Anda di halaman 1dari 107

Combustion

1. Introduction to Combustion
Combustion is a physical and chemical phenomenon during which the chemical energy is
released as thermal energy as the result of the reaction between fuel and oxidiser. Fuel may
contain many kinds of chemical species that consist of many kinds of chemical elements.
More than 90 % of fuel used as energy source is based on carbon and hydrogen and more than
90 % of oxidiser used in combustion is the oxygen of the ambient air. However, there are
some reactions between other kind chemical species or elements that we may also call
combustion. For example, reaction between hydrogen and flor:
H+F

HF

The common property of the reactions that we may consider as combustion is a net heat
release. This heat release is due to the positive difference of the chemical bond energy
between the initial species (fuel + oxidiser or reactants) and the final species (products) *.
Some combustion reactions and the usage places are shown in Table 1.
Table 1 Some fuel and oxidiser pairs
Fuel

State

Oxidiser

State

Products

Usage Place

H2

gas

O2

gas

H2O

Rockets

solid

O2

gas

B2 O3

Rockets

CH4

gas

Air

gas

CO2+ H2O+ N2

Heating

solid

Air

gas

CO2+ N2

Heating

C7.5H17

liquid

Air

gas

CO2+ H2O+ N2

Automotive

C3.H5 N3 O9

liquid

CO2+ H2O+ N2

Detonation

Combustion begins from the initial state (of reactants) and proceeds in time to the final state
(of products) via various intermediate reactions and species. However, for the first step, it is
more practical to assume the reactions to occur with infinite rate and consider only the
reactants and the final products. This approach gives us the opportunity to analyse the
combustion system in a much easier way, using only the classical conservation laws and the
algebraic equations between temperature and species concentrations. However, it is of course
necessary in many cases to analyse the real physical combustion process using the governing
differential equations of time, concentrations and temperature.

There are some reactions between fuel and oxidiser (for example reaction between hydrogen and air) during
which not heat but free electrons and ions are produced as energy carrier. These reactions are called cold
combustion. The device that makes the direct generation of electricity possible from the cold combustion is
called fuel cell.

Let us consider the combustion of general hydrocarbon fuel (CnHm) and oxygen (O2) mixture
and calculate oxygen to fuel ratio for a complete combustion. In case of hydrocarbon fuels,
complete combustion means that all carbon atoms and hydrogen molecules react with oxygen
to form carbon dioxide (CO2) and water (H2O).
The reaction describing the complete combustion of a hydrocarbon fuel molecule is:
m
CnHm + (n + 4 ) O2 
This means that (n +

m
n CO2 + 2 H2O

(1)

m
4 ) moles of oxygen are required to burn one mole of fuel (CnHm)

completely.
If the hydrocarbon fuel consists of the elements in the following mass fractions *,
Carbon
Hydrogen
Oxygen
Nitrogen
Sulphur

:
:
:
:

c
h
o
n
:

Then**
c+h+o+n+s=1

(2)

The molecular weight (Mf) of such a fuel can directly be calculated with the following
formula:
Mf = 12.010 c + 1.008 h + 14.007 n + 36.060 s
To burn the c kg carbon of the 1 kg hydrocarbon fuel completely,
C +

O2

CO2

1 kmole CO2

O2

44 kg

CO2

O2

22.4 m3

CO2

12 kg

C + 6.0236 1023**** O2

6.0236 1023

CO2

1 kg

C +

32
12 kg

O2

44
12 kg

CO2

c kg

C +

8c
3 kg

O2

11c
3 kg

CO2

1 kmole

C +

1 kmole

12 kg ***

C +

32 kg

12 kg

C +

22.4 m3

O2

(8c/3) kg O2 and to burn the h kg of hydrogen to H2O

For example c kg carbon in 1 kg fuel


There are of course other materials such as water, metals and metal oxides (ashes), although in very low
concentrations.
***
Molecular weights are rounded.
****
Avagadro number is defined as the number of molecules in one mole mass of the species
**

H2

1
2

O2

H2O

1 kmole H2 +

1
kmole O2
2

 1 kmole H2O

2 kg

H2

16 kg

O2

 18 kg H2O

h kg

H2

8h kg

O2

 9h kg H2O

8h kg O2 is required. Similarly, s kg O2 is needed to burn the sulphur in the fuel according to


the following complete combustion reaction:
S

O2 

SO2

1 kmole S

1 kmole O2 

1 kmole SO2

32 kg S

32 kg

s kg

s kg

O2
O2

64 kg

SO2

2s kg

SO2

Then the minimum total mass of oxygen (mO2 min) required for complete combustion of 1 kg
hydrocarbon fuel is:
8c
mO2 min = 3 + 8h + s o

(3)

(kg-O2)

If air is used as oxidiser instead of pure oxygen then the minimum air mass is:
mair min =

mmin O2
0.232

(kg-air)

(4)

Here, the nitrogen of the air or in fuel is assumed not to react with oxygen, although this not
completely true for the most cases.
Not that oxygen in fuel is taken into account of the oxygen source lowering the total amount
of oxidiser required.
Example:
The minimum air mass required for complete combustion of 1 kg ethanol (C2H5OH)
Molecular weight of ethanol is MC2H5OH = 46 and the mass fractions of the elements forming
ethanol are:
mcarbon =

24
46 ,

6
mhydrogen = 46 ,

16
moxygen = 46

Minimum air mass can be calculated from (3) and (4) using the mass fractions:
c = mcarbon = 0.522,

h = mhydrogen = 0.130, o = moxygen = 0.348

mair min = 8.99 kg-air/kg-ethanol


Minimum number of air moles (Nmin air ) for complete combustion of 1 mole ethanol is
Nair min = 3*4.762 = 14.29 moles - air/ mole ethanol
according to the reaction below*:
*

Assuming that 1 mole air consists of 0.21 mole O2 and 0.79 mole N2

C2H5OH + 3(O2 + 3.762 N2)

2 CO2 + 3H2O + 11.29 N2

(5)

2. Some Basic Definitions


2.1. Gaseous systems
Most combustion systems consist of the gaseous components especially when products are
concerned. Even most of the liquid or solid fuels burn in gaseous form after vaporisation and
mixing with the oxidiser. Because of that, before going further in the combustion phenomena
it will be useful to remember some descriptions relating to the multi component gaseous
system.

Mass fraction
The mass fraction (xi) of species i is defined as the mass of species i (mi ) in a
homogenous gas mixture to the total mass of the mixture (m):
mi
xi = m

(6)

Then, according to the definition,


n

mi = m

(i = 1,2,,3..n)

i=1
and
n

xi = 1

(i = 1,2,3..n)

i=1

Mole fraction
The mole fraction (yi) of species i in a n component homogenous gas mixture is the ratio
of the number of moles of species i (Ni) the total number of moles (N) in the mixture :
Ni
yi = N

(7)

Similar to the mass fraction,


n

Ni = N

(i = 1,2,3..n)

i=1
and
n

yi = 1

(i = 1,2,3..n)

i=1
Mass fraction and mole fraction can be related each other as follows:
The mass of species i in the mixture is,
mi = Mi Ni

whereas the total mass of mixture (m) may be expressed via mean molecular weight (M) for
the mixture of total number of moles of N :
m= NM
Here
n
M= Mi yi
i=1

(8)

is the mean molecular weight defined for the mixture of total number of n components. Then
the relation between mole fractions and mass fractions is:
xi =

Mi
y
M i

(9)

Partial pressure
Partial pressure (pi) of the component i in a mixture of ideal gases is the pressure, if only
this component had occupied the same volume (V). Hence, the sum of the partial
pressures of all species is the pressure of the mixture*.
n

pi = p
i=1
By definition, partial pressure and mole fraction may be related to each other using the
equation of state of the ideal gases:
pi =

Ni T
V

p=

NT
V

pi
Ni
=
p
N = yi

(10)

Partial pressure has the same value as mole fraction.

Concentration (partial molar density)


Concentration ([ ]) is defined as the number of moles of the specified species in unit
volume of the mixture.
Ni
[]: V

(number of moles / m3)

(11)

Concentration may be related to mole fraction by the following equality:


Ni
Ni p
yi p
[]: V =
=
NT
T

(12)

The Daltons law

Similar to concentration, the partial density (i) may be defined as the mass of the
specified species per unit volume of the mixture.
i =

mi
= xi
V

(kg / m3)

(13)

2.2. Stoichiometric, rich and lean mixtures


The mixture of fuel and oxidiser is called stoichiometric if one unit of fuel is mixed with
minimum amount of oxidiser as to burn the fuel completely. In case of ethanol-air mixture, 1
kg ethanol and 8.99 kg air forms 9.99 kg stoichiometric fuel-air mixture. Similarly, 1 mole
ethanol and 14.29 moles air forms 15.29 moles (stoichiometric) fuel-air mixture.
The oxidiser that takes part in combustion may be more or less than the minimum
(stoichiometric) amount. In case of oxidiser excess the mixture is called lean (lean with
respect to fuel) and in case of lack of oxidiser rich mixture. To express lean or rich mixture
mathematically, air excess coefficient () is introduced which is the ratio of amount of air
(mair) ready for combustion to the stoichiometric amount of air (mair stoich.= mair min). So the air
excess coefficient is
mair
= m
air stoic.

(14)

By definition:
If

=1

then the mixture is stoichiometric,

If

>1

then the mixture is lean,

If

<1

then the mixture is rich.

Air excess coefficient and air to fuel ratio (A/F), the amount of air (mair) ready for
combustion of 1 unit mass of fuel, are conventionally used for the air-fuel mixtures. A more
general parameter used for this purpose is equivalence ratio or fuel excess ratio which
is reciprocate of air excess ratio, i.e. the ratio of fuel participated in combustion per unit of
mass of oxidiser to the stoichiometric amount of fuel.
In the global reaction (5) describing the combustion of ethanol,
C2H5OH + 3 O2 + 11.29 N2 

2 CO2 + 3 H2O + 11.29 N2

the coefficients 1 of C2H5OH, 3 of O2, 11.29 of N2, 2 of CO2 and 3 of H2O are called
stoichiometric coefficients.
The reaction (5) may be expressed using the general stoichiometric coefficients ( ) for
reactants and () for products.
F C2H5OH + O2 O2 + N2 N2  CO2 CO2 + H2O H2O + N2 N2
(reactants)
(products)

(15)

The stoichiometric coefficients are here:


F = 1,

O2 = 3,

N2 = 11.29

CO2 = 2,

H2O = 3,

N2 = 11.29

Stoichiometric coefficients are the number of moles of species and do not need to have the
values of stoichiometric mixture. However, stoichiometric mixture (reactants) requires that
6

the number of moles of oxidiser (NO2) to the number of moles of fuel (NF ) is equal to the ratio
of stoichiometric coefficient:
NO2
O2
N stoich. =
F
F
or in terms of mass fractions

xO2
O2 MO2
stoich. =
=
xF
F MF

(16)

Here MO2 and MF are the mole mass of oxygen and fuel respectively and is the
stoichiometric mass ratio or stoichiometric air to fuel ratio (A/F). In case of lean or rich
mixture the air excess coefficient is,
NO2
=
O2

(from

N O2 F
N F O2

for NF = F = 1)

or equivalence ratio is:


NF O2
(from N
F
O2

NF
=
F

for NO2 = O2 = 1)

As a remark, we must point out that we need to take into account only oxygen representing the
combustion air, because nitrogen does not take part in oxidation reaction *.
Stoichiometry can be defined more generally for any kind of reaction on the basis of
conservation principle of the elements during reaction.
Each of the equations such as
1
H2 + 2 O2

H + O2

 OH + O

H2O

(17a)
(17b)

can be cast into the following form :


n

'i Mi =

i=1

"i Mi

(18)

i=1

where 'i and i are stoichiometric coefficients of species i for reactant (left hand) and
product (right hand) sides respectively and Mi stands for the chemical symbol for of species i.
Not that the number of species (n) is assumed same for both sides. The non-existence of a
species, for example k, at any sides is expressed with the stoichiometric coefficient, which is
equal to zero (k=0 or k=0). Then the net stoichiometric coefficient is:
i = i - I

(19)

Nair
air
For air as oxidiser (16) may be expressed
stoich. =
NF
F

with Nair = NO2 + 3.76 NN2 and air = O2 + 3.76 N2

This coefficient is positive for products and negative for reactants and zero for inert elements.
System of r elementary reactions (17) may be expressed using this coefficient in the short
form given below:
n

ik Mi = 0

(k = 1,2,3,r)

(20)

i=1
Here ik is the net stoichiometric coefficient (or net change) of species i in reaction k .
The ratio of change of the number of moles of species (i) to that of species, for example (1)
can also be expressed as the ratio of the net stoichiometric coefficients of these species:
dni
i
=
dn1
1
Then the ratio of changes in partial masses is:
dmi
i Mi
=
dm1
1 M1

(21)

Since, the total mass of the reactants and products are same and do not change during the
reaction (which is of course not true for the total number of moles), the same equality as (21)
will be valid for mass fractions too:
dxi
i Mi
dx1 = 1 M1

(22)

Integrating (22) for the stoichiometric reaction (5) between the initial unburnt state (subscript
u) and any later state, with i : O2 , 1 : F and 1O2 = 0, F = 0,
xO2 - xO2u 'O2 MO2
xF - xFu = 'f MF
(23)
or using the definition (16), the stoichiometric mass ratio , this may be written as,
xF xO2 = xFu xO2u

(24)

What effect the equivalence ratio or air excess coefficient on combustion has, can be seen in
the following analyses:
2.3. Mixture fraction
For a two component (mixture of oxidiser and fuel) combustion system mixture fraction (Z)
is defined as the ratio of unburnt fuel mass (mF) to the total mass of unburnt fuel and oxidiser
mixture (mF + mO). This is the mass fraction of fuel in the unburnt mixture.
mF
Z = m +m
F
O

(25)

Mixture fraction may be related to the mass fractions of fuel and oxidiser as follows:
From (25) we obtain
mF
Z
=
mO
1-Z

Since, according to this, the fuel mass in this mixture would change proportional to Z and the
oxidizer mass to (1-Z), the mass fractions would also change with the same proportionality:
xF = Z xF 0

(26)

xO2 = (1-Z) xO2 0

(27)

and
Fuel or oxidizer itself may be a mixture of various species and here xF 0 and xO2 0 are base
(initial or unburnt) values of mass fractions of fuel and oxygen in the mixtures. If oxidizer is
air (mixture of oxygen and nitrogen), then the initial mass fraction of oxygen is xO2 0 = 0.232.
The dependence of fuel and oxygen mass fractions in unburnt mixture on mixture fraction can
be seen in Figure 1.

X O2 0

1,00

1,00

0,75

0,75

0,50

0,50

0,25

0,25

0,00
0,00

XF 0

0,00
0,25

0,50

0,75

1,00

Mixture Fraction

Figure 1. The dependence of fuel and oxygen mass fractions in unburnt mixture on mixture fraction

Mixture fraction may be written in term of mass fractions introducing (26) and (27) into
equation 24 ( xF xO2 = xFu xO2u) :
Z =

xF xO2 + xO2 0
xF 0 + xO2 0

(28)

For stoichiometric mixture, that is, for xF = xO2


Zst =

xO2 0
xF 0 + xO2 0

(29)

If Z > Zst, that means xO2 > xF and fuel is deficient (lean mixture). Then, combustion
terminates when all the fuel is consumed. That means there is no fuel in the burnt gas mixture
(xF b=0). The remaining oxygen in the burnt gas mixture may then be calculated from (28),
substituting xF = 0, as
Z
xO2 b = xO2 0 (1 - Z )
st
Similarly, if Z > Zst, that means xO2 < xF and oxidiser is deficient (rich mixture). Then,
combustion terminates when all the oxidiser is consumed (xO2,b=0). The remaining fuel the
burnt gas is found, substituting xF = 0 in (28), as
Z - Zst
xF,b = xF,0 1 - Z
st

On the other hand, for hydrocarbon fuel (CnHm), the element mass fractions (for C and H) in
the unburnt mixture are
Mc
xc = n M xF,u ,
F
(= mass of carbon in fuel x mass of fuel in mixture = mass of carbon in mixture)

xH = m

MH
xF,u
MF

and in the burnt gas


xc = n

Mc
Mc
x +
MF F,b
MCO2

xCO2,b

(= mass of carbon in burnt mixture + mass of carbon in CO2 x mass of CO2 in burnt mixture)

MH
Mc
xH = m M xF,b + M
F
H20

xH20,b

For lean mixture (Z > Zst ), using (26) and with xF,b = 0 (that means there is no fuel in the
burnt gas mixture) we have:
xCO2,b = xCO2,st

Z
,
Zst

xH2O,b = xH2O,st

Z
Zst

And for rich mixture (Z < Zst ), with xO2,b = 0 (no oxygen in the burnt gas mixture) we have:
Z - Zst
xCO2,b = xCO2,st 1 - Z ,
st

Z - Zst
xH2O,b = xH2O,st 1 - Z
st

Here, the mass fraction of carbon dioxide in burnt gas mixture is calculated as follows:
xCO2,st = xF Zst n

MCO2
MF

(mass of carbon dioxide in total stoichiometric mixture = mass of fuel in total x CO2 in fuel)

MH2O
xH2O,st = xF Zst m 2M
F
The dependence of fuel, oxidizer and some products mass fractions in burnt mixture of
ethanol on mixture fraction is seen in Figure 2.

X F0 , X o2 0

XF 0

0,2

0,75

0,15

H 2O
0,50

0,1

XO20

0,25

0,05

0,00

X H2 O , X CO 2

CO 2

1,00

0,00

0,2

Z st =0.1

0,4

0,6

0,8

M ixture Fraction (Z)

Figure 2. The dependence of fuel, oxygen and products mass fractions in burnt mixture on mixture fraction

10

2.4. Heat value of the fuel


The first law of thermodynamics describes the balance between different kind of energies
(internal energy (e), enthalpy (h), heat energy (dq)) and work (w) for a thermodynamical
system:
de + pdv = dh vdp = dq + dwR
or
dh = de + d(pv)
For a multi component system
n

e=

xi ei ,

h=

i=1

x i hi
i=1

As the specific heats (cv , cp) of a real gas depend on the temperature, the specific internal
energy (kJ/kg or kJ/kmole) and specific enthalpy (kJ/kg or kJ/kmole) at temperature T must be
expressed by the following integral:
T
cv (T) dT + e0
e (T) =
0

(18)

T
cp (T) dT + h0
h (T) =
0

(30)

and

Here e0 and h0 are the specific internal energy and enthalpy at absolute zero (T=0). However,
as by definition:
h = e + pv = e + T

(kJ/kmole)

( = 8.314 kJ/kmole K, the universal gas constant).


e0 will be equal to h0 (e0 = h0) for T=0.
There are some methods to calculate the enthalpy or internal energy at a given temperature.
One method is to use the mean specific heats for a given temperature interval (i.e. from the
reference temperature T0 to the temperature T ). In this case, cp is given in the form
cp = a + bT + cT +
where a, b, c, are constants. Then the enthalpy difference between two temperatures can be
calculated as follows:
h (T2) - h (T1) = cpT2 (T2 - T0) - cpT1(T1 -T0)
However, the dependence of the specific heats on temperature is generally expressed in
polynomial form with coefficients determined experimentally. If cp is given with the polynom
below:
cp = a0 + a1 T + a2 T2 + a3 T3 + a4 T4
Then the specific enthalpy takes the form

11

(kJ/kmole K)

(31)

T
2
3
4
5
h (T) =
cp (T) dT = a0 T + a1 T + a2 T + a3 T + a4 T
0
or more conveniently,
h (T)
= a0 + a1 T + a2 T2 + a3 T3 + a4 T4
T
Similarly, with e(T) = h(T) + T, for internal energy
e (T)
= -1 + a0 + a1 T + a2 T2 + a3 T3 + a4 T4
T
In these polynoms, the value of the coefficients ( a0, a1, ) are generally so given that e and h
are obtained in kJ/kmole. To get the internal energy or enthalpy in kJ/kg, the result must be
divided by molecular weight of the species or element. It must be pointed out that the
coefficients of the polynoms are only for real gases and do not give us the changes in internal
energy or enthalpy due to the phase changes. If gas liquefies or solidifies in the temperature
interval considered then the latent heat of vaporization or sublimation must be included
manually in the calculation. For example, the enthalpy of water at 100 Co is 2300 kJ/kg (latent
heat of vaporisation of water at 0.1 Mpa) less than the water vapor at the same temperature.
However, there is some published enthalpy data for gases that include the additional terms for
phase changes.
The specific internal energy or enthalpy (e0 , h0 ) at absolute zero is zero for elements and has
the value of heat of formation for the substances at this temperature. According to the sign
convention heat of formation is negative for the substances during which formation heat is
released. For water, for example, the heat of formation at absolute zero is h0 = - 2.391 105
kJ/kmole. That means this amount of heat is released during the formation of 1 kmole (18 kg)
H2O from its elements (H and O), according to the reaction given below:
1
H2 + 2 O2 H2O + 2.391 105 kJ
Such a reaction is called exothermic. Reactions during which heat is absorbed are called
endothermic.
Heat of Reaction:
During combustion the released heat is absorbed by the products increasing their temperature.
The heat of reaction (net heat released) and the temperature of the end products can be
calculated if heat of formations of all compounds and the enthalpy (or internal energy) of all
species, depending on the temperature, are known. The total enthalpy of reactants (subscribt
R) and products (subscribt P) may be expressed in two parts, a temperature dependent part and
a constant part, as follows:
n

hR =

[mi (hi (T) + hi0)]R =

hR(T) + h0R

i=1
n

hP =

[mi (hi (T) + hi0)]P

hR(T) + h0P

i=1
(for non existing species mi = 0)
12

The heat of reaction Qp(T) at constant pressure and at the temperature T is the difference
between the enthalpy of products and that of reactants:
Qp(T) = hP(T) hR(T) + (h0P - h0R)
or
Qp(T) = h (T) + h0 = h
Heat of reaction depends on the temperature as it may be seen from Fig 3.
Similarly, the heat of reaction at constant volume may be found using internal energy instead
of the enthalpy:
Qv(T) = eP(T) eR(T) + (e0P - e0R)
Reac tants

Entha lpy

P roduc ts

h R (500)

h OR
h0

h P (500)

h OP

Q p(500)

0
0

1000

2000

T= 500

3000

Te m pe ra ture (K)

Figure 3. Heat of reaction at constant pressure

Example:
For the reaction between carbon monoxide and oxygen,
CO + O2

CO2 +

1
O
2 2

28 kg CO + 32 kg O2 60 kg reactant
44 kg CO2 + 16 kg O2 60 kg product
mCO =

28
= 0,47,
60

mO2 =

44
mCO2 = 60 = 0,73,

32
= 0,73
60

16
mO2 = 60 = 0,27

(h0)CO = - 0,0406 108 j/kg


(h0)CO2 = - 0,0894 108 j/kg
(h0)O2 = 0
h0 =

(mihi0) P - (mihi0) R = -0,73 x 0.0894 108 0.47 (- 0,0406 108) h0 = -

i=1
i=1
8
0,047 10 J/kg-reactant
The heat of reaction at constant pressure and at T = 500 K is then:

13

Qp = 0,73 x 402750 + 0,27 x 460340 (0,47 x 519170 0,53 x 460350) 0,047 108
Qp = 0,0473 108 J/kg-reactant
The heat of reaction is the heat value (calorific value) of CO as a fuel at 500 K.
Lower and higher heat values:
If there is water (H2O) in the products two kind of heat values must be considered:
If water is in liquid form (steam) in the reaction below,
CH4 + 2O2 2H2O + CO2 + (- 8.776 105 kJ)
the higher heat value of CH4 is then

8.776 105
= 54850 kJ/kg-CH4
16

If water is in vapour form (steam) in the reaction


CH4 + 2O2 2H2O + CO2 + (- 8.776 105 kJ + 2 x 0.375 105 kJ)
the lower heat value can be calculated, using 0.375 105 kJ/kmole for the latent heat of
8.027 105
vaporisation of water, as
= 50168 kJ/kg-CH4
16
Since the water is in the vapour form in many combustion applications, including the exhaust
gases of internal combustion engines, the lower heat values are given mostly as the heat
content of the fuels.
2.5. Adiabatic combustion temperature
Assuming that there is no heat transfer from or to the system, whole heat released causes the
product temperature to increase. That means, that the total enthalpy of the reactants at the
initial temperature (of the reactants) TR , will be equal to the total enthalpy of the final
temperature (of the products) TP. The temperature TP meeting this condition is called
adiabatic combustion (or flame) temperature at constant pressure. Indeed, if write the first
law of thermodynamic,
dh vdp = dq + dwR
for constant pressure (dp=0) and isentropic (dq=0, dWR=0) process
dh = 0
which means,
(hR)T=TR = (hP)T=TP
or
hP(TP) hR(TR) + h0 = 0

(32)

14

Reac tants

Entha lpy

P roduc ts

h OR

h R (500) = h P(T P)

h OP
0
0

T= 500

1000

2000

3000

TP

Te m pe ra ture (K)

Figure 4. Adiabatic Combustion temperature at constant pressure

Similar to the heat of reaction at constant volume, the adiabatic combustion temperature at
constant volume is calculated from
eP(TP) eR(TR) + e0 = 0

(33)

(see Fig 5.)


TP (or Tad) can be found only by iteration from (32), since enthalpy hP(TP) and composition of
products itself depend on the temperature (Fig. 4). However, adiabatic combustion
temperature may be calculated directly for one-step global reaction under some simplifying
assumptions:
The equality (32) from which the product temperature TP is calculated by iteration may be
rewritten in term of heat of reaction,
hP(TP) - hP(TR) = hR(TR) - hP(TR) - h0
or
hP(TP) - hP(TR) = Qp(TR)

(34)

Inte rna l
Ene rgy

Reactants
P roduc ts

e OR

e R (500) = e P(T P)

e OP
0
0

1000

2000

TP

3000

Te m pe ra ture (K)

Figure 5. Adiabatic Combustion temperature at constant volume

Here, the temperature depended part of the enthalpy of the products or reactants are is to be
T

calculated by integrating the specific heats at constant pressure (h (T) =


cp (T) dT .). The
0

specific heat of products and reactants are the sum of the specific heats of the components:
15

[xi Cpi (T) ]P

(Cp)P =

i=1
n

(Cp)R =

[xi Cpi (T) ]R


i=1

First assumption to be made for simplicity is to take the specific heats constant:

For combustion with air the contribution of nitrogen is dominant in (Cp)P. Its specific heat is
about (Cp)N2 =1.3 kJ/kg K at temperatures around 2000 K. The value of (Cp)CO2 is somewhat
larger and that of O2 is smaller, while the specific heat of H2O is twice as large as that of O2.
So, as a first approximation the specific heat of the products (especially for lean and
stoichiometric mixture) may be assumed (Cp)P =1.4 kJ/kg K . With the assumption above the
left hand side of (33) may be expressed as follows:
hP(TP) - hP(TR) Cp(TP - TR)

(35)

As a second approximation we may neglect the temperature depended part of heat of reaction
and assume that this quantity consists of only heat of formation h0 at T = 0 or generally, href at
a reference temperature Tref. Then,
n

Qref =

(xi R - xi P)hi ref = ihi ref

(36)

i=1
On the other hand, heat of reaction written in terms of mass fractions of all species may be
expressed globally in term of fuel mass fraction, by integrating (22) between initial (R ) and
final (P) states:
xiR - xiP
i Mi
=
xFR - xFP
F MF
n

iMi hi ref
Qref = (xFR - xiFP)

i=1

(37)

FMF

For lean or stochiometric mixture there will be no fuel in products


xFP = 0 or F = 0

F = F - F = F

and

and introducing (35) and (37) in (34) we obtain for TP,


n

iMi hi ref
TP = TR + xFR

i=1
CP 'FMF

(38)

or
TP = TR +

Q ref xFR
Cp 'FMF

(39)

(Here the dimension of Q ref is obtained in kJ/kmol when the dimension of hi ref is kJ/kg. Then
Cp must be used in kJ/kg K)

16

In case of rich mixture, with


xO2P = 0 or O2 = 0 ,

O2 = O2 - O2 = O2 and

xiR - xiP
xO2R - xO2P

i Mi
O2 MO2
we obtain
TP = TR +

Q ref xO2R
Cp 'O2MF

(40)

Equations(39) and (40) may be expressed in term of the mixture fraction by introducing,
xFR = Z xF 1
xO2R = (1- Z) xO2 1
and by specifying the temperature of the unburnt mixture (reactants) consisting of components
of different temperature,
TR (Z) = TO2 + Z (TO2 TF)
we obtain
TP (Z) = TR (Z) +

Q ref xFR
Z
Cp 'FMF

(for Z<Zst)

(41)

TP (Z) = TR (Z) +

Q ref xO2R
(1 Z)
Cp 'O2MF

(for Z>Zst)

(42)

Maximum flame temperature is obtained for Z = Zst (Fig 6.):


Q ref xFR
Z
Cp 'FMF st

TP (Zst) = TR (Zst) +

(43)

(for Z=Zst)

T e m p e r atur e

TP

TO2

TR
TF

0
0,00

0
0,2

0,4

0,6

0,8

Zst= 0.1 M ixtu r e Fr actio n (Z)

Figure 6. The dependence of adiabatic combustion temperature on mixture fraction

3. Chemical Reactions with Finite Reaction Rate and Chemical


Equilibrium
3.1 Reaction Rate
While writing the global combustion reaction (5) of ethanol

17

C2H5OH + 3O2 + 11.29 N2

2 CO2 + 3H2O + 11.29 N2

(5)

we have assumed that the one step reactions


C + O2 CO2
H2 +

1
O H2 O
2 2

are:

unidirectional

completed with in an infinite small time interval with infinite reaction rates.

But in the physical world, all chemical reactions are bidirectional, occur with finite rate and
no complete combustion can be realised. That means some species of the reactant side
(unburnt mixture) will exist in the product side (burnt mixture) too. This is due to the
simultaneous reactions in the product side creating initial species of the reactant side.
Backward reactions creating initial species by decomposition of the product species are also
called dissociation of the combustion products. For example, the reaction that is a step of
the combustion of hydrocarbon fuel:
CO + H2 O

kf

kb

H2 + CO2

(44)

occurs in two directions, with finite rate of reactions vf and vb, forward (direction to create
products) and backward (direction to create reactants) respectively. Here, the reaction rates vf
and vb are described as the change in total number of reactants (or products) molecules per
unit volume (molar density or concentration) per unit time and proportional to the forward and
backward rate coefficients (vf kf and vb kb). The forward reaction rate of the reaction (44)
may be expressed in the mathematical form as follows:
vf =

d(H2-CO2)
:
dt

(number of moles of reactants / m3 s)

(45)

In many combustion phenomena the reaction rates are depend on the concentrations of the
reacting species. This is very reasonable, because, for a possible reaction, two (or more) atoms
or molecules must meet together firs of all. The reaction rate will then be proportional to the
frequency of collision between the reacting atoms, hence to the concentrations. The
collision frequency (Y) between the molecules of a species A may be expressed
mathematically by using the kinetic theory of gases, for a first approximation, as follows:
Let us assume that molecules are hard spheres of diameter d and a collision would occur when
the distance between the centers of two molecules are smaller than 2d (Fig. 7).

18

Figure 7. The collision model of two molecules

Then the volume scanned per second by the molecule during its travel with a mean velocity of
c is d2 c (m3/s). If n is the number of molecules in unit volume (concentration) then the total
number of collisions will be nd2 c . One must use of course, the mean velocity instead of the
absolute velocity. Then the collision frequency of the molecules of one species is
YAA =

1
2

2 n2 d2 c

We must introduce the (1/2) factor, otherwise a collision between two molecules is counted
twice. The mean velocity c of the molecule is given, according to the kinetic theory of gases
as function of temperature:
c=

8 kT
m

R
Universal Gas Constant
where k = N =
is the Boltzman Constant and m is the mass of
Avagadro Number
the molecule. Then,

YAA =

2
2

8 kT
m

n2 d2

and in case of reaction between two different molecules (A) and (B) the collision frequency
will be
YAB =

2
2

nA nB d2AB

mA + mB
8 kT m m
A
B

Here, nA , nB are the concentrations, mA , mB the moll masses of the species and dAB =

dA + dB
2

the linearly averaged diameter of the molecules.


However, the number of collisions calculated by the collision theory is not the real value
and it is possible to obtain better results using the statistical theory in calculating the
thermodynamical properties of gases. But, both approaches show clearly the concentration
dependence of the reaction rates. The dependence of reaction rate on concentration and time
may be seen from Fig. 8 for reaction A B.

Conce ntra tion

V A = d[A ]/dt = dec rease rate of initial spec ies

V B = d[B ]/dt = increas e rate of final s pec ies

0
0

Tim e

Figure 8. The dependence of reaction rate on concentration and time

19

However, one must bear in mind that the rate of a reaction may have different values
depending on how this rate is defined. For the reaction below, for example,
N2 + 3H2

2NH3

the rate of formation of NH3 molecules is twice as much as the decay rate of N2 molecules,
vNH3 =

d [NH3]
d [N2]
= - 2vN2 = - 2 dt
dt

and the rate of decrease of H2 is three times as much as the decrease rate of N2 ,
vN2 =

d [H2]
d [N2]
= 3vN2 = 3 dt
dt

3.2 Order of reaction


The order of a reaction is the sum of the power of the concentrations of the participating
species. For example:

Decomposition of Ethan to Ethilen and Hydrogen


C2H6 C2H4 + H2
vC2H6 = k [C2H6]

is the first order.

Decomposition of Iodhydrogen
2HJ

H2 + J2

or the backward reaction


H2 + J2 2HJ
is 2. order
d [HJ] [HJ]
d [HJ]2
vHJ = k
= k dt
dt
v(H2+J2) = k [H2] [J2]

The reaction
2NO + Cl2

2NOCL
2

v(NO+CL2) = k [NO] [Cl]


is third order.
The rate or the order (n) of a reaction between many species can generally be expressed:

v = k C + C +
A
B
n=++
However, when the rate of reaction cannot be expressed explicitly in terms of concentrations,
such as the reaction given below, it is not possible to define an order for the reaction.

20

1
1
H
2+
2
2 Br2 = HBr
d [HBr] k [H2] [Br]1/2
=
dt
k'[HBr]
1 + [Br]

3.3 Rate coefficient


The proportionality factor k in the rate expressions is called rate constant or rate
coefficient. Rate coefficient is a better description for this factor, since it is actually not a
constant but varies to some extent depending on the conditions of the experiment, as this
coefficient is obtained experimentally.
Activation energy and Arhenius model for the reaction rate:
Another main factor which affect the reaction rate, other than the concentrations, is the energy
level of t he reacting molecules. Because, only the collisions of the molecules that have a
certain (minimum) energy level may results in a reaction. The dependence of the rate
coefficient (k) on both collision number (collision frequency) and energy level of the molecule
is given by S. Arhenius (1885) in the following form:
E
ln k = ln A - RT
or
- E

k=Ae

RT

where A is the frequency factor which gives the collision probability without effect of
concentrations, i.e.,
YAB
AAB [A][B]
or
YAA
AAA [A]2
E is the activation energy or heat of activation which denotes the energy that the
molecule must acquire before it can take part in the reaction. Only the collisions of such
molecules can results in a reaction to build a different molecule.
Although it seems to be possible to calculate the activation energy theoretically and to obtain
accurate results for some reactions, there is not a general method to calculate it precisely.
Instead, this quantity, involving the frequency factor too, is obtained experimentally.
For example, for the reaction
AB
with the initial concentrations:
[A] = a and

[B] = x = 0

the formation rate of B is:

21

dx
dt = k (a - x)
Separation of variables leads:
dx
=kx
a-x
for x = 0,
ln

- ln (a x) = kt + constant

at t = 0

constant = - ln a

a
a-x =kt

- E

x = a (1 - e

RT

and
a
ln a - x
k=
t

(47)

is obtained. By measuring the concentration x of substance B during the development of the


reaction one can calculate an average value for k (Fig. 9). Such an approximation is possible,
since the temperature dependence of collision number can be neglected, although this is not
true for real reactions. Indeed, with
- E

v = k [A] [B],

k = AAB e

RT

YAB
, AAB [A][B]

and with,
YAB =

2
2

[A][B] d2AB [8 kT

mA + mB 1/2
]
mA mB

m +m

AAB d2AB [8 kT mA m B ]1/2


A
B
ln k = (B +

1
E
ln T) 2
RT

ln a/(a-x)

The experimentally obtained linear relationship between ln k and 1/T shows that the
temperature dependence of collision number can be neglected (Fig. 10).

0
0

Time
Figure 9. Experimentally obtained rate coefficient

22

ln k
0
0

1/T
Figure 10. Experimentally obtained correlation between temperature and rate coefficient

3.4 Chemical equilibrium


Even if the chemical reactions occur in two directions, i.e. forward and backward, there is
always a dominant direction that is determined by the greater reaction velocity at the
temperature considered. Chemical equilibrium will be established at the temperature and
concentrations for which forward and bacward rates are equal and the concentrations of the
products do not change any more. For the reaction (44), for example,
kf
CO + H2 O
kb

H2 + CO2

when equilibrium is reached, the formation rate of the products will be equal to the formation
rate of the reactants by backward reactions.
kf [CO][H2 O] = kb [H2][CO2] = Rf

(46)

Here, Rf is the one way equilibrium rate.


The ratio of forward rate coefficient kf to the backward rate coefficient kb at the equilibrium
state is the equilibrium constant( Kc) written in terms of concentrations.
kf
[H2][CO2]
Kc = k = [CO][H O]
f
2
Equilibrium constant may also be expressed in terms of mole fractions (KN). For the reaction,
for examle,
kf
1
CO + 2 O2
kb

CO2

[CO2]
Kc = [CO][O ]1/2
2

(47)

yCO2
KN = y y 1/2
CO O2

(48)

and

and with (12)

23

Ni
Ni p
yi p
[ ]= V =
=
NT
T
Kc = KN (

T 1/2
)
p

pi
Or, with yi = p , equilibrium constant in terms of partial pressures (Kp):
pCO2
pCO pO2 1/2

(49)

Kc = Kp ( T)1/2

(50)

KN
Kp = p1/2

(51)

Kp =
and

Dissociation
The inversion (or decomposition) of the products molecules to the initial reactant molecules
due to the backward reactions is generally called dissociation. Dissociation may be
expressed by a simple way for relatively simple reactions. For example, the bidirectional
reaction (44)
kf
CO + H2 O
kb

H2 + CO2

may be cast in two reactions (forward and backward) such as:


1
CO + 2 O2

CO2

(52a)

1
CO + 2 O2
_____________________________________________
CO2

1
CO + 2 O2

(52b)

1
(1- ) CO2 + CO + 2 O2

(52)

Here, is the degree of dissociation or dissociation fraction and (34) expresses the two
reactions (50a and 52b) in form of one reaction.
The (backward) dissociation reactions are due to the collision of the product molecules with
each other or with an other body such as an other atom or an inert. This is expressed generally;
A + A + Mi

A2 + Mi

A2 + Mi

A + A + Mi

It is seen from (47) that a part () of the reactant CO is always present in the product side. The
mole fractions of the products may be given in term of
yCO2 =

1-
,

(1 + 2 )

yCO =

(1 + 2 )

Using these in 48 and 50 one obtain,

24

yCO2 =

/2

(1 + 2 )

Kp =

1 - 2 + -1/2 1 -1/2
( ) ( )

(53)

Equilibrium constants are only the function of temperature and pressure. This dependence is
given in thermodynamical tables for the reference pressure, p = 1 atm. For the simple case
above, can be calculated from (53) using Kp found from the tables for the given temperature
and pressure. For example for,
p = 1 atm.,

T = 2000 K,

equilibrium constant is Kp = 748,7 and the composition of the products is:


[CO] = 1.5 % ,

[CO2] = 96.7 % ,

[O2] = 1.8 %

As Kp is function of temperature, the composition of products changes with temperature.


Indeed, for:
p = 1 atm.,

T = 2500 K,

the composition of the products is:


[CO] = 12 % ,

[CO2] = 82.1 % ,

[O2] = 5.9 %

Because of dissociation there is always CO and O2 in the exhaust gases of the internal
combustion engines, even if the mixture is lean and complete combustion is possible.
The effect of dissociation is not only on exhaust gas composition but also on heat of reaction.
Because, some fuel (CO in this case) is not converted into the nominal products (CO and
H2O, for hydrocarbon fuel) the heat value of fuel decreases. This effect may be seen from Fig.
11.
R

(1 - ) Q p

Qp

(with dis s oc .)

(without dis s oc .)

T (K)

0
0

TR

TP

T P (withou t dissociation)

Figure 11. The effect of dissociation on heat of reaction and adiabatic combustion temperature.

The heat of reaction of the reactions with dissociation is lower than that of without
dissociation by Qp, which is the heat, absorbed by the endothermic dissociation reactions.
This corresponds to the sum of the heat of formations of the dissociation products and taken
from the mixture of products. The achieved temperature is then lower than that of reaction
without dissociation.
Whereas the dissociation of CO2 and H2O to the initial species CO, H2 and O2 has an increasing
effect on CO emission and decreasing effect on released heat, the dissociation of NO to N2
according to the reaction below has great effect on NO emission but practically no effect on
heat release.

25

3.5 Methods for calculating equilibrium constant


State of Equilibrium:
A system is sad to be in a state of equilibrium if there is not a spontaneous change in the
state of the system. Equilibrium may be stable or unstable. By stable equilibrium we
mean that, if we slightly disturb the state of the system, it will return to its original state.
T1 , A
T2 , B

Figure 12. The isolated thermodynamical system consisting of a system and its surroundings.

There is a direct relation between the entropy and state of equilibrium of a thermodynamic
system; from the second law of thermodynamics we know that the entropy of a system and the
surroundings (an isolated system) can only increase. Assuming that there is no change in
volumes and temperatures of system B and of surroundings A (Fig. 12), total change in
entropy, after a heat transfer of Q from surroundings to system B, will be:
dQ dQ
dQ
dS = dSA + dSB = T - T = T T (T1 T2)
2
1
2 1
Since T1 > T2 , by assumption, net change in entropy will be positive (dS > 0). But even if T2
> T1 ,
dS = dSA + dSB =

dQ
(T T1)
T2 T1 2

same result (dS > 0) is obtained.


Let us consider the limit case T2 = T1 = T. If we can make it possible to rise the temperature
of A by dT2 and to drop that of by dT1 by transfering the heat dQ from B to A , then,
assuming that the total heat capacity (mCp) of both system are equal, dT1 = dT2 = dT and
dQ
dQ
dT
dS = dSA + dSB = T + dT - T - dT = - 2 dQ T2
Same result would had been obtained when we had raised the T2 (d T2 > 0) and decreased T1
(d T1 < 0). This result indicates a decrease in entropy.
Since, according to the second law of thermodynamics, the entropy of an isolated system can
only increase, this results also implies that a change in temperature would not be possible (dT
= 0) and that maximum entropy is attained when both system B and surroundings A has the
same temperature (where dS = 0).
As the volume of the system is also fixed (as initial constraint) no other variations in state are
possible*. This may be commented that a perturbation of state (in this case change in
temperature by dT) would not bring the system to an unstable position. The system will return
to its original state immediately. This is the stable equilibrium state.
*

dQ dW = dE and

dW = pdV = 0

gives

dQ = TdS = dE

26

or

dE = dEA + dEB = 0

Generally:

If the properties of an isolated thermodynamical system change spontaneously, there will


be an increase in entropy of the system ([dS]E > 0) and the system is not in equilibrium
state and spontaneous changes may occur.

When the entropy of the isolated system is at a maximum ([dS]E = 0), the system is in
equilibrium and no spontaneous change in properties is expected.

If all possible variations in state of the system lead to a negative change in the system
entropy ([dS]E < 0), the system is in stable equilibrium.

In the last case, of course, we may say that the possible variation cannot take place due to the
2. law of thermodynamic, only after we have tested the system for its stability.
We have analysed the equilibrium conditions for a system and surroundings. Now the
equilibrium of
system B (a combustion system for example) immersed in a surroundings may be expressed as
follows:
The condition,
dSsystem + dSsurroundings 0
and the assumption there is a heat flow from surroundings, i.e. dQ < 0
dQ
dSsurroundings = - T
implies
dQ
dSsystem = T 0

(54)

Integration (54) and dQ dW = dE , the first law of thermodynamics, between initial (1) and
final (2) states gives:
T(S2 - S1) 0
Q W = E2 - E1
and with Q = TdS
T(S2 - S1) (E2 - E1) W

(55)

is obtained. With the definition of system property F = (E TS), the Helmoltz potential or
Helmoltz free energy function (55) may be expressed
(F1 F2) W

(56)

This means that the maximum work obtained above process (heat and work interactions
which may take place in the system B) is less than or equal to the decrease in Helmoltz free
energy function. Of course, this decrease in Helmoltz potential is associated with the increase
in entropy, so, instead of maximum entropy minimum Helmoltz free energy function can be
used as equilibrium condition.
The criterion for stable equilibrium at constant temperature and constant volume (dW = 0) is
then:
(dF)T < 0

or

F2 F1

(only for reversible processes F2 = F1)

27

If the system is at constant temperature and constant pressure but of variable volume, there
will be work done by the system to maintain constant pressure. The total work done by the
system can be cast in two parts. one part being the work of expansion or compression at
constant pressure:
W = W + p (V2 - V1) = Q (E2 - E1)
(56)
Similar to the case of constant volume process, the criterion for stable equilibrium at constant
temperature and constant pressure
T(S2 - S1) (E2 - E1) - p (V2 - V1) W

(57)

is obtained. With the definition of system property G = (U + pV TS) = H - TdS, the Gibbs
free energy function, (57) may be expressed
(G1 G2) W

(58)

The criteria for equilibrium at constant temperature and constant pressure are then:

(dG)p,T < 0), the system is not in equilibrium state and spontaneous changes may occur.

(dG)p,T = 0), the system is in equilibrium and no spontaneous change in properties can
occur.

(dG]p,T < 0), the system is in stable equilibrium.

Chemical potential
The relations obtained from the first law of thermodynamics dQ dE = pdV , between the
system properties are,
h = e + pV

(enthalpy)

(59)

g = h Ts

(Gibbs function)

(60)

f = e Ts

(Holmoltz function)

(61)

and
du = Tds pdV

(62)

dh = Tds Vdp

(63)

df = sdT pdV

(64)

dg = Vdp sdT

(65)

For a constant composition system consisting of N moles the properties will be a function of
not only other properties but also the number of the moles N1 , N2 , N3 , of the species i = 1,
2, 3, . Internal energy for example,
u = u (s,V, N1, N2 , N3,.)
du = [

u
u
u
u
]V,N1,N2,.. ds + [
]s,N1,N2,.. dV + [
]V,s,N2,.. dN1 + [
]
dN + ..
s
V
N1
N2 V,s,N1,.. 2

From (59)

28

u
]
=T,
s V,N1,N2,

u
]
=p
V s,N1,N2,..

so
k

du = Tds pdV +

[ Ni ]V,s,Nj dNi

(66)

i=1
Here
i = [

u
]
Ni V,s,Nj

is called chemical potential. Similarly, the chemical potential for Gibbs free energy
function is
i = [

g
]
Ni p,T,Nj

and
k

dg = Vdp sdT +

i dNi

(67)

i=1
On the other hand, with (30)
h = h(T) + h0
and with
dh
dp
ds = T - R
p
obtained from (63) using the state equation of gaseous systems, we obtain for the entropy
h dh
p
s = T - R ln p + s0

0
0
and with reference pressure p0 = 1 atm. ,
s = s(T) - R ln p + s0
Using these in (60)
g = (h0 - T s0) + h (T) T s(T) + RT ln p
and with
g0 = h0 - T s0,

g(T) = h(T) Ts(T),

g0 = g(T) + g0

we obtain
g = g0 + RT ln p

(68)

Calculation of equilibrium composition and equilibrium constants:


Now we can calculate the equilibrium composition of reaction (52) by applying the
equilibrium conditions, that is by maximising the Gibbs free energy function:

29

1
CO + 2 O2

1
CO2 + (1-) CO + 2 (1-) O2

To find the equilibrium value of = 1-, degree of reaction we must first find the
equilibrium constant Kp for this reaction. This constant will be found from the condition, that
Gibbs function of the mixture (of products) will have a maximum at equilibrium:
[

G
] =0
p,T

(69)
1
(1-) O2
2

G = CO2 CO2 + CO (1-) CO + O2

(70)

(Here , is the specific Gibbs energy )


The equilibrium value of can be found from (69) using (70). Or, for a more general mixture:
1
1
x [COCO + O2 2 O2] x CO2 CO2 + x (1-) CO CO + 2 (1-) O2 O2 + y O2 + z CO2
Here, y and z are excess O2 and CO2 added to the products, = 0 means no reaction and = 1
means no dissociation.
The changes in the number of the species of the product side are:
NCO2 = x CO2 + z

dNCO2 = x CO2 d

NCO = x (1-) CO

dNCO = - x CO d

1
NO2 = x 2 (1-) O2 + y

dNO2 = - x O2 d

or the equations of constraints of the chemical system is:


dNCO2 dNCO
dNO2
=== xd
CO2
CO
O2
On the other hand, for constant pressure (dp = 0) and temperature (dT = 0) process (67),
k

dg = Vdp sdT +

i dNi
i=1

becomes
k

dg =

i dNi
i=1

and with (71)


(dG)p,T = (CO2 CO2 - CO CO - O2 O2) xd
is obtained. As x 0,
[

G
] =0
p,T

gives
CO2 CO2 = CO CO + O2 O2
30

(71)

Putting = 0 + RT ln p (68) for the ideal gases we obtain:


1
0
0
0
CO2 ln pCO2 - CO ln pCO - O2 ln pO2 =
RT (CO2 CO2 - CO CO - O2 O2)

(72)

If we define the equilibrium constant with the right hand side of (72):
Kp =

1
(
0 - 0 - 0 )
RT CO2 CO2 CO CO O2 O2

then
lnKp = ln [

(pCO2) CO2
]
(pCO) CO (pO2) O2

(73)
For the stoichiometric reaction (52), CO2 = 1,

CO = 1,

O2 = and

pCO2
Kp = p p
CO O2

(74)

For a more general reaction given below:


a A + b B

c C + d D

at any stage of the reaction,


a A + b B

(1 - ) a A + (1 - ) b B + c C + d D

the number of the molecules


Na = (1 - ) a + C1
Nb = (1 - ) b + C2
Nc = c + C3
Nd = d + C4
0
- GT = a 0a + b 0b - c 0c - d 0d

(75)

and
c

lnKp = ln [

pc pd
]
pa a pb b

0
GT
= - RT

(76)
0
Since GT is a function only of temperature, the equilibrium constant is function only of
temperature (Kp = Kp (T)) too. (75) can be expressed in terms of mole fractions by
introducing pa = ya , pb = yb, pc = yc , pd = yd,
yc c yd d a +b -c -d
]p
= KN pi
ya a yb b

(77)

[ ] c [ ] d
]
[ ] a [ ] b

(78)

Kp = [
or
Kc = [

31

Kc = Kp (RT)a +b -c -d
(79)

= KN pi

Equilibrium constant may be written in a more general form:


Kp = [

(p )products
]
(p )reactants

(80)

3.6 A model for combustion of hydrocarbon fuels


Combustion of a hydrocarbon fuel with general formula CnHm may be modeled by 12 species
as components of the products. These twelve species are the selected species which are of
most significant importance, if < 3. Otherwise we must also consider some unburned fuel
molecules and radicals, such as C2H2, CH4, CH3 etc.
The combustion reaction written for one mole (28,9 g) air is:
CnHm + 0.21 O2 + 0.79 N2

1CO2 + 2H2O + 3N2 + 4O2 + 5CO + 6H2 +


+ 7H + 8O + 9OH + 10NO + 11N + 12NO2

(81)

Here is the equivalence ratio (i.e. 1/) and is a coefficient to be calculated to have a
stoichiometric mixture of air and fuel specified with n and m. If we write the reaction (1)
describing the complete combustion of a hydrocarbon fuel for 1 mole
m
m
m
m
)O2 + 3.76(n +
)N2  nCO2 +
H O + 3.76 (n +
)N2
4
4
2 2
4

CnHm + (n +
and for mole:
CnHm + (n +

m
m
m
m
)O2 + 3.76(n +
)N2  nCO2 +
H O + 3.76 (n +
)N2
4
4
2 2
4

(82)
A comparison (81) and (82) gives:
=

0.21
m
(n + 4 )

Conservation of the number of atoms of the elements yields the following equations:
For

n = N (y1 + y5)

(83)

For

m = N (2y2 + 2y6 + y7 + y9)

(84)

For

0.42 = N (2y1 + y2 + 2y4 + y5 + y8 + y9 + y10 + 2y12)

(85)

For

1.58 = N (2y3 + y10 + y11 + y12)

(86)

Here,
12
N=

i :

(total number of moles of products / 1 mole air)

i=1

32

i
yi = N :

(mole fractions of the species of the products)

and
12

yi = 1

(87)

i=1
Reactions to be considered for the formation of the selected 12 species and the corresponding
equilibrium constants in term of partial pressures are:
1
H 
2 2

K1 =

y7
p1/2
y61/2

(88)

1
2 O2 

y8
K2 = y 1/2 p1/2
4

(89)

1
1
2 O2 + 2 H2 

OH

y9
K3 = y 1/2 y 1/2
4
6

(90)

1
1
2 O2 + 2 N2 

NO

y10
K4 = y 1/2 y 1/2
4
3

(91)

1
H2 + 2 O2 

H2O

y2
K5 = y 1/2 y p -1/2
4
6

(92)

CO2

K6 =

y1
1/2

(93)

CO +

1
O 
2 2

1
2 N2 

1
2 N2 + O2 

NO2

y4

y5

p -1/2

y11
K7 = y 1/2 p1/2
3

(94)

y12
K8 = y y 1/2 p -1/2
4 8

(95)

With:
C1 = K1 / p1/2 ,
C5 = K5 / p

1/2

C2 = K2 / p1/2 ,
C6 = K6 / p

1/2

C3 = K3 ,

C7 = K7 / p

C4 = K4
1/2

C8 = K8 / p1/2

we obtain from (88) to (95),


y7 = C1 y61/2

(96),

y8 = C2 y41/2

(97)

y9 = C3 y41/2y61/2

(98),

y10 = C4 y31/2y41/2

(99)

y2 = C5 y41/2y6

(100),

y1 = C6 y41/2y5

(101)

y11 = C7 y31/2

(102),

y12 = C8 y4y31/2

(103)

from (83) and (84),


m
2y2 + 2y6 + y7 + y9 - n (y1 + y5) = 0

(104)

from (83) and (85),

33

2y1 + y2 + 2y4 + y5 + y8 + y9 + y10 + 2y12 -

0.42
(y + y5) = 0
n 1

(105)
and from (83) and (86),
2y3 + y10 + y11 + y12 -

1.58
(y + y5) = 0
n 1

(106)

From the 12 equations (96 to 106) and (87) we can obtain 12 unknown mole fractions (y1 to
y12). If we rearange the equations we get four equation for 4 unknown y3, y4, y5, y6:
f1(y3, y4, y5, y6) = 0

(107a)

f2(y3, y4, y5, y6) = 0

(107b)

f3(y3, y4, y5, y6) = 0

(107c)

f4(y3, y4, y5, y6) = 0

(107d)

We obtain other concentrations, y1, y2, y7, y8, y9, y10, y11, y12 from equations (88 to 95).
These four equations (107) are non-linear and can be solved by Newton-Raphson iteration.
For application the iteration procedure equations (107) must be expanded into Taylor series
about a known point.
Newton-Raphson iteration:
The Taylor expansion of a function f(x), about the point x, neglecting the second order
derivatives is:
df(x)
f(x + x) = f(x) + dx x
If x is the solution and x(1) is the first approximation about the solution x then f(x) = 0 and with
x(1) = x + x
df(x)
f(x(1)) = 0 + dx x
and the error is
f(x(1))
x(1) = df(x(1))
d x(1)
The second approximation is then x(2) = x(1) + x. The procedure is repeated until the error is
smaller than a convergence criterion (x<)*.
Taylor expansion of the equations (107) may be expressed using indices notation:
*

Example:

with the first guess

f(x) = x2 - 1 = 0
x(1) = 1.1,

f(x(1)) = 0.21,

df(x(1))
= 2.2,
d x(1)

x(1) =

f(x(2)) = 0.008,

df(x(2))
= 2.008,
d x(2)

x(2) =

0.0954
x(2) = 1.1 0.0954 = 1.004,
0.008
= 0.004
2.008
x(3) = 1.004 0.004 = 1

34

0.21
=
2.2

fJ +

fJ
fJ
fJ
fJ
y3 +
y4 +
y5 +
y6 = 0
y3
y4
y5
y6

(J = I, II, II, IV)

(108)

or in matrix form:
fI
y3

fI
y4

fI
y5

f
y6

y3

fII
y3

fII
y4

fII
y5

fII
y6

y4

fIII
y3

fIII
y4

fIII
y5

fIII
y6

y5

fIII

fIV
y3

fIV
y4

fIV
y5

fIV
y6

y6

fIV

fI
=

fII

Beginning with the first approximations y3(1), y4(1), y5(1), y6(1) the errors yi (1) (i=3,4,5,6) are
obtained from the solution of the equations matrix until the convergence criterion
|yi(n-1) |
is satisfied and yi(n) is found after n iterations. Here the absolute error |yi| is taken as
yi(n-1)
convergence criterion. If relative error ( = y (n-1) ) were used as convergence criterion the
i
condition << 1 could not be provided for all species. Because of the truncation errors yi(n-1)
can be much more smaller than yi(n-1) itself.

4. Non-equilibrium processes
We have seen that the equilibrium constant Kp is function of Gibbs free energy for a constant
pressure and temperature process. Gibbs free energy, on the other hand, is function of
temperature and so the Kp . That means, that the composition of a reacting mixture in
equilibrium state is only the function of the temperature. The composition of the combustion
product will depend on the final temperature, namely the adiabatic combustion temperature.
Since the adiabatic flame temperature itself depends on the composition, it can be determined
only by iteration. However, both dependence of the equilibrium composition on temperature
and the dependence of final temperature on final product composition are only true if there is
enough time to reach the equilibrium state at a given temperature or to reach the adiabatic
flame temperature at the last stage of combustion. The equilibrium constant is the ratio of
forward and backward reaction coefficients and the reaction rate is determined by these rate
coefficients. For example, the rate of formation of (A), as a result of the dissociation and
recombination process below, is:
A2 + Mi
A + A + Mi

kf

kb

A + A + Mi
A2 + Mi

d[A]
2
dt = [A2] kf [Mi] - [A] kb [Mi]

35

(109)

One of the rate coefficients is determined experimentally and the other is calculated from Kc =
kf / kb. Whether the concentration of A, [A], is the equilibrium value at any stage of the
reaction depends only on the velocity of formation of (A), i.e. whether tis reaction is fast
enough. According to the discussion above three kinds of processes are possible:

Processes that reach the equilibrium state (if relaxation time is relatively short enough) at
temperature and pressure fixed.

Processes at constant pressure or volume, that reach the final temperature and accordingly
equilibrium composition (if there is enough time).

Processes that are frozen at any stage and the compositions that do not correspond to the
equilibrium composition at the current temperature and pressure.

The last case is encountered in the supersonic (or hypersonic) nozzle flows with chemical
reactions, such as rocket nozzles. In rocket nozzles the flow is sonic at throat, becomes
supersonic in the divergent section of the nozzle and reaches very high velocities of over 1000
m/s (Fig 13).
AThroat

A Nozzle

Figure 13. Rocket chamber and nozzle

Under these conditions, the flow from the combustion chamber may sustain its high
dissociation degree further in the nozzle and at the exit. This, of course, causes the
temperature and the pressure to drop and reduce the performance of the rocket or ramjet (Fig
14.).
De g r e e o f
Dis s o ciatio n
1,00
0,90
0,80
Full f roz en f low
0,70
0,60
0,50
0,40
Partially f roz en f low

0,30
0,20

Equilibrium

0,10
0,00
1

10

100

1000
10000
T h r o at Ar e a / No z z le Ar e a

Figure 14. Dissociation degree for hypersonic flow through a rocket nozzle

36

Another example for the rate-controlled reaction with final composition, which does not
correspond to equilibrium composition, is the NOx formation in the IC engines.
During the combustion in the cylinder of IC engine reaction between N2 and O2 molecules of
combustion air takes place at high temperature. These reactions result in formation of NOx
(NO, NO2). If reactions were in equilibrium nitric oxide concentration would be insignificant
at the temperatures at the end of expansion stroke. However, nitric oxide concentration
exceeds the equilibrium value. This is because of the relatively slow decomposition of NOx to
N2 and O2 during the expansion. The frozen value of the NOx concentration can be calculated
assuming equilibrium concentrations for other substances. The method is given below for a
general dissociation reaction.
Partially equilibrium assumption:
Let as consider the reaction,
A+B

kf

kb

C+D

Here the rates of formation of the species are:


d[A]
dt = - kf [A] [B] + kb [C] [D]
d[B]
= - kf [A] [B] + kb [C] [D]
dt
d[C]
dt = - kf [A] [B] - kb [C] [D]
d[D]
= - kf [A] [B] - kb [C] [D]
dt
At equilibrium
d[A]
d[B]
d[C]
d[D]
=
=
=
dt
dt
dt
dt = 0
or
kf [A]e [B]e - kb [C]e [D]e = Rf
(110)
Here, subscribt e denotes the equilibrium stage of reaction and Rf is the one way equilibrium
rate as defined before (46). If the concentrations at any stage are [A] , [B] , [C] ,[D] , then
with
[A]
= [A] ,
e

[B]
= [B] ,
e

[C]
= [C] ,
e

d[A]
dt = - kf [A]e [B]e + kb [C]e [D]e
or with (110),
d[A]
= (- + )Rf
dt

37

[D]
= [D]
e

or more generally:
d[ ]
= ( )Rf
dt

(111)

We may apply the model above to the nitric oxide formation model given below:
NO + N

kf

kb

N2 + O

The rate of heat releasing reactions in a combustion chamber are sufficient fast, so that
burned gas may be assumed to have their equilibrium state. Then the chamber pressure and
temperature may be calculated from the equilibrium data. As combustion products, N2 and O
(besides CO, CO2, H2O, H2, O2 and OH) are very close to their equilibrium value. Then we
may assume:
[NO]
[N]
= [NO] , = [N] ,
e
e

[N2]
= 1,
[N2]e

[O]
= [O] = 1
e

and calculate the rate of formation of NO from


d[NO]
dt = (- + 1)Rf
with
Rf = kf [N]e [NO]e - kb [N2]e [O]e
The NOx concentration in the exhaust gas of an internal combustion engine calculated both for
equilibrium and rate controlled mechanism is shown in Fig. 15.
5000

Engine Speed = 3000 rpm

Nitr ic Oxid e (p p m )

4000
Rate c ontrolled
3000

2000

Ignition

1000
Equilibrium

0
-20

20

40

60

80

100

120

140

160

Cr ank An gle afte r T DC (d e g r e e )

Figure 15. NOx concentration in the exhaust gas of an internal combustion engine

It is possible to set up more complex models for the formation of nitric oxides in exhaust
gases, such as:

38

NO + N

N2 + O

N2 + O

N2 + O

N + OH

NO + H

H + N2O

N2 + OH

O + N2O

N2 + O2

O + N2O

NO + NO

N2O + M

N2 + O + M

Or Zeldowich model
NO + N

N2 + O

N + O2

NO + O

N + OH

NO + H

5. Calculation of adiabatic flame temperature


Heat of reaction was given by
Qp(T) = hP(T) hR(T) + (h0P - h0R) = hP hR

(j/mole-products)

and for a exothermic reaction hP < hR and Qp >0 (Fig. 3).


We may write the heat of reaction in terms of the enthalpy of the species:
n

Qp(T) =

[i hi (T)]P -

i=1

[i hi (T)]R +

i=1

[ihi0]P -

i=1

[hi0]R

i=1

(112)
or with
h0 =

[ihi0]P -

i=1

[hi0]R

i=1

and with mole fraction yi = i /N (N: total number of moles of reactant or product side),
n

Nh0 = Qp(T) +

[yi hi (T)]R -

i=1

[yi hi (T)]P

(113)

i=1

When the adiabatic combustion temperature is reached then Tp = Tad and Qp = 0, since, hP =
hR or
n

[yi hi (Tad)]P -

i=1

[yi hi (TR)]R + Nh0 = 0

(114)

i=1

at this point.
If the difference of the heat of formations h0 is not known, but the heat of reaction Qp(Ts) at
the temperature Ts is known the equation (113) may be written;
39

{yi [hi (Tad) - hi (Ts)]}P - {yi [hi (TR) - hi (Ts)]}R + Qp(Ts) = 0


i=1
i=1
(115)
If there is no dissociation the equation above can be expressed in the simple form
f (T) B = 0
Since f(T) = h(T), Tad can be calculated from,
a0 T + a1 T2 + a2 T3 + a3 T4 + a4 T5 = 0

(116)

by Newon-Raphson iteration:
df(T)
f(T) + dT T = 0

f(T)
T = df(T)
dT

T(new) = T(old) + T

In case of the presence of dissociation, not only the enthalpy but also the mole fractions of the
species will change with temperature. The derivative of (113) with respect to temperature is,
df(T)
d
=
dT
dT

[yi hi (T)]P =
i=1

dhi(T)
dyi
[yi dT + hi (T) dt ]P

(117)

i=1

To be able to calculate (117) derivatives of the concentrations with respect to temperature (dyI
/ dT) must be obtained first. Derivatives of y3, y4, y5, y6 are calculated from the derivative of
(108). Derivatives of the other concentrations are obtained from the other relations described
in part (3.6):
dfJ
fJ dy3
fJ dy4
fJ dy5
fJ dy6
dT = y3 dT + y4 dT + y5 dT + y6 dT = 0

(J = I, II, II, IV)

or in matrix form:
fI
y3

fI
y4

fI
y5

f
y6

dy3/dT

dfI /dT

fII
y3

fII
y4

fII
y5

fII
y6

dy4/dT =

dfII /dT

fIII
y3

fIII
y4

fIII
y5

fIII
y6

dy5/dT

dfIII /dT

fIV
y3

fIV
y4

fIV
y5

fIV
y6

dy6/dT

dfIV /dT

5.1. Calculation of the specific enthalpy and specific heats


Specific enthalpy and specific heats of the species (i) are generally expressed with such
polinoms given below:

40

hi (T)
A2i T
A3i T2
= A1i + 2
+ 3
T

A4i T3
4

A5i T4
A6
+
5
T

(118)

or
hi (T)
A2i T2
= A1i T + 2

A3i T3
3

A4i T4
4

Specific enthalpy is calculated then by hi (T) = [

A5i T5
+ A6
5

(119)

hi (T)
] T in j/mole.
T

Here = 8.314 (kJ/kmole) is the universal gas constant.


The polinom for the specific heat at constant pressure (in form
Cpi (T)
= A1i + A2i T + A3i T2 + A4i T3 + A5i T4

d
hi (T)
Cpi (T)
(
)=
) is:
dT

(120)

The sixth coefficient A6i of the enthalpy polinom is the formation enthalpy of the species and
is zero for the elements as mentioned before.
For example this coefficient is for water:
A6H2O = 0.29 105 (K)
and the formation enthalpy of water at absolute zero is
h0H2O = 8.314 x 0.29 105 = 2.4 105 kJ/kmole
This is the heat content of 2 kg hydrogen as fuel. The heat value of hydrogen is then 1.2 105
kJ/kg-H2.

6. Combustion Kinetics and Sensitivity Analyses


We have seen that the chemical reactions occur with a finite rate, in two directions (forward
and backward), between initial end final species. Since all chemical reactions occur as result
of the collisions between the atoms or the molecules in the mixture, many kinds of species
may be born as a result of the many possible reactions. That means that the final species are
obtained no directly from the reactants but through many intermediate (reaction) steps. The
mathematical model of such a multiple component reaction mechanism consist of nonlinear,
first order, ordinary differential equations (ODE).
In this chapter we will deal with:

Combustion kinetics problem encountered in homogenous, static (e.g. system without


heat, energy and matter transport) and gas phase systems. Chemical kinetic means here
dealing with the time depended concentrations and system properties, such as pressure
and temperature due to the finite rate reactions.

Method of sensitivity analyses, to determine what effect the changes (and


uncertainties) of input parameters has on solution. Such an analyses will provide us to
have better understanding of the mechanism and the parameters on which we muss
focus.

6.1. Combustion Kinetics Equations

41

Combustion kinetics equations express the rate of production or consumption of the chemical
species in terms of temperature (T), mass density () and the concentrations of the other
species, which take place in reaction.
d[i]
= fi ([k], T, )
dt

(120)

i, k = 1, .NS

Here NS is the total number of the chemical species in reaction and i is the mass
concentration (moles/g-mixture). Right hand side of (120), the net rate of formation species (i)
is then:
fi =

NR

(ij - ij ) (R j R j )

j=1

m3 moles
g m3 s

moles
)
gs

(121)

Here :
NR

: Total number of the distinct chemical reaction j .

ij , ij

: Stoichiometric coefficient of species (i) in reaction (j) in reactant and product


sides respectively.

R j , R j

: Molar forward and reverse rates (for species i ) in reaction (j) per unit volume
(moles / m3 s).
NS

R j = kj ( k) ij

j = 1, .NR

i=1

NS

R -j = k -j ( k) ij

j = 1, .NR

i=1

k j , k j

: Rate coefficients written in general form involving the temperature


dependence too.
k j = A j Tnj exp (-E/RT)
k j = k j / Kcj

(Kcj : equilibrium constant)

From NR reactions (120) we may calculate the temporal evaluation of temperature and
concentrations.
However, to calculate the temperature we must make use of the equation of the energy
conservation. This equation is, for example, given for an adiabatic, constant pressure process
as follows:
NS

i hi = constant

moles
J
J
=
)
g
moles
g

i=1
The temperature at any stage of the reaction is :

42

(122)

NS

i hi

i=1
T=
NS

mole J
mole K g
= K)
g mole
J
mole

(123)

Cpi i

i=1
and the time derivative of the temperature is (with

d[i]
= fi ) :
dt

NS

fi hi

dT
i=1
=
dt
NS

K
)
s

(124)

Cpi i

i=1
The differential equations both for concentrations (120) and for temperature (124) are first
order (only first derivatives are present), ordinary (no partial derivatives) and non-linear (
because of the concentration and temperature dependence of the right hand sides).
For a non-adiabatic system a term that models the heat transfer must be introduced to the
equation (124). If also the density is variable (independent of p and T) an additional ODE or
an algebraic equation must be added to the system of equations. For a constant pressure,
adiabatic process T may be calculated either from (123) by employing an iterative technique
or by adjustment of T until (122) is satisfied. It is of course possible to obtain T by solving
ODE (124) too.
6.2. Solving the OD equations
The governing ODEs can be written in vector form,

. dyi
yi = dt = f (y)

(125)

for the time interval,


t0 t tfinal
and the given initial conditions
yi = yi0

at

t = t0

Here yi is an N (= NS +2) dimensional vector consisting:


yi = i

i = 1,. NS

yNS+1 = T
yNS+2 =
Since, as mentioned before, in chemical reactions (even in the very simple ones) there may be
many intermediate steps associated with numerous chemical species, even if in very low
concentration. Such systems must be modeled by somehow complicated reaction mechanism,
which requires validation by theoretical analyses. This analysis, on the other hand, requires
repetitive and reliable solution techniques for the set of ODEs. For reacting flow problems

43

this technique must also be very fast and economic, because the rate equations are to be
integrated at several grid points of the solution field in such problems.
The stiff differential equations
The change of species concentrations and temperature are illustrated In Fig. 16 depending on
the reaction time in an ignition and subsequent combustion process at constant pressure of 10
atm. and for the initial temperature of 1000 K. Reactants are CO, H2 and air mixture and
mechanism consist of 12 reactants of 11 species.

Figure 16.Speicies mole fractions and temperature profiles for constant pressure combustion of CO-H2- Air
mixture (p= 10 atm. Tinitial = 1000 K)

Three distinct phases, namely, induction, heat release and equilibration are apparent. Reaction
starts by adding some heat to or passing a shock wave through the system to raise the
temperature to a level enough for initiating the reaction.
During the induction period concentrations of the chemically active species (O,H,OH)
increase very rapidly to a level sufficiently high to initiate oxidation reactions of H2, CO and
C. Heat release period begins when induction period ends.
During heat release temperature and concentrations of the main products increase very rapidly.
Temperature reaches 89-90 % of its maximum value at the end of the heat release phase.
Active intermediate species reach their maximums.
The last phase is the equilibration period during which all variables approach asymptotically
toward their equilibrium values.
The set of ODEs may be solved by classic numerical methods such as explicit fourth order
Runge-Kutta method. In general, numerical methods replace the ODEs with difference
equations
y
yn - yn-1
. dy
y = dt
= t -t
t
n
n-1
and solution is obtained by step-by-step integration.

yn = yn-1 + y ( tn - tn-1)
(n : step number)

44

Main steps of the solution procedure are as follows:


1- Solution starts with the initial conditions at the time t0
y0i = yi (t0)

i = 1, . N

2- The approximate solution to ODE is generated at discrete points in time (tn , n=1, ).
Since the differential equations are not linear iteration is necessary. The approximate solution
approach to exact solution as the iteration procedure proceeds.
3- Interval between two discrete points in time (hn = tn - tn-1) is called step size or step
length.
The step length must vary generally from one step to another to obtain both an economical
and an accurate solution. The step size used in solving the problem given before by explicit
fourth-order Runge-Kutta method is given in Fig. 17. The step size must be small during
induction and heat release period to obtain an accurate solution because of the very steep
increase in concentrations and temperature in these phases. It is expected that the step size can
be chosen larger during the late heat release and equilibration phases. However, the step size
must still be kept as small as in the regimes with sharp changes, to obtain an accurate and
stable solution. This requirements cause to increase the total number of steps and the CPU
time of the computer. For this reasons faster and more economical special techniques are
developed for the solution of ODEs arising in chemical kinetics.

Figure 17. The dependence of step length on the stages of combustion

Curtiss and Hirschfelder (1952) used the term stiff to describe some properties of ODEs in
chemical kinetics. Although stiffness does not have a simple definition in general:

Stiff equations are characterised by time constant of widely varying magnitude.

Stiff problems force the classical ODEs solution methods to use unacceptable small step
size to maintain numerical stability, e.g., an asymptotic approach to a solution. However,
special programs developed for stiff problems, such as LSODE, need considerably less
number of steps and shorter CPU time using variable step size. Number of step is 90 and
CPU time is 0.35 s in LSODE instead of 10.000 and 60 s as in Runge-Kutta method.

Another requirement for a problem to be stiff is that ODEs are stable in the sens that
different solutions approach to each other for increasing (t).

45

In the early heat release phase the concentrations and the temperature increase
exponentially and the ODEs are unstable. In this phase small time steps must be used to
get stable solution. During late heat release and equilibration phase however, the equations
are stable, these regimes are stiff and larger time steps can be used.

Besides being stable and economic, the numerical method must also be accurate, so that the
unstable modes can also be tracked. Accuracy of a numerical method is related with the
truncation and discretion errors introduced in a single step.
Solution Methods for Stiff ODEs
The solution methods for ODEs can generally be cast into two types:
1- Single step methods:
These methods approximate the solution at the current time (tn) in a single time step using the
information of the previous point in time (tn-1). The simplest method is the Euler method
(Euler straight forward integration).
The ordinary, first order and non-linear differential equation
dy
dt = f (y, t)
is written in finite difference form
y
yn - yn-1
= f (yn-1, tn-1) =
tn - tn-1
t
and integrated numerically
yn = yn-1 + h f ( yn-1, tn-1)

yn

starting from the initial conditions


y = y0

at

yn-1

t = t0

tn-1

tn
h = t

Estimation of the total error


If yn is the numerical approximation and n the exact analytical solution the total error is,
n = yn - n
Expanding n in Taylor series gives:
d
h2 d2
(tn) = (tn-1) + h ( dt )n-1 + 2 ( dt2 )n-1 + .
If yn-1 = n-1 the total error in y consists of the three component:
1- the discretion error, due to the first order approximation,
h2 d2
Ed,n = 2 ( dt2 )n-1
2- the truncation error Et,n
yn = yn-1 + h f ( yn-1, tn-1) + Et,n
3- the conducted error.
Then the sum of the error at nth step is

46

n = n-1 + h [f ( yn-1, tn-1) - f ( n-1, tn-1)] + Et,n + Ed,n


or
df
n = n-1 [1 + h ( dy )n-1] + Et,n + Ed,n
Here the first term is the conducted error from the previous step
Error
Dis c retion error

Trunc ation error

0
0

Figure 18. The dependence of errors on step size

As (h) decreases the discretion error decreases also. However, sum of the truncation error may
increase in this case, as a result of the increased number of steps (Fig 18).
Minimising the error by extrapolation of the solution with respect to h is possible. In case of
the error is proportional to h linearly, by extrapolating of two solutions for two different step
lengths (h1 and h2), a more accurate result can be obtained for h = 0 (Fig. 19).

y
y1
y2
y
h2

h1

Figure 19. Extrapolation of the solutions

The single step methods are approximation of 1. order and there is no guaranty that the
solution will become more accurate as step size decreases. Methods of higher order must be
used to obtain more accurate solutions by using shorter time steps.
Linear multi step methods
General solution formula is given by,
yn =

K1

K2

k=1

k=0

.
nk yn-k + hn nk yn-k
.

The method is called linear because yn and yn-k occurs linearly.

The method is multi step for K1 >1 or K2 > 1.

47

nk and nk are coefficients selected appropriately to account the weight of each step in
solution.

For example, if K1 = 1 and K2 q 1


yn = n1 yn-1 + hn

q-1

.
nk yn-k

k=0

is the implicit Adam methods of order q.


-

The method is implicit, because it uses the as yet not known yn to compute yn and
therefore requires iteration for the solution.
.

For q = 1

(k = 0  0)

yn = n1 yn-1 + hn n0 yn

For q = 2

(k = 0  1)

yn = n1 yn-1 + hn (n0 yn + n-1 yn-1)

Here, if n0 = 0, the method becomes explicit because we do not need to know yn but we use
.
the known yn-1.
Backward differential formula
For K1 = q and K2 = 0 we obtain Backward differential formula:
q

yn =

.
nk yn-k + hn n0 yn
k=0

Implicit method and backward differentiation formula is widely used to solve ODEs. This
method needs larger time step but provides more accurate results.
Explicit methods are only used for the prediction of the unknown value of yn which then will
yn - yn-1
.
be used for the approximation of yn = t - t . Implicit calculation correct this initial guess
n
n-1
by iteration and provides better approximation.
The coefficients nk and nk are constant if time step length (hn) is constant. If step size is
changed at some or every step the coefficients must be calculated at these steps.
The solution is gained by iterative methods, for example based on Newton-Raphson
technique, from the iteration matrix (G) of equations of NxN dimension for each reaction.
G [yn

(m+1)

(m)

+ yn ] =

nk yn-k + hn n0 f ( yn(m), tn)


k=1

Here m is the number of iteration.


Iteration techniques to solve ODEs require long computer time. After the convergence is
achieved, an estimate for the local truncation error is made. If this is not of acceptable value,
then both step size and order of the solution formula is varied for minimum computer time
and prescribed accuracy.
Another alternative for the solution methods is the single step implicit Runge-Kutta method.
This method is very suitable for chemical kinetics problems because of its stability and
accuracy. RK method is single step but works with intermediate stages within a step. General
solution formula is:

48

yn = yn-1 + hn

bk Kk
k=1
r

Here

Kk = f (yn-1+

akj Kj , tn-1+ cihn).


j=1

For example:
yn = yn-1 + hnf (y*n- 1 + tn- 1 )
2

y*n- 1 = yn-1 +

where

hn
f (yn-1, tn-1).
2

6.2. Sensitivity Analyses


It is often necessary to know what effect the variations of input parameters, such as initial
conditions or rate coefficients, have on solution. Rate coefficients are determined
experimentally and the data to estimate the parameters A, n, E are not sufficient to know
these coefficients with great precision.
With sensitivity analyses it is possible to determine the effect of uncertainties in the rate
coefficients or errors in initial conditions on the model prediction. Thus, sensitivity analyses
helps to identify the parameters that have to be determined accurately, because they have great
effect on the solution. Reaction rates that have negligible effect on the result need not to be
determined with great precision. Especially in determining of a reaction mechanism such as
analyses is very useful to eliminate the intermediate steps that do not affect the result.
Perhaps, the most significant utility of sensitivity analyses is the increased understanding it
provides of a complex reaction mechanism.
Two methods, local and global, are used for application of sensitivity analyses:
-

In local methods the effect of small changes in the input parameters or in initial conditions
dyi
about their nominal value are investigated. To do this [
], the first order sensitivity
dj
coefficient is evaluated for species concentration yi ,temperature T and density about
the nominal values of rate coefficient parameters (A, n, E) and the initial value of a
species. The sensitivity matrix is then:
Sij = (

dyi
)
dj

j = 1,.M

Here yi is the N dimensional (i = 1,N) vector of dependent variables (concentrations,


temperature and density) and M is the total number of independent parameters (parameters of
rate coefficients, initial conditions)
To obtain ODEs for the first order sensitivity analyses we may write (125) in terms of the
concentrations and the parameters of the rate coefficients.
dyi
dt = f (y, n, E)

(126)

For an isothermal reaction the rate coefficients will be time invariant and (126) may be
written:

49

dyi
dt = f (y, k)

(127)

Differentiation of (126) with respect to j gives derivatives of Sij with respect to time.
dSij
d dyi
d
df dyi
df
=
(
)
=
f
(y,

(
)
+
j) =
dt
dyi dj
dj dt
dj
dj

(128)

First term of the right hand side of (128) gives the implicit dependence of Sij on j through yi
and the second term the explicit dependence on j.
dSij
df
= Jmn Sij +
dt
dj

(129)

dfm
Here Jmn = dy (m,n = 1,N) is the jacobien matrix of NxN.
n
The M vector equations of (129) are independent of each other. In addition, they are linear,
even if (126) or (127) are highly non-linear.
Local method can also be used to examine effect of simultaneous variation of two parameters
dyi
by using the second order sensitivity coefficient
, if changes are very small.
dj dj
However, global methods can account for simultaneous parameter changes of arbitrary
magnitude. Global methods provide an average effect of all uncertainties examined.

7. Reduction of Chemical Reactions


In a complex reaction mechanism consisting of the time dependent species formation or
consumption it is always to select some reactions, which can be assumed in steady state
relative to the other reactions. This means the concentration of the species resulting from the
steady state reaction do not change with time. The assumed steady state reactions can then be
expressed in form of algebraic equations instead of the time dependent ones. Some times, of
course, these algebraic expressions may involve other steady state species. In this case these
equations are non-linear (of higher order) and the solution is not unique; the right one
(physically possible) must be selected out.
A way to overcome such difficulties is the truncation of some steady state reaction, e.g.,
reducing the reaction mechanism to a simpler one.
7.1. Justification of steady state approximation
If an intermediate species consumed with a rate much faster than its production, then we may
assume that its concentration will always be much smaller than that of the initial reactants and
final products. Since its concentration is small and stays small, its time derivatives (equation
120) also stay small compared to the time derivatives of the other species.
d[i]
dt = fi ([k], T, )

i, k = 1, .NS

The steady state approximation is expressed mathematically by equating the rate of


production of the species (121) to zero implying that the concentration of the species does not
change.

50

1
fi =

NR

(ij - ij ) (R j R j ) = 0

j=1

For example, the Zeldowich mechanism for the formation of NO

is

O + N2  N + NO

(130a)

N + O2  O + NO

(130b)

If the concentration of O is determined (as equilibrium value for example) by the oxidation
reactions in a combustion system and the reaction (130b) is faster than (130a), then N is in
steady state (its concentration does not change) because it is consumed by (130b) much faster
than produced by (130a), so we can add these two reaction cancelling N to obtain the global
one step reaction:
N2 + O2 2NO

(131)

In this case O are also cancels, but this is fortuitous.


The rate of formation of NO is determined by (130a), because (130b) is very fast. That means
we may conclude that the time change of NO is
d[NO]
= 2 kf [O][N2]
dt

(132)

even if the assumed reduced mechanism is (131).


This result may also be obtained in a more systematic way. The rates of formation of NO and
consumption of N written according to (130a) and (130b) are:
d[NO]
= kf [O][N2] + kb [N][ O2]
dt

(133)

d[N]
= kf [O][N2] - kb [N][O2]
dt

(134)

With the non-dimensional values,


[NO]
[NO] = [NO] ,
ref

[N]
[N] = [N] and
ref

t kf [O]ref [N]ref
for the time,
[NO]ref

and if we assume that the temperature and the concentrations of O2, O and N2 have their
reference values, then [N] ref must be
[N]ref =

kb [O]ref [N2]ref
kf [O2]ref

(135)

in order to obtain the non-dimensional equations:


d[NO]'
= 1 + [N]
d

(136)

d[N]'
= 1 - [N]
d

(137)

[N]ref
kf [O]ref [N2]ref
= [NO] = k [O ] [NO]
ref
b
2 ref
ref
Here is a small parameter because of:

51

1. [N] is a intermediate species and very small compared to the product [NO],
2. kb >> kf
The solutions of (136) and (137) are:
[N] = 1 exp (- /)

(138)

[NO] = 2 [exp (- /) 1]

(139)

The solutions show that there are two kind of time scales, and / .
If  0, that means, if [NO] >> [N] then:
[N] = 1

and

[NO] = 2

and from (135),


[NO] = 2 t kf [O]ref[N2]ref

(140)

(140) is equivalent to (132).


Solution of (138), (139) is plotted in Fig 20. It is seen that results depend on . This method
for the analyzing of the effect of the steady state assumption may be complicated in case of
the system with intermediate reactions (decision to select ). Instead, it is more convenient to
analyze the numerical solution of the complete system and compare the magnitudes of the
concentrations of the intermediates to the initial species. Practically it is acceptable to assume
an intermediate is in steady state if its concentration is less than 10% of the initial or final
concentration.

Figure 20. Non-dimensional concentrations of N and NO as a function of non-dimensional time.

7.2. A reduced mechanisms for Hydrogen-Oxygen Combustion


To obtain a reduced mechanism for a full mechanism of a reacting system, generally, the first
step is to define a suitable starting mechanism from a more comprehensive full
mechanism. Starting mechanism should contain only those elementary reactions that are
necessary to reproduce a characteristic quality such as the burning velocity, ignition delay etc.,
within about less than 5 % accuracy. Numerical solutions obtained from the full mechanism
52

and a sensitivity analyses helps to identify the influence of each individual reaction. For
hydrocarbon flames typically about fifty elementary reactions are necessary to reproduce the
burning velocity over the whole range (both lean and reach side) of equivalence ratio and
pressure up to 50 atm with reasonable accuracy. For lean-to-stoichiometric mixtures flame the
skeletal (starting) mechanism of 25 elementary reactions given in the Table 2 is sufficient.
However, because these reactions only contain hydrocarbons of the C1-chain it is insufficient
to reproduce reach-mixture where higher hydrocarbons of chain more than C1 expected. In this
table the reaction rate coefficients are given in the form
kj = AT exp (-E/RT)

(141)

with the units, moles/cm3 s for kj, cal/mole for E.


Let us use the first 8 reactions in the table to model the Hydrogen-Oxygen combustion
mechanism.

Tablel 2 Methane-Air reaction mechanism

53

1)

H + O2  O + OH

2)

O + H2  OH + H

3)

H2 + OH  H2O + H

4)

OH + OH  H2O + O

5)

H + O2 + M  HO2 + M

6)

H + HO2  OH + OH

7)

H + HO2  O2 + H2

8)

OH + HO2  H2O + O2

(142)

The balance (conservation) equation of the species considered are:

54

d[H]
dt

= - k1[H][O2] + k2[O][H2] + k3[OH][H2] k5[H][O2][M] k6[HO2][H] k7[H][HO2]

d[OH]
= k1[H][O2] + k2[O][H2] k3[OH][H2] 2k4[OH][OH] + 2k6[H][HO2] k8[HO2] [OH]
dt
d[O]
dt

= k1[H][O2] - k2[O][H2] + k4[OH][OH]

d[H2]
dt

= - k2[O][H2] k3[OH][H2] + k7[H][HO2]

d[O2]
dt

= - k1[H][O2] k5[H][O2][M] + k7[H][HO2] + k8 [HO2] [OH]

(143)

d[H2O]
= k3[OH][H2] + k4[OH][OH] + k8 [HO2] [OH]
dt
d[HO2]
= k5[H][O2][M] - k6[H][HO2] - k7[H][HO2] - k8 [HO2] [OH]
dt

At the left hand side there may exist not only time derivatives but also convective and
diffusive terms, if the system is non-homogenous.
d[ ]
We know that the steady state assumption ( dt = 0) for species i leads to an algebraic
equation between reaction rates. So we may each of these equations to eliminate the rates in
the remaining balance equations for the non-steady state species.
Since the stochiometrie of the resulting balance equation defines the global (reduced)
mechanism between the non-steady state species. Therefore, the global mechanism (if this
should reflect the real mechanism) depends on the choice of the species that are to be
eliminated. However, the general rule is to chose the species, which are consumed relatively
faster than others.
Let us chose the species OH, H and HO2 of which rate are to be set zero:
d[H]
dt

=0

d[OH]
=0
dt
d[HO2]
=0
dt

After choosing the steady state elements the consumption rates (the relatively higher ones) of
which must be set zero, k2[O][H2] = 0, k3[OH][H2] = 0, k7[H][HO2] =0. As these rates will no
longer appear on the right hand side of the of the balance equations (143), recombination of
these equation will be:

55

d[H]
d[OH]
d[O] d[HO2]
+{
+2
}
dt
dt
dt
dt

= 2k1[H][O2] 2k5[H][O2][M] + 2k6[HO2][H]

d[H2]
d[OH]
d[O] d[HO2]
dt + {- dt - 2 dt + dt } = - 3k1[H][O2] + k5[H][O2][M] - 3k6[HO2][H]
d[O2]
d[HO2]
+{
}
dt
dt

= - k1[H][O2] k6[HO2][H]

d[OH]
d[H2O]
d[O]
+
{
+
2
dt
dt
dt }

= 2k1[H][O2] + 2k6[HO2][H]

Since the rates in the braces are chosen to be zero for the assumed steady state species O, OH
and HO2 we get equation for four species:
d[H]
dt

= 2 (k1[H][O2] k5[H][O2][M] + k6[HO2][H] )

d[H2]
= - 3(k1[H][O2] + k5[H][O2][M] - k6[HO2][H]
dt
d[O2]
dt

) + k5[H][O2][M]

(144)

= - (k1[H][O2] + k6[HO2][H])

d[H2O]
= 2(k1[H][O2] + k6[HO2][H])
dt

The stoichiometry of these four balance equations corresponds to the global (reduced)
mechanism:
I)

3H2 + O2

2H + H2O

II)

2H + M

H2 + M

With the new rates:


WI = k1[H][O2] + k6[HO2][H]

(145)

WII = k5[H][O2][M]
M (third body) in the equation (II) comes from the equation (5).
The equation (I) represents the (global) chain branching reactions, while the equation (II)
illustrate the rate of chain breaking* reactions, where only remaining radical H is being
consumed. Alternatively to choosing k7[H][HO2] to be eliminated in the steady state equation
for HO2, one could have chosen k6[H][HO2], which is faster than k7[H][HO2] at high
temperatures. Employing the same procedure as before, this would result in the alternate
global steps:

During the chain branching reactions radicals such as H, OH are produced with and increasing rate, which then
causes to form the end products such as H2O.
*
During the chain braking reactions radicals such as H, OH are consumed, which causes to decelerate the rate of
formation of the end products such as H2O.

56

I)

3H2 + O2

2H + H2O

II)

2H2 + O2

2H2O

(146)

With the rates:


WI = k1[H][O2] - k7[HO2][H] - k8 [HO2] [OH]
WII = k5[H][O2][M]
This mechanism seems to be valid in the high temperature region where the first mechanism
represents the auto ignition mechanism of hydro carbon fuels, where the chain breaking
reactions play great role.
7.3. A reduced mechanisms for Methane-Oxygen Flames
As mentioned in 7.2 for lean-to-stoichiometric methane flames the skeletal mechanism of
25 elementary reactions given in the Table 2 is sufficient. The second step is to determine the
steady state species of which concentration will be considered time independent. A numerical
analyses of the premixed methane flames based on the whole starting mechanism shows that it
is necessary to retain H as non-steady-state species the in the 1. reaction (H + O2  O + OH).
It competes with the reaction (H + O2 + M  HO2 + M) as the most important chain breaking
reaction.
It is now possible to eliminate 8 reactions rate from the system using the 8 steady state
conditions. On the other hand, we want to eliminate the fastest reactions that consume each
selected steady state species. These reactions are then called main chain. From a sensitivity
analyses it is found that the main chain for the oxidation of CH4 via CH3, CH2O and CHO to
CO is:
11)

CH4 + H  CH3 + H2

13)

CH3 + O  CH2O+ H

14)

CH2O + H  HCO + H2

17)

HCO + M  CO + H + M

We can then use the steady state reactions

(147)

d[CH3]
d[CH2O]
d[HCO]
=0
,
=
0
and
= 0 to
dt
dt
dt

eliminate the rates k13[CH3][O], k14[CH2O][H] and k17[HCO][M].


On the other hand, the main chain for the chain branching reactions is:
1)

H + O2  O + OH

2)

O + H2  OH + H

3)

H2 + OH  H2O + H

and we use the steady state reactions

(148)

d[O]
d[OH]
dt =0 and
dt = 0 to eliminate the rates

k2[H2][O], k3[H2][OH].
Similarly the main chain for the chain breaking reactions to eliminate k7[H][HO2] (using the
d[HO2]
steady state relation dt ) is:

5)

H + O2 + M  HO2 + M

7)

H + HO2  O2 + H2
57

(149)

Finally, the main chain for the conversion of CO to CO consists of the two reactions:
8)

OH + HO2  H2O + O2

3)

H2 + OH  H2O + H

(150)

By adding the reactions in (147), (148), (149) and (150) and canceling the steady state species
(using reaction 3 twice in 147) one obtains the following global four step mechanism for the
oxidation of methane:
I)

CH4 + 2H + H2O

CO + 4H2

II)

H2O + CO

CO2 + H2

III)

H+H+M

H2 + M

IV)

O2 + 3H2

2H + 2H2O

(151)

Unlike the more systematic and comprehensive procedure used for the H2 O2 system the
consideration of the main chains do not provide the rates of the global reactions of the reduced
mechanism as in (145). It only provides the rate determining steps (as principal rates). They are the
first one of each secuences (147) to (151), namely k11[CH4][H] for (I), k11[CO][OH] for (II),

k5[H][O2][M] for (III) and k1[H][O2] for (IV).

8. Ignition of Homogenous Mixtures


If a spatially homogenous premixed mixture of fuel (vapor) and air is preheated to a
temperature of 750 K and higher the chain branching reactions accelerate and as a result, the
amount of the radicals and the heat released increases to an amount, which subsequently leads
to a thermal explosion of the mixture. This process, called auto-ignition is a very
important phenomenon in combustion in gasoline or Diesel engines. The ignition delay, the
time between the sudden temperature increase and the commence of ignition, depends only on
the kinetic properties of the reaction process and on the initial temperature and pressure. There
are lower limits for the pressure and the temperature, where auto-ignition is not possible even
if there exists enough time. These conditions are called explosion limits. Heat loss may also
prevent auto-ignition by decreasing temperature and pressure.
The auto ignition may proceed either under constant pressure or volume (or density).
8.1.

Auto-ignition at Constant Volume

Let us consider a closed vessel (Fig 21.) at constant volume (V) containing a given mass (m)
of mixture. This vessel is adiabatic and the total mass stay conserved during the ignition
process so that also the density ( = m/V) does not change.

Figure 21. Closed volume containing fuel-air mixture.

Ignition process in such a volume can be analyzed by writing down the energy balance
between the energy sources. The first law of thermodynamics (derived from the relation
between enthalpy and internal energy h = e + pv = e + T) describes the relation between
different forms of energy:

58

de + pdv = dh vdp = dq + dwR

(152)

where de, dh, dq, dwR are the changes in internal energy, enthalpy, transferred heat and
frictional work respectively. For a constant volume adiabatic process (dq=0, dv=0) and if no
frictional work is present (dwR=0) there will be no change in internal energy:
de
=0
dt

(153)

With the definition of the internal energy and the enthalpy of a gas mixture (yi being mass
fractions),
n

e=

yi ei ,

yi hi

h=

i=1

(154)

i=1

and time change of mole fraction of the species i (see eq. 120 and 121)
NR
dyi 1
m3
= Mi (ij - ij )(w j w j ) ( kg
dt
j=1

kg
mole

mole
)
m3 s

i = 1, .NS *

(155)

and with = Const.


n

yi dei + ei dyi

de =

i=1

(156)

i=1

If we assume constant specific heats (average value in the temperature range considered) then
we may express (156)
de = (

i=1

i=1

yi dCvi ) dT + ei dyi

or
n
n
de
dT
dyi
dt = ( yi dCvi ) dt + ei dt
i=1
i=1

(157)

with (153) we have


dT
Cvi dt

NR
=

(ij - ij ) (w j w j ) ei

j=1

NS : Total number of the species i., NR : Total number of the distinct chemical reaction j .
ij , ij : Stoichiometric coefficient of species (i) in reaction (j) in reactant and product sides
respectively.
w j , w j : Molar forward and reverse rates (for species i ) in reaction (j) per unit volume (moles/m3
s).
Mi : Molar mas of the species i

59

or with the heat of reaction at constant volume,


NS
Qvj =

(ij - ij ) ei

j = 1, .NR

i=1
We obtain the temperature change in constant volume ignition process:
NR

dT
Cvi dt

Qvj (w j w j )

(158)

j=1

8.3 Auto-ignition at Constant Volume


Let as, for simplicity, assume a single reaction, between fuel and oxidizer, with a large
activation energy E and the reaction rate:
w = B (

yf

Mf

)(

yO

MO

E
)exp (- RT )

The governing equations (155 and 158) then simplify to


dyi
dt = Mi (i - i) w

(159)

dT
= Qv w
dt

(160)

Cv

To determine the temperature change during the ignition delay (159) and (160) must be solved
simultaneously.
However, under some simplifying assumptions it is possible to have an idea about the
temperature rise. With the modified reaction rate
E
w = Bexp (- RT ),

B = B (

yf

Mf

)(

yO

MO

the temperature equation (16) becomes


dT
Qv
E
dt = Cv Bexp (- RT )

(161)

Integrating (159) and (160) and introducing one into other, with i - i = i gives
Cv(T-T0)
(yi-yi0)
=
Qv
i Wv

(161)

where T0 and Yi0 are the initial temperature and initial mole fraction respectively. We will
first assume that (Qv/Cv) is large compared to (T-T0) during the ignition process. Second
assumption is to neglect the reactant consumption, which allows us to set (yi = yi0). This
assumption is justified, because, small amount of fuel is consumed during the ignition period.
Under these assumptions we may deduce that the temperature increase will not be so high and
we may express the temperature with
T = T0 (1+ Y)
(162)
where is a small parameter.
60

On the other hand the Taylor expansion of the inverse of the temperature
1
1
T- T0
1
=
- T2 =
(1- Y)
T
T0
T0
0
leads to an expansion of the exponential term of temperature equation (161)
E
E
E
exp (- RT ) = exp (- RT ) exp (- RT Y)
0

which suggest the definition of as ( =

RT0
). This implies, again, the activation energy is
E

large as it was assumed above.


When the expansion (162) is introduced into (161) one obtains, for the temperature increase
dY
1
=
dt
ti exp (Y)

(163)

where
ti =

RT02 Cv
E
exp
(
E QB'
RT )

(164)

is the ignition delay time.


The solution of (163) is with the initial condition Y = 0 at t = 0 and using the transformation
X = e-Y
Which leads to dX/dt = -1/ti and x = 1-1/ti
t
Y =- ln (1 - t )
i

(165)

Temperature

Temperature change obtained from the solution (165) is plotted in Fig. 22.

Tim e

Figure 22. Thermal explosion Closed volume containing fuel-air mixture.


8.2.
Auto-ignition of Hydrogen-Oxygen mixture
In order to analyze the auto-ignition process of hydrogen-oxygen (or air) mixture we will
follow the simplified mechanism given below
1)

H + O2

O + OH

2)

O + H2

OH + H

3)

H2 + OH

H2O + H

61

(166)

5)

H + O2 + M 

HO2 + M

7b)

O2 + H2

H + HO2

Knowing that the reactions (2) and (3) are very fast and thereby O and OH are in steady state
we can combine the first three reactions to the global step
I)

3H2 + O2

kI


2H + 2H2O

(167)

The heat of reaction of this reaction is Qv= 48 kJ/mol. Whereas reaction (7b) is slightly
endothermic with Qv= - 21 kJ/mol, reaction (5) releases most of the heat with Qv= 197 kJ/mol.
So only this contribution can be retained in the temperature equation (160)
dT
Cv dt = Qv5 w5

(168)

and the balance equation for the formation of H-radicals is (remember

dyi
dt = Mi (i - i)

wj)
dyH
dt = MH (2wI + w7b w5)

(169)

It will be assumed, as before, that the reactants and the rate coefficients remain constant
during (the first stage of) ignition. Under these assumption, introducing non-dimensional
variables
X=

yH
(yH2)0

Y=

yH
(yH2)0

= tk1(T0)0YO20/MO2
= k5(T0)z5p/ k1(T0)R T0
= k7b(T0)/ k1(T0)
q = Qv5(YH2)0/Cv MH
(168) and (169) becomes
dX
d = (2 )X + 1

(170)

dY
d = X

(171)

and the solutions are, with the initial conditions X=0, Y=0 at = 0
X=

exp[(2- )] - 1
2-

Y= 2-

(172)

exp[(2- )] - 1
-]
2-

(173)

The non dimensional temperature Y is plotted against the non dimensional time ,for = 0.1,
= 2 and = 10, in Fig 23. This shows that temperature increases exponentially for < 2 ,

62

that means for higher initial temperature, quadratically = 2 and linearly for > 2. The H
radical, on the other hand, increases linearly for = 2 and reaches a constant value X = /(
1) for > 2.

Figure 23. Non-dimensional temperature increase, according to (173), during the thermal explosion
of Hydrogen-Oxygen mixture

t
While (165), the simpler expression (Y =- ln (1 - t ) already dictates a definition of the
i
ignition delay time (t=ti) as Y  , such a definition is not evident from (173). However, for
the temperature increase (T T0) not more than RT02/E1, where E1 is the activation energy of
the first reaction, same definition may be valid. For larger temperature increase the
assumption of constant rate coefficients is not valid any more and chain branching effect of
the 1. reaction increases very rapidly so that the constant concentration assumption for the
reactants would not be valid as well.
For a stochiometric homogenous mixture of hydrogen and oxygen at 0.1 bar the onset of
ignition was calculated using 19-step mechanism and a reduced 3-step mechanism. Ignition
delay time for the cases are compared at different initial temperature in Fig. 24. It is seen that
the ignition delay time increases with the decreasing temperature almost exponentially up to
800 K. From 800 K (cross-over temperature Tc) on ignition delay time increases suddenly to
very large values, making ignition in technical systems (such as systems with heat loss) quasi
impossible. The condition = 2 defines a cross-over temperature Tc between 1. and 5.
reaction

63

2k1(Tc) =

k5(T0)z5p
R TC

(174)

where z5
z5 = XH2,0 +0.4 XO2,0 + 0,4XN2,0
is the effective third body efficiency of the reactants. It is around 0.6 for hydrogen-air flames
and around 0.8 for hydrogen-oxygen flames. For p=1 atm Tc is around 1000 K . Below this
temperature the chain breaking effect of the reaction 5 dominates as compared to the chain
branching effect of reaction 1.

Figure 24. Ignition delay times for stoichiometric hydrogen-oxygen mixture


(solid line : 19-step mechanism, dashed line: 3-step reduced mechanism)

8.3. Explosion limits for hydrogen-oxygen mixtures


Explosion limits of a fuel-air mixture in a closed vessel depend on additional influences in
practice. Especially, heat loss to the walls, coatings of the wall and thereby on the dimension
of the vessel. Fig. 25 shows the explosion limits of the hydrogen-oxygen mixture as a function
of temperature and pressure. For the temperature range between 750 K and 840 K there are
three branches separating explosive and non-explosive regions. The middle branch is

64

essentially determined by the competition between the reactions (1) and (5) while the the low
and high pressure regions are influenced by the consumption of chain carrier on the walls.
Different coatings give different explosion limits in these regions. The branch agrees well vith
Tc evaluated from (174).
The lower explosion region growth as the temperature decreases. This is due to the increasing
effect of the (first) trimolecular reactions, on the initial low temperature chain branching,
H2 + O2 + M

H2O2 + M

H2O2 + M

OH + OH + M

with increasing pressure and decreasing temperature.

Figure 25. Explosion limits for stoichiometric hydrogen-oxygen mixture in a spherical vessel
(solid line : Numerical calculation using full kinetic mechanism by Stahl and Warnatz and data
points are referenced therein, dashed line: from (173) with z5 = 0.7)
For pressure above the middle branch in Fig. 25 chain breaking reaction (5)
5)

H + O2 + M  HO2 + M

leads to a massive formation of HO2 which then reacts with H and OH according the reactions
6)

H + HO2  OH + OH
65

7)

H + HO2  O2 + H2

8)

OH + HO2  H2O + O2

and the last two of them are chain breaking. When large amounts of HO2 are being formed,
however, there is a chain branching effect via reactions
9)

H2 + O2 + M

H2 O2 + H

10)

H2O2 + M

OH + OH + M

Whether chain breaking by the reactions (9) and (10) or chain breaking by (6) and (7)
dominates depends on the ratio of H and OH to H2 and on temperature. Additionally, in
vessels with large volume the removal of the radicals due to the wall effect is lesser and the
possibility of the third explosion is higher.

8.3.

Ignition of higher hydrocarbon-oxygen mixtures

The main chain of combustion of higher hydrocarbon-oxygen mixture may be modeled by the
reaction given below:

here,

R = C7H15 ,

R = C7H14

R = C7H13

Ignition delay times obtained by full kinetic mechanism and various reduced mechanisms are
showed in figures 26, 27 and 28.

66

Figure 26. Ignition delay times for stoichiometric n-heptan-oxygen mixture


(full mechanism is consist of 1000 reaction)

Figure 27. Ignition delay times for stoichiometric n-heptan-oxygen mixture


(full mechanism is consist of 1000 reaction)

67

Figure 28. Ignition delay times for stoichiometric n-heptan-oxygen mixture


(full mechanism is consist of 1000 reaction)

9. Physics of Combustion
In physical combustion systems the first step is to bring fuel and oxidizer together as to form a
proper mixture (as end step fuel vapor-air mixture) of a broad oxidizer-fuel ratio that then can
burn efficiently and with minimum amount of pollutant. There are mainly two kinds of
oxidizer-fuel mixture, which broadly used in practice:

In case of premixed mixture fuel vapor and oxidizer are mixed before they burn.
Although in some cases this mixture ignites spontaneously if auto-ignition conditions
are present, it is ignited by a heat source like spark plug or a similar device. To form
of a homogenous mixture is essential. However, in some cases we try to have a nonhomogenous premixed mixture, as in case of stratified charged gasoline engines (Fig.
29), where a slightly reach gasoline vapor-air mixture is desired near the spark plug to
ignite it easily, whereas lean mixture is preferred in the remaining combustion
chamber to have high effective efficiency. Combustion process is controlled
essentially by the rates of the reactions as, if the mixture is not highly nonhomogenous, a combustible mixture is already present.

=1

Flame front

= 1.2

Figure 29. Homogenous and non-homogenous mixture in gasoline engine chambers.

As the mixture is homogenous flame (propagation) velocity is same in every direction


resulting in a spherical flame front.

Ignition ( = 1)
Injector

Air flow

68

Figure 30. Heterogeneous spray combustion in gas turbine or Diesel engine chambers

Second form of oxidizer-fuel mixture is non-premixed mixture. In this case fuel and
oxidizer are brought together as they enter into the combustion chamber. Fuel may
readily be in vapor or gaseous form as, for example combustion of gaseous fuels.
However, in many cases (combustion chambers of gas turbines or Diesel engine
chambers) fuel is injected in liquid form and atomize into droplets to form a fuel spray.
In this case a heterogeneous mixture is present in the chamber, because of the nonhomogenous distribution of the fuel droplets and fuel vapor in the chamber (Fig 30).

In this case combustion is controlled firstly by the rate of mixture formation both in liquid and
vapor phases. Since fuel vapor from the fuel droplets or wall films needs time to diffuse into
the surrounding air diffusion plays great role in burning of heterogeneous fuel-air mixtures
(Fig. 31).

Figure 31. Shapes of diffusion flames surrounding a burning spherical fuel droplet:
a)non-spherical, b) spherical

Another classification of the combustion phenomenon may be made according to the flow
type of the reactants or products. Type of the flow affects considerably the transfer parameters
of the chemical species and the heat affecting the nature of the flame. Classification may be
made as follows:
Laminar Flames:
Laminar premixed flames
Laminar diffusion flames (Fig . 32)

69

Figure 32. Laminar premixed flame and laminar diffusion flame

Turbulent Flame

Turbulent premixed flames


Turbulent heterogeneous (diffusion) flames
(Fig 33)

Figure 33. Turbulent diffusion flame

It is possible to say that analyzing of a combustion system shall cover the themes of

fluid mechanics,
thermodynamics,
heat and mass transfer
chemical reactions.

The main task in analyzing of a physical combustion system is to find temperature, pressure
and chemical species distribution. This process requires the main equations of a physicalchemical system, namely, the conservation of

mass,
momentum,
energy
chemical species.

Except the equations (such as the conservation of the turbulence kinetic energy or dissipation)
that model the turbulence they are the main equations to be solved simultaneously, since the
source terms of the equations couple them closely each other. Sources are:
1- For mass conservation equation:

evaporation from liquid fuel droplets or films or condensation of fuel vapor,


evaporation from solid fuel particles or surfaces as positive source for mass
conservation,

70

mass (fuel or oxidizer in solid, liquid or vapor form) transfer from the boundary of the
system.
2- For momentum

momentum exchange between fuel spray (droplets) and gas phase due to the phase
changes in heterogeneous systems

3- For energy

heat release due to the chemical reactions

3- For chemical species

consumption or production of the species due to the chemical reactions

In the following section form of these equation will be overviewed briefly.

9.1. Conservation equations for reacting systems


The conservation equation of mass, momentum, energy and chemical species for a control
volume dV = dxdydz in a reacting flow system may be evaluated by applying to it the balance
of quantities and the second law of Newton which states that the momentum change, both in
time and in space, of the volume is balanced by the net force acting upon it.
9.1.1. Conservation equation for mass
For the control volume in Fig. 34 the balance of mass may be written in cartesian coordinates
as follows:
The net change of the total mass (m = dxdydz ) of the volume element dV = dxdydz is
D
dm = Dt (m) = mass
Where mass is the mass per unit volume per unit time produced or consumed in the volume
and the operator D/Dt
D


= +u +v v ,
Dt t
x
y z
is called total derivative and represents the infinitesimal changes in a quantity in X direction
with respect to time and to space as the fluid flows through the control volume with the
velocities u, v and w in X, Y and Z directions
Evaluating the right hand side in X direction, for example, gives
D

Dt ( dxdydz) = t dxdydz + t dydz + u x dxdy dz


x
x
u
With dV = dxdydz , t = u, or t dydz = x dV one obtain

71

dV
+
t
x (u)dV = mass
Written in all direction is the continuity equation:

+
(u) +
(v) +
(w)= mass
t
x
y
z

kg
( 3 )
m s

(175)

y
dy

[mx]x+dx

[mx]x

dz
dx

q=
x

z
Figure 34 - Mass balance of the control volume

Continuity equation may be expressed in vector form

t + div (V) = mass

kg
(m3 s )

(176)

where V is the velocity vector.


9.1.2. Conservation equation for momentum
Similarly the conservation equation of momentum for the control volume dv = dxdydz in a
reacting flow system can be evaluated by applying to it the second law of Newton which states
that the momentum change, both in time and in space, of the volume is balanced by the net
force acting upon it. For the control volume in Fig. 35 this balance in X direction, for
example, is as follows:

dFx =

D
(m.u )
Dt

1
m
kgm
( s kg s = s2 = N)

Evaluating the right hand side gives


dFx =

( .dx.dy.dz.u ) + u ( .dx.dy.dz.u ) + v ( .dx.dy.dz.u ) + w


( .dx.dy.dz.u )
t
x
y
zz

and by substituting dV for dx.dy.dz we obtain

72

[yx]y+dy

dy

[p+xx]x+dx

[p
+xx]x
dz

dx

[yx]y
x
dq
z

Figure 35 - Momentum balance of the control volume

x
y
y
+ u.dV
+ u. dy.dz + u. dx.dz + u. dx.dy
t
t
t
t
t
y
y
x
u

.......... + u. .dV
+ u.u.dV
+ u.u. dy.dz + u.u. dx.dz + u.u. dx.dy
x
x
x
x
x
y
y
x
u

.......... + v. .dV
+ u.v.dV
+ u.v. dy.dz + u.v. dx.dz + u.v. dx.dy
z
y
y
y
y
u

x
y
y
.......... + u. .dV
+ u.u.dV
+ u.w. dy.dz + u.w. dx.dz + u.w. dx.dy
z
z
z
z
z
dFx = .dV

If we substitutes

x
y
x
y
= u ,....... = v,....... = 1,........ = 0,...............etc
t
t
x
x

for the derivatives and

V
V
V
V
u
V
v
V
v
= dy.dz ,...
= dx.dz ,....
= dx.dy....
du =
dV ,.....
dv =
dV ....
dw = dV
x
y
z
x
x
y
y
z
z
for other products we obtain
73

u
v
w
+ u.dV
+ u. .dV
+ u. .dV
+ u. .dV
+
t
t
x
y
zz
u

V
w
.......... + u. .dV
+ u.u.dV
+ u.u.
+ u.u .dV
+
x
x
x
zz
u

V
w
.......... + v. .dV
+ u.v.dV
+ u.v.
+ u.v .dV
+
y
y
y
zz
u

V
w
.......... + w. .dV
+ u.w.dV
+ u.w.
+ u.w .dV
z
z
z
zz
dFx = .dV

u
u
v
w

+u
+ v + w ) + u( + u
+v
+ w ) ]+
z
t
x
y
z
t
x
y
u v w
v
v
w
.......... + u. ( +
+ )dV + u. (u + v + w )dV
x y z
x
y
z
dFx = dV [ (

dFx = (

u
u
v
w

u
v
w
+u
+v +w
)dV + u ( + u
+v
+w
+
+
+ )dV
t
x
y
zz
t
x
y
z
z
x
y

dFx = (

u
u
v
w

+u
+ v + v )dV + u[ ( + ( .u ) + ( .v) + ( .w) ]dV
t
x
y
z
t x
y
z

With the continuity equation (175), if there is no mass production in the volume,

+ ( .u ) + ( .v) + ( .w) = 0
t x
y
z
the equation of momentum in X direction becomes
dFx
u
u
u
u
= ( + u
+v
+w )
dV
t
x
y
z

(177)

In case of incompressible(1) flow the force, in X direction, acting upon the control volume
consist of static pressure force and shear stress forces only, because no normal stress force
exists for incompressible fluid:
dFx = dp.dy.dz + d xx .dy.dz + d yx .dx.dz + d zx .dx.dy

(178)

The viscous tangential stresses (shear stresses) are proportional to velocity gradient for a
Newtonian fluid:

The compression of the fluid is due to the inertia effect of the accelerating or decelerating flow. The compressibility is depend on the Mach
number, the ratio of flow velocity to sound velocity (M=v/a), through the formula
1
0
k 1 2 k 1
M )
= (1 +

For the velocity less than v = 100 m/s the Mach number is M=100/330=0.3 and the increase in density is less than 4.5%.

74

yx =

du
,
dy

yx =

du
,
dx

yz =

du
dz

If we put the relations above in (178) and 178 in the right hand side of (177) we obtain the
following "Navier-Stokes equations" for an incompressible, viscous fluid in Cartesian
coordinates for X, Y and Z directions:

2u 2u 2u
u
u
u
u
p
+u
+v
+ w ) = + ( 2 + 2 + 2 )
t
x
y
z
x
x
y
z

(179)

2v 2v 2v
v
v
v
v
p
+ u + v + w ) = + ( 2 + 2 + 2 )
t
x
y
z
y
x
y
z

(180)

2v 2v 2v
w
w
w
w
p
+u
+v
+ w ) = + ( 2 + 2 + 2 )
t
x
y
z
y
x
y
z

(181)

Momentum equation may also be expressed in vector form


dV
dt + div ( V V) + grad p - div (gradV) = mom

kg m
m3 s

1
N
= 3)
s
m

(182)

where mom is the source term which may arise for a multi component system, due to the
different body force (gravitational force per unit volume) acting on each component and if
continuity equation has a source term and gradV is the viscous stress tensor with
V=(u2+v2+w2)/2.

9.1.3. Conservation equation for energy


Total enthalpy, h = e + p/ + V2/2+ k , is the energy to be conserved (k : turbulence kinetic
energy).
D
(h) + div (Jh) + div (Jk) = eneryy
Dt

1 kg kJ
kJ
=
)
s m3 kg m3 s

and the conservation equation of energy written in vector form is


d (h)
+ div (V h) + div (Jh) + div (Jk) - div (gradV) = energy
dt
(183)
where div (V h) represents the energy of the in- and out-flow, div (Jh) the energy transported
by heat transfer, div (Jk) the energy transported by turbulence kinetic energy transfer, div
(gradV) viscous dissipation energy and energy is the source term mainly due to the chemical
reactions. Transferred energies fluxes are given in the form:
Jh = h eff grad (e + p/ )
Jk = k eff grad (k)

Jh = h eff grad (CpT)

that is
3

kJi/m s

75

where h eff is the effective (in case of turbulent flow) exchange coefficient of heat (thermal
conductivity divided by the Cp and k eff are effective exchange coefficient of the kinetic
energy of the turbulence. Temperature can than be calculated from the known enthalpy values.

9.1.4. Conservation equation for chemical species


Total mass of each chemical species must also be conserved according to the balance equation
below, written in terms of mass fractions (xi).
D
(mi) + div (Ji) = i
Dt
and the conservation equation of the species (i) written in vector form is
d (xi)
+ div (V xi) + div (Ji) = i
dt

1 kg (kg)i
(kg)i
= 3 )
s m3 kg
m s

(184)

where div (V h) represents the difference of the mass of species of the in- and out-flow, div
(Ji) the diffusive mass transfer and i is the source term (production or consumption of (i) due
to the chemical reactions). Transferred mass flux is given in the form:
Ji = i eff grad (xi)

(kg)i/m3s

where i eff is the effective (in case of turbulent flow) exchange coefficient of species (i)
(diffusion coefficient times mixture density).
The source term i is (as given by 120 and 121)

i =

NR
d(xi)
= Mi ij w j
dt
j=1

(kg)i (kmol)i
(kg)i
=
3
(kmol)i m s
m3 s

(185)

9.1.5. Solution of a multi-component reacting flow


If we have a system of N species we have then N + 6 unknown. They are
x1, x2, x3, ., T, p u, v, w
For the solution of the system N + 6 equations below will be made of use
1 overall continuity equation

Eq. (176)

3 momentum equations

Eq. (182)

1 energy equations

Eq. (183)

N-1 species equations

Eq. (184)

1 equation of state

p = RT

The equation relating all xi

x1+ x2 + x3 + .= 1

76

First N+4 equations are partial differential equations. As they coupled to each other through
source terms and are highly non-linear, with exception of some simple cases, no analytical
solution is possible. Numerical methods are developed for solution.

9.2. Laminar premixed flames


Premixed flames are used whenever intense combustion within a small volume is required.
Examples to this case are household devices or spark ignition engines (Fig.29, 30, 33). Large
combustion devices such as furnaces or big gas-burners, on the contrary, operate with nonpremixed diffusion flames since the premixed mixture of air and fuel represents a serious
safety problem (Fig 34).
Premixed mixture burns with a blue to bluish-green color due to the chemiluminescence of
some exited species such as C2 and CH radicals. Diffusion flames, on the other hand, radiate
in bright yellow. Radiating soot particles causes this color. Radiation effect dominates over
chemiluminescence which is also present in diffusion flames. However, if diffusion flame is
highly stretched blue flame appears again since the local residence time is too short for soot
particles to be formed.

9.2.1. Laminar burning velocity


The classical device to generate a laminar premixed flame is the Bunsen burner shown in
Fig. 36. This burner operates with gaseous fuel. Fuel-air ratio can be controlled by orifice
adjustment screw. Mixing chamber of the burner must be long enough to ensure a full and
homogen mixing. If the flow velocity at the exit is slightly greater than the laminar burning
velocity to be defined below, a flame cone establish itself at the top of the burner body. This
corresponds to the case, a steady state flame propagating into the unburnt mixture with the
velocity SL normal to the cone surface.
The kinematic balance of this process is illustrated for a steady oblique flame in Fig. 37. Here
vu is the flow velocity of the unburnt mixture, vt,u and vn,u are tangential and normal
components. Similarly vb is the flow velocity in the burnt side, vt,b and vn,b are tangential and
normal components. Due to the thermal expansion within the flame front normal component
of the flow velocity at the burnt side increases so that the continuity equation is satisfied.

77

Figure 36 The Bunsen burner

Figure 37 Kinematic balance of a steady oblique flame

(vn)u = (vn)b

(186)

and
v n ,b = v n ,u

u
b

(187)

The tangential velocity components are not affected by the gas expansion and remain the same
vt,u = vt,b

(188)

78

The stability condition of the flame front is


SL,u = vt,u

(189)

SL,u = vu sin

(190)

or

where SL,u is the laminar flame velocity with respect to unburnt mixture. This relationship
allows to determine the burning velocity by measuring the cone angle () under the condition
that the flow velocity vu is uniform at the pipe exit. This is in general not the case and
therefore the cone angle varies along the flame front.
The tip of the cone is in general closed for hydrocarbon flames (but not, for example, for lean
hydrogen flame). This means that the burning velocity is equal here to the flow velocity,
which is otherwise larger than the burning velocity by a factor 1/sin . This is because of the
higher temperature revealed here due to the increased heating effect of the flame front.
Similarly, increased heat loss near the rim of the burner causes the temperature so to drop that
a sustained flame is not possible any more and flame detaches here.
Another example for an experimental device to measure the laminar flame velocity is the
combustion bomb (Fig. 38)

Figure 38. Homogenous combustion in a spherical combustion bomb

A central spark ignites the homogenous mixture content of the bomb. A spherical flame front
propagates with a laminar velocity SL. This velocity may be found from the mass continuity at
the flame front
drf
drf
b (vb - dt ) = u (vu - dt )

(191)

drf
is the flame propagation velocity. The kinematic relations between propagation
dt
velocity, flow velocity and the burning velocity on the other hand are
where

drf
= vu + SL,u
dt

(192a)

79

drf
= vb + SL,b
dt

(192b)

In case of spherical bomb the flow velocity in the burnt side is zero (vb=0) because of the
symmetry. This leads with (191) and (192a) to

u
drf
=
dt
u - b vu = vu + SL,u

(193)

from which
SL,u =

drf b
dt u

(194)

drf
the laminar burning velocity SL,u
dt
with respect to unburnt zone can be calculated. Similarly burning velocity SL,b with respect to
burnt zone is
is found. By measuring the flame propagation velocity

SL,b =

drf
dt

(195)

since vb = 0.

9.3. Governing equations for steady premixed flames


Let us consider a planar steady state (time independent) flame which propagates in x direction
with the unburnt mixture at x => - and the burnt gas at x => + .
The governing conservation equations in this case are
d
(u) = 0
dx
(u)

dmi
dji
=+ i
dx
dx

(continuity)

(196)

(species)

(197)

NS
NS
dT
d
dT
dT
(uCp)
=
(
) - hi i - j i C p
dx
dx
dx
dx
i=1
i=1

(energy)

(198)

Here mi is used for the mass fraction instead of mi in order to avoid to confuse it with x
coordinate. Energy equation is the modified form of (183)
d (h)
+ div (V h) + div (Jh) + div (Jk) - div (gradV) = energy
dt
using for dh, and Jh

80

NS

hi dm i

dh = CpdT +

i=1
NS
Jh = - grad T +

hi j i

i=1
or
NS
Jh = - grad T +

dT

Cp ji dx

i=1
NS

hi m i

energy =

i=1
and neglecting the viscous dissipation term and the turbulence kinetic energy as the flow is
laminar. Additionally the pressure term in the momentum equation may be neglected since the
propagation velocity is much smaller than the sound velocity and at the velocities valid in this
system the flow can be assumed incompressible. As the u velocity can be calculated from the
continuity equation use of the momentum equation is not necessary.
The continuity equation may be integrated to yield

u = u S L

(199)

Here the burning velocity SL is an eigenvalue of the system, which must be determined as a
part of the solution.
System of equations (196, 197, 198) can be solved numerically using a kinetic reaction
mechanism. In Fig. 39 burning velocities calculated with the starting C1 mechanism give in
Table 2 and a starting C2 - mechanism that includes the species C2H6, C2H5, C2H4, C2H3,
C2H2, C2H, are shown in Fig. 39

81

Figure 39. Burning velocities calculated with the starting C1 mechanism give in Table 2 and a
starting C2 - mechanism that includes the species C2H6, C2H5, C2H4, C2H3, C2H2, C2H

9.4. Premixed flames based on one-step asymptotics


A classical thermal flame theory, which considers a single one-step reaction and describes the
structure of a premixed flame is given by Zeldowich and Frank-Kamenetzki in 1938. Single
step reaction is
FF - O2O2 =>
products
(200)
The reaction rate, which is first order with respect to fuel and oxygen and suitable for lean
mixture is
w = B (

yf

)(

yO

Mf MO

) exp (-

E
)
RT

(201)

The activation energy E is assumed to be large. Arhenius type of the temperature dependence
is present. However, frequency factor B and activation energy E are just adjustable parameters
and cannot be deduced from elementary kinetic data. Flame structure is shown schematically
in Fig. 40.
The conservation equations (197) for fuel and oxygen, with (199) are
dmF
d dmF
u SL dx = - dx ( C
) - F MF w
dx
p

(202)

dmO2
d dmO2
u SL dx = - dx ( C
) - O2MO2w
dx
p

(203)
NS

and the temperature equation (198) written in the modified (with i = Qpi wi and Qpj = i
i=1
hi) form is

82

dT
d
dT
Q
u SL dx = dx ( C dx ) + C w
p
p

(204)

Figure 40. Flame structure of a lean flame with one-step asymtotics

Combining these yield for mass fractions, integrated from the burnt state to any state within
the flame
mF = -

F MF Cp
(T Tb) + mF,b
Q

mO2 = -

(205)

O2 MO2 Cp
(T Tb) + mO2,b
Q

(206)
The mass fractions are a function of temperature only and the reaction rate is too. This implies
that only the temperature equation (204) needs to be solved. With dimensionless coordinate x*
temperature T* and reaction rate
x
Cp
x * = u S L

dx

T* =

T Tu
Tb Tu

w* =

Q w(T)
u S LC2p (Tb - Tu)
2

the temperature equation may be written


dT* dT*2
=
+w
dx
dx2

(207)

83

This equation may be rewritten using new stretched coordinate = x/ and the expanded
temperature around its maximum value T* = 1 y the new temperature equation is
d2 y
= (y + by2) exp(-y)
d 2

(208)

where is a small parameter, is the modified burning velocity eigenvalue. Numerical


solution of (208) is plotted in Fig. 41.

Figure 40. Numerical solution of (208)

9.5. Non-premixed laminar diffusion flames


Combustion systems where fuel and oxidizer enter into the chamber separately are diffusion
controlled systems. Mixing then take places by convection and diffusion. Diffusion is on
molecular level in case of sufficiently low velocities ensuring that flow is laminar. A typical
example for laminar diffusion flame is the candle flame (Fig.42). Air enters into the
combustion zone by natural convection (buoyancy effect) while paraffin that evaporates from
the wick diffuses into the air. In the system shown in Fig. 32, on the other hand, gaseous fuel
is injected through a nozzle into the surroundings air and mixing occurs by forced convection
and molecular diffusion. Air can also be send into the system as forced flow as shown in Fig.
33. If the jet velocities are low enough than flow is laminar.

84

Figure 42. The candle flame as example for laminar non-premixed diffusion flame

In combustion zone, only where fuel and oxidizer are mixed within a near stochiometric ratio
chemical reactions can occur. The reaction mechanism may then be best described by a
perturbation around the limits of complete combustion with very fast chemistry. The
perturbations in the air-fuel ratio are result in the perturbation in adiabatic flame temperature.
The nonlinear perturbation mechanism around the limits will then take the finite rate
chemistry and flame quenching due to the insufficient mixing into account. The non-premixed
laminar flow combustion may be formulated by a one-dimensional so called flamelet model
by using the mixture fraction as an independent variable in place of all reacting scalars in the
limits of sufficiently fast chemistry.

9.5.1. Flamelet structure of a laminar diffusion flame with a one-step large activation energy
reaction mechanism
Since the flamelet model is based on the complete combustion depending on the mixture
fraction distribution the first task is to determine the surface of the stoichiometric mixture as a
function of spatial and time coordinates. This may be expressed by
Z(x,t) = Zstoich
To do this we may express the conservation of the chemical species (184)
d (xi)
+ div (V xi) + div (Ji) = i
dt
in term of mixture fraction, where i, the production or consumption of each species (the
source term) was given by (185)
NR
(kg)i (kmol)i
(kg)i
i = Mi ij w j
(kmol)i m3 s = m3 s
j=1

85

Mass fractions of elements


If we denote by mk the total mass of the atoms of element k contained in all species and if aik
is the number of atoms of element k in a molecule of species i
NS
mk =

k
aik mi Mi

(209)

i=1

where mi is the partial mass of species i .


The mass fraction of the element k (=1,,..NE) is then
NS
Mk NS
mk
Mk
zk =
= aik xi
=
aik yi
m
Mi
M
i=1
i=1

(210)

While the mass of species changes due to the chemical reactions the mass of the elements is
conserved through the reactions. That means the conservation equation of the element k will
have no source term.
NS
Mk
d (zk)
+ div (V zk) + div ( aik yi
J )=0
dt
Mi k
i=1
The chemical source term k (185 written for elements) vanishes because the summation over
all reactions (j =1..,NR) and all species (i =1..,NS) gives null

k = Mk

NS NR

NR

NS

i = 1j = 1

j=1

i=1

aikij w j = Mk wj aikij

since the last sum vanishes for each reaction.


The diffusion term simplifies if one assume that the diffusivities (exchange coefficients i) of
all chemical species and temperature (h) are same (i = h= ) and Jk = grad (zk).
d (zk)
+ div (V zk) + div (Jk) = 0
dt
(211)
The assumption that the diffusivities (exchange coefficients i eff) of all chemical species are
same is realistic in case of hydrocarbon flames but not valid for hydrogen flames.
Conservation equation for mixture fractions
mF

A similar equation can also be derived for the mixture fraction Z= m + m (25), which is
F
O
described as the fuel fraction in the fuel oxidizer mixture. The local fuel mass fraction (26) in the
combustion chamber is proportional to mixture fraction and to fuel fraction in the original fuel stream.

(xF = Z xF0).
The mass fraction of the fuel is, on the other hand, the sum of the element mass fractions of
the fuel

86

NE
mF
xF =
= zk,F
m
k=1

(212)

where (similar to 210)


zk,F = aF,k xF,u

Mk
MF

(213)

With (26) the mixture fraction may then be expressed as a sum of element mass fractions
NE

zkF

Z=

k=1
xF0

And the conservation equation for mixture fraction is


d (Z)
+ div (V Z) + div (grad Z) = 0
dt

(214)

If we choose the fuel and oxidizer as the species that are to be taken into account and under
the assumption that the diffusivities are same (if= h= and Ji = grad (xi), 184 may be
written for i=F and O
d (xF)
+ div (V xF) + div (JF) = F
dt
(215)
d (xO)
+ div (V xO) + div (JO) = O
dt

(216)

Here the source terms for a one-step reaction between fuel and oxidizer may be expressed as
follows with the reaction rate w and for complete combustion

F = MF F w ,

O = MO O w

Dividing (215) by MFF , (216) by MOO and subtracting yields a source free conservation

xO
O MO
equation for the combination of (with x stoich= . =
)
F MF
F

xO
xF xO
xF
)=
MF F
MO O
'0 M 0

which is a linear function of the mixture fraction

mF
xF xO + xO 0
Z = m + m or Z =
xF 0 + xO 0
F
O

In this case (215) and (216) can be combined in one equation, i.e. the equation 214, written in
terms of mixture fraction
When the mixture fracture field is found from the solution of (214) fuel and oxidizer fraction
can be obtained from (26) and (27)
xF = Z xF 0

87

xO2 = (1-Z) xO 0
Combustion mechanism within a thin reaction layer
Combustion occurs in a thin layer in the vicinity of the surface Z(x,t) = Zstoich, where heat
release and heat loss are in balance. This is of course only possible if the local mixture
fraction gradient is sufficiently high as to maintain a step temperature profile. Otherwise flame
extinction may occur due to the insufficient heat release and increasing heat loss. The
governing equations written in tensor form are (i = x1 , x2 , x3 Cartesian coordinates)

Z
Z
t + Vi x
i

Z
(
xi
xi ) = 0

T
T
t + Vi x
i

NS Q
QR

T
k
xi ( xi ) = Cp wk + Cp
k=1

(217)

(218)

The temperature equation is derived from the energy equation (183) under the assumptions of
Q

constant specific heats and pressure. The last term ( CR ) is the source term due to the internal
p
radiation and heat transfer from surroundings.
Since the flamelet model is based on the relation between the mixture fraction and the
temperature field in the combustion system it is more convenient to apply a coordinate
transformation, replacing the coordinate x by the mixture fraction. Then the temperature can
then be expressed as a function of Z (Crocco type coordinate transformation). This allows us
an easier analytical analyses.
By definition, the new coordinate, Z, is locally normal to the surface of stoichiometric
mixture. Using the other independent variables Z2 = x2, Z3 = x3, and t* = t, with
transformation rules

Z
=
+
t
t*
t Z

(219)

Z
x1 = x1 Z

(221)

=
xk
Zk

Z
xk Z

(k=2,3)

(222)

we obtain the temperature equation in the form

T
T
T
( ) T
( ) T
( t * + v2 Z + v3 Z ) - x
Z2
x3 Z3
2
3
2
- [(

Z 2 2 T
Z
T
Z
T
2 T
2 T
)
+2
+
2 +2
2 +
xk
Z
x2 Z Z2
x3 Z Z3
Z2
Z32 ] =

NS Q

k
R
Cp wk + Cp

(223)

k=1

88

For a thin flamelet in Z direction (normal to Zsoich surface) an order-of-magnidute analyses


shows that the second derivative with respect to Z is the dominating term on the left-hand side
of (223). The term containing time derivative may also be retained since it is important in case
of very rapid changes, such as flame extinction, occur. This can be formally shown by
introducing the stretched coordinate and the fast time coordinate

= (Z-Zstoich)/

= t*/2

where is a small parameter representing the width of the reaction zone.


If only time derivative and second derivative terms are retained the time dependent onedimensional temperature equation is

T
t

-
2

NS Q
2 T
QR
k
Z2 = Cp wk + Cp
k=1

(224)

Similar equations may be derived for the chemical species instead of mixture fraction. In
equation (224)
Z 2
= 2 (
)
(225)
xk
is the instantaneous scalar dissipation rate of the mixture fraction (species). It has the
dimension 1/s and may be interpreted as the inverse of the characteristic diffusion time. This
term implicitly incorporates the influence of convection and diffusion normal to the surface of
stoichiometric mixture. In the limit  0, the flame sheet model (zero flame thickness) is
obtained.
If the time dependent term is neglected, (224) is an ordinary differential equation that
describes the structure of a steady state flamelet normal to the surface of stoichiometric
mixture. We may also assume the heat loss term QR to be negligible.
In the limit of complete combustion chemical reactions are confined to infinitely thin sheet at
Z = Zstoich. Assuming infinite rate reaction and constant Cp, the temperature profile is given by
(41, 42), the Burke-Schuman solution
TP (Z) = TR (Z) +

Q ref xFR
Z
Cp 'FMF

(for Z<Zst)

TP (Z) = TR (Z) +

Q ref xO2R
(1 Z)
Cp 'O2MF

(for Z>Zst)

TR (Z) = T2 + Z(T2 T1)


with the maximum flame temperature obtained for Z = Zst (Fig 6.)
TP (Zst) = TR (Zst) +

Q ref xFR
Z
Cp 'FMF st

(for Z=Zst)

89

T e m p e r atur e

TP

TO2

TR
TF

0,00

0,2

0,4

0,6

0,8

Zst= 0.1 M ixtu r e Fr actio n (Z)

The dependence of adiabatic combustion temperature on mixture fraction

Fuel, oxidizer and products mass fractions profiles are piecewise linear function of Z (Fig 2)
and the coupling relations (205, 206) give the corresponding profiles for the mass fractions.
F MF Cp
(T Tb) + xF,b
Q

xO2 = -

O2 MO2 Cp
(T Tb) + xO2,b
Q

CO 2

X F0 , X o2 0

1,00

XF 0

0,2

0,75

0,15

H 2O
0,50

0,1

XO20

0,25

0,05

0,00

X H2 O , X CO 2

xF = -

0,00

0,2

Z st =0.1

0,4

0,6

0,8

M ixture Fraction (Z)

The dependence of fuel, oxygen and products mass fractions in burnt mixture on mixture fraction

In case of one-step reaction and finite reaction rate of form (201)


w = B (

xF

)(

xO

MF MO

) exp (-

E
)
RT

temperature equation (224) can describe the diffusion flame quenching. Expanding
temperature, fuel and oxidizer mass fraction and the exponential term in the reaction rate
around Zstoich
T = Tstoich (Tstoich TR (Zst)) y
xF = xF0 (Zst(y - a) + )

(226)

90

xO = xO0 ((1 - Zst)(y - a) - )


exp (-

E
E
) = exp () exp (Ze y)
RT
RTstoich

with
a = (T2 T1)/ (Tstoich TR (Zstoich))
and with Zeldowich number
Ze = E (Tstoich TR (Zstoich))/ RT2stoich
(224) can be written in terms of y, and

2 y
3
2 = 2 Da (Zstoich (y - a) + )( (1-Zstoich )(y - a) - ) exp (- Ze y)
where (Da)
Da =

Bstoich O2 xF0
E
exp ()
RT
stoich MF (1-Zstoich )

(227)

(228)

is the Damkhler number, which is a criterion for the effect of heat release and dissipation
rate. With the further transformation
z = 2 y ( (1-Zstoich ) Zstoich

= 2 Zstoich (1+ a(1-Zstoich ) )- 1

(229)

= Ze /(2 Zstoich (1-Zstoich ) )


we get

2 y
3 2
2
2 = Da (z ) exp (- (z+ ))

(230)

The expansion parameter may be defined in two ways, either setting = 1 or by setting Da
3 = 1. The first approximation is called large activation energy expansion since it points
out the effect of heat release rate while the second points out the effect of diffusion rate and
called Damkhler number expansion. Both formulations are interrelated with a
distinguishing limit

= Da/3
(231)
-1/3
which is to be order of one. With this definition (229) becomes (with = Da = -1/3/)
2 y
2
2
-1/3
2 = (z ) exp (- (z+ ))

(232)

which is called Linans equation for the diffusion flame regime. Boundary conditions are,
obtained by matching the outer flow solution

z
= 1 for 

(  0)

z
= -1 for -

As is assumed to be order of one the equation (231) is written under the approximation of
large activation energy limit. The essential property of this equation, as compared to large
Damkhler number limit (  ), is that the exponential term remains. This form of the
91

transformed temperature equation allows extinction to occur if the parameter decreases


below a critical value d. For small values of Zstoich, which are typical for hydrogen-air or
hydrocarbon-air flames, extinction occurs at a temperature TP = Tstoich (1 -R Tstoich/E) that
corresponds to transition to the premixed regime . The critical value of .is then

d = e (1- ||)
The solution of the full structure of the laminar diffusion flamelets consists of the BurkeSchuman solution TBS for the complete combustion and the first-order expansion T1 in the
outer flow, which takes the relations between the reaction rate and the heat loss into account.
T (Z) = TBS (Z) T1 (Z, Z2, Z2, )

(233)

The outer solutions for the left and right sides of the thin flame zone are
T-1 (Zstoich) = (Tstoich TR (Zstoich)) (y + ) -
T+1 (Zstoich) = (Tstoich TR (Zstoich)) (y - ) +

(234)

The quantities (y + ) - and (y - ) + are obtained from the numerical solution of (232).
The solution in the outer field are then
T-1 (Z) = Z/Zstoich T-1 (Zstoich)
T+1 (Z) = ((1-Z)/(1-Zstoich )T+1 (Zstoich)
Characteristic profiles for the temperature over Z are schematically shown in Fig 43. There is
a limiting temperature profile Tq (Z) corresponding d. Under the conditions below this profile
flamelet would be extinguished locally.

Figure 43. Temperature profiles over mixture fraction for diffusion flamelet at increasing

Damkhler numbers
The equations (234) define a maximum instantaneous dissipation rate d at the surface of the
stoichiometric mixture.

92

d =

8Bstoich O2 xF0 Z3stoich (1-Zstoich )2


E
exp ()
3
RTstoich
q MFZe

(235)

The instantaneous dissipation rate stoich, defined by (225) = 2 ( x )2 , may be considered


k

as the inverse of a characteristic diffusion time. If stoich is large, i.e. diffusion time is long,
heat will be conducted to both sides of the flame at such rate that is not balanced by the heat
release due to the chemical reactions. Thus the flame temperature will decrease until the
flemelet is quenched at the value stoich= d. For large activation energy, a plot of the
maximum temperature with respect to (1/stoich) will have the well-known S shaped ignition
limits curve (Fig. 44)

Figure 44. Ignition and quenching limits depending on diffusion time

Beyond the upper branch of the curve a stable burning is possible. If stoich is increased
temperature drops, the curve is traversed to the left until d is reached. Below the temperature
corresponding to d (point Q) no self-sustained reaction is guaranteed and the flamelet is
quenched. There is an unsteady transition from burning to quenching from point Q up to the
lower branch of the curve. Auto-ignition, which would correspond to an unsteady transition
from the point I to the upper branch, is unlikely to occur in open diffusion flames, since the
required very large residence time (corresponding very small values of stoich ) are not reached.
This is why a long ignition delay time is required in Diesel engine combustion system, where
inter-diffusion of the fuel vapor from the fuel spray with the surrounding hot air leads to
continuously decreasing mixture fraction gradients and therefore to decreasing scalar
dissipation rates. This mechanism corresponds to a shift on the lower branch of the S-shaped
curve up to the point I of ignition.

9.5.2. Flame time and length scales in diffusion flames


A chemical time scale at extinction may be defined as
tc = (1-Zstoich )2 Z2stoich/d

(236)
93

This leads, using (234), to

d
4Bstoich O2 xF0 e-1 RTstoich 3
E
=
(
) exp ()
2 2
3
(1-Zstoich ) Z stoich
E
RTstoich
MF(Tstoich TR)

(236)

where the right hand side is proportional to the inverse of the equivalent chemical time of a
stoichiometric premixed flame. This indicates that there is a fundamental relation between
premixed and diffusion flames at extinction. However, while in a diffusion flame the scalar
dissipation rate and so the characteristic flow time and the diffusion time are independent
variables and may be chosen arbitrary, the flow time in a premixed flow is fixed by the
burning velocity, which is an eigenvalue of the problem. Therefore combustion in diffusion
flame offers an additional degree of freedom; the ratio of convective time to reactive time,
represented by the Damkhler number (228). That makes the non-premixed combustion to be
better controllable and stable as in case of Diesel engine combustion, which is more robust
and less fuel quality dependent than that in spark ignition engines.

9.5.3. Diffusion flame structure of methane-air flame


The one-step model with large activation energy is able to predict the important features such
as extinction limits of the diffusion flame, but for small values of Zstoich it does not predicts
fuel or oxidizer through the reaction zone truly. A four step reduced mechanism has better
results with this respect. An asymptotic analysis based on the four-step model shows a close
correspondance between premixed and diffusion methane flames as shown in Fig.45a and
45b. The outer structure of the flame is the classical Burke-Schumann structure governed by
the overall one-step reaction CH4 + 2O2 =>CO2 + 2H2O with the flame sheet positioned at
Z = Zstoich .

Figure 45-a. Flame structure of a methane diffusion flame with reduced four-step asymptotic The
dotted line is the Burke-Schumann solution.

94

Figure 45-b .Flame structure of a premixed methane-air flame.

The inner structure on the other hand consists of a H2 CO oxidation layer of thickness O()
toward the lean side and a thin inner layer of thickness O() toward the rich side of Z = Zstoich.
Beyond this layer the rich side is chemically inert because all radicals are consumed by the
fuel. The rich part of the diffusion flame corresponds to the upstream preheat zone of the
premixed flame while its lean part corresponds to the downstream oxidation layer.
The maximum temperature which, corresponds to the upper branch of the S-shaped stability
curve is shown in Fig 46. The calculations agree well with numerical and experimental data.

Figure 46 .Inner zone flame temperature as function as a function of stoich .

95

9.6. Turbulent Diffusion flame structure


9.6.1. The structure of turbulence
The main feature of turbulence is velocity fluctuations (denoted by u) around a main (mean)
value (denoted by ).
U(x,t) = (x,t) + u(x,t)

(237)

The fluctuations are irregular (chaotic) in nature. Because of the irregularity it is impossible to
define the flow in all details as a function of time and space. But it is certainly possible to
describe the turbulence by law of probability. In turbulent flow various quantities show
random variations with respect to time and space, so that statistically distinct average values
can be discerned. The average values exist, because
- At a given point of the flow field a distinct flow pattern is repeated more or less regularly in
time
- At a given instant a distinct flow pattern is repeated more or less regularly in the flow field
This more or less regularity (quasi periodicity) imposes that a scale in time and space may be
defined for the turbulence. For turbulent flow through a pipe through a pipe, for instance, one
may a time scale proportional to the ratio between pipe diameter and flow velocity and a
length (space) scale proportional to the pipe diameter. Besides these average maximum
quasi periodicity there are many smaller quasi periodicities. In general, turbulence consists
of many superimposed quas-periodic motions. The motions are called quasi periodic,
since they are not permanent in time and space. This feature of turbulence is the irregular side
of turbulence. The dimension of the quasi periodic motions in space (eddies) has its limits,
because of the viscous dissipation. The smaller an eddy the greater is the velocity gradients in
the eddy and the greater is the viscous shear stress that dissipates the turbulence energy. This
mechanism causes the turbulence to become more homogenous in space; The intensity of the
1

turbulence shall decreases in time, if the source of the turbulence kinetic energy ( 2 u2) are not
enough to meet the dissipation rate. Thus, in each turbulent flow, there will be a statistical
lower limit to the size of the smallest eddy; and, there is a minimum scale of turbulence that
corresponds to a maximum frequency in the turbulent motion. Within the smallest eddy the
flow is no longer turbulent but laminar and molecular effects are dominant. The dimension of
the smallest eddy is about 1 mm for a moderate flow velocity of 100 m/s. This means that the
turbulent flow may be treated as laminar flow in the space of such a dimension2 . Velocity
fluctuations in the several places in the wake of a cylinder inserted into the flow in a wind
tunnel are shown in Fig. 47.
The average, both in time and space, of the fluctuations is zero ( = 0) by definition. However
there are some difficulties in defining the averaging process. This is because of the
dependence of turbulence on both time and spatial coordinates. In case of a homogenous
turbulence flow field averaging with respect to space can be used. If the turbulence flow field
is steady on the average (or quasi-steady), the time mean is defined by
1+ t
1
(x) = lim { t
(238)
U(x,t)dt }
1
t 
2

That may lead to an alternative solution method of the turbulent flow field using well known viscous flow
theory.

96

Figure 47. Oscillograms of turbulence in a wind tunnel and in the weak of a cylinder.

Here t must be long enough compared to the time scale of turbulence, which may be
considered as the main period of changes in flow pattern, so that enough number of
fluctuations can be involved in the averaging. On the other hand t must be small compared
with the period of any slow variation in the flow field that we do not wish to regard as
belonging to the turbulence. The average value should be independent of the time, provided t
< T2. That means /t must be either zero or negligibly small.
An ensemble average, including non-homogenous and instationary flows, may be defined by
n
Uk(x,t)
(x) = lim {
}
(239)
n

k=1
n 
where Uk(x,t) are instantaneous values measured. Instantaneous values are the measured
velocities at the same point and times during repeated (n times) measuring cycles. A typical
diagram indicating the fluctuations and the mean value is shown in Fig. 48.
If we deal with the Eulerian description of the flow field (stationary coordinates) we may use
one of the three averaging methods. However, it is more convenient to use Lagrangian
97

description (moving coordinate system) of the paths of the separate fluid particles, if we want
to deal with turbulent transport processes. In this case, if the flow field is homogenous on the
average, averaging may be carried out on the large number of the particles that have same
starting time but different origins. If the flow field is quasi-steady averaging is taken on the
particles with the same origin but different starting time.
Intensity of turbulence
Average of the absolute values of the fluctuation | u | may be chosen as a measure of the
turbulence. However, it is much more convenient to use root-mean-square value u =
2 (square root of the average of the fluctuations squares) as a measure of turbulence. The
u'
).
(relative) intensity (degree of turbulence, turbulence level) will then be defined by (

U
U(x,t)

U(x,t0)

(x,t)

Figure 48. Velocity fluctuations around a mean value in turbulent flow.


Equations of motion for turbulent flow; Reynolds stresses

Although the governing equations of the turbulent flow are well known Navier-Stokes
equations written for the Newtonian fluids, the dependent variables such as velocity, pressure
and density are not unique functions of time and space coordinates but average values
described by probability laws. The conservation equations of mass (175), written in tensor
notation,

+
t
xi (Ui) = 0

(240)

must also hold in turbulent flow at any instant and on average. The instantaneous value of the
velocity in a turbulent flow, for example, consists of the mean velocity and random turbulence
velocity components superimposed:
Ui = i + ui

98

The average value of the instantaneous velocity

= +u=+=

since = 0 by definiton.

Similarly for density

=+
and

= ( + u)( + ) = + u
The conservation equations of mass for turbulent flow is then

+
(i + ui) = 0
t
xi

(241)

It is seen that an additional term (ui) arise in the conservation equation for turbulent flow if
there is a correlation between the density fluctuations and the velocity fluctuations, i.e. the
product ui is not zero. Because of that the turbulence velocity components are random there
is always the possibility that a certain amount of mass is transferred, for example through the
surface dxdy of control volume dxdydz. in Fig. 34. with the velocity component v'. This mass
enters into a medium, which has an other momentum level because of the turbulence
component u' is not zero at that time. This change in momentum results in a force
proportional to the product u'.v' on the dxdy surface. A turbulence viscosity can then be
imagined due to the shear or normal stresses (called Reynolds stresses) as a result of the
momentum exchange between main flow and turbulence component. Reynolds suggested that
turbulence shear stress is proportional to the time average of the product u'.v'. This may be
seen explicitly if we apply the procedure above to the momentum equation written in tensor
form for incompressible, constant viscosity fluid.

Ui
Ui
P

Ui
+ Uj
=+
(
)
t
xj
xi xj xj

(242)

Introducing Ui = i + ui , = + and P = P + p gives

i
i
ui
ui
i
ui
ui
P
2i
t + j x + t + uj x + uj x + j x + uj x = - x + x x
j
j
j
j
j
i
j j
i
i
P
ui
i
ui
2i
ui
ui
( t + j x ) = - x + x x - t - uj x - uj x - j x - uj x
j
i
j j
j
j
j
j

(243)

or by applying the equation of continuity (241) and rearranging, we obtain


(

i
i
P
2 i

+ j
)=+
- ( u +
u u +
u +
u +
u u )
t
xj
xjxj t i xj i j
xi
xj j i xj i j xj i j

(244)

The effect of the turbulence on momentum equation is determined by the three correlations
ui, uiuj and uiuj.. As we have assumed the flow to be incompressible (up to M=0,2 the
effect of compressibility on density is about M2, which is less than %4) four of the five
turbulence terms can be neglected. The neglected terms are the terms that expresses the
normal stresses (pressures) due to the correlation of the turbulent fluctuations. The equation
(244) simplifies than to

i
i
P

i
( t + j x ) = - x + x ( x - uiuj)
j
i
j
j
99

(245)

The additional turbulence term uiuj (Reynolds stresses) and express the effect of the turbulent
momentum transfer, which causes additional shear and normal stresses. Normal Reynolds
stress (u2n) is obtained by putting i = j.
Turbulence stresses are much more greater than the laminar viscosity stresses

t = - uiuj

>>

Ui Uj
= l ( x + x )
j
i

According to Boussinesq, under certain conditions, a turbulence (or eddy) viscosity t can be
defined by which the turbulence stress may be expressed proportional to velocity gradients.

t = - uiuj = t (

i j
+
)
xj
xi

The effective viscosity is then e = t + l and the effective viscous stress is

i j
e = e ( x + x )
j
i

(246)

Of course, if we define such a scalar eddy viscosity, the normal Reynolds stress (u2n) can not
be related to the velocity gradient in direction ( i ) for incompressible flow, since the equation
ui ui = 2 t

i
xi

does not hold for compressible flow. A more correct procedure would be to introduce the
turbulence pressure
Pt = (u2n) =

1
ui ui (average over all n)
3

and to define the turbulent viscosity as follows


- ui uj +

1
i j

ui ui ij = t (
3
xj + xi )

where ij is the Kronecker delta.


The difficulties in handling with turbulent flows arise, in general, from the difficulties in
defining the correlation between turbulence velocities u, v and z. This correlation is somewhat
depend on the flow type and the intensity of turbulence.
A rather useful model for the calculation of turbulent viscosity is the k model. This semi
1

empiric model proposes the conservation of the turbulence kinetic energy k = 2 ui2 and the
dissipation . The turbulent viscosity is calculated by
k2
t = CD
with CD = 0,09.
The conservation equations for k and are

100

k
k

k
j
( t + j x ) = x (t x ) - uiuj x -
j
j
j
i
(247)
(

t
j
+ j
)=
(
) C1 uiuj
- C2
t
xj
xj xj
xi
k

(248)

where = 1.3, C1 = 1.44 and C2 = 1.9 in the standard k model.


To complete the set of equations conservation equations of mixture fraction and mixture
fraction variance must be included.

Z
Z

t Z
(
+ j
)=
(
)
Sc xj
t
xj
xj

(249)

Z'
Z'

t Z'
t Z'
( t + j x ) = x ( Sc x ) + 2 Sc ( x )2 -
j
j
j
j

(250)

The mean scalar dissipation is modeled as

= C k Z2
Here the Sc is the Schmidt number, which relates the turbulent diffusion of the mixture
fraction to viscosity. Sc and C are typically Sc = 0,7 and C = 2.0.
Turbulent length, time and velocity scales

A turbulent flow field may locally be characterized by the formerly described root-meansquare of the velocity fluctuations (u), turbulent macroscale (l) and the time scale (= l/u).
These may be related to k and with the expression given below:
u' = (2k/3)1/2, (: m2/s2 = m/s),

l = u3/,

( =

dk
: m2/s3) and
dt

= k/ (s)

In terms of kinematic viscosity and dissipation rate of turbulence kinetic energy


Kolmogporof length, time and velocity scales may be defined as follows:

= (3/ )1/4,

t = (/ )1/2,

v = ()1/4

According to Kolmogorof there is an energy transfer from large eddies, characterized by the
integral length scale (l), to smaller and smaller eddies. The dissipation rate, defined as the
energy transfer per unit turnover time, is then
u'2 u'3
= = l

(251)

The energy transfer procedure continues in a discrete sequence of eddies within the (inertial)
range
l = l/an > ,

n=1,2,.

101

where an is an arbitrary number (a>1). Since energy transfer rate between eddies is assumed
to be constant within the inertial eddy dimension range, dimensional analysis relates the
turnover times n, velocities un and eddy dimension, similar to (251), as
u'n2 u'n3
=
=
n
ln

(252)

This relation holds for the viscous dissipation scales (minimum eddy scale called Kolmogorof
length scale) too

u'2 u' 3
=

(253)

9.6.2. Regimes in turbulent combustion


Before going into the detail of turbulent combustion some basic definition will be given here:

Laminar flame speed; depends only on thermal and chemical properties of the mixture
Turbulent flame speed; depends on flow conditions as well as mixture properties
Flame surface is assumed to be represented by the time-mean value as instantaneous
position of the high temperature zone fluctuations
Flame velocity is determined from reactant flow rates measured

Regime criteria are:

Smallest scale of turbulence (Kolmogorov microscale); smallest eddies, which rotate


rapidly and have high vorticity, resulting in the dissipation of the fluid kinetic energy
into heat (fluid friction results in a temperature rise of the fluid)
Integral scale; largest eddy size
Basic structure of turbulent flame is governed by the relationship of integral scale and
Kolmogorof scale to laminar flame thickness
Laminar flame thickness, characterizes the reaction zone thickness which is controlled by
molecular (not turbulent) transport of heat and mass
In case of turbulence three flame regimes are possible:

Wrinkled and corrugated flames


When flame thickness is much smaller than smallest scale of turbulence, turbulent
motion can only wrinkle (distort) the thin laminar flame zone (this criterion is also
referred to as Williams-Klimov criterion)

Flamelets in eddies and distributed reactions


If all scales of turbulent motion are smaller than the reaction zone thickness, then
transport within the zone is no longer governed by molecular processes only, but is
controlled/influenced by turbulence also (this criterion is called Damkhler criterion)

Slow chemistry in well-stirred reactor


When reactions are slow compared to mixing rates

When chemical reaction rates are fast in comparison to fluid mixing rates (fast chemistry
regime) the first two regime, otherwise the third regime are valid.

102

The shape of turbulent flame front is somehow different than the laminar case (Fig. 49).

Figure 49. Flame borders in turbulent and laminar flow.

9.6.2.1. Premixed turbulent combustion


In analyzing the premixed turbulent flow combustion it is useful to define a reference
kinematic viscosity as a product of laminar burning velocity and flame thickness

ref = SL lf

(254)

With this viscosity the turbulent Reynolds number is


Rel =ul/ SL lf

(255)

turbulent Damkhler number as the ratio of characteristic flow time (turbulent time scale) to
characteristic reaction (flame) time
Dal = /tF = SLl/ ulf

(256)

and Karlovitz number, as the inverse of the Damkhler number defined with Kolmogorof time
scale
Ka = Da-1 = tF /t = l2F/2 = v2/SL2

(257)

The relations between the ratios u/SL (turbulence velocity to flame speed) and l/lF
(turbulence length scale to flame thickness) can be expressed in terms of Re, Da and Ka as
u/SL = Re (l/lF)-1
u/SL = Da-1 (l/lF)
u/SL = Ka2/3 (l/lF)1/3
Different regimes of premixed turbulent flame can be analyzed with these three number. The
regimes are shown in Fig. 48 (Borghi diagram). In this diagram coordinates are (u/uF ) and
(l/lF) in logarithmic sclae.
The regimes are:

Laminar flames regime for Re<1


Wrinkled and corrugated flames belong to the flamelet regime, which is characterized
by Re>1, turbulence, Da > 1 with fast chemistry and Ka < 1, with sufficiently weak
flame strech.

103

When flame thickness is much smaller than smallest scale of turbulence, turbulent
motion can only wrinkle (distort) the thin laminar flame zone (this criterion is also
referred to as Williams-Klimov criterion)(Fig. 50). Chemical reactions occur in thin
sheets at Da>1 and depending on Re the reaction sheet regime is characterized by fast
chemistry. Burning in spark ignited IC engine is this type. Typical values for engine
data are
Da 500,

Re 100 and u SL

Figure 50. Phase diagram showing different regimes in premixed turbulent flow

The wrinkled (laminar) flame regime of turbulent combustion can be assumed as


flamelets propagating with a velocity of a laminar flame. However, turbulence
wrinkles the flame resulting in an increase in flame area. The increase of the flame
front area increases the heat and mass transfer to the unburnt side also, increasing the
burning velocity. Therefore the ratio of turbulent to laminar flame speed will supposed
to be the ratio of wrinkled flamelet area (Afalmelet) to the time-mean flame area () ,
turbulent burning rate (Fig. 51).

m& = u A St = u A flamelets S L

St / S L = A flamelets / A

Figure 51. Flame front distortion due to the wrinkled flame regime

104

Wrinkled and corrugated regimes are separated by the equality of turbulence velocity
to flame velocity (SL = 1). In the wrinkled regime u<SL. This means, since u may be
interpreted as the turnover velocity of the large eddies, flame propagation is
dominating and flame displacement with SL is faster than displacement by turbulence
velocity u.
In case of corrugated regime flame speed is u > SL > v, according to (257). Since the
velocity of large eddies is greater than burning velocity, these eddies distort the flame
front. However, the smallest eddies, having a turnover velocity less than the burning
velocity, will not be able to wrinkle the flame front. Size of the eddy that interacts
locally with the flame front can be found by setting turnover velocity vn equal to the
burning velocity SL . This is the Gibson scale.
lG = S3L/

(258)

Figure 52. Turbulent flame speed depending on turbulence intensity

The Gibson scale is the size of the burnt pockets that move into the unburnt mixture.
This pockets are supposed to grow there due to the advance of the flame front of the
pocket, but are reduced in size again by newly arriving eddies of size lG. The unburnt
pockets, on the other hand, penetrate into burnt gas and consumed there by the flame
front advancment. Using (251) we may also write (258) in the form
lG / lt = (SL/u) 3

(259)

The mean thickness of the turbulent flame will increase with SL.

For Ka =1, the boundary to the distributed regimes, the flame thickness is equal to
Kolmogorof scale. Here the kinematic viscosity is still dominant in transport
processes; Flame time is equal Kolmogorof time and burning velocity is equal to
Kolmogorof velocity.

The distributed regimes is characterized by Re >1, Da>1, Ka>1. The last inequality
indicates that flame stretch is strong and smaller eddies can enter into the flame

105

structure, since < lF. Turbulence effect is dominant. However, small eddies produce
large strain rates that may lead to local extinction of the flamelet.

The well-stirred reactor regime is characterized by Re>1, Ka>1 but Da<1, which
indicates that the chemistry is slow compared with turbulent dissipating. A perfect
stirring is possible before chemical reactions occur and end.

9.6.2.2. Non-premixed turbulent combustion


In a non-homogenous mixture field the reaction zones are attached to the readily high
temperature regions close to the stoichiometric mixture and is advected and diffused with the
mixture field. Different from the premixed combustion, there is not a burning velocity, by
which the flame front moves relative to its previous position. Although there is time scale and
chemical reaction time there is no meaningful length scale such as flame thickness. Therefore
the velocity fluctuation u is not a relevant parameter in defining the regimes in non-premixed
combustion. Instead, since the mixture field fixes the flame position and the mixture fraction
fluctuations Z are responsible for corrugations of the flame surface, these fluctuations are be
considered in defining the different regimes.
In a non-homogenous turbulent mixture field reaction rate is large only where the mixture
fraction is at the stoichiometric value. For small mixture fraction variances, which corresponds to a
rather homogenous mixture, due to either intense mixing or partial premixing, a situation arises where
reactions are sufficiently fast everywhere and reaction zones are connected. If, on the other hand, the
mixture fraction fluctuations are large enough, diffusion flame structure may consist of separated
reaction zones. A criterion based on the flame thickness in the mixture fraction space (Z)F, as in the
case of laminar diffusion flame, is used to distinguish partially premixed and separated flamelet
regimes. In case of very large Damkhler numbers, when flamelet structure approaches local
equilibrium, the distinction between the two regimes becomes irrelevant since there is chemical
equilibrium.

Figure 53. Phase diagram showing different regimes in non-premixed turbulent flow
The second criterion is based on time scale and consider, as in premixed combustion, the ratio of
Kolmogorof scale to to the chemical time. Kolmogorof time (t = (/ )1/2) is the period of the smallest
eddies and therefore the shortest characteristic time of turbulence. If the combustion is fast with
respect to this time scale, then it can be consisered as quasi-steady and the diffusion flamelet concept
is valid in each eddy and for all mixture field. If however, the Kolmogorof time is of the same order or

106

shorter than the chemical time first the regions of intense mixing and then reaction will appear. This
regime is called distributed reaction zones regime (Fig. 53)

107

Anda mungkin juga menyukai