You are on page 1of 13

THERMAL CRACKING OF PROPANE

Kinetics and Product Distributions

A L F O N S G. B U E K E N S A N D G I L B E R T F. F R O M E N T
Rijksuniuersiteit Gent, J. Plateaustraat 22, Gent, Belgium

The thermal cracking of propane was studied in a flow apparatus between 625' and 850' C. and a t
atmospheric pressure. Fairly complete product distributions, including those for C4, C5, and Ca hydro-
carbons, were established and a reaction scheme was deduced. The order of the propane decomposition
was determiined by comparing experiments with different degrees of feed dilution and found to vary with
conversion and temperature. When the rate was fitted by means of a first-order kinetic expression, the
rate coefficient decreased with increasing conversion. This so-called "inhibition" was expressed mathe-
matically by considering the rate coefficient to b e a hyperbolic function of conversion. The activation energy
of the first-order rate coefficient increases with conversion. Finally, rate equations based on the radical
nature of the process are discussed.

HE main products of the thermal cracking of propane are


Tethylene and methane on one hand, propylene and hydro- Table 1. Literature Data
gen on the other. These products may be considered to be Activa-
formed by two parallel decomposition reactions of propane. tion
Energy, Temp.
Ethane, butenes, butadiene, and aromatics are also formed. Frequency Gal./ Range,
All these products are most important building blocks for the Authors Order Factor Mole C.
petrochemical industry, so that steam cracking of propane (or Paul and Marek
light naphtha) has become the major key process of this in- (1934) 1 3.98 Y 10'6 74,850 550-650
Engel et al. (1957) ... 71,000 500-590
dustry. The plant capacities are growing steadily. Plants Martin et al. (1964) 1 .i l l .3 67,000 545-600
producing 250,000 tons of ethylene per year are now in oper- Laidler et al. (1962) 2 . 5 8 ' X l O I 4 67,100 530-670
ation and bigger units are under way. The necessity of in- ti.5 8.50 X 10'3 54,500
Steacie and Pud-
creasing the ratio of ethylene to propylene in the cracker dington (1938) 1 2.88 X 10la 63,300 551-602
effluent has led to increased severity in cracking conditions. De Boodt (19621 ... ... 53.000 700-750
Kershenbaurn '
T h e design of such units requires precise kinetic equations and (1967) 1 2.40 X 10" 52,100 800-1000
detailed product distributions.
T h e literature on propane cracking is rather extensive, but
the majority of the published studies present a physicochemical
character, concerned with the nature of the elementary radical range encountered in industrial practice (650' to 825' C.) is
steps from which the over-all process is built up. They are striking.
generally carried out under static conditions, a t temperatures This work was undertaken to explain the inconsistencies and
below 650' C. and a t reduced pressure. contradictions found in the literature. Experimental data
Partial product distributions were published by Frey and were obtained in a flow apparatus for very wide ranges of
collaborators (1928, 1933) and Steacie and Puddington temperatures and residence times, covering those encountered
(1938). Kinney and Crowley (1954) were mainly concerned in industrial operation. From the product distribution a
with the formation of aromatics a t nearly complete propane reaction scheme is derived. T h e over-all kinetic equation is
conversion. Schutt (1 947) gives product distributions derived established. Finally, radical mechanisms are discussed in an
from industrial data. T h e most comprehensive product dis- attempt to explain some of the observed facts.
tributions were published in 1931 by Schneider and Frolich
(1931) with analytical techniques which bear no comparison Apparatus
with present-day mass spectrographic and gas chromatographic The apparatus is represented schematically in Figure 3.
techniques.
T h e rate of propane cracking is generally expressed as: T h e flow rate of propane is measured by the calibrated
capillary, C, and controlled by means of the valve, Kr. I n a
certain number of experiments the propane was diluted with
r = k exp( -j&)Cn nitrogen. The flow rate of nitrogen was indicated by the
flowmeter, r, and determined from the total flow rate measured
by the wet-gas flowmeter and the analysis of the exit gases.
Some of the most representative kinetic data are summarized in T h e propane and nitrogen are mixed in a packed tube, D,
Table I and represented in Figures 1 and 2. before entering the tubular reactor, the upper part of which
I n most studies the order was found to be 1, but Martin, serves as a preheater. Upon leaving the reactor the gases are
Dzierzynski, and Niclause (1964) and Laidler, Sagert, and rapidly cooled, K , and analyzed in an on-line gas chromato-
Wojciechowski (1962) found an order higher than 1. Table I graph. Their flow rate is measured in a wet-gas flowmeter, G .
T h e reactor tube, 163 cm. long, is made of chromium steel
shows considerable spread in activation energy and frequency (16% Cr, no nickel) and has an internal diameter of 4 mm.
factor. T h e absence of reliable kinetic data in the temperature Short bends permit the insertion a t different heights of four

VOL. 7 NO. 3 JULY 1 9 6 8 435


.
0

A
a
DE BOODT
PAUL
STEACIE
LAIDLER
9 KERSCHENBAUM

BUEKENS
x = 30%
x x = 60%
a x = 90%

log Ti [E :(set)-']

1000 900 '90 LOO

40 10.0 11,o lZ,O


'I0 $ (OK)-'

Figure 1 . Arrhenius diagram


Integral valuer of rate coefficient, i

very thin thermocouples, Philips 2ABI 10, external diameter


1 mm., so that the temperature of the gas stream itself is mea- Table II. Columns Used
sured. T h e geometry is such that the value measured is close Length, Tzmp., Products
to the mean cross-sectional temperature. T h e furnace is Column Material M. C. Anahred
cylindrical and has a length of 160 cm. and an outside diameter
of 36 cm. I t consists of two hinged half cylinders, made of
fireproof concrete and surrounded by a steel casing. There Silica gel 2 25 Hz, 5 3 3 4
are two sections: a preheat and a reaction section. The Carrier gas, NZ
preheater is a metallic cylinder insulated by a mica sheet and Dimethylsulfolane on Celite
wound with resistance wire. Its heating capacity is 3.5 kw. 545 (60/100 mesh) 12 40 From C1 to C4
T h e reaction section (heating capacity 6 kw.) holds 12 vertical Squalane on Chromosorb P 8 100 All HC up to
( 6 0 / 8 0 mesh) toluene
porcelain elements upon which resistance wire is wound a t a
variable speed. Six of these elements have their maximum
heating capacity on the top, the other six on the bottom.
Both types alternate along the cylinder wall. Both series of the cylinder. Variable transformers are built in the control
six elements are separately connected and controlled. TO circuits in order to attenuate the power input during operation.
compensate for heat losses at the exit a radiant resistance wire T h e temperatures are recorded by means of a Leeds &
with a heating capacity of 1.5 kw. is placed in the bottom of Northrup potentiometric recorder, Type G.

436 l&EC PROCESS DESIGN A N D DEVELOPMENT


x DEBOODT
T PAUL
a LAIDLER
o KERSCHENBAUM
A STEACIE

BUEKENS
8 KO
1:x. 30%
~ 60%
2 : :
3:x 8 90%

"C

99
1
Figure 2. Arrhenius diagram
Point values of rate coefficient, k

Analysis Propane flow range, 0.4 to 10 moles per hour, and thus a
residence time of 0.04 to 1 second and a Reynolds number of
T h e reactor effluent stream was analyzed on a Wilkens
8 to 200.
Moduline 202 gas chromatograph with hot wire detector. Dilution factor, 6 (molar ratio of nitrogen to propane),
T h e columns used are described in Table 11. 0 to 10.
The carrier gas was hydrogen, unless otherwise mentioned. The analysis of the propane was as follows: propane 98.9
The peak surfaces were measured by means of a mechanical mole %, ethane 0.3 mole %, propylene 0.2 mole %,2-methyl-
integrator. Calibration factors determined for nitrogen, propane 0.6 mole %, sulfur 10 p.p.m. Before a series of experi-
methane, ethane, ethylene, acetylene, propane, propylene, ments the reactor surface was deactivated with CS2 at tem-
and benzene were in excellent agreement with those published peratures between 450' and 550' C.
by Rosie and Grob (1957).
The material balance of the experiments checked within 1%.
Experimental Results The results are presented in two ways: in conversion us.
The experiments covered the following range of variables : V/F, diagrams and in product distribution or selectivity dia-
Temperature, 625' to 850' C. grams. The diagrams of the first type are more directed
Pressure, atmospheric; maximum pressure, 1.3 atm. toward a kinetic analysis, those of the second type toward a
absolute. study of the reaction mechanism.

VOL 7 NO. 3 JULY 1968 437


V(

Figure 3. Flow sheet of apparatus

Figure 4. Propane conversion vs. V / F ,

438 I&EC P R O C E S S D E S I G N A N D D E V E L O P M E N T
' 0 ° 1 800°C

s-
x

50,

25

Figure 5. Total conversion of propane and conversion to primary products as a function of V / F , at 800" C.

Figure 4 shows t.he x us. V/F, curves for temperatures


ranging from 625' to 850' C. Vis the equivalent reactor volume Table 111. Primary Product Distribution at Zero Conversion
discussed in detail below. x represents the total propane (Moles formed per 100 moles of propane cracked)
conversion. Figure 3 shows, by way of example, the conversion 675/ 7.251 775/ 825/
us. V/F, curves for propane and the principal reaction products 700°C. 750" C. 8GUOC. 850OC.
a t 800' C. Methane 47 47 49 48
T h e propylene curve goes through a maximum, resulting Ethylene 48 49 50 50
from the opposing effects of production and decomposition. Ethane 0.7 1.1 1.4 3.2
Hydrogen 52 52 51 53
Figures 6 and 7 show, by way of example, the product distri- Propylene 52 51 49 47
butions as a function of the total propane conversion for two
temperature ranges: 725' to 750' and 825' to 850' C. The
temperature dependence is not very pronounced.
+ propylene; at 800' C. 63 and 37%. Experimentally the
Product Distribution proportion is practically 50 to 50 a t 800' C.
I t follows from Table I11 that there is a slight trend in
Primary Products,. T h e primary products are methane, ethylene and propylene selectivity with temperature. These
ethylene, hydrogen, and propylene, which initially are formed trends may be explained by the Rice rules, according to which
in nearly equal molar quantities. Ethane is also partly a the formation of n-propyl radicals requires 1.2 cal. per mole
primary product, but it is produced in much smaller quantities. more than the formation of isopropyl radicals. Ethylene and
Table I11 gives the primary product distribution a t zero methyl radicals are the decomposition products of n-propyl
conversion for several temperature ranges. radicals; propylene and hydrogen radicals originate from iso-
T h e selectivity with respect to a given product is defined propyl radicals. T h e temperature dependence of ethane
below as the ratio of the number of moles of this product selectivity is more pronounced. This may be explained by the
formed to 100 moles of propane decomposed. T h e hydrogen growing importance of initiation and termination with
selectivity a t zero conversion was calculated from a hydrogen temperature.
balance. T h e initiation can be represented by
The classical Rice rules for radical reactions (1934) would
predict at 600' C. 6057, methane + ethylene and 40% hydrogen CaHs * CH3' + CzHs'
VOL. 7 NO. 3 JULY 1968 439
725-75Q.C
9 4

IO
Figure 6. Selectivity diagram for temperature range 725" to 750" C.

The ethyl radicals may then form ethane. methylacetylene, and propadiene (or allene) are dehydrogena-
One possible termination reaction is the recombination of tion products of ethylene and propylene. The single curve
two methyl radicals drawn for these products is to be regarded as a mean curve,
since the dehydrogenations are temperature-dependent. T h e
2CHsO + C2He rate of these reactions is lowered when the reactor wall is pre-
treated with CS2. At a conversion of 95% the selectivity for
which also produces ethane. acetylene amounts to 1.8%-the equilibrium value-without,
Figures 6 and 7 show how the selectivity for ethylene and but only to 0.6% with pretreatment. The activation energies
methane increases with increasing conversion, while that for are estimated to be 90,000, 80,000, and 70,000 cal. per mole
propylene and hydrogen decreases. T h e ethylene-propylene for acetylene, allene, and methylacetylene formation. T h e
molar ratio rises from 1.0 a t zero conversion to 3.7 a t 85% and temperature effect on the 1-butene formation is undeniable.
6 a t 95% conversion. The selectivities go through a maximum for a propane con-
I n the technical literature the product distribution is often version of 40 to 50%; below that conversion 1-butene is the
referred to the amount of ethylene produced. When propane most important secondary product. The decrease in selectivity
is cracked to 90% conversion under industrial conditions, 33 kg. beyond the conversion is probably due to isomerization into
of propylene are produced per 100 kg. of ethylene (Burke and 2-butene and 2-methylpropene and also to dehydrogenation
Miller, 1965). The figure obtained in this study is 36.6. into 1,3-butadiene.
For methane the figures are 61 and 62; for the secondary There are several possibilities for the formation of 1-butene.
products, 7 and 7 (1.3-butadiene) and 4 and 3 (butenes). One possibility is:
Secondary Products, T h e selectivities of the principal
secondary products are shown in Figure 8. Acetylene, 2C2H4 S CH&H2CH=CHz (1)

440 I&EC PROCESS DESIGN A N D DEVELOPMENT


801 *
825 850%
'
-

40

30 -

20 -

IO .

.
I
T . '.. T _c -- T T.
' T
.
T T
TT 'ZH6,

10 20 30 40 x59. 60 70 80 90 1 0

Figure 7. Selectivity diagram for temperature range 825" to 850" C.

Several objections may be formulated against this reaction traced back to the temperature dependence of the radical
as the principal source of 1-butene. Indeed, the conversion to concentrations. When compared at the same inhibition level,
1-butene is higher than the equilibrium conversion of Reaction these concentrations vary with temperature according to an
1. Furthermore, the selectivity for 1-butene increases with Arrhenius-type relation, so that an activation energy may be
temperature, whereas the equilibrium conversion decreases used to characterize their temperature dependence. From
with temperature. Finally, Reaction 1 was found by Krauze the simulation of the cracking of propane on a digital computer
et al. (1935) to have an activation energy of 37,700 cal. per a value of about 48,000 cal. per mole may be derived for the
mole, considerably less than that for total propane decom- "activation energy" of the methyl radical, while that for the
position. A comparison of selectivities at equal conversion for allyl radical must be between 20,000 and 30,000 cal. per mole
different temperatures indicates that the activation energy of (Buekens, 1967). I t follows that the apparent activation
1-butene formation is approximately 15,000 cal. per mole energy for 1-butene formation according to Reaction 2 must lie
higher than that of propane decomposition, which varies between 70,000 and 80,000 cal. per mole, in agreement with
between 52,000 and 64,000 cal. per mole. the experimental results.
For all these reasons, a more plausible way for the pro- As mentioned above, 1-butene isomerizes partially into
duction of 1-butene is the recombination between methyl and 2-methylpropene and 2-butene. The selectivity for 2-butene
allyl radicals: does not exceed 0.3%. 2-Methylpropene could not be
ki2
separated completely from I-butene by gas chromatography.
CH3" C3Hb0 --t C4H3 (2) For propane conversions of 90 to 95% the amount of 2-methyl-
propene was estimated to represent some 40% of the sum of
where the allyl radicals are formed according to: 1-butene and 2-methylpropene.
C3Hrj + R" -+ C3Hb" + RH (3)
The 1,3-butadiene selectivity, also represented in Figure 8,
was found to be practically independent of temperature. If
Reaction 2 requires no activation energy. Therefore, the 1,3-butadiene were formed by dehydrogenation of 1-butene,
temperature dependence of the 1-butene formation has to be the selectivity would decrease at high propane conversions,

VOL. 7 NO. 3 JULY 1968 441


2.6 .

2.4 .
s
2.0 -
*
t
2
t-
1.6 -
u
W
-I
W
* 1.2 -

0.8.

~ H c / / CH

Figure 8. Selectivity diagram for secondary products

whereas the opposite is observed. I t is therefore likely that 2 CH3' * CzHs


1,3-butadiene is mainly formed according to:

2C2H4 + CHFCH-CH=CHZ + Hz
CzHs'
CH3'
++ RC3H5'
H
- 7
I-butene
J.
-
+ 2-butene +
1
1,3-butadiene +
Hz
2-methylpropene
a conclusion set forward by Schneider and Frolich (1931). C2&' + C3H5' I-pentene

-{
-+
The Csand Cg hydrocarbons are mainly dienes and aromatics
and are probably formed by condensation reactions of ethylene C3H70+ C3H50 4-methyl-1-pentene
l-hexene
and propylene. The following products were identified :
pentadiene, isoprene, methylpentadienes, cyclic dienes such as
cyclopentadiene and probably cyclohexadiene; olefins such as Kinetic Study
I-pentene, 1-hexene, and 4-methyl-I-pentene; cyclic olefins Equivalent Reactor Volume. One of the main problems
such as cyclopentene and cyclohexene; and aromatics such as encountered in the derivation of rate equations for homogene-
benzene and toluene. ous gas-phase reactions from experiments in tubular flow
The formation of aromatics increases rapidly a t high con- reactors is the longitudinal temperature profile. Indeed, it
versions. I t is strongly dependent on wall effects. is not possible to distinguish sharply between preheat and
Reaction Scheme. To conclude this discussion of the reaction sections as can be done in fixed-bed catalytic reactors.
product distribution, the following reaction scheme may be If the rate is to be determined a t a reference temperature,
set u p : say TR, and if the reaction volume is counted from the point
Hz + CzHz where TR is reached, an error is introduced by neglecting the

CHI
7
+ CzH4
\
-
7
'/z(Hz -t G H s ) +
conversion accomplished in the section below TR. A similar
situation occurs at the outlet, where the conversion continues
to some extent at a temperature lower than TR. To
C3HS 4 H z + CsHs 3 correct for this and to compare experimental data at the same
reference temperature TR,use can be made of the equivalent
L /
Hz + C3Hs * '/2(Hz + Ce"o) reactor volume concept introduced by Hougen and Watson
(1947). The equivalent reactor volume, V, is defined as that
volume which, at reference temperature TR, would give the
same conversion as the actual reactor, with its temperature
profile. I t follows that
T h e reactions of this scheme probably proceed over radicals
almost entirely. An additional set of reactions is required to TT
dV = - dV' (4)
explain the formation of ethane and olefins: ~ T R

442 I&EC PROCESS DESIGN A N D DEVELOPMENT


and
Table IV. Variation of Order
LvLxp( - A ) d V t 625' 650' 675' 700- 775' 825'
V =
c. c. C. 750" C. C. C.

exp( -")
RT R
Conversion
Order
0.022
1.36
0.058
1.21
0.067
1.14
. . . 0.238 0.412
1 0.99 0.96

T h e usefulness of this concept was discussed and demonstrated


by Froment, Pijcke, and Goethals in their study of the thermal
used in this work is based upon the integral method of analysis.
cracking of acetone (1961). T o be practical its utilization
The formulas used to calculate & for orders of 0.5, 1, 1.5, and 2
requires a good evaluation of the activation energy, E, prior to
are given in the Appendix. I n these equations the expansion
the knowledge of the rate coefficient, k.
factor is taken to be 2 over the complete range of conversion, as
T h e experimental data of the present work were corrected by
observed experimentally. Figure 9 illustrates how the correct
means for this concept, using a n estimated value of 50,000 cal.
order is obtained from the intersection of two curves related to
per mole for the activation energy. This value was sufficiently
the experiments being compared.
close to the final value obtained from an Arrhenius diagram to
The semilogarithmic representation used leads to a sharper
make iteration superfluous. I n some cases the true volume was
intersection. Table I V shows the mean values of the order a t
reduced by a factor of 2.
several temperatures.
Determination of O r d e r a n d Rate Coefficient, I n this
Table IVdoes not make it possible to conclude unambiguously
work it was attempted, as is usual in the study of cracking
whether the order varies with conversion, with temperature,
reactions, to describe the rate by a simple equation like
or with both. Indeed, the conversions used to determine the
r = kCn (5) order at the higher temperatures are much higher than those
a t the lower temperatures. I t has been shown, however, that
When the differential method of analysis is applied, the rate is the order decreases mainly with increasing conversion-Le.,
obtained as the slope of the tangent to the curves of x us. V/Fo. with increasing inhibition (Buekens 1967).
When the integral method of analysis is used, Rate Equation 5 First-Order Kinetics. I n the preceding section the order
has to be substituted in the continuity equation of the reaction, although not constant, was found to be close
F,dx = rdV to 1. With a fixed value of 1 for the order it becomes possible
to determine a n activation energy and a frequency factor for
and the resulting equation integrated. T h e data points x us. propane decomposition. When the first-order & values, deter-
V/F, are then used to calculate k. mined by the integral method, are plotted in a n Arrhenius
I t was found using both methods of analysis that the rate diagram (Figure l ) , a definite trend with respect to conver-
coefficient decreases with increasing conversion, a t least for sion is observed. T h e following activation energies and
orders between 1 and 1.5. Using the terminology of the frequency factors are obtained:
cracking literature the reaction is "inhibited by its reaction
products or by intermediate species." Another way to put it is 30% conversion
that Rate Equation 5 does not adequately describe the rate of
the complex cracking phenomenon. Yet, in this section, the
kinetic treatment is based upon Equation 5 , whose simplicity
130 = 2.33 X 10l2exp( -T) 58,000
(second-')

is attractive for practical purposes. Moreover, this treatment 60% conversion


leads to considerable insight into more complex mechanisms.
T h e variation of the rate coefficient, k, defined by Reaction 5 ,
with conversion necessitates a distinction between point and
= 8.62 X lo1* exp (--6yio) (second-')

integral values. T h e differential method leads to point values; 90% conversion


the integral method to integral values, E.
I t also follows that the correct order of the reaction can be
obtained only by comparison of data having the same level of
= 1.64 X 1013exp (--':io) (second-')

inhibition, which is mainly determined by the relative rates of This shift in activation energy may explain the spread of the
reaction of the chain-carrying radicals with propane and values reported in the literature, probably obtained a t different
propylene. For this reason k/k,, the ratio of the point value levels of inhibition.
of the rate coefficient a t conversion x to the point value at zero I t is of importance to present point values of the rate co-
conversion, may be written: efficient, too. Indeed, most of the k values reported in the
k k(R") (C3Hd literature are point values. Furthermore, if a relation is to be
- k(Ro)(C3Hd + k"(Ro)(C3Hd
- E
established between the over-all rate coefficient and the inti-
k,
mate nature of the reaction, point values are of more direct
Since the selectivity for propylene is almost independent of interest than integral values. Equation 6 suggests a hyper-
the pressure, the extent of inhibition is in first approximation bolic law for the variation of point values k, with conversion:
determined by the conversion, so that Equation 6 is obtained:
k0
k=-
1 + ax (7)

In Equation 7, a is a n empirical inhibition coefficient and k, is


the value of k a t zero conversion and therefore a rather artificial
Therefore, the determination of the order requires compari- quantity. Indeed, k should vary with conversion in the same
son of experiments carried out a t different ratios of nitrogen to way as the radical concentration. The latter goes through a
propane but leading to the same conversion. T h e procedure maximum: I t increases as the reaction starts, but decreases

VOL. 7 NO. 3 J U L Y 1 9 6 8 443


+:

+I

-1

-2

i ORDER
1/2 1 I20 3/2
Figure 9. Determination of reaction order
tR = 675OC. V / F , = 7.86 x = 0.0253, dilution factor 6 = 0
V / F o = 13.50 x = 0.0259, dilution factor 6 = 6.43

afterward, as a result of the formation of allyl radicals. Such


a maximum was observed experimentally a t the very low con- Table V. k,, a VI. Temperature
versions which were measured in the temperature range 625’ Temg ., kQ,
to 675’ C. The hyperbolic law of Equation 7 does not con- O c. Sec. -1 a

sider the drop in k as zero conversion is approached, of course. 625 0.0548 7.345
650 0.1819 11.720
T h e “initial value,’’ k,, and the inhibition coefficient, a, the 675 0,3298 3.696
constants of the empirical hyperbolic law describing the in- 700 0.9452 6.505
hibition, were determined in the following way. 725 1 ,7800 5.896
750 2.2140 2.371
When the expansion factor equaIs 2, the combination of 775 4,0760 1.128
Equation 7 with the continuity and rate equations leads to: 800 9.3580 1.799
825 13.9500 1.375
-dx- - kOCt(1 - x )
d-
v (1 + ax)(l + x)
(8)

FO
With an equation of this form k, and a may be calculated
Upon integration and rearrangement, Equation 8 gives: from a set of experimental x us. C,V/Fo values by linear re-
gression. T h e results are shown in Table v.
V
C,k, -= -a 2 In (1 - x) + 2x + - - 2 In (1 - + x
X2
2) Figure 10 shows a k us. V/Fo or x curve and compares experi-
FO 2 mental and calculated E values at 775’ C. The calculated E
(9) values were obtained from:

444 l & E C P R O C E S S DESIGN AND DEVELOPMENT


I rV/Fo. .v 7 7 5%
k =-vJ , k d -
FLl
FO
and the k values were related to V/Fo through Equations 7 and
9.
I t follows from Figure 10 that the hyperbolic law (Equation
7 ) permits a n excellent description of the inhibition.
I t is seen from Table V that the values of a are less in line a t
the lower temperatures of 625' to 675' C. a is very sensitive
to the spread on the data. T h e maximum in the k us. V/Fo
curves also influences the results. From an Arrhenius diagram,
the following temperature dependence is derived for a:

a = 3.01 X lO5exp
(2:3
___

If a were exactly equal to the ratio kN/k', the temperature


dependence would be less pronounced and amount to no more
than 3000 cal. per mole. T h e value of 23,100 may be ex-
plained by the role played by the allyl radical, C ~ H S ' ,which Figure 10. Comparison of experimental & values with
becomes more effective as a chain carrier a t higher temperature. those calculated from point values
T h e following temperature dependence is found, by linear
regression, for k,:

k, = 4.10 X 10" exp (- 'y:)


- (second-')
I n 1934 Rice proposed the following sequence to describe the
thermal cracking of propane under initial conditions:

I t now becomes possible to calculate point k values a t


different conversions-e.g., 30, 60, and 90%. These values
are plotted in a n Arrhenius diagram, as shown in Figure 2 .
T h e following activation energies and frequency factors are
found by linear regression:

30y0 conversion:

k30 = 5.10 X 1OI2 exp (- 'y;)


~ (second-')

6OYc conversion :

k60 = 1.98 x loi3 exp (- 6y:o)


__ (second-')
With this system of reactions the rate of propane decompo-
sition may be written:

90% conversion:

kgo = 3.84 x 1013 exp (- 6y:)


~ (second-')

I.'
T h e point values are lower than the integral values, of where 0 = C1 -
course, while the activation energies are higher, as may be FO
foreseen from the trend observed in Figure 1 .
The inaccessible radical concentrations may be eliminated
The activation energies given here are lower than those
from this expression with the help of the so-called steady-state
usually reported. No effect of heat transfer limitations may be
approximation. According to this concept, introduced by
involved: T h e thermocouples are inserted in the gas stream
Bodenstein, the radical concentration does not vary rapidly
itself, not on the wall. However, De Boodt (1962) obtained a
beyond the initial startup period. A balance on the pro-
value of 53,300 cal. per mole in a flow apparatus of different
duction and consumption of CH3' and H' then leads to ex-
construction than the one used in this study, while Kershen-
pressions for the concentrations of these radicals as a function of
baum and Martin (1967) came to an expression for the first-
the propane concentration. After substitution of these ex-
order rate coefficient in the temperature range 800' to 1000' C.
pressions into Equation 10 the following rate law, which is of
(Table I) which is in remarkable agreement with the equation
first-order with respect to propane, is obtained, provided no
given here for k,.
distinction is made between n- and isopropyl radicals and
Radical Mechanisms. T h e over-all order of propane de-
composition depends upon the conversion and temperature.
ki << k3.

Values ranging from 1.3 to 1 were obtained. This clearly


indicates that a simple power law like Equation 5 is not ade-
quate for the description of the rate of propane decomposition
under widely varying conditions. A shift in reaction order, as experienced in this work, has
More satisfactory equations have to be based upon a more been reported by Laidler, Sagert, and Wojciechowski (1962).
detailed consideration of the nature of the reaction. According to these authors the order 1.5 they observed would

VOL. 7 NO. 3 J U L Y 1 9 6 8 445


be due to a bimolecular initiation and a termolecular termi- work which would not rely upon the steady-state approxi-
nation involving two CHaOor two H o and a third body, while mation. This requires the integration of a set of ordinary
order 1 would be due to bimolecular initiation and termolecu- differential equations describing the evolution of the different
lar termination involving CH3" and C3H7' or HO, CsH7', and species in the reactor. Such a computer simulation of the
a third body. There appears to be little justification for such thermal cracking of propane is being investigated a t present
hypotheses. Indeed, initiation has always been considered in our laboratory.
to be unimolecular. I t is also widely accepted that combin-
ations of two CH3' or of CH3' and C3H7O do not require a Appendix. Calculation of Integral Rate Coefficient for
third body. There is, however, no need to resort to such
Different Orders and E 2. =
hypotheses to explain the shift in the order: I t suffices to add a 1
n = 0.5.2 =
second termination reaction like V
-
FO
dZ
+ (Arc sin 2x
m +-
6j
to the scheme proposed by Rice to come to a rate equation
which is much more complex than Equation 11 and which
leads to a n order varying bettween 1.5 and 1.O. Termination Arc sin -)]
6
according to Reaction 7 must be preponderant over that 2+6
proposed by Rice, because very little butane is formed in the
thermal cracking of propane.
The Rice scheme, eventually completed with termination
Reaction 7 , may give an excellent description of the initial order
and product distribution, but it does not account for the for- n = 1.5. =
mation of secondary products, nor for the inhibition. From
the discussion of the product distribution it appears that the
allyl radical plays an important part in the cracking of propane.
I2(2
1
+ 6) 1 d"f"
I-?.
L '-
- 1 - dl+s
'. 4
1+
This radical may be formed by hydrogen abstraction from 1 [ d(1- x ) ( l + 6 + x ) - d i x ]-
2% + 6
propylene according to:
( 3 + :)[Arc
C3He + Ho 8 -+ + HP
C3H5' ( sin 2+6 - Arc sin
T I
~

2+6

CsH6 + CH3' 5 C3Hb0 + CH4


I n turn allyl radicals may abstract hydrogen from propane 1
according to: n=2. E = -V 1 - %
- cz
+ C3H6' 2 CsH7' + C3H6
C3Hs
If allyl radicals were incapable of H-abstraction, Reactions 8
and 9 would have to be considered as terminations. Ter-
Fo x + 2(2 + 6) In (1 - x )
1
minations sensu stricto involving allyl radicals could be : Ac knowledgment
One of the authors (A.G.B.) is grateful to 1.W.O.N.L.-
I.R.S.I.A. for a research grant over the period 1965 to 1967 and
C3H5O + CH3O C4H8 to the Robert De Keyser Foundation of the Belgian Shell Co.
T h e experimental and computational assistance of P. Martin
13
2C3H5' + products and R. Van Slembrouck is also acknowledged.
Abstractions 8 and 9 and terminations 11, 12, and 13 would
be responsible for the inhibition. Nomenclature
A classical steady-state treatment applied to the Rice scheme a inhibition coefficient, dimensionless
leads to the rate equation: c concentration, moles/cc.
ct total concentration, moles/ cc.
r =
ks(C3Ha)
2kik3(CsHs)'
+ ~ ~ C I ~ ( CX ~8kikdCaHs)
H B ) ~
+ 2ki(C3Hs) E
FO
activation energy, cal./mole
molar feed rate of propane moles/sec.
k rate coefficient, sec. -1 cu. m.(n-') mole'-n
provided CH30 is the only chain-carrying radical, Reaction 7 rG rate coefficient, as determined by integral method
ko initial rate coefficient
the only termination, and Reaction 9 the only inhibition. rate coefficient for dehydrogenation of ethylene into
ko
This equation leads to a variable order that becomes one at acetylene
high conversions, when kg(C3H6) predominates. Under these k' rate coefficient for abstraction on propane
conditions the activation energy would be E = El + -E3 E9 k" rate coefficient for abstraction on propylene
= 82,000 + 8500 - 7700 = 82,800 cal. per mole, much higher n
r
order of reaction
rate of reaction, kmole/cu. meter/sec.
than the experimental values. This may throw some doubt TT rate of reaction, a t temperature T
upon the proposed reaction scheme, but it is much more likely ~ T R rate of reaction at reference temperature TE
that the application of the steady-state approximation is not R gas constant, cal. mole-' O C.-l
justified in the case of an inhibited system, which is character- (RO) = concentration of Ro radicals, mole/cu. meter
T = temperature, O C. or O K .
ized by a decrease in concentration of active radicals, as a result T E = reference temperature, O C. or O K.
of their reaction with the products. I t therefore seems neces- x = conversion, dimensionless
sary to attempt a quantitative explanation of the results of this v =equivalent reactor volume, cu. meters

446 l & E C PROCESS D E S I G N A N D DEVELOPMENT


V' = actual reactor volume, cu. meters Kershenbaum, L. S., Martin, J. J., A.Z.Ch.E. J . 13, 148 (1967).
e = expansion factor Kinney, C. R., Crowley, D. J., Znd. Eng. Chem. 46,258 (1954).
6 = dilution factor Krauze, M. V., Nemtzow, M. S., Soskina, E. S., J . Gen. Chem.
= C, V / F , reciprocal space velocity, sec. USSR 5 , 343 (1935).
6 Laidler, K. J., Sagert, N. H., Wojciechowski, B. TY., Proc. Roy.
q = selectivity, % Sac. A270, 242 (1962).
Martin, R., Dzierzynski, M., Niclause, M., J . Chim. Phys. 61,
literature Cited 286 (1964).
Messner, A. E., Rosie, D. M., Argabright, P. A., Anal. Chem. 31,
Buekens, A. G., Ph. D. thesis, Rijksuniversiteit Gent, 1967. 230 (1959).
Burks, P. D., Miller, R., Chem. Week 18, 64 (1965). Paul, R. E., Marek, L. F., Znd. Eng. Chem. 26,454 (1934).
De Boodt, H., Ingeniersthesis, Rijksuniversiteit Gent, 1962. Rice, F. O., Herzfeld, K. F., J . A m . Chem. Soc. 56, 284 (1934).
Engel, J., Combe, A , , Letort, M., Niclause, M., Compt. Rend. Rosie, D. M., Grob, R. L., Anal. Chem. 29, 1263 (1957).
244. 453 (1957). Schneider, V., Frolich, P. K., Znd. Eng. Chem. 23, 1405 (1931).
Frey, b. E., Heppj H. J., Znd. Eng. Chem. 25,441 (1933). Schutt, H. C., Chem. Eng. Progr. 43, 103 (1947).
Frey, F. E., Smith, D. F., Znd. Eng. Chem. 20, 948 (1928). Steacie, E. W. R., Puddington, T. E., Can. J . Res. B 16, 411
Froment. G. F., Piicke, H., Goethals, G., Chem. En,<. Sci. 13, 173, (1938).
180 (1961).
Hougen, 0. A., TVatson, K. M., "Chemical Process Principles," RECEIVED
for review October 19, 1967
Vol. 111, TViley, New York, 1947. ACCEPTEDJanuary 26, 1968

REMOVAL OF NITROGEN FROM ARGON


W I T H TITANIUM-METAL SPONGE
M A R T I N L. K Y L E , R . D E A N P I E R C E ,
L E S i T E R F . C O L E M A N , A N D J O H N D. A R N T Z E N
Chemical Engineering Diuision, Argonne National Laboratory, Argonne, Ill. 60439

The rate of reaction of titanium-metal sponge with nitrogen in argon-nitrogen gas mixtures was studied
a t 900" C. The reaction rate is dependent on the partial pressure of nitrogen in the gas phase. At least
three titanium-nitrogen solid phases are formed as the reaction proceeds, and the rate-controlling mech-
anism is believed to b e the diffusion rate of atomic nitrogen through the TiN,(G) phase. A single relation
has been developed which describes the titanium-nitrogen reaction kinetics of argon-nitrogen mixtures.
Additional mathematical relationships were developed to permit estimation of the required size and useful
lifetimes of titanium-sponge gettering beds designed to remove a nitrogen impurity from otherwise pure
argon.

increasing utilization of large inert-gas enclosures has made of the kinetics of the titanium sponge-nitrogen reaction
Tied to interest in purification systems capable of removing
HE
with nitrogen present in otherwise pure argon. A reaction
oxygen, nitrogen, and water vapor contaminants from large temperature of 900" C . was chosen for this study asacompromise
volumes of inert gas, particularly helium and argon. Chemical between good reaction kinetics and ease of containment of the
gettering methods are often employed to maintain a pure titanium sponge. Previous work (Wasilewski and Kehl,
atmosphere. I n a system of this type, a n activated sorbent 1954) showed that the titanium-nitrogen reaction rate increases
such as molecular sieves (product of the Linde Division, Union with temperature from 750" to 1400' C. Holvever. the exist-
Carbide Corp.) can be used to remove water vapor, and a rela- ence of a titanium-nickel liquid phase (Hansen and Anderko,
tively low-temperature gettering material (such as manganous 1958) above about 955°C. prevents containment of titanium
oxide a t 150" C.) can be used to remove oxygen. T h e removal sponge in stainless steels or other nickel alloys a t the higher
of nitrogen presents a greater problem because of its lower temperatures
chemical reactivity.
Interest in the use of titanium as a nitrogen-gettering material Equipment
has stemmed from several advantages that titanium possesses A diagram of the experimental system is shown in Figure 1.
over gettering materials now commonly used. Titanium-metal An argon-nitrogen mixture of carefully controlled composition
sponge, a high-surface-area form suitable for gettering opera- was continuously circulated through a small bed of titanium
sponge. This bed was composed of eight removable sections
tions, is available a t a relatively low cost and requires no pre- hereafter called sample beds. Each sample bed consisted of
treatment before use. Neither the oxide nor nitride reaction 1 to 10 grams of -8 +IO-mesh titanium sponge (see Table I )
products of titanium are pyrophoric, and they can be handled contained in a 2-inch 0.d. by 0.5-inch high circular tray, the
in air. Both titanium and its reaction products have low bottom of which was constructed of two layers of 70-mesh
stainless steel (Type 310) screen. Figure 2 shows the titanium
toxicity. sponge in a sample bed. Lava (aluminum silicate) spacers
T o investigate the possible utilization of titanium as a get- were used between the individual sample beds to facilitate their
tering material for nitrogen impurities in argon, a study was removal.

VOL. 7 NO. 3 JULY 1968 447

You might also like