Kinetics05 Tutorial
Kinetics05 Tutorial
Condensed-Phase Thermal
Decomposition
Alan K. Burnham
PhD, Physical Chemistry
November, 2011
Revised October, 2014
1
Outline of Presentation
2
Chemical Reactions
All life and many manufacturing processes involve
chemical reactions
• Reactants Products
Chemical reactions proceed at a finite rate
The rate of virtually all chemical reactions varies with
time and temperature
Chemical kinetics describe how chemical reaction rates
vary with time and temperature
3
Why Study Chemical Kinetics?
4
Unimolecular Decomposition
k
reactant products
5
The Energy Barrier
Endothermic reactions Exothermic reactions
Er Ef Er
Ef
products reactants
k = Ae-E/RT
k is the rate constant
A is the pre-exponential factor or frequency factor
(units are reciprocal time for unimolecular reactions)
E is the activation energy
R is the universal gas constant (1.987 cal/molK)
T is the absolute temperature (Kelvins)
8
The Arrhenius Law is Approximate
[See Burnham and Braun, Energy & Fuels 13, p. 3,1999, for specific examples]
9
Transition State Theory
Increase at Decrease at
lower pressures higher pressures
probably due to due to some type
autocatalysis of hindrance
2
[exp( x ) / x ]dx
x
where x = E/RT
• Several hundred papers have addressed the temperature
integral and its solution by various approximations
[see, for example, J. H. Flynn, Thermochimica Acta 300, 83-92, 1997] 14
Tabulations Exist For Isothermal Models
S. Vyazovkin and C. A. Wight, Annual Rev. Phys. Chem. 48, 125-49, 1997
15
Two basic approaches to kinetic analysis
Model fitting
• Do some type of numerical comparison of selected
models to determine the best model
Isoconversional fitting
• Assume that the reaction is infinitely sequential, i.e.,
that the same reactions occur at a given extent of
conversion independent of temperature
Both approaches can be used to make predictions
for other thermal histories, including with Kinetics05
16
Model Fitting is Often Done Poorly
17
One Example of Why Fitting to a Single
Heating Rate Doesn’t Work
30
Frequency Factor, ln(A/min-1)
20
10
0
0 100 200 300
Activation Energy, kJ/mol
22
Back to basics to get it right
First, get accurate data over a range of thermal histories
Next, look at the reaction profile to understand its
characteristics—you can narrow the choices considerably
Reactions can be (1) accelerating, (2) decelerating, or
(3) sigmoidal
Decelerating reactions are the
most common type for fossil fuel
kinetics
Sigmoidal reactions are the
most common type for
decomposition of energetic
materials and crystalline solids
23
Data requirements
The data should cover as wide a temperature range as
possible, as that helps constrain the model parameters
• Enough temperature change to cause a tenfold change in
reaction rate for isothermal experiments (~40 oC)
• At least a factor of 10 change for constant heating rates
Multiple heating schedules can include constant heating
rate, isothermal, and arbitrary thermal histories
• Having both nonisothermal and isothermal histories is
advantageous, because they are sensitive to different aspects
of the reaction
• Having methods that can analyze arbitrary heating rates are
advantageous, because ideal limits are hard to achieve in
practice
• Using sinusoidal ramped thermal histories is a promising but
untapped approach
24
Sidebar: Example of using a single sinusoidal
ramped heating rate C:\Kinetics\MEBrown\oscillint.otd
A = 9.2074E+12
Create synthetic data for a reaction with Derived kinetics using
the discrete E model
A = 1e13 s-1 20
E = 40 kcal/mol 15
Percent
Gaussian s = 4%
reaction order = 2 10
reacting over this thermal history
5
450 0
30 32 34 36 38 40 42 44 46 48 50 52 54 56
C:\Kinetics\MEBrown\oscillint.otd
Activation Energy, kcal/m ol
Temperature, C
400
350
300 0.8
0.6
250
0.4
200
0.2
0 5 10 15 20
Time, min 0.0
2 4 6 8 10 12 14 16 18 20
AKB 8/99 Presented by A. Burnham at UC Davis, Oct 11,1999 Tim e, m in
This approach was mentioned by J. Flynn, Thermochimica Acta 300, 83-92, 1997
It is different from modulated DSC, which is designed to separate reversible and
25
non-reversible contributions to the heat flow in DSC
Comparing Reaction Profiles to First-Order
Behavior is Instructive
0.1
Nth-order
nth-order
n = 0.5
nth-order • nth-order and
Reaction rate, s-1
n = 1.0
distributed reactivity
Reaction rate
0.01 n = 1.5
n=2
models are both
0.001 n = 0.67
n = 2.0 deceleratory
n = 0.5
n=1 • Is the rate of
0.0001
0 25 50 75 100 125
deceleration greater
0.1
Time, s
375 425 475 525
m = 0.45
m = 0.30 Temperature, C
575 or less than a first-
Modified Prout-Thomkins
Nucleation-growth m = 0.15 order reaction?
Reaction rate, s-1
m = 0.0
0.01
Reaction rate
m = 0.0
Nucleation-growth
m = 0.15 • Nucleation-growth
0.001
m = 0.30 models are sigmoidal
0.0001
m = 0.45
• m is a model growth
0.1 0 25 50 75 100 125 parameter, which is
375 425 475 525 575
Gaussian Distribution
GaussianTime,
Distribution
s of E
in E
Gaussian Temperature, C
Gaussian
sigma = 0% zero for a first-order
sigma = 2%
Reaction rate, s-1
sigma = 2% reaction
Reaction rate
0.01
sigma = 4%
0.001
sigma = 4% A. K. Burnham and R. L.
sigma = 0% Braun, Energy & Fuels
13, 1-39, 1999
0.0001
0 25 50 75 100 125 375 425 475 525 575 26
Time, s Temperature, C
The Model Selection Process Can Be
Formalized
Preliminary analysis
-inspect reaction profiles for multiple reactions
-check constancy of E by isoconversional analysis
-examine profile shape for acceleratory,
deceleratory, or auto-catalytic character
Single Complex
Reaction reaction
28
Kissinger’s method
2
ln( H r / RTmax ) E / RTmax ln( A / E )
Hr is the heating rate 0.1 1.0 10 oC/min
1.0
shifted /RT))
ln(Hrr/RT
2
-16
slope is -E/R
-18
negligible for nth-order, nucleation-
growth, and distributed reactivity -20
1.35 1.4 1.45 1.5 1.55 29
models 1000/T, K
-1
Shapes of nonisothermal reaction profiles
2.5 2.0
2.0 1.6
Weibull
Weibull nth-order
Asymmetry
Asymmetry
nth-order
1.5 1.2
Gaussian
1.0 0.8 nucleation
Gaussian
nucleation
The left-hand plot is for rate data, and the right-hand plot
is for fraction-reacted data
Finding where the sample in question is located on one of
these plots helps define the correct model 30
Another Way to Estimate Reaction Order is to
Plot Rate Versus Conversion
Fractional order reactions
are skewed to high 1.0
conversion
0.8
Higher order reactions are
d /dt (normalized)
skewed to low conversion
0.6
The three linear polymers
shown all have profiles 0.4
0.5 order
narrower than a first-order 1st order
2nd order
reaction polystyrene
0.2
LD polyethylene
Consequently, an nth- HD polyethylene
order nucleation-growth 0.0
model is appropriate 0.0 0.2 0.4 0.6 0.8 1.0
(Avrami-Erofeev or Prout-Tompkins
—more on that later)
Vyazovkin et al., Thermochim. Acta 520, 1-19 (2011)
31
Isoconversional Kinetics are Instructive and
Useful for Predictions, Also
Assumes simply that an Arrhenius
plot of the ith extent of conversion
gives a true value of E and Af() at
that extent of conversion ln(d / dt ) ,i E / RT ,i ln[ A f ( )]
Miura’s formula
• Designed to take activation energy distributions into account
to derive more fundamental A and E pairs
• For Friedman-like analysis:
AMiura ln(0.58) H r E /( RT 2e E / RT )
• For integral isoconversional analysis:
AMiura A ln(0.58) / ln(1 ) 33
Review:
Things to look for when picking a model
Is the reaction deceleratory or sigmoidal for isothermal
conditions?
• If sigmoidal, use a nucleation-growth model
Do A and E change with conversion for isoconversional
analysis?
• If an increase, use an E distribution model
• If a decrease, the reaction is probably autocatalytic
For constant heating rates, are their multiple peaks or
inflection points that suggest multiple reactions?
• If so, use parallel reactions or independent analyses
For constant heating rates, is the reaction
• Narrower or broader than a first order reaction? (see p.30)
• Is it skewed more to high or low temperature compared to
a first-order reaction? (see p.30)
34
Model fitting by nonlinear regression
Involves minimizing the residuals between measured and
calculated curves
• The minimization can be accomplished by a variety of
mathematical methods
The function to be minimized, hence the answer, will be
slightly different for analyzing rate or fraction-reacted data
• Minimizing to the actual function (rate for EGA and DSC and
fraction reacted for TGA) has some advantages
• It is possible and even desirable to minimize both
simultaneously
Minimizing to measured values is preferable to
mathematically linearizing f() or g() and using linear
regression, which usually weights the error-prone small
values too heavily
35
Models available in Kinetics05
dx d
kxn k (1 ) n
dt dt
Reaction order has a completely different
interpretation for decomposition of 8E-02
materials than in solution and gas kinetics n = 0.5
6E-02 n = 1.0
Reaction orders of 2/3 and 1/2 apply to
Reaction rate
n=1.5
shrinking spheres and cylinders, n=2.0
4E-02
respectfully
Zero-order kinetics have been observed 2E-02
for a moving planar interface
0E+00
Reaction orders greater than 1 generally 100 150 200
reflect a reactivity distribution, as a Temperature, C
gamma distribution in frequency factor is
equivalent to an nth-order reaction 37
nth-order Gaussian distribution
D( E ) (2 ) 1/ 2 s E1 exp[ ( E E0 ) 2 / 2s E2 ]
The model originated in the coal literature in the 1970s
Although the model is often described in continuous mathematical
distributions, the implementation is actually as a discrete distribution
with weighting factors approximating a Gaussian distribution
dx / dt i wi ki xi
n
39
Weibull distribution of E
D( E) ( / )[(E ) / ] 1 exp{[( E ) / ] }
is a width parameter, is a shape parameter, and is the
activation energy threshold
The Weibull distribution is very flexible and can approximate
Gaussian and nth-order distributions
Up to three parallel Weibull reactions are allowed
A Weibull distribution in E is completely different from a Weibull
distribution in temperature advocated by some
• the later is useful only for smoothing data, not deriving kinetics
Although the model is described as a continuous mathematical
distribution, the implementation is actually as a discrete distribution
with weighting factors approximating a Weibull distribution
40
Discrete E distribution
41
Nucleation-Growth Models
These were developed for solid-state reactions and linear polymer
decomposition more than 50 years ago
• Don’t be a Luddite and ignore them
Different variations emerged from the Prout-Tompkins and Avrami-
Erofeev (or JMA) approaches for solid-state reactions
They are equally applicable to organic pyrolysis reactions
• Initiation is analogous to nucleation
• Propagation is analogous to growth
• It is an approximation to the autocatalytic reaction A B; B + A 2B
We use the extended Prout-Tompkins formalism
dx / dt kxn (1 qx) m
• x (=1-) is the fraction remaining
• q is a user selectable initiation parameter (default in Kinetics05 is 0.99)
• m is a growth parameter
• n is still the reaction order 42
Equilibrium-Limited Nucleation-Growth
(1) Initiation P 2R P R
(2) Recombination/disproportionation R + R P + 2E R P + E
2.0 1.6
Kinetic parameters Weibull nth-order
Asymmetry
Asymmetry
1.5 1.2
Isoconversional analysis: Gaussian
E ~ 42 kcal/mol; A ~ 1012 s-1 1.0 0.8 nucleation
approximately independent of conversion nucleation
0.5 0.4
1st-order
Extended Kissinger analysis:
E, kcal/mol 43.3 0.0 0.0
A, s-1 4.0 1012 0.5 1.0 1.5 2.0 2.5 0.5 1.0
C:\Kinetics\Biom as s\fibcell.otm
rel. width 0.71 } FWHH
asym. 0.64 oC/min
approx. m 0.48
Normalized reaction rate 0.94 6.7 48
1.0
0.8 0.8
Fraction reacted
Fraction reacted
0.6 0.6
0.4 0.4
Note
0.2 sigmoidal 0.2
shape!
0.0 0.0
0 200 400 600 800 1000 1200 1400 500 550 600
Tim e, m in Temperature, C
dx/dt =-kx0.9(1-0.99x)0.9
Nonlinear regression k = 7.9x1012exp(-28940/T) s-1 [57.5 kcal/mol]
48
A. K. Burnham, J. Thermal Anal. Cal. 60, 895-908, 2000
Frejus Boghead Coal is a Good Example of a
Well-Preserved Algal Kerogen
50
The Nucleation-Growth model Fits the Frejus
Fluidized-Bed Data Very Well, Also
52
Pittsburgh #8 Coal Pyrolysis Requires a
Distributed Reactivity Model
Isoconversion analysis Approximate kinetic
Activation energy, kcal/mol
120 parameters
Friedman
100 E50% 57 kcal/mol
Modified Coats- A50% 4.1x1014 s-1
Redfern
80 Width relative
to 1st-order rxn 3.6
60
Asymmetry 2.4
(skewed to high T)
40
0 0.2 0.4 0.6 0.8 n from width 5.1
Fraction reacted Gaussian s 9.8%
53
The Discrete E Model Easily Provides the
Best Fit to Pittsburgh #8 Volatiles
C:\Kinetics\M iura coal kinetics\M iuraPitt8.otg C:\Kinetics\M iura coal kinetics\miur apitt8.otd
1.0
0.8 0.8
Fraction reacted
Fraction reacted
0.6 0.6
Gaussian Discrete
0.4 rss =0.799 0.4 rss =0.011
0.2 0.2
0.0 0.0
300 400 500 600 700 800 300 400 500 600 700 800
Temperature, C Temperature, C
C:\Kinetics\M iura coal kinetics\miur apitt8.otn
30
-1
A = 4 .3 8 E + 1 4 s
P ercent of reaction
0.8
Fraction reacted
20
0.6
nth-order
0.4 rss =0.195
10
0.2
0.0 0
300 400 500 600 700 800
43
49
55
61
67
73
79
85
91
Temperature, C
A c tiva tio n e ne rg y, k c a l/m o l 54
The Extended Discrete Model Fits Slightly Better
and Agrees Better With Isoconversional Analysis
1.E+27 1.000
Miura isoconversion Model
1.E+24 LLNL isoconversion Discrete
Frequency factor
0.800
extended discrete CR-Miura
Reaction rate
1.E+21 lnA=a+bE
0.600
1.E+18 1000 K/s
0.400
1.E+15
1.E+12 0.200
1.E+09
0.000
30 50 70 90 110
425 525 625 725
Activation energy, kcal/mol Temperature
40 1.000
ln(A) = 21.68 + 2.33E-04*E Model
Discrete
% of total reaction
30 0.800
CR-Miura
Reaction rate lnA=a+bE
20 0.600
3 K/m.y.
10 0.400
0 0.200
36926
49482
62038
74594
87149
99705
112261
124816
137372
0.000
55
50 75 100 125 150 175 200
Activation energy, cal/mol Temperature
The Ability of the Discrete Model to Model
Residual Activity Has Been Tested
1 7 50 oC/min
Measure Pyromat kinetics 1.0
Pyromat
1.2
0.8
Fraction reacted
1.0
0.6
0.8
0.6 0.4
0.4
0.2
0.2
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 100 200 300
Fraction reacted Tim e, m in
Sublimation is
Kinetic parameters -1
endothermic
-3
Rxn 1 27% of total Rxn 2 73% of total Heat flow
-5 Decomposition
A 1.36x107 s-1 A 4.19x106 s-1 is exothermic
E 22.8 kcal/mol E 26.6 kcal/mol -7
experiment
m 1.00 m 0.00 nonisothermal kinetics
-9
n 1.96 n 0.28 isothermal kinetics
Decomposition: Sublimation: -11
nucleation-growth receding interface 200 250 300 350 400 62
450
Temperature, C 62