Ind. Eng. Chem. Res.

2010, 49, 1991–1994 1991

Pretreatments of Coils to Minimize Coke Formation in Ethylene Furnaces

Zhaobin Zhang
Beijing Research Institute of Chemical Industry, SINOPEC, Beijing 100013, China

Lyle F. Albright*
Forney Hall of Chemical Engineering, Purdue UniVersity, West Lafayette, Indiana 47907

Numerous techniques have been considered on how to reduce coke formation and/or coke collection on the
inner surfaces of the coils in furnaces producing ethylene, propylene, etc. Such reduction would lead to several
beneficial events including much reduced production costs. The techniques tested by Nova Chemical in several
of their furnaces have resulted in a major reduction of coke. They have tested several pretreatments of the
coils in their furnaces using several oxidizing gases at 800-1200 °C. For example, hydrogen/steam mixtures
have been employed. Various pretreatments have been employed in the current study that suggest improved
pretreatments are now possible.

Introduction In 1984, Horseley and Cairns6 publicized a technique for

All ethylene furnaces until about 10 years ago had to be
decoked after 30-60 days to remove the coke that collected in
the coil. Such coke has several adverse effects. First yields of
ethylene and other desired hydrocarbons are reduced. Second
resistances to heat transfer from the hot combustion gases to
the hydrocarbon gases being cracked increase. Hence, energy
demands for operating the furnaces increase. Third, resistances
to the gas flow in the coils increases as the thickness of the
coke increases. Fourth when the furnaces are decoked, produc-
tion of the desired products is stopped for significant time
periods.
Three different coking mechanisms have been identified in
the coils and transfer line exchanges of the ethylene furnaces.1,2
Iron and nickel catalyze the formation of carbon (or coke)
filaments, via mechanism 1. These filaments promote coke
growth via mechanisms 2 and 3. These filaments act as
collection sets for tiny tar droplets suspended in the gas phase
(mechanism 2). These droplets then decompose to form coke
and release hydrogen. In mechanism 3, the free radicals on the
surfaces of the filaments react with hydrocarbon free radicals
in the gas phase. As a result, the filaments grow in diameter
and to some extent in height.
Albright3 reported an example of the compositional changes
with time (and use) on the inner surfaces of coil of an industrial
furnace. The chromium content had increased sometimes by a
factor as high as two. Meanwhile the iron and nickel contents
at the surface had decreased substantially. At that time, no
attempt was made to measure the manganese concentration; only
later was it discovered that this concentration had increased
several fold. Obviously, the compositions of the inner surfaces
of the coils are of significant importance since the rates of coke
formation and/or collection are considerably higher in new coils.
The following coils have been found to have reduced coke
formation in industrial furnaces: an aluminized surface produced
by the Alon Processing Company4 and a siliconized surface
developed by the British Petroleum Company.5 Unfortunately,
the longevities of both coils were relatively short so neither was
judged suitable for industrial applications.
* To whom correspondence should be addressed. E-mail:
albright@purdue.edu.
10.1021/ie900271q  2010 American Chemical Society
Published on Web 01/15/2010
pretreating stainless steels in the presence of a gaseous mixture
of hydrogen and steam at high temperatures. As a result, the
surfaces of the steel were enriched in Cr and Mn, but depleted
in Fe and Ni. They indicated that oxides of Cr and Mn formed
at and near the metal surface. Such oxides formed Cr-Mn
spinels. Simultaneously Fe and Ni atoms migrated inwardly
forming sublayers enriched with these elements. The findings
suggested such pretreatments might be beneficial to coils to be
employed in ethylene furnaces. Szechy, Luan, and Albright7
and Luan8 confirmed that such pretreatments were promising
for ethylene furnaces. Shell Chemical Company9 also confirmed
the potential benefits.
Nova Chemical Company has tested pretreated coils in 41
commercial ethylene furnaces.10,11 With a newly treated coil,
the time that the furnace can be operated before decoking often
varied from 420 to 505 days. Shorter times occurred being about
150-200 days after the second through fourth decokings. As a
comparison with nonpretreated coils, the times before decoking
were about 40 days. Nova has obtained a significant number of
patents on pretreated coils. The exact pretreating procedure that
they employ is unknown, but it seems to be similar to ones
investigated at Purdue University. Nova indicates in their
publications that spinels having the composition of MnxCr3-xO4
are desired on the surface layer. Simultaneously Fe and Ni atoms
migrate inward forming a sublayer enriched in these two
elements.
Analysis of Surface Layers of Pretreated Stainless Steels
The calculation procedure developed by Luan et al.12 was used
to determine the amounts of Cr, Mn, Fe, and Ni transferred to or
from the surface of a pretreated steel. Energy dispersive X-ray
analyzer (EDAX) readings at 10-35 kV levels were employed to
determine first the average weight percent of each metal element
for several layers between 0 and about 3 µm below the surface.
Then a curve was drawn for that element to predict the composition
as a function of depth. The EDAX results sometimes indicate that
minor amounts of Si, Al, and Ti are sometimes present. To help
in drawing the curve for a specific element, the concentration of
these four to seven elements should add to 100% at any specific
depth below the surface.
The amount (A) of each element transferred to or from the
surface layer can be calculated with the following equation:
1992 Ind. Eng. Chem. Res., Vol. 49, No. 4, 2010

∫ (F)(% metal)(d micrometers)

x
A) 0

Where, x is the thickness in micrometers of the enriched or

depleted surface layer, F is the initial density of the original
stainless steel expressed in grams per cubic centimeter, and %
metal is the change of the weight percent at a given depth.
Graphical integration of the curves constructed is used to
calculate values of A for each metal element for the pretreated
surface. The actual grams transferred per unit surface area (cm2)
can be determined as follows:

g/cm2 ) A/(F × 106)

In making these calculations, two assumptions are made. First

the distance that the X-rays penetrate below the surface for a
given kilovolt value does not change. Yet it is known that Cr Figure 1. Amounts of Cr, Fe, and Ni transferred at temperatures from 760
and Mn oxides (and spinels) form near the surface. Second, the to 1000 °C for Incoloy 800: used H2/H2O ) 50 for 4 h pretreatments.
densities of the surface layers do not change. Although both
assumptions are questionable, comparison of A values as
calculated are thought to be relatively quantitative.
Luan et al.12 earlier reported the results for numerous
pretreatments of Incoloy 800. In the present investigation,
considerably more results are reportedsfor additional stainless
steels, for pretreatment times varying from 0.25 to 24 h, and
for several different pretreating gases. For several pretreated
surfaces, coke failed to adhere to the surface during subsequent
coking experiments. This finding suggests that such a pretreat-
ment might result in a much improved coil in an ethylene
furnace.

Pretreatments of Different High-Alloy Steels

Figures 1 and 2 show comparisons of the calculated value of
A for pretreated Incoloy 800 and HK 40. The pretreatments were
made using a 50:1 molar mixture of H2/H2O at four temperatures
Figure 2. Amounts of Cr, Mn, Fe, and Ni transferred at temperatures from
varying from 750 to 1000 °C with a pretreatment time of 4 h. 760 to 1000 °C for HK 40: used H2/H2O ) 50 for 4 h pretreatments.
For Incoloy 800, considerably more Cr, Fe, and Ni was
transferred at a given temperaturesoften by factors of 3 or more. Table 1. Composition of Stainless Steels Prior to Pretreatments
For this stainless steel, the relative amounts of Mn transferred elements
were lower, in the 1.1-1.8 range. The A values for Incoloy
stainless steel Al Si Ti Cr Mn Fe Ni
800 are considered more accurate since five kilovolt values were
available for the calculations as compared to four with HK 40. Incoloy 800 0.7 2.0 0.4 22.8 0.2 42.7 30.3
Even though the A values for Incoloy 800 are considerably HK 40 0.2 4.7 25.2 0.9 48.6 18.6
HP 40 0.2 4.1 26.7 1.5 33.9 32.8
higher, the compositions of the two stainless steels were fairly
similar prior to pretreatments, as shown in Table 1. of the enrichment and depletion zones for a given pretreatment
The results with Incoloy 800 at 760, 850, 950, and 1000 °C differ somewhat.
indicate that plots of ln A verses 1/T(K) are essentially straight For Mn enrichment at and near the surface, obviously most
lines for each of the four metal elements, i.e. Arrhenius-type of the Mn diffuses from relatively large depths. For Incoloy
plots occur. The results with HK 40 although less reliable also 800, the Mn concentration near the surface sometimes was as
can be represented by similar straight line correlations having high as 30% whereas the concentration in the untreated steel
essentially identical slopes as compared to Incoloy 800. was about 0.7%. In such a case, the Mn concentration increased
The thicknesses of the enrichment layers for Cr and Mn were by a factor of over 40. It is estimated that part of the Mn had
also calculated using the EDAX results. These layers for Cr diffused from depths of 75-100 µm.
and Mn are defined as those having higher concentrations of A preliminary estimate can be made of the amounts of metals
Cr or Mn than were present in the nontreated steel. These transferred to and from the surfaces of the stainless steel, HP
thicknesses increased in each case as the amounts of Cr and 40; a H2/H2O ) 50 mixture was used at 950 °C. With this
Mn transferred increased. For Incoloy 800, the thickness for stainless steel, only three kilovolt values were employed,
Cr was about 3.5 µm for the pretreatment at 1000 °C. The permitting analyses at depths up to only about 2 µm. For such
thickness of the enrichment layer was found to be essentially pretreatments, the amounts of metals transferred with HP 40
directly proportional to the amount of the metal transferred. are approximately equal to those for Incoloy 800.
Calculating the thicknesses of the depletion layer near the
surface for Fe and Ni is considered to be less accurate as
Times of Pretreatment
compared to the thicknesses of the enrichment layer. These
thicknesses of the depletion layer are however also probably Figure 3 and 4 indicate the amounts of Cr, Mn, Fe, and Ni
directly proportional to the amounts transferred. The thickness transferred in Incoloy 800 and HK 40 respectively as the time
 Ind. Eng. Chem. Res., Vol. 49, No. 4, 2010 1993
Table 2. Transfer Amounts Using Several Gases for Pretreatments
of Incoloy 800 at 950 °C for 4 h
pretreating gas Cr Mn Fe Ni
air 120 12 72 57
CO2 117 11 73 57
CO 76 11 52 39
CO2/H2O ) 50 105 13 75 41
H2/H2O ) 50 58 13 35 29

Table 3. Transfer Amounts Using Several Gases for Pretreatments

of HK 40 at 950°C for 12 h
pretreating gas Cr Mn Fe Ni
CO 27 6 20 10
H2/H2O ) 3 39 10 33 15
H2/H2O ) 50 72 14 49 23

a given time of pretreatment varied in the following decreasing

Figure 3. Amounts of Cr, Mn, Fe, and Ni transferred in Incoloy 800 at
pretreatment times up to 24 h; pretreating gas used H2/H2O ) 50, at 950
°C.
order: Cr, Fe, Ni, and Mn.
These pretreatments provide new information on the phe-
nomena occurring at and near the surfaces of the steels. Initially,
the transfers of Cr and Mn toward the surface are through
relatively pure metal atoms. Both of these transfers are appar-
ently relatively rapid. Meanwhile, the oxide layer formsschromium
oxides are initially predominant building up to concentrations
of 50-70%. These oxides are both heavier and physically larger
in size. Transfers of elemental metals in this layer are apparently
slower. Eventually the Cr concentration builds up to about 70%
and then decreases as the manganese concentration increases
to 30%. Elemental Mn is apparently diffusing through the oxide
layer. At the surface, the oxides of chromium and manganese
react to form spinels.10,11 Diffusion of oxygen from the
pretreating gases to sublayers below the surfaces of the steels
is apparently of little or no importance.

Different Pretreating Gases

Figure 4. Amounts of Cr, Mn, Fe, and Ni transferred in HK 40 at
pretreatment times up to 24 h; pretreating gas used H2/H2O ) 50, at 950
°C.
of pretreatment increased to 24 h. These pretreatments were
with a 50:1 molar ratio of hydrogen to steam at 950 °C. Very
large changes of the metal composition occur at and near the
surface in the first hour. For the remainder of the run, the rates
of metals transferred were much slower and relatively constant.
For these experiments, much larger amounts of metals trans-
ferred with Incoloy 800. The amounts of metal transferred for
Comparative runs were made to determine the amounts of
metal transferred in Incoloy 800 using five different pretreating
gases at 950 °C for 4 h. The amounts transferred sometimes
differed substantially as indicated in Table 2. For Cr, Fe, and
Ni, the amounts transferred varied by a factor of over 2 for air
(having the most transfer) and H2/H2O ) 50 (having the least).
Yet the A of Mn transferred varied only between 11 and 13 for
the five gases (air, CO2, CO, CO/H2O ) 50, and H2/H2O )
50). Air obviously contained the largest concentration of oxygen
or oxygen free radicals. The similar transfer of Mn can be
explained in part at least by the much greater distances that it
is transferred in Incoloy 800. Such transfer is apparently a rate-
controlling step.
Three gases were used to pretreat the stainless steel HK 40
at 950 °C for 12 h. As indicated in Table 3, the largest transfers
of all four metals occurred with H2/H2O ) 50 whereas the
smallest transfers occurred with CO. H2/H2O ) 3 resulted in
intermediate transfer amounts. Yet, larger transfer amounts
occurred with CO pretreatments of Incoloy 800, as shown in
Table 2.
Coking Experiments
Several pretreated coupons were first weighed and then
positioned in the laboratory tubular furnace for coking experi-
ments. There is a 3:1 molar ratio of ethane and steam was passed
over each coupon for 1.5 h at 850 °C causing some coke to
collect on the coupon. Various coupons were tested as follows:
(a) Each coupon was weighed after coking. The weight gain
after coking was considered to be the weight of the coke
1994 Ind. Eng. Chem. Res., Vol. 49, No. 4, 2010
collected on the coupon. Knowing the dimensions of the coupon,
the amount of coke collected per squared centimeter of the
coupon was calculated. The coupon was handled carefully to
ensure that no coke was lost during handling.
(b) The coupon was then brushed with a soft nylon brush.
Sometimes 80-100% of the coke was removed due to brushing.
Visual observations suggest that more coke deposited on the
top side of the coupon as compared to the bottom. Scanning
electron microscopy (SEM) photographs indicated that globular
coke particles were prevalent on the top side. Apparently tar
droplets suspended in the gas phase collected primarily on the
top side. These droplets after being deposited on the top surface
then decomposed forming globular coke deposits. Apparently
more coke on a relative basis could be brushed from the top
side as compared to the bottom. During brushing, some metal
particles can spall from the surface, but for the coupons
examined, spalling was likely of minor importance.
(c) Several coupons were further tested. After brushing, the
coupon was returned to the furnace and contacted with a 3:1
molar mixture of H2/H2O at 900 °C for 2 h. Such a pretreatment
acted to further decoke the coupon as indicated by a decrease
in weight of the coupon.
(d) For several coupons, coking followed decoking was
repeated. For one coupon, this sequence was repeated four times.
In this case, the fraction of coke that could be removed by
brushing increased to about 100%.
These coking tests indicated several promising results. First,
the pretreatments of the coupon decreased the amount of coke
deposited on the surface. Second the fraction of the coke that
was removed by brushing increased toward 100% as the coupon
was repeatedly coked then decoked. Third pretreating the coupon
with CO at 850 °C for 24 h resulted in essentially no adherence
of coke to the coupon. The lack of adherence suggests that
filamentons coke formation is small at most. Such a finding is
expected when Fe and Ni concentrations on the surfaces to the
steel are small. The above results suggest that in industrial
furnaces coke collection on the surfaces of the coil may be
negligible with the proper pretreatments. The high velocity gases
in industrial coils may remove all the coke from the surfaces.
The resulting entrained coke or coke precursor would be
transferred to and through the transfer line exchangers to the
following scrubbing tower.
Conclusions
Although the current results are promising, additional research
is still needed to determine the most promising pretreatment
technique. There is still the need to determine why the rates of
metal transfer during a pretreatment sometimes vary so much
for different high alloy steels. The size and character of the
metal grains in the steel may be a factor. Determining the
preferred composition of the Cr-Mn spinels is also still needed.
More data are yet needed in the importance of so-called trace
metals (Ti, Si, Al, etc.) in the stainless steel. Finally, the current
results suggest that improved decoking procedures can be
developed that will minimize coke formation and/or collection
after decoking.
Literature Cited
(1) Baker, R. T. K.; Harris, F. S. Chemistry and Physics of Carbon;
Walker, P.L., Thrower, P.A., Eds.; Marcel Dekker, Inc.: New York, 1978;
Vol. 14, Chapter 1.
(2) Albright, L. F.; Marek, J. C. Mechanistic Model for Formation of
Coke in Pyrolysis Units Producing Ethylene. Ind. Eng. Chem. Res. 1988,
27, 755.
(3) Albright, L. F. Metal Diffusion from Furnace Tubes Depends on
Location. Oil Gas J. 1988, (Aug 15), 69–75.
(4) Albright, L. F.; McGill, W. A. Aluminized Ethylene Furnaces Tubes
Extend Operating Life. Oil Gas J. 1987, (Aug 31), 46.
(5) Brown, D. E.; Clark, J. T. K.; Foster, A. J.; McCarroll, J. J.; Sims,
M. L. Inhibition of Coke Formation in Ethylene Steam Cracking. In Coke
Formation on Metal Surfaces; Albright, L.F., Baker, R.T.K., Eds.; ACS
Symposium Series, no. 202; American Chemical Society: Washington, DC,
1982; Chapter 2, p 23.
(6) Horsely, G. W.; Cairns, J. The Inhibition of Carbon Depositor on
Stainless Steels by Prior Selective Oxidation. Appl. Surf. Sci. 1984, 18,
273–286.
(7) Szechy, G.; Luan, T. C.; Albright, L. F. Pretreatment of High Alloy
Steels to Minimize Coking on Ethylene Furnace. In NoVel Production
Methods for Ethylene, Light Hydrocarbons, and Aromatics; Albright, L.F.,
Crynes, B.L., Nowak, S., Eds.; Marcel Dekker: New York, 1992; Chapter
18, pp 341-359.
(8) Luan, T. C. Reduction of Coke Deposition in Ethylene Furnaces.
Ph.D. Thesis, Purdue University, West Lafayette, IN, August 1993.
(9) John, R. G.; Ngan, D. Y. Heat Treatment of High-Temperature Steels.
U.K. Patent Application 2,234,530A, June 28, 1990.
(10) Gyorffy, M.; Benum, L.; Sakamoto, N. Increased Run Length and
Furnace Performance with Kobota and Nova Chemicals ANK 40 Anticoking
Technology. AIChE 18th Ethylene Producers’ Conference, AIChE Spring
Meeting, Orlando, FL, April 23-27, 2006.
(11) Saunders, R.; Gyorffy, M. ANK 400 Anticoking Technlogy.
Seminar, Houston, TX, Nov 30, 2006.
(12) Luan, T. C.; Eckert, R. E.; Albright, L. F. Gaseous Pretreatment
of High Alloy Steels Used in Ethylene Furnaces: Pretreatment of Incoloy
800. Ind. Eng. Chem. Res. 2003, 42, 4741–4747.
ReceiVed for reView February 18, 2009
ReVised manuscript receiVed September 25, 2009
Accepted November 24, 2009
IE900271Q