1
ZR = R Z L = jLω ZC = = −j
jCω Cω
Addition of Impedance
Z1
Z1 Z2
Z2
Z = Z1 + Z 2 1
Z=
1 1 1
Y= +
1 1 Z1 Z 2
+
Y1 Y2 Y = Y1 + Y2
Circuit Components: Resistor

-5
Zj / Ω cm2
Re 0
Z = Re Re
5
0 5 10 15 20
Zr / Ω cm2
Circuit Components: Capacitor
Re Cdl
1 1
Z =+
Re =−
Re j
jωCdl ωCdl
j= −1
Re
1
Simple RC Circuit f RC =
2π Rt Cdl
Cdl Rt
158.5 Hz
2 -40
Zj / Ω cm2
-20
Re
0
Rt 0 20 40 60 80 100 120
Re Zr / Ω cm2 Re + Rt
1
= Re +
Z
1
+ jωCdl
Rt
Note:
= Re +
Rt
ω = 2π f
1 + jω Rt Cdl
ω ≡ rad / s or s-1
Rt ω Rt 2Cdl f ≡ cycles / s or Hz
=
Re + −j
1 + (ω Rt Cdl ) 1 + (ω Rt Cdl )
2 2
120
Impedance 100
real
imaginary
Zr or −Zj / Ω cm2
80
Cdl
60
40
20
Re
0
Slope = 1 10-3 10-2 10-1 100 101 102 103 104 105 106
Rt
f / Hz
102
real
imaginary
158.5 Hz 101
Zr or −Zj / Ω cm2
-40
Zj / Ω cm2
100
-20
10-1
0 Slope = -1
-2
10
0 20 40 60 80 100 120
Zr / Ω cm2 10-3 10-2 10-1 100 101 102 103 104 105 106
f / Hz
Bode Representation
-90
-75
2
10
-60
= Z r + jZ j
φ / degrees
Z
|Z| / Ω cm2
-45
Z r =| Z | cos (φ )
Z j =| Z | sin (φ ) -30
-15
0
101
10-3 10-2 10-1 100 101 102 103 104 105 106
f / Hz
Modified Phase Angle

 Zj 
φadj −1
= tan  
Input & output are in phase
 r
Z − Re 
102 -90
|Z|adj -75
|Z| Note:
φ / degrees
101 -60
φadj
|Z| / Ω cm2
The modified phase angle yields

φ excellent -45
insight, but we need an
accurate estimate for the solution
-30
resistance.
100
-15
0  Zj 
φ = tan  
−1
10-1
10-3 10-2 10-1 100 101 102 103 104 105 106
 Zr 
f / Hz
Input & output are out of phase
Constant Phase Element

CPE
Z − Re 1
Re = +
1 + ( jω ) Rt Q
α
Rt
Rt
-0.6
α=1
α=0.9
α=0.8
α=0.7
-0.4 α=0.6
α=0.5
Zj / Rt
-0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
(Zr−Re) / Rt
Slope = CPE: Log Imaginary Slope =

+α −α
100
10-1
|Zj| / Rt
10-2
-3
α=1
10 α=0.9
α=0.8
α=0.7
10-4 α=0.6
α=0.5
10-5 -5 -4 -3 -2 -1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10
ω
ωτ/ τ
Note: With a CPE (α≠1), the asymptotic slopes are no longer ±1.
CPE: Modified Phase Angle
-90
α=1
φadj / degrees
α=0.9
-60 α=0.8
α=0.7
α=0.6
α=0.5
-30 Note:
The high-frequency asymptote for
the modified phase angle depends
0 on the CPE coefficient α.
10-5 10-4 10-3 10-2 10-1 100 101 102 103 104 105
ωωτ
/τ
Example for Synthetic Data
Rt + zd ( f )
Z ( f=
) Re +
1 + ( j2π f ) Qdl ( Rt + zd ( f ) )
α
tanh j2π f τ
zd ( f ) = zd ( f )
j2π f τ
M. Orazem, B. Tribollet, and N. Pébère, J. Electrochem. Soc., 153 (2006), B129-B136.
Traditional Representation
400 α =1
α = 0.7
−Zj / Ω cm2
α = 0.5
200 1 Hz 0.1 Hz
10 Hz 10 mHz
0 100 Hz 1 mHz
0 200 400 600 800 1000 1200
Zr / Ω cm2
α =1 -90
α = 0.7 α =1
α = 0.5
-75 α = 0.7
103
| Z | / Ω cm2
α = 0.5
φ / degree
-60
-45
-30
102
-15
10-3 10-2 10-1 100 101 102 103 104 105 106 10-3 10-2 10-1 100 101 102 103 104 105 106
f / Hz f / Hz
Re-Corrected Bode Phase Angle

 Zj 
φadj = tan 
−1
 φ (∞) =−90(α )
 Z r − Re,est  adj
-90 α =1
α = 0.7
-75
α = 0.5
φadj / degree
-60
-45
-30
-15
10-3 10-2 10-1 100 101 102 103 104 105 106
f / Hz
Re-Corrected Bode Magnitude

( )
2
Z adj
= Z r − Re,est + Z 2
j
Slope =
−α
103
| Z |adj / Ω cm2
slope = −α
102
1
α =1
10 α = 0.7
α = 0.5
100
10-1
10-3 10-2 10-1 100 101 102 103 104 105

f / Hz
Slope = α Log(|Zj|)
103 Slope =
α= 1 −α
10 2 α = 0.7
α = 0.5
2
| Zj | / Ω cm
101
slope=α
0
10
slope=−α
10-1
10-2
10-3 10-2 10-1 100 101 102 103 104 105 106
f / Hz
Effective Capacitance or CPE Coefficient

1  απ  1
Z CPE = Q = − sin  
QCPE ( j2π f )  2  ( 2π f ) Z (f )
α α
eff
j
100 α =1
α = 0.7
Qeff / Qdl
α = 0.5
10
10-3 10-2 10-1 100 101 102 103 104 105 106
f / Hz
Alternative Plots
• Re-Corrected Bode Plots (Phase Angle)
– Shows expected high-frequency behavior for surface
– High-Frequency limit reveals CPE behavior
• Re-Corrected Bode Plots (Magnitude)
– High-Frequency slope related to CPE behavior
• Log|Zj|
– Slopes related to CPE behavior
– Peaks reveal characteristic time constants
• Effective Capacitance
– High-Frequency limit yields capacitance or CPE
coefficient
Application
Mg alloy (AZ91) in 0.1 M NaCl
-200
-100 65.5 Hz
2
Zj / Ω cm
0.011 Hz
0
15.5 kHz 0.0076 Hz
time / h
100 0.5
3.0
6.0
200
0 100 200 300 400 500 600
Zr / Ω cm2
Proposed Model
Mg →
k1
Mg + ( )ads + e −
(Mg ) +
ads + H 2 O →
k2
Mg 2+
+ OH −
+ 12 H 2 “negative difference effect”
(NDE)
→
k3
(Mg ) +
ads
←
Mg 2+ + e −
k −3
→
k4
Mg 2+ + 2 OH − Mg(OH) 2
←
k −4
→
k5
Mg(OH) 2 MgO + H 2 O
←
k -5
G. Baril, G. Galicia, C. Deslouis, N. Pébère, B. Tribollet, and V. Vivier, J. Electrochem. Soc.,

154 (2007), C108-C113
Normalized Plot
G. Baril, G. Galicia, C. Deslouis,

N. Pébère, B. Tribollet, and V.
Vivier, J. Electrochem. Soc., 154
(2007), C108-C113
Mg → Mg
k1
( +
) +e − Physical Interpretation
ads
diffusion of Mg2+
→
k3
(Mg )
+
ads
←
Mg 2+ + e −
k −3 -200
-100 65.5 Hz
Zj / Ω cm2
0.011 Hz
0
15.5 kHz 0.0076 Hz
time / h
100 0.5
3.0
6.0
200
0 100 200 300 400 500 600
Zr / Ω cm2
relaxation of (Mg+)ads.
intermediate
Log(|Zj|
Slope =
−α
102
2
|Zj| / Ω cm
101
time / h
0.5
3.0
6.0
100
10-3 10-2 10-1 100 101 102 103 104 105
f / Hz
Effective CPE Coefficient

 απ  1
Qeff = − sin  
 ( 2π f ) Z j (f )
α
 2
105
Qeff / (MΩ-1 cm-2) sα
time / h
0.5
104 3.0
6.0
103
102
101
10-2 10-1 100 101 102 103 104 105

f / Hz
Re-Corrected Magnitude
 Zj 
φadj (∞) =−90α φadj −1
= tan 
Z −R 
 r e ,est 
103
2
| Z |adj / Ω cm
102
time / h
0.5
3.0
6.0
101
100
10-3 10-2 10-1 100 101 102 103 104 105
f / Hz
-200
Physical Parameters
-100 65.5 Hz
Zj / Ω cm2
0.011 Hz
0
15.5 kHz 0.0076 Hz
time / h
100 0.5
3.0
6.0
200
0 100 200 300 400 500 600
Zr / Ω cm2
Application for Local EIS
layer
∆Vapplied
Z local =
∆ilocal
Global Impedance
Mg alloy (AZ91) in Na2SO4 10-3 M at the corrosion potential after 1 h of immersion.
Electrode radius 5500 µm.
J.-B. Jorcin, M. E. Orazem, N. Pébère, and B. Tribollet, Electrochimica Acta, 51 (2006), 1473.
Local impedance
Local Impedance
α = 1 to 0.92
Results of Graphical Analysis
1.0
0.9
resistance
0.8
α
0.7
capacitance
0.6
0.5
0.0 0.2 0.4 0.6 0.8 1.0
r / r0
The local impedance has a pure RC behavior at center. Current

distribution affects value at edge.
Graphical Representation of Impedance Data
• Expanded Range of Plot Types

– Facilitate model development
– Identify features without complete system model
• Suggested Plots
– Re-Corrected Bode Plots (Phase Angle)
• Shows expected high-frequency behavior for surface
• High-Frequency limit reveals CPE behavior
– Re-Corrected Bode Plots (Magnitude)
• High-Frequency slope related to CPE behavior
– Log|Zj|
• Slopes related to CPE behavior
• Peaks reveal characteristic time constants
– Effective Capacitance
• High-Frequency limit yields capacitance or CPE coefficient
Guidelines for Nyquist

Plots -40
Zj / Ω cm2
-20
• Use symbols, not lines
• Label characteristic
0
frequencies
0 20 40 60 80 100 120
• Use same scale for real and Zr / Ω cm2
imaginary axes
158.5 Hz
-40
Zj / Ω cm2
-20
0
0 20 40 60 80 100 120
Zr / Ω cm2
Choice of Representation
• Plotting approaches are useful to show governing

phenomena
• Complement to regression of detailed models
• Sensitive analysis requires use of properly weighted
complex nonlinear regression
Chapter 5. Development of Process Models page 5: 1

Spectroscopy
Chapter 5. Development of Process Models
• Use of Circuits to guide development

• Develop models from physical grounds
• Model case study
• Identify correspondence between physical models and
electrical circuit analogues
• Account for mass transfer

Electrical Circuits
Method to Combine Impedances

Electrically Equivalent Circuits

Use circuits to create framework

Addition of Potential
Addition of Current
Equivalent Circuit at the Corrosion Potential

Equivalent Circuit for a Partially Blocked

Electrode
ZF
= Re +
Z
1 − γ + jω ( γ C + (1 − γ )Cdl ) Z F
Implications: Normalized Nyquist

Implications: Normalized Impedance

Equivalent Circuit for an Electrode Coated

by a Porous Layer
Equivalent Electrical Circuit for an Electrode

Coated by Two Porous Layers
εε 0
=
C 2 =; ε 0 8.8452 × 10−14 F/cm
δ2
R 2 = δ2 / κ2
 ZF 
 R + 
R 2  1 + jωCdl Z F 
Z=
Re + +
1 + jω R 2 C 2  ZF 
1 − γ + jωγ C  R + 
 1 + jωCdl Z F 
Scaling and Corrosion of a Steel Electrode
L Bousselmi, C. Fiaud, B. Tribollet, and E. Triki, Electrochimica Acta, 44 (1999), 4357-4363.

Use kinetic models to determine

expressions for the interfacial impedance
Approach
√ identify reaction mechanism
√ write expression for steady state current contributions
√ write expression for sinusoidal steady state
√ sum current contributions
√ account for mass transfer
√ calculate impedance
General Expression for Faradaic Current
i f = f (V , ci ,0 , γ i )
~ j ωt
i = i + Re i e { }
 ∂ f   ∂f   ∂f 
if =  + ∑  ci ,0 + ∑   γk
  V

 ∂V ci ,0 ,γ k i  ∂ci ,0 V , c j , j ≠i ,γ k k  ∂γ k V , c γ
j j , j≠k
Reactions Considered
• Dependent on Potential
• Dependent on Potential and Mass Transfer
• Dependent on Potential, Mass Transfer, and Surface
Coverage
• Coupled Reactions
Irreversible Reaction:
Dependent on Potential
Az+
A A A A A A A A A
A → A +e + -
A A A A A A A A A A
A A A A A A A A A
• Potential-dependent heterogeneous reaction

• Two-dimensional surface
• No effect of mass transfer
Impedance Model Development

steady-state
 αA F  iA = K A exp ( bAV )
iA = nA FkA exp  V
 RT 
oscillating component
∂ iA 
V K AbA exp ( bAV )  1 
=iA = = V V
∂V Rt ,A
1
Rt ,A =
K AbA exp ( bAV )
β A = 2.303/ bA βA
iA =
2.303Rt ,A
β A = 2.303Rt ,A iA
Significance of Rt I
Simple Approach for Model

Development
 ∂ iA  1 
=iA = V V
∂V Rt ,A
V
Z F = = Rt ,A
iA

dV
=
i iA + Cdl
dt
i= i + Re { ie jωt }
V= V + Re{Ve jωt }

i + Re ie {
jωt
= } 
iA + Re iA e jωt
{ d
dt
}
+ Cdl (V + Re{Ve jωt })
{ }
i + Re ie jωt = { }
iA + Re iA e jωt + Cdl Re{ jωVe jωt }
Re { ie }
= jωt
{ }
Re iA e jωt + Cdl Re{ jωVe jωt }
 iA + jωCdlV
i=
 iA + jωCdlV
i=
V
=i + jωCdlV
Rt ,A
 1   1 + jω Rt ,A Cdl 
= V  + jωCdl  = V  
R   Rt ,A
 t ,A   
V Rt ,A
= ZF
i 1 + jω Rt ,A Cdl
= iRe + V
U
U + Re{Ue jωt } = ( i + Re{ie jωt }) Re + V + Re{Ve jωt }
=
U iRe + V
=
U  iR
 e + V
U V
= Re + = Z A
i i
Cdl
U V
Z= = Re +
A i i
Rt ,A
= Re + Re
1+jω Rt ,ACdl
Rt,A
Interpretation Models for Impedance

Spectroscopy
• Models can account rigorously for proposed kinetic and
mass transfer mechanisms.
• Circuits can provide a framework for model development.
• Models can be expressed in terms of circuits.
Chapter 6. More on Development of Process Models page 6: 1

Spectroscopy
Chapter 6. More on Development of Process Models
• Use of Circuits to guide development

• Develop models from physical grounds
• Model case study
• Identify correspondence between physical models and
electrical circuit analogues
• Account for mass transfer

Approach
√ identify reaction mechanism
√ write expression for steady state current contributions
√ write expression for sinusoidal steady state
√ sum current contributions
√ account for mass transfer
√ calculate impedance
General Expression for Faradaic Current
i f = f (V , ci ,0 , γ i )
~ j ωt
i = i + Re i e { }
 ∂ f   ∂f   ∂f 
if =  + ∑  ci ,0 + ∑   γk
  V

 ∂V ci ,0 ,γ k i  ∂ci ,0 V , c j , j ≠i ,γ k k  ∂γ k V , c γ
j j , j≠k
Reactions Considered
• Dependent on Potential
• Dependent on Potential and Mass Transfer
• Dependent on Potential, Mass Transfer, and Surface
Coverage
• Coupled Reactions
Dependent on Potential
Az+
A A A A A A A A A
A → A +e + -
A A A A A A A A A A
A A A A A A A A A
• Potential-dependent heterogeneous reaction

• Two-dimensional surface
• No effect of mass transfer
Cdl
U V
Z= = Re +
A
i i
Rt,A
= Re + Re
1+jω Rt,A Cdl
Rt,A
Dependent on Potential and Mass Transfer
O R
O + ne − → R
• Irreversible potential-dependent heterogeneous reaction

• Reaction on two-dimensional surface
• Influence of transport of O to surface
Current Density
steady-state
− K O cO,0 exp ( −bOV )

iO =
oscillating component
=iO K O bO cO,0 exp ( −bOV )V − K O exp ( −bOV ) cO,0

V
= − K O exp ( −bOV ) cO,0
Rt ,O
1
Rt ,O =
K O cO,0 exp ( −bOV )
Mass Transfer
dcO
iO = −nO FDO
dy 0
iO= iO + Re {iO e jω t }
iO = −nO FDO dcO

dy 0
c
iO = −nO FDO O,0 θ′ ( 0 )
δO
Combine Expressions

− K O exp ( bOV ) cO,0
V
=iO
Rt ,O
iOδ O
cO,0 = −
n FD θ′ ( 0 )
O O
Current Density
V
iO =
δO 1  1 
Rt ,O + − 
nO FDO cO,0 bO  θ ′(0) 

V
= =
Z F Rt,O + zd,O
Rt ,O + zd ,O
δO 1  1 
− 
bO  θ ′(0) 

zd ,O
nO FDO cO,0
1
Rt ,O =
K O cO,0 exp ( −bOV )
Simple Approach for Model

Development
V
iO =
Rt ,O + zd ,O
V
Z= = Rt ,O + zd ,O
F
iO
Mass Transfer
Film Diffusion
∂ ci  ∂ 2 ci 
= Di  2 
∂t ∂ y 
steady state
ci → ci (∞) as y → δf
=ci c=
i (0) at y 0
y
=
ci ci (0) + ( ci (∞) − ci (0) )
δf
Film Diffusion
∂ ci  ∂ 2 ci 
= Di  2 
∂t ∂ y 
c=
i ci + Re { ci e jωt }
jωt d 2
c d 2
ci jωt
jω
= 
ce Di i
+ Di e y
dy 2
dy 2
ξ=
δf
d ci jωt
2
 jωt
jω ce = Di e ωδ f2 d 2θi  =
dy 2
Ki = − j K θ 0
dξ 2 i i
Di
d ci
2
c
jω c = Di θi = i
d y2 ci (0)
Warburg Impedance
d 2θi
− jK iθi =
0
dξ 2
= (
θi A exp ξ jK i + B exp −ξ jK i) ( )
=θi 0=
at ξ 1 θi = 0 at ξ = ∞
=θi 1=
at ξ 0 =θi 1=
at ξ 0
1 tanh jK i 1 1
= − = −
θi′(0) jK i θi′(0) jK i
1 1
jωδ 2
= −
θi′(0)
f
1
tanh
Di jωδ f2
= − Di
θi′(0) jωδ f2
Di
1.2
t=0.5nπ/K
1
Concentration 0.8
t=0, 2nπ/K
c(ξ)/c(∞)
t=1.5nπ/K
0.6
0.4
t=nπ/K
0.2 K=1
0
0 0.5 1 1.5 2
ξ/δ
1.2
1
t=0, 2nπ/K
0.8 t=0.5nπ/K
c(ξ)/c(∞) 0.6
t=1.5nπ/K
0.4
t=nπ/K
0.2 K=100
0
0 0.5 1 1.5 2
ξ/δ
Rotating disk electrode

Concentration Profile
Coupled Diffusion Impedance
Di δ f ,i
ZD + Z D,f
Di,f δ N,i
Z D,net =
 δ f2,i  Di δ f ,i
Z D,f Z D  jω  +
 Di,f  Di,f δ N,i
Dependent on Potential and Adsorbed Intermediate
B P
B 
→ X+e k1 -
X X X X
X → P+e k2 -
• Potential-dependent heterogeneous reactions

• Adsorption of intermediate on two-dimensional surface
• Maximum surface coverage
Steady-State Current Density
reaction 1: formation of X
(
K1 (1 − γ ) exp b1 (V − V1 )
i1 = )
reaction 2: formation of P
=i2 K 2γ exp b2 (V − V2 ) ( )
total current density
i = i1 + i2
Steady-State Surface Coverage

balance on γ
d γ i1 i2
Γ = −
dt F F
=0
steady-state value for γ
γ =
(K1 exp b1 (V − V1 ) )
K exp ( b (V − V ) ) + K
1 1 1 2 exp ( b (V − V ) )
2 2
Steady-State Current Density
( )
i = K1 (1 − γ ) exp b1 (V − V1 ) + K 2γ exp b2 (V − V2 ) ( )
where
γ =
( K1 exp b1 (V − V1 ) )
K exp ( b (V − V ) ) + K
1 1 1 2 exp ( b (V − V ) )
2 2
Oscillating Current Density
( ) (
i = K1 (1 − γ ) exp b1 (V − V1 ) + K 2γ exp b2 (V − V2 ) )
 
R  2 2 ( 2 1 (1 )
i =  1 + 1 V + K exp b (V − V ) − K exp b (V − V ) γ
R 1 ( ))
 t ,1 t ,2 
( )
−1
Rt=  K1b1 (1 − γ ) exp −b1 (V − V1 ) 
,1  
( )
−1
Rt ,2  K 2 b2γ exp −b2 (V − V2 ) 
=
 
Need Additional Equation

balance on γ
1 1 1 
Γjωγ
=  −
F  Rt ,1 Rt ,2 
( ( )
V − K1 exp b1 (V − V1 ) + K 2 exp b2 (V − V2 ) γ ( ))
Rt−,11 − Rt−,21
γ = V
( ( )
F Γjω + K1 exp b1 (V − V1 ) + K 2 exp b2 (V − V2 ) ( ))
Impedance
1
=
1 
+
( ) ( )
 K 2 exp b2 (V − V2 ) − K1 exp b1 (V − V1 )   Rt−,11 − Rt−,21 
 
Z F Rt ( ( ) (
F Γjω + F K1 exp b1 (V − V1 ) + K 2 exp b2 (V − V2 ) ))
1 A
= +
Rt jω + B
where
1 1 1
= +
Rt Rt , M Rt , X
Cdl
Re
Zf
Corrosion of
Steel
Fe → Fe2+ + 2e-
i f = iFe + iH 2 + iO 2
O2+2H2O + 4e- → 4OH-
2H2O + 2e- → H2+2OH-

Corrosion: Fe → Fe2+ + 2e-

iFe = K Fe exp ( bFeV )
iFe = K FebFe exp ( bFeV )V
~
~ V
iFe =
Rt ,Fe
1
Rt ,Fe =
K FebFe exp ( bFeV )
H2 Evolution: 2H2O + 2e- → H2+2OH-
iH 2 =
− K H 2 exp −bH 2V( )
=iH 2 K H 2 bH 2 exp −bH 2V V ( )

~
~ V
iH 2 =
Rt ,H 2
1
Rt,H 2 =
K H 2 bH 2 exp −bH 2V( )
O2 Reduction: O2+2H2O + 4e- → 4OH-
iO2 = (
− K O2 cO2 ,0 exp −bO2V )
=iO2 K O2 bO2 cO2 ,0 exp −bO2V V ( )

− K O2 exp −bO2( V ) cO2 ,0
O2 Evolution: O2+2H2O + 4e- → 4OH-

mass transfer: in terms of dimensionless
gradient at electrode surface
dcO2
iO2 = −nO2 FDO2
dy 0
cO2 ,0 
= −nO2 FDO2 θ ′ (0)
δO 2
O2 Evolution: O2+2H2O + 4e- → 4OH-

coupled expression
V
iO2 =
δO 1  1 
Rt ,O2 + 2
 − θ′(0) 
nO2 FDO2 cO2 ,0 bO2  
V
=
Rt ,O2 + zd ,O2
δO 1  1 
=zd ,O2 2
 − θ′(0) 
nO2 FDO2 cO2 ,0 bO2  
1
Rt,O2 =
K O2 bO2 cO2 ,0 exp −bO2V ( )
Process Model
U
Z= = Z r + jZ j
i
1
= Re +
 1 
Re
1 1
 +  + + jωCd
 Rt,Fe Rt,H2  Rt ,O2 + zd ,O2
1
= Re + Zd,O2
1 1
+ + jωCd Rt,Fe Rt,H2
Reff Rt ,O2 + zd ,O2
Cdl
Rt,O2
1 1 1
= +
Reff Rt,Fe Rt,H2
Development of Impedance Models
 identify reaction mechanism

 write expression for steady state current contributions
 write expression for sinusoidal steady state
 sum current contributions
 account for mass transfer
 calculate impedance
Interpretation Models for Impedance

Spectroscopy
• Systematic method to account for proposed kinetic and
mass transfer mechanisms.
• Model development can be expressed in terms of circuits.
Chapter 7. Constant Phase Elements page 7: 1

Spectroscopy
Chapter 7. Constant Phase Elements
• CPEs can arise from surface or axial

distributions
• CPE parameters can be interpreted in
terms of capacitance, depending on type
of distribution

Constant Phase Element (CPE)

Semi-Circle
Rt
= Re +
Z
-Zj 1 + jωC0 Rt
Depressed Semi-Circle
Rt
= Re +
Z
1 + ( jω ) QRt
α
Re Zr Re+Rt
Re Re Re Re
CPE caused by
Rt Rt Rt Rt distribution of time
C0 C0 C0 C0 constants
Intentionally left blank

Intentionally left blank
Semi-Circle
CPE: Constant Phase Element Rt
= Re +
Z
1 + jωC0 Rt
Depressed Semi-Circle
Rt
Z= Re +
1 + ( jω ) QRt
α
Proposed Origins of a CPE

1. distribution of capacitance
2. electrode roughness
3. distribution of reaction rate constants
4. nonuniform current distributions
5. coupled reactions
6. distribution of resistivity
Usual Use of a CPE

1. improve fit for a circuit model
2. explanation added as needed
Recent Developments
in our Understanding of the CPE
Types of Distributions
• Surface
• Axial
Interpretation in Terms of Capacitance

1. Direct use of Q
Ceff,Q = Q
2. Surface Distribution
(1−α )/α G. J. Brug, A. L. G. van den Eeden, M. Sluyters-
1/α  Re Rt  Rehbach, J. H. Sluyters, J. Electroanal. Chem., 176
Ceff,surf =Q  
 e
R + Rt  (1984), 275.
3. Normal (Axial) Distribution

C. H. Hsu, F. Mansfeld, Corrosion, 57 (2001), 747.
Ceff,norm = Q1/α Rf(1−α )/α
4. Power-Law Normal Distribution

Ceff,PL = Q ( ρδ εε 0 )
1−α
gB. Hirschorn, M. E. Orazem, B. Tribollet, V. Vivier,
I. Frateur, and M. Musiani, J. Electrochem. Soc.,
g 1 + 2.888(1 − α )
= 2.375
157 (2010), C452.
Surface Distribution
From characteristic
frequency for
admittance
(1−α )/α
1/α  Re Rt 
 e
R + Rt 
B. Hirschorn, M.E. Orazem, B. Tribollet, V. Vivier, I. Frateur, and M. Musiani,

Electrochim. Acta, 55 (2010), 6218.
Normal (Axial) Distribution
From characteristic
frequency for
impedance
Ceff,surf = Q1/α Rf(1−α )/α
B. Hirschorn, M.E. Orazem, B. Tribollet, V. Vivier, I. Frateur, and M. Musiani,

Electrochim. Acta, 55 (2010), 6218.
Power-Law Model
ρδ
εε 0
Ri ( x) = ρ ( x)dx dx Ci =
dx
ρ0
δ
ρ ( x)
Zf = ∫
1
dx Z cpe =
0
1 + jωρ ( x)εε 0 ( jω )α Q
• Assume dielectric constant is uniform

• Is there a resistivity distribution that leads to a CPE?
Model Development
−γ
 regression with sequential RC circuit ρ = ρδ ξ
 conversion in terms of resistivity
−1
 inferred power-law model ρ  ρδ  ρδ  γ 
= + 1 − ξ
ρδ  ρ 0  ρ 0  
δρδ1/γ 1
= g (γ ) =
( jωεε 0 )(γ −1)/γ ( jω )α Q
α 1−α
γ −1 (εε0 ) g
α= Q=
γ gδρδ1−α 1 + 2.888(1 − α ) 2.375
g=
Interpretation in Terms of Capacitance

1. Direct use of Q
Ceff,Q = Q
2. Surface Distribution
(1−α )/α G. J. Brug, A. L. G. van den Eeden, M. Sluyters-
1/α  Re Rt  Rehbach, J. H. Sluyters, J. Electroanal. Chem., 176
 e
R + Rt  (1984), 275.
3. Normal (Axial) Distribution

C. H. Hsu, F. Mansfeld, Corrosion, 57 (2001), 747.
Ceff,norm = Q1/α Rf(1−α )/α
4. Power-Law Normal Distribution

1−α
gB. Hirschorn, M. E. Orazem, B. Tribollet, V. Vivier,
I. Frateur, and M. Musiani, J. Electrochem. Soc.,
g 1 + 2.888(1 − α )
= 2.375
157 (2010), C452.
Impedance of Human Skin
Split-thickness human cadaver skin (300-400 µm thick)

7 samples
E. White, Characterization of the Skin Barrier to Chemical Permeation by
Impedance Spectroscopy, PhD dissertation, Colorado School of Mines, 2011.
Pin-Hole Experiment
Q, α
Re
Rskin
Q, α
Re
Rskin
−1
 1 1 
=
Reff  + 
Rhole  skin
R Rhole 
Impedance Response
before performation
30 after performation
characteristic frequency
−Zj / kΩ cm2
fbefore = 141 Hz
20
10
fafter = 938 Hz
0 10 20 30 40 50
Zr / kΩ cm2
Impedance Response
50
10
before perforation before perforation
40
−Zj / kΩ cm2
Zr / kΩ cm2
1
30
20 0.1 after perforation

after perforation
10
fbefore fafter
0.01
0
10-2 10-1 100 101 102 103 104 105 10-2 10-1 100 101 102 103 104 105
f / Hz f / Hz
Average Regressed Parameters (7 Samples)
Before Pinhole After Pinhole

Rt =
118 ± 54 kΩcm 2 =
Rt 16.2 ± 5 kΩcm 2
α 0.80 ± 0.01
= α 0.83 ± 0.01
=
Q = 74 ± 12 nFsα −1 / cm 2 = 67 ± 10 nFsα −1 / cm 2
Q
f=
c 69 ± 39 Hz f c = 685 ± 266 Hz
f after
= 13 ± 8
f before
Hypotheses: Skin Unaffected by Pinhole
-50 εε 0
Cskin = Ceff,surf =
-40 δ
Zj / kΩcm2
-30 1
fc =
-20
2π RC
-10
f after Cbefore Rbefore
0 =
f before Cafter Rafter
0 20 40 60 80 100 120
Cbefore = Cafter
Zr / kΩcm2
f after R
= before
f before Rafter
M. E. Orazem, B. Tribollet, V. Vivier, S. Marcelin, N. Pébère, A. L. Bunge, E. A. White, D. P. Riemer,

I. Frateur, and M. Musiani, ECS Transactions, (2012), in press.
Hypotheses: Skin Unaffected by Pinhole

Constant Q and α
1−α
g
1 + 2.888(1 − α ) 2.375
g= fc =
1
2π ( RQ )
1/α
f after ( RbeforeQbefore )
-50 1/α
-40 =
( Rafter Qafter )
1/α
Zj / kΩcm2
f before
-30
-20 Q = constant
-10 α = constant
0 1/α
f after  Rbefore 
0 20 40 60 80 100 120 = 
f before  Rafter 
Zr / kΩcm2
Test of Hypotheses
Capacitance obtained from characteristic

frequency not a property of the skin
Hutchinson Technology: Tailoring Stainless Steel

Surface Chemistry
Application: Floating heads for hard drives
Douglas Riemer http://www.htch.com

Principal Scientist
Hutchinson Technology
Hutchinson, MN 55350-9784
303 free machining steel
http://www.htch.com
Impedance Data: Free-Machining Steel
-500 0.025 Hz
Model parameters
−5 α-1 2
-400 Q=
eff 1.10 × 10 Fs / cm
= 11 μFsα-1 / cm2
Zj / kΩ cm2
-300
α = 0.91
-200
Re = 15.2 Ωcm2
0.1 Hz
-100 Rp 2.33 × 106 Ωcm2

=
0 Assumed values
0 100 200
ε = 12
Zr / kΩ cm2 ρδ = 450 Ωcm
Data courtesy of Douglas Riemer, Principal Scientist, Hutchinson
Technology, Hutchinson, MN 55350-9784
Analysis of Data
From EIS (CPE) Data From XPS
Ceff ,
Approach Distribution δ, nm δ, nm
µF/cm2
Ceff,Q = Q - 11 0.97 6.3
(1−α )/α
1/α  Re Rt 
Ceff,surf =Q   surface 4.9 2.2 6.3
 e
R + Rt 
surface,
Ceff,surf = Q1/α Re (1−α )/α 4.9 2.2 6.3
Rt  Re
Ceff,norm = Q1/α Rf (1−α )/α normal 15 0.72 6.3
Ceff,PL = Qg ( ρδ 0 )
1/α
normal 1.8 5.9 6.3

1 + 2.888 (1 − α )
g=
2.375
Test of C=Q (normal distributions)

Test of C=Q (surface distributions)

Constant Phase Elements

• Can arise from surface or axial distributions
• Parameters can be interpreted in terms of capacitance,
depending on type of distribution
• C≠Q
• C≠Ceff,norm
• C= Ceff,surf or C= Ceff,PL , depending on type of distribution
B. Hirschorn, M. E. Orazem, B. Tribollet, V. Vivier, I. Frateur, and M. Musiani, J. Electrochem.

Soc., 157 (2010) C458-C463.
B. Hirschorn, M. E. Orazem, B. Tribollet, V. Vivier, I. Frateur, and M. Musiani, J. Electrochem.
Soc., 157 (2010) C452-C457.
Chapter 8. Regression Analysis page 8: 1

Spectroscopy
Chapter 8. Regression Analysis
• Regression response surfaces

– noise
– incomplete frequency range
• Adequacy of fit
– quantitative
– qualitative

Nonlinear Regression
∂Z 1 p p ∂2 Z
Np N N
p) Z (p 0 ) + ∑
Z (= ∆p j + ∑∑ ∆p j ∆pk + 
=j 1 ∂p j 2 =j 1 =k 1 ∂p j ∂pk
p0 p0
N dat
( Z i − Z (ωi | p)) 2
χ =∑
2
i =1 σ i2
β= α ⋅ ∆p
Z i − Z ( xi | p) ∂Z (ωi | p)
N dat
βk = ∑
i =1 σ 2
∂pk
i
Derivative of
1  ∂Z (ωi | p) ∂Z (ωi | p) 
N dat
function with
α j ,k ≈∑ 2  
i =1 σ i 
respect to
 ∂p j ∂pk 
Variance of data parameter
Methods for Regression
• Evaluation of derivatives
– method of steepest descent
– Gauss-Newton method
– Levenberg-Marquardt method
• Evaluation of function
– simplex
Effect of Noisy Data

Add noise: 1% of modulus
1.5
1.000
75.75
150.5
τ/ s
225.2
1.0
Sum of Squares
300
300.0
200
0.6
100
0.8
0
1.4 1.0 0.5
1.2
1.0 1.2 τ/s 0.5 1.0 1.5
R / Ω cm2 0.8
0.6
1.4 R / Ω cm2
• response surface is smooth and parabolic

• value at minimum greater than zero
Effect of Noisy Data:

Test Circuit with 3 Time Constants
• R0 = 1 Ω cm2
• R1 = 100 Ω cm2 C1 C2 C3
• τ1 = 0.001 s
• R2 = 200 Ω cm2 (1) (2) (3)
R0
• τ2 = 0.01 s
• R3 = 5 Ω cm2 R1 R2 R3
• τ3 = 0.05 s
Note: 3rd Voigt element contributes only 1.66% to DC cell impedance.

Effect of Noisy Data

no noise noise: 1% of modulus
103 103
Sum of Squares
102 102
Sum of Squares
101 101
100 100
10-1 10-1
-4 -4
10-2 4 -3 10-2 4 -3
3 -2 3 -2
2 -1 2 -1
1 0 1 0
2 0 0
log10(R/Ω cm ) 1 log10(τ / s) 1
-1
-2 2 log10(R/Ω cm2) -1
2 log10(τ / s)
-2
Note: use of log scale for parameters
All parameters fixed except R3 and τ3

Resulting Spectrum
-150
2
Zj / Ω cm
-100
-50
0
0 50 100 150 200 250 300 350
Zr / Ω cm 2
Model with 1% noise added

− Model with no noise
Effect of Frequency Range:

Test Circuit with 3 Time Constants
• R0 = 1 Ω cm2
• R1 = 100 Ω cm2 C1 C2 C3
• τ1 = 0.01 s
• R2 = 200 Ω cm2 (1) (2) (3)
R0
• τ2 = 0.1 s
• R3 = 100 Ω cm2 R1 R2 R3
• τ3 = 10 s
Note: 3rd Voigt element contributes 25% to DC cell impedance. The time
constant corresponds to a characteristic frequency ω3=0.1 s-1 or f3=0.016 Hz.
2
Zj / Ω cm Resulting Test Spectra
-100
0
0 100 200 300 400
Zr / Ω cm 2
0.01 Hz to 100 kHz

1 Hz to 100 kHz
Effect of Truncated Data

0.01 Hz to 100 kHz 1 Hz to 100 kHz
103 103
102
Sum of Squares
Sum of Squares
102
101
101
100
-2 -2
-1 -1
0 0
100 5 1 10-1 5 1
4 4 2
3 2 3
2 2 3 log10(τ / s)
1 3 log10(τ / s) 1
0 0 4
log10(R / Ω cm2) -1
4
log10(R / Ω cm2) -1
All parameters fixed except R3 and τ3

Conclusions from Test Spectra

• The presence of noise in data can have a direct
impact on model identification and on the confidence
interval for the regressed parameters.
• The correctness of the model does not determine the
number of parameters that can be obtained.
• The frequency range of the data can have a direct
impact on model identification.
• The model identification problem is intricately linked
to the error identification problem. In other words,
analysis of data requires analysis of the error
structure of the measurement.
When Is the Fit Adequate?
• Chi-squared statistic
– includes variance of data
– should be near the degree of freedom
• Visual examination
– should look good
– some plots show better sensitivity than others
• Parameter confidence intervals
– based on linearization about solution
– should not include zero
Test Case: Mass Transfer to a RDE
Rt + zd (ω )
Z (ω=
) Re +
1 + ( jω C ) ( Rt + zd (ω ) )
Single reaction coupled with mass transfer. Consider

model for a Nernst stagnant diffusion layer:
tanh jωτ
z d (ω ) = z d (ω )
jωτ
Evaluation of χ2 Statistic
σ/|Z(ω)| 1 0.1 0.05 0.03 0.01
χ2 0.0408 4.08 16.32 45.32 408
χ2/ν 0.00029 0.029 0.12 0.32 2.9

Comparison of Model to Data

Impedance Plane (Nyquist)
80
60
−Zj / Ω
40
20
0
0 20 40 60 80 100 120 140 160 180 200
Zr / Ω
Value of χ2 has no meaning without accurate assessment
of the noise level of the data
Modulus
100
|Z| / Ω
10
10-2 10-1 100 101 102 103 104

f / Hz
Phase Angle
φ / degrees -45
-30
-15
0
10-2 10-1 100 101 102 103 104
f / Hz
Real
200
160
120
Zr / Ω
80
40
0
10-2 10-1 100 101 102 103 104
f / Hz
Imaginary
80
60
−Zj / Ω
40
20
0
10-2 10-1 100 101 102 103 104
f / Hz
Log Imaginary
100
Slope = -1
for RC
−Zj / Ω
10
1
10-2 10-1 100 101 102 103 104
f / Hz

-90
-75
 Zj 
φ * = tan 
−1

φ∗ / degrees
-60  r
Z − Re 
-45
-30
-15
0
10-2 10-1 100 101 102 103 104
f / Hz
Real Residuals
0.06
0.04
0.02
εr / Zr
-0.02
±2σ noise level
-0.04
-0.06
10-2 10-1 100 101 102 103 104
f / Hz
Imaginary Residuals
0.3
0.2
0.1
εj / Zj
-0.1
±2σ noise level
-0.2
-0.3
10-2 10-1 100 101 102 103 104
f / Hz
Plot Sensitivity to Quality of Fit

• Poor Sensitivity
– Modulus
– Real
• Modest Sensitivity
– Impedance-plane
• only for large impedance values
– Imaginary
– Log Imaginary
• emphasizes small values
• slope suggests new models
– Phase Angle
• high-frequency behavior is counterintuitive due to role of solution resistance
– Modified Phase Angle
• high-frequency behavior can suggest new models
• needs an accurate value for solution resistance
• Excellent Sensitivity
– Residual error plots
• trending provides an indicator of problems with the regression
Test Case: Better Model for Mass

Transfer to a RDE
Rt + zd (ω )
Z (ω=
) Re +
1 + ( jω ) Q ( Rt + zd (ω ) )
α
Consider 3-term expansion with CPE to account for

more complicated reaction kinetics:
1( ) 2( )
 1 Z p Sc 1/3
Z p Sc 1/3

zd (0)  −
zd = + + 
 θ 0′ ( pSc )
1/3
Sc 1/3
Sc 2/3

 
χ2/ν=4.86
Comparison of Model to Data

Impedance Plane (Nyquist)
80
60
−Zj / Ω
40
20
0
0 20 40 60 80 100 120 140 160 180 200
Zr / Ω
Phase Angle
-45
-30
−Zj / Ω
-15
0
10-2 10-1 100 101 102 103 104
f / Hz

-90
-75
 Zj 
φ * = tan 
−1

φ / degrees
-60  r
Z − Re 
-45
*
-30
-15
0
10-2 10-1 100 101 102 103 104
f / Hz
Log Imaginary
100
−Zj / Ω
10
1
10-2 10-1 100 101 102 103 104
f / Hz
Real Residuals
0.006
0.003
εr / Zr
-0.003
±2σ
-0.006
10-2 10-1 100 101 102 103 104
f / Hz
Imaginary Residuals
0.02
0.01
εj / Zj
-0.01
±2σ
-0.02
10-2 10-1 100 101 102 103 104
f / Hz
Confidence Intervals
• Based on linearization about trial solution

• Assumption valid for good fits
– for normally distributed fitting errors
– small estimated standard deviations
• Can use Monte Carlo simulations to test assumptions
Regression of Models to Data

• Regression is strongly influenced by
– stochastic errors in data
– incomplete frequency range
– incorrect or incomplete models
• Some plots more sensitive to fit quality than others.
• Quantitative measures of fit quality require
independent assessment of error structure.
• The model identification problem is intricately linked
to the error identification problem.
Chapter 9. Error Structure page 9: 1

Spectroscopy
Chapter 9. Error Structure
• Contributions to error structure

• Weighting strategies
• General approach for error analysis
• Experimental results

Contributions to Error Structure
• Sampling Errors
• Stochastic Phenomena
• Bias Errors
– Lack of Fit
– Changing baseline (non-stationary processes)
– Instrumental artifacts
Time-domain  Frequency Domain
1 mHz 10 Hz
100 Hz 10 kHz
Error Structure
ε resid
= (ω ) Z obs (ω ) − Z model (ω )
=ε fit (ω ) + ε bias (ω ) + ε stochastic (ω )
fitting error
inadequate model noise (frequency domain)
experimental errors: inconsistent

with the Kramers-Kronig relations
Weighting Strategies
( Z r ,k − Z r ,k ) 2 ( Z j ,k − Z j ,k ) 2
J=∑ +∑
k σr ,k 2
k σ j ,k 2
Strategy Implications
No Weighting σr=σj
σ=α
Modulus Weighting σr=σj
σ = α|Ζ|
Proportional Weighting σr≠σj
σr = αr|Ζr|; σj = αj|Ζj|
Experimental σr=σj
σ = α|Ζj| + β|Ζr| + γ|Ζ|2/Rm
Refined Experimental σr=σj
σ = α|Ζj| + β|Ζr-Rsol| + γ|Ζ|2/Rm+δ
Assumed Error Structure often Wrong
Data obtained by T. El Moustafid, CNRS, Paris, France

Reduction of Fe(CN)63- on a Pt Disk, 120 rpm

-300
I/ilim
1/4
1/2
-200
3/4
Zj, Ω
-100
0
0 100 200 300 400 500
Zr, Ω
Reduction of Fe(CN)63- on a Pt Disk,

Imaginary Replicates @ 120 rpm, 1/4 ilim

Real Replicates @ 120 rpm, 1/4 ilim

Error Structure @ 120 rpm, 1/4 ilim

Error Structure @ 120 rpm, 1/4 ilim
Interpretation of Impedance Spectra

• Need physical insight and knowledge of error structure
– stochastic component
• weighting
• determination of model adequacy
• experimental design
– bias component
• suitable frequency range
• Approach is general
– electrochemical impedance spectroscopies
– optical spectroscopies
– mechanical spectroscopies
Chapter 10. Kramers-Kronig Relations page 10: 1

Spectroscopy
Chapter 10. Kramers-Kronig Relations
• Mathematical form and interpretation

• Application to noisy data
• Methods to evaluate consistency

Contraints
• Under the assumption that the system is
– Causal
– Linear
– Stable
• A complex variable Z must satisfy equations of the

form: ∞
2ω ⌠ Z r ( x) − Z r (ω )
Z j (ω ) = −
 dx
π ⌡ x −ω
2 2
0
∞
2ω ⌠ − xZ j ( x) + ωZ j (ω )
Z r (ω ) = Z r (∞) + −
 dx
π ⌡ x −ω
2 2
0
Use of Kramers-Kronig Relations

• Concept
– if data do not satisfy Kramers-Kronig relations, a
condition of the derivation must not be satisfied
• stationarity / causality
• linearity
• stability
– interpret result in terms of
• instrument artifact
• changing baseline
– if data satisfy Kramers-Kronig relations, conditions of
the derivation may be satisfied
For real data (with noise)

Z ob (ω ) = Z (ω ) + ε (ω )
= ( Z r (ω ) + ε r (ω )) + j ( Z j (ω ) + ε j (ω ))
where ε (ω ) = ε r (ω ) + jε j (ω )
E (Z (ω ) ob ) = Z (ω ) If and only if E (ε (ω ) ) = 0
Kramers-Kronig in an expectation sense

∞ 
2ω ⌠ Z r ( x) − Z r (ω ) + ε r ( x) − ε r (ω )
E (Z j (ω ) ) =

−
E  dx 
π ⌡ x2 − ω 2
0 
 ∞ − xZ ( x) + ωZ (ω ) − xε ( x) + ωε (ω ) 
2 
E (Z r (ω ) − Z r (∞) ) = E  ⌠ −
j j j j
 dx 
 π ⌡ x 2
− ω 2
 0 
The Kramers-Kronig relations can be

satisfied if
 2ω ∞ ε ( x) 
 ⌠ =0
E (ε (ω ) ) = 0 and E − 2

r
dx 
 π ⌡ x −ω 2
 0 
• This means
– the process must be stationary in the sense of replication at every
measurement frequency.
– As the impedance is sampled at a finite number of frequencies, εr(x)
represents the error between an interpolated function and the “true”
impedance value at frequency x. In the limit that quadrature and
interpolation errors are negligible, the residual errors εr(x) should be
of the same magnitude as the stochastic noise εr(ω).
 2ω ⌠∞ ε ( x ) 
Meaning of E  − 2 r 2 dx =0
 π ⌡ x −ω 
 0 
(Zr(x)-Zr(ω)) / (x2-ω2)
Correct Value
-100
Interpolated Value
0 -200
(Zr(x)-Zr(ω)) / (x2-ω2)
-200
-300
-400
-400
-600 0.2 0.5 1
f / Hz
-800
-1000
10-4 10-3 10-2 10-1 100 101 102 103 104 105
f / Hz
Use of Kramers-Kronig Relations
• Quadrature errors
– require interpolation function
• Missing data at low and high frequency
Methods to Resolve Problems of Insufficient

Frequency Range
• Direct Integration
– Extrapolation
• single RC
• polynomials
• 1/ω and ω asymptotic behavior
– simultaneous solution for missing data
• Regression
– proposed process model
– generalized measurement model
Chapter 11. Use of Measurement Models page 11: 1

Spectroscopy
Chapter 11. Use of Measurement Models
• Generalized model
• Identification of stochastic component of error structure
• Identification of bias component of error structure

Approach
• Experimental design
– quality (signal-to-noise, bias errors)
– information content
• Process model
– account for known phenomena
– extract information concerning other phenomena
• Regression
– uncorrupted frequency range
– weighting strategy
• Error analysis - measurement model
– stochastic errors
– bias errors
Error Structure
ε res
Z obs − Z model =
Z obs − Z model =ε fit + ε bias + ε stoch
• Stochastic errors
• Bias errors
– lack of fit
– experimental artifact
• Kramers-Kronig consistent
• Kramers-Kronig inconsistent
Measurement Model
• Superposition of lineshapes
n
Rk
= R0 + ∑
Z
k =1 1 + jω Rk Ck
Identification of Stochastic Component

of Error Structure
How? Why?
• Replicate impedance scans. • Weight regression.
• Fit measurement model to • Identify good fit.
each scan: • Improve experimental
– use modulus or no- design.
weighting options.
– same number of
lineshapes.
– parameter estimates do
not include zero.
• Calculate standard deviation
of residual errors.
• Fit model for error structure.
Replicated Measurements
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
1 Voigt Element
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
2 Voigt Elements
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
3 Voigt Elements
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
4 Voigt Elements
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
5 Voigt Elements
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
6 Voigt Elements
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
10 Voigt Elements
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
0.03
imaginary
Residual
εj / Zj
0
Errors
-0.03
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
0.005
real
εr / Zr
-0.005
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
Residual Sum of Squares

1000000
100000
10000
Σ(ε /σ )
2
1000
2
100
10
0.1
1 3 5 7 9 11
# Voigt Elements
Fit to Repeated Data Sets
100
-Zj / Ω
50
0
0 50 100 150 200
Zr / Ω
Residual Errors
Real
0.005
εr / Zr
-0.005
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
Standard Deviation of Residual Errors

Imaginary
0.02
εj / Z j
-0.02
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
Standard Deviation
Comparison to Impedance
Identification of Bias Component

of Error Structure
How? Why?
• Fit measurement model to real • Identify suitable
component: frequency range for
– predict imaginary component. regression.
– Use Monte Carlo calculations to • Improve experimental
estimate confidence interval. design.
• Fit measurement model to
imaginary component:
– predict real component.
– Use Monte Carlo calculations to
estimate confidence interval.
1st of Repeated Measurements

Imaginary
1st of Repeated Measurements

Real
Fit to Imaginary Part

100
-Zj / Ω
50
0
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
Fit to Imaginary Part

0.03
εj / Z j
-0.03
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
Predict Real Part

200
190
150
170
Zr / Ω
100
150
0.01 0.1
50
0
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
Predict Real Part

0.03
εr / Zr
-0.03
0.01 0.1 1 10 100 1000 10000 100000
Frequency / Hz
Validation of Measurement Model

Approach
The error structure identified should

be independent of measurement model
used
N
Ri
Voigt: Z (ω=) R0 + ∑
i =1 1 + ( jω )τ i
Transfer
∑ b ( jω )
i
ni
Z (ω ) = i =0
n = 1/ 2
Function:
P
∑ j
a
j =0
( jω ) nj
b0 + b1 ( jω )1/ 2 + b2 ( jω ) +  + bM ( jω ) M / 2 =M P=
Z (ω ) = ; ; aP 1
a0 + a1 ( jω ) + a2 ( jω ) +  + ( jω )
1/ 2 P/2
Convective Diffusion to a RDE

Reduction of Fe(CN)63- on a Pt Disk: i/ilim = ¼; 120 rpm
-80
0 8.93
-60 0.33 9.25
0.66 9.58 0.15 Hz
Zj / Ω
1.0 9.91
-40 1.34
1.67
1.99
-20 2.33
0.0.022 Hz
0
0 40 80 120 160 200
Zr / Ω
Convective Diffusion to a RDE

Reduction of Fe(CN)63- on a Pt Disk: i/ilim = ¼; 120 rpm
real part imaginary part

-80
200 0 8.93 -70 0 8.93
0.33 9.25 0.33 9.25
160 0.66 9.58 -60 0.66 9.58
1.0 9.91 1.0 9.91
-50
1.34 1.34
Zr / Ω
120
Zj / Ω
1.67 -40 1.67
1.99 1.99
80 2.33 -30 2.33
-20
40
-10
0 0
10-2 10-1 100 101 102 103 104 105 10-2 10-1 100 101 102 103 104 105
f / Hz f / Hz
Data Sets for Analysis

-86
1 2 measurement 3
-87 before impedance scan
after impedance scan
-88
Current / mA
-89
-90
-91
-92
0 2 4 6 8 10
time / h
Regression Results for Set #1:

0-1 h
• Data: -86
– 74 frequencies 1 2 3
-87
– 148 data points
• Voigt Model: -88
Current / mA
– 10 elements + Re -89
– 21 parameters
-90
• Transfer Function:
– 5 ai + 6 bi -91
– 11 parameters
-92
0 2 4 6 8 10
time / h
Stochastic Error Structure for Set #1:

Voigt
100
10-1
σZ , σZ / Ω
j
10-2
r
real imag
10-3 direct
VMM
10-2 10-1 100 101 102 103 104 105

f / Hz

Voigt & Transfer Function
100
10-1
σZ , σZ / Ω
j
10-2
r
real imag
direct
10-3 VMM
10-2 10-1 100 101 102 103 104 105

f / Hz
Regression Results for Set #3:

8.9-9.9 h
• Data: -86
– 74 frequencies 1 2 3
-87
– 148 data points
• Voigt Model: -88
Current / mA
– 9 elements + Re -89
– 19 parameters
-90
• Transfer Function:
– 5 ai + 6 bi -91
– 11 parameters
-92
0 2 4 6 8 10
time / h

Voigt
100
10-1
σZ , σZ / Ω
j
10-2
r
real imag
10-3 VMM
10-2 10-1 100 101 102 103 104 105

f / Hz

Voigt & Transfer Function
100
10-1
σZ , σZ / Ω
j
10-2
r
real imag
TMM
10-3 VMM
10-2 10-1 100 101 102 103 104 105

f / Hz
Identification of Stochastic Component

of Error Structure
• Direct evaluation of standard deviation includes
significant contributions from changes in system
behavior
• The error structure identified was independent of
measurement model used
• Selection of measurement model
– Ease of identification of initial parameters → Voigt
– Number of parameters → Transfer Function
Evaluation of Consistency with the

Kramers-Kronig relations
• Approach
– Fit measurement model to imaginary component
– Use experimental variance to weight regression
– Predict real component
– Compare prediction to measured data
– Use Monte Carlo calculations to estimate confidence interval
• Assumption
– Measurement model satisfies the Kramers-Kronig relations
Voigt Measurement Model: Data 1

Fit to Imaginary Prediction of Real
0.02 0.04
(Zr-Zr,mod) / Zr,mod
(Zj-Zj,mod) / Zj,mod
0.01 0.02
0 0
-0.01 -0.02
-0.02 -0.04
10-2 10-1 100 101 102 103 104 105 10-2 10-1 100 101 102 103 104 105
f / Hz f / Hz
• 20 parameters, guessed value for Re

• Tight confidence interval at low frequency
• Justify deleting 5 data points at lowest frequency
Voigt Measurement Model: Data 5

Fit to Imaginary Prediction of Real
0.010 0.04
(Zr-Zr,mod) / Zr,mod
(Zj-Zj,mod) / Zj,mod
0.005 0.02
0 0
-0.005 -0.02
-0.010 -0.04
10-2 10-1 0
10 1
10 10 2 3
10 4
10 5
10 10-2 10-1 100 101 102 103 104 105
f / Hz f / Hz
• 20 parameters, guessed value for Re

• Justify keeping all data points
Interpretation of Impedance Spectra

• Physical insight and knowledge of error structure
– stochastic component
• weighting
• determination of model adequacy
– bias component
• suitable frequency range
• Measurement model approach is general
– electrochemical impedance spectroscopies
– optical spectroscopies
– mechanical spectroscopies
Chapter 12. Conclusions page 12: 1

Spectroscopy
Chapter 12. Conclusions
• Overview of Course
• Credits

Overview of Course
• Electrochemical measurement of macroscopic properties
• Example of a generalized transfer-function measurement
• Can be used to extract contributions of
– electrode reactions
– mass transfer
– surface layers
• Can be used to estimate
– reaction rates
– transport properties
• Interpretation of data
– graphical representations
– regression
– process models
– error analysis
Integrated Approach
Acknowledgements
• Students
– University of Florida
• P. Agarwal, M. Durbha, S. Carson, M. Membrino, P. Shukla, V. Huang, S. Roy, B. Hirschorn,
S-L Wu
– CNRS
• T. El Moustafid, I. Frateur, H. G. de Melo
– CIRIMAT
• J-B Jorcin
• Collaborators
– University of South Florida
• L. García-Rubio
– University of Florida
• O. Crisalle
– CNRS, Paris
• B. Tribollet, C. Deslouis, H. Takenouti, V. Vivier
– CIRIMAT, Toulouse
• N. Pébère
– CNR, Italy
• M. Musiani
• Funding
– NSF, ONR, SC Johnson, CNRS, NASA, ExxonMobil, BP
• Other
– The Electrochemical Society
– The Fuel Cell Seminar
Chapter 13. Suggested Reading page 13: 1

Spectroscopy
Chapter 13. Suggested Reading
• General
• Process Models
• Orazem group work on Fuel Cells
• Measurement Models
• CPE
• Plotting

Suggested Reading
General
• J. R. Macdonald, editor, Impedance Spectroscopy Emphasizing Solid
Materials and Analysis, John Wiley and Sons, New York, 1987.
• E. Barsoukov and J. R. Macdonald, editors, Impedance Spectroscopy:
Theory, Experiment, and Applications, 2nd Edition, John Wiley and
Sons, New York, 2005.
• C. Gabrielli, Use and Applications of Electrochemical Impedance
Techniques, Technical Report, Schlumberger, Farnborough, England,
1990.
• D. Macdonald, Transient Techniques in Electrochemistry, Plenum
Press, NY, 1977.
• M. E. Orazem and B. Tribollet, Electrochemical Impedance
Spectroscopy, John Wiley and Sons, New York, 2008.
• Other Sources
– See instrument vendor websites for for application notes.
Suggested Reading
Process Models
• A. Lasia, “Electrochemical Impedance Spectroscopy and Its
Applications,” Modern Aspects of Electrochemistry, 32, R. E. White, J.
O'M. Bockris and B. E. Conway, editors, Plenum Press, New York,
1999, 143-248.
• C. Deslouis and B. Tribollet, Flow Modulation Techniques in
Electrochemistry, Advances in Electrochemical Science and
Engineering, 2, H. Gerischer and C. W. Tobias, editors, VCH,
Weinheim, 1992, 205-264.
Suggested Reading
CPE
• G. J. Brug, A. L. G. van den Eeden, M. Sluyters-Rehbach, and J. H.
Sluyters, “The Analysis of Electrode Impedances Complicated by the
Presence of a Constant Phase Element,” Journal of Electroanalytic
Chemistry, 176 (1984), 275-295.
• V. Huang, V. Vivier, M. Orazem, N. Pébère, and B. Tribollet, “The
Apparent CPE Behavior of an Ideally Polarized Blocking Electrode: A
Global and Local Impedance Analysis,” Journal of the Electrochemical
Society, 154 (2007), C81-C88.
• B. Hirschorn, M. E. Orazem, B. Tribollet, V. Vivier, I. Frateur, and M.
Musiani, “Constant-Phase-Element Behavior Caused by Resistivity
Distributions in Films: 2. Applications,” Journal of The Electrochemical
Society, 157 (2010) C458-C463.
• B. Hirschorn, M. E. Orazem, B. Tribollet, V. Vivier, I. Frateur, and M.
Musiani, “Constant-Phase-Element Behavior Caused by Resistivity
Distributions in Films: 1. Theory,” Journal of The Electrochemical
Society, 157 (2010) C452-C457.
• M. E. Orazem, B. Tribollet, V. Vivier, S. Marcelin, N. Pébère, A. L.
Bunge, E. A. White, D. P. Riemer, I. Frateur, and M. Musiani,
“Interpretation of Dielectric Properties for Materials showing Constant-
Phase Element (CPE) Impedance Response,” Journal of The
Electrochemical Society, 160 (2013), C215-C225.
Suggested Reading
Measurement Models
• P. Agarwal, M. E. Orazem, and L. H. García-Rubio, "Measurement

Models for Electrochemical Impedance Spectroscopy: 1.
Demonstration of Applicability," Journal of the Electrochemical
Society, 139 (1992), 1917-1927.
• P. Agarwal, Oscar D. Crisalle, M. E. Orazem, and L. H. García-Rubio,
"Measurement Models for Electrochemical Impedance Spectroscopy:
2. Determination of the Stochastic Contribution to the Error Structure,"
Journal of the Electrochemical Society, 142 (1995), 4149-4158.
• P. Agarwal, M. E. Orazem, and L. H. García-Rubio, "Measurement
Models for Electrochemical Impedance Spectroscopy: 3. Evaluation of
Consistency with the Kramers-Kronig Relations,” Journal of the
Electrochemical Society, 142 (1995), 4159-4168.
• M. E. Orazem “A Systematic Approach toward Error Structure
Identification for Impedance Spectroscopy,” Journal of
Electroanalytical Chemistry, 572 (2004), 317-327.
Suggested Reading
Plotting
• M. Orazem, B. Tribollet, and N. Pébère, “Enhanced Graphical

Representation of Electrochemical Impedance Data,” Journal of the
Electrochemical Society, 153 (2006), B129-B136.
Spectroscopy
Chapter 14. Notation
• Roman
• Greek
Chapter 14. Notation page 14: 2
Roman
a coefficient in the Cochran expansion for velocity, a = 0.51023
b coefficient in the Cochran expansion for velocity, b = -0.61592
ci concentration of reacting species i , mol/cm3
ci steady-state value of the concentration of reacting species i , mol/cm3
ci Oscillating component of the concentration of reacting species i , mol/cm3
ci ,o concentration of species i on the electrode surface, mol/cm3
ci ,o steady-state value of the concentration of species i on the electrode surface, mol/cm3
ci ,o Oscillating component of the concentration of species i on the electrode surface, mol/cm3
c∞ bulk concentration of the reacting species, mol/cm3
Cd double layer capacitance, F/cm2
C dl double layer capacitance, F/cm2
Di diffusion coefficient of species i , cm2/s
f arbitrary function, e.g., i f = f (V , ci )
F Faraday’s constant, C/equiv

i Total current density, A/cm2
i Steady-state total current density, A/cm2

i Oscillating component of total current density, A/cm2
if Faradic current density, A/cm2
if Steady-state Faradic current density, A/cm2
if Oscillating component of Faradic current density, A/cm2
i0 Exchange current density, A/cm2
j imaginary number, −1
kA rate constant for reaction identified by index A (units depend on reaction stoichiometry)
1 1
ω  9ν  3
ω 9  3
K =
dimensionless frequency, Ki =  2 
1/3
 2  Sci
Ω  a Di  Ωa 
Mi notation for species i
n number of electrons produced when one reactant ion or molecule reacts

R universal gas constant, J/mol/K
Re Ohmic resistance, Ω cm2
Rt , A Charge-transfer resistance associated with reaction A, Ω cm2
r radial coordinate, cm
r0 radius of disk, cm
si stoichiometric coefficient for species i , ( si > 0 for a reactant and si < 0 for a product)
Sci Schmidt number, Sci = ν / Di
T electrolyte temperature, K
t time, s
vr , v z radial and axial velocity components, respectively, cm/s
V potential, V
V steady-state potential, V
V oscillating contribution to potential, V
Z impedance, Ω cm2
z axial position, cm
zd diffusion impedance, Ω cm2
zi charge for species i
Greek
α coefficient used in the exponent for a constant-phase element. When α = 0 , the element behaves as an ideal capacitor in
parallel with a resistor.
αA apparent transfer coefficient for reaction A
βA Tafel slope for reaction A, V
γi Fractional surface coverage by species i

Γ Maximum surface coverage
δi characteristic diffusion length for species i
θ homogeneous solution to the oscillating dimensionless convective diffusion equation
θ′(0) derivative of the solution to the oscillating dimensionless convective diffusion equation evaluated at the electrode surface
η surface overpotential, V
κ∞ solution conductivity, (Ω cm)-1
 =
characteristic length for a finite-length diffusion layer,  2.598ν 1/ 3 Di2 / 3 / Ω
µ viscosity, g/cm s
ν kinematic viscosity,ν = µ / ρ , cm2/s
ρ density, g/cm3
τ finite-length diffusion time constant, τ =  2 / Di
ξ normalized axial position, ξ = z / δ i
Φ ohmic potential, V
ω frequency of perturbation, rad/s
Ω rotation speed, rad/sec

Advanced EIS Orazem May 2014

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advanced EIS Orazem May 2014

Uploaded by

Copyright:

Available Formats

Advanced Electrochemical Impedance

 Mark E. Orazem, 2000-2013. All rights reserved.

Advanced Electrochemical Impedance

• How to think about impedance spectroscopy

 Mark E. Orazem, 2000-2013. All rights reserved.

June 16-21, 2013

The Blind Men and the Elephant

It was six men of Indostan

The First approached the Elephant,

Interrogate Electrochemical Cell

• IUPAC Convention: Z ' and Z '', i

Advanced Electrochemical Impedance

 Mark E. Orazem, 2000-2013. All rights reserved.

• Yield information on state after transient is completed

Transient Response to a Step in Potential

current response 10-2

0 2 4 6 8 10 10-5 10-4 10-3 10-2 10-1 100

Transient Response to a Step in Potential

1.5 Short times/high frequency

current response long times/low frequency

(i-i0) / max(i-i0) 0.0

-1.0 -0.5 0.0 0.5 1.0

• Information sought is the same

System with Large Ohmic Resistance

1000 1020 1040 1060 1080 1100

Advanced Electrochemical Impedance

 Mark E. Orazem, 2000-2013. All rights reserved.

D. Macdonald, Transient Techniques in Electrochemistry, Plenum Press, NY, 1977.

Phase Sensitive Detection

V (t ) = V0 cos(ωt ) Fourier Analysis:

1.0 f = 100 Hz 1.0

Influence of Nonlinearity on Impedance

Influence of Nonlinearity on Impedance

Guideline for Linearity

0.01 0.1 0.2 1

Influence of Ohmic Resistance

Influence of Ohmic Resistance

Overpotential as a Function of Frequency

2-electrode 3-electrode 4-electrode

Working Working Working

Counterelectrode Counterelectrode Counterelectrode

• Seek uniform current/potential Electrolyte

– reduce frequency dispersion

Working Electrode Seat

Primary Current Distribution

I  V / Z   V  I Z  

Reduce Stochastic Noise

• Current measuring range

Reduce Non-Stationary Effects

Reduce Instrument Bias Errors

Improve Experimental Design

Improve Experimental Design

- Use more cycles during integration.

Improve Experimental Design

- Use the Kramers-Kronig relations to see if data are self-

- Use the Kramers-Kronig relations to see if data are self-

Can We Perform Impedance on Transient

• Timeframes for measurement

Time per Frequency for 200 rpm

0.01 0.1 1 10 100

Time per Frequency Measured

10-2 10-1 100 101 102 103 104 105

Advanced Electrochemical Impedance

• Electrical circuit components