Ind. Eng. Chem. Res.

2009, 48, 2801–2809 2801

Influence of EDC Cracking Severity on the Marginal Costs of Vinyl Chloride

Production
Reinhard Schirmeister,*,† John Kahsnitz,‡ and Michael Träger†
VESTOLIT GmbH, P.O. Box 102360, D-45753 Marl, Germany, and EVonic Degussa GmbH,
D-45753 Marl, Germany

The selectivity of the EDC (1,2-dichloroethane) cracking process in vinyl chloride monomer (VCM)
manufacturing strongly depends on the cracking severity and the EDC feed quality. A strong decline of
selectivity is observed for EDC feed purities below 99.5 wt % or cracking rates above approximately 60%.
The selectivity of the cracking process also deteriorates in the course of the plant’s operating time as a result
of accelerated coke formation. Reduced selectivity in the process of EDC cracking results in increased marginal
costs (additional costs per each additional tonne of VCM). This is especially relevant for a VCM plant running
near full load. The marginal costs can be expressed as a function of the cracking rate and the resulting costs
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

for processing the uncracked portion of EDC. With a new and simplified EDC cracking simulation model for
the energy input and conversion, together with a simplified reaction model for the definition of the corresponding
product spectrum, the cost impact of the aforementioned influences on the cracking process can be estimated
in advance based on the current production rate and plant performance.
Downloaded via UNIV DE LIEGE on October 24, 2024 at 09:32:01 (UTC).

1. VCM Manufacturing Process These deposits are visible on the walls of the ends of cracker
tubes, both on the connecting tubes outside the radiation zone
To manufacture vinyl chloride monomers (VCMs) economi- and in the subsequent heat recovery section. Lower cracking
cally through the thermal cracking of 1,2-dichloroethane (EDC), rates improve the yield but also result in lower VCM production,
hydrogen chloride (HCl) produced in the cracker must undergo higher specific energy costs, and higher specific losses of raw
further processing, normally by oxychlorination to yield EDC. material (amounts of residues and waste gas). Higher specific
Under practical conditions, the cracking process is run with a energy costs are due to higher residual amounts of EDC that
residence time of 20-30 s in the gaseous phase at 480-520 require reprocessing steps.
°C. It is typically conducted until an EDC cracking rate of about Experience shows that regular downtimes lasting over several
55-63% is reached in order to avoid a strong increase of days are inevitable after every 1-2 years of operation for
byproducts. The VCM yield after the cracking furnace is at about cleaning of the cracker. To avoid expensive storage of inter-
98-99% based on cracked EDC. Volatile and high-boiling mediate products, it is important for the entire production line,
byproducts are subsequently removed from the residual amount including all associated units, that these downtimes are planned
of unconverted EDC as much as possible by distillation. The reliably in advance. The undesirable effects already mentioned
latter is recycled back to the cracker along with the EDC from
other process steps (direct chlorination and oxychlorination Table 1. Compounds, Radical Species, and Short Forms Used in the
units), resulting in a typical EDC feed purity of >99.5 wt %. Reaction Model in Figure 2 and Table 2
The overall selectivity of the entire VCM manufacturing process no. compound short form
including reprocessing is hence even lower.
In an ethylene-based integrated balanced process (see ref 1, 1 1,2-dichloroethane EDC
p 290) such as that used in VESTOLIT’s production plants 2 vinylchlorid VCM
3 hydrogen chloride HCl
(Figure 1), HCl production in the cracker and HCl consumption 4 trichloromethane CHCl3
in other production units must be balanced. The “EDC mix” in 5 tetrachloromethane CCl4
Figure 1 represents the fresh feed to the crackers, resulting from 6 ethylchloride EC
EDC from direct chlorination, EDC from oxychlorination, and 7 1,1-dichloroethane 1,1
8 1,1,2-trichloroethane 1,1,2
EDC from reprocessing.
9 1,1,1,2-/1,1,2,2-tetrachloroethane 1,1,1,2
The byproduct formation in the EDC cracking process 10 1,1-/cis-/trans-dichloroethylene Di
essentially determines the raw materials lost during distillation 11 trichloroethylene Tri
product reprocessing and, hence, the VCM yield and the 12 1-/2-chloroprene CP
resulting VCM manufacturing costs. A high EDC throughput 13 acetylene C2H2
14 benzene C6H6
at a concurrently high cracking rate can result in a high VCM 15 3,4-dichlorobutene C4H6Cl2
production; however, it also increases the amounts of undesired 16 soot/coke C
byproducts. These are, in particular, chlorinated C1, C2, C4, and Radical
C6 hydrocarbons and unsaturated hydrocarbons such as acetylene
1 Cl• R1
and benzene. The availability of the plant is strongly reduced 2 CH2ClsCH2•/CH3sCHCl• R2
in the course of such a procedure as a result of coke deposits. 3 CH2ClsCHCl• R3
4 CHCl2sCH2• R4
* To whom correspondence should be addressed. E-mail: 5 CHCldCH•/CH2dCCl• R5
reinhard.schirmeister@vestolit.de. 6 CH2ClsCCl2•/CHCl2sCHCl• R6
†
VESTOLIT GmbH. 7 CHCl2sCCl2•/CCl3sCHCl• R7
‡
Evonic Degussa GmbH. 8 CCl3• R8

10.1021/ie8006903 CCC: $40.75  2009 American Chemical Society

Published on Web 02/13/2009
2802 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009

Figure 1. EDC/VCM production at VESTOLIT.

can counteract this schedule with a dramatic increase toward 2.1. Reaction Model. Concerning the endothermic reaction
the end of the operating time.
It is practically impossible to monitor changes in cracked 1,2-EDC f VCM + HCl ∆HoR ) 71 kJ/mol ) 0.717 kJ/kg
products by means of online analytics (e.g., gas chromatography (1)
or mass spectroscopy) at the outlet of the cracker coils because
of the extreme conditions encountered there (approximately 12 the conversion inside the tubular reactor depends on the local
bar pressure and 500 °C). Solids (soot and coke) and the energy input by means of heat radiation on the heating side
condensation of high-boiling substances do not permit a correct and local heat transfer. The local energy input is consumed for
withdrawal of samples. For this reason, a VCM production plant heating and for the reactions. For the calculation of the EDC
was modeled with all of the information from the upstream and conversion along the nonisothermal and nonisobaric tubular
downstream process and all available analytics. Simplified reactor, additional information about the axial temperature
kinetic/reactor models were designed to calculate the effects of profiles of the cracker coil is needed.
changes in the overall performance rate. Depending on the cracking conditions (feed, conversion,
heating) and operating time, these temperature profiles change
with locally accumulating soot/coke layers, which reduce the
2. Kinetics and Cracking Model
local heat-transfer coefficient. As a consequence, a rising
The radical chain mechanism of thermal cracking is well- temperature level on the heating side compensates for the related
known in principle. Experimental studies and rigorous heat flux reduction. Otherwise, the total conversion would
mathematical models of the complex EDC cracking reaction decrease. This temperature is limited by material constraints of
can be found in the literature (e.g., refs 2-7). Up to 135 the tube material (e.g., Incoloy 800H with a working temperature
compounds and radical species and more than 800 reactions up to 700 °C).
have been reported for comprehensive calculations.4 In view To simulate the spectrum of byproducts analyzed in the
of the accuracy of the data and optimization of the expen- technical plant, a simplified reaction model was designed on
diture for model adjustments, simplifications must be made. the basis of the published radical chain mechanisms, including
In our model, this resulted in 31 reactions describing all data on bond energies, collision coefficients, and activation
relevant products, intermediates, and byproducts (cf. Tables energies of each partial reaction step.2-6,12–18 If no data were
1 and 2). Benzene formation and the very complex soot and available, analogous data of the same reaction type were applied
coke formation via acetylene, benzene, chloroprene, and for the parameters of the reaction rate constants.
polycyclic aromatics (cf. refs 8-11) are considered by Table 1 lists all 16 “stable” substances and 8 radicals and
reactions 30 and 31, which represent the summed pathways their brief descriptions used in this kinetic model (cf. Table 2).
of model chain reactions. We also found more than 15 The following assumptions were made for this model: (1) The
unknown high-boiling species in the GC analysis but reaction start by radicals via homolytic CsCl bond cleavage
neglected them in our model because they amounted to less from EDC and CCl4 (reactions 1 and 2). (2) The chain reactions
than 0.1 wt % of the total mass flow. The component 3,4- proceeds (a) by H abstraction from the stable compound to form
dichlorobutene was used as a key component for high boilers a new radical (reactions 3-9, 11, and 19-23); (b) by decay of
with close boiling points. radicals, forming Cl• and unsaturated stable compounds (reac-
 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009 2803
Table 2. Kinetic Model Reactions Kinetic Data (Ahrrenius Law Expression Constants from the Literature and Estimated)
frequency factor ko
no. reaction [(cm3/mol)n-1 s-1] n activation energy Ea (kJ/mol) ref
1 EDC f R1 + R2 5.9 × 10 15
1 342 3, 5
2 CCl4 f R1 + R8 2.2 × 1012 1 230 12

3 EDC + R1 f HCl + R3 1.3 × 1013 2 7 4, 5

4 EDC + R5 f VCM + R3 1.2 × 1013 2 34 4
5 EDC + R2 f EC + R3 1.0 × 1012 2 42 4
6 EDC + R4 f 1,1 + R3 5.0 × 1011 2 45 est
7 EDC + R6 f 1,1,2 + R3 2.0 × 1011 2 48 est
8 EDC + R7 f 1,1,1,2 + R3 1.0 × 1011 2 56 est
9 EDC + R8 f CHCl3 +R3 1.0 × 1012 2 63 4
10 VCM + R1 f R4 9.1 × 1010 2 0 5
11 VCM + R1 f HCl + R5 1.2 × 1014 2 56 13
12 VCM + R5 f CP + R1 5.0 × 1011 2 31 4
13 VCM + R4 f C4H6Cl2 + R1 2.0 × 1010 2 30 est
14 VCM + R2 f EC + R5 3.0 × 1011 2 61 4

15 R3 T VCM + R1 2.1 × 1014 1 84 5, 14

16 R5 T C2H2 + R1 5.0 × 1014 1 90 est
17 R6 T Di +R1 2.0 × 1013 1 70 est
18 R7 T Tri + R1 2.5 × 1013 1 70 15

19 EC + R1 f HCl + R2 1.7 × 1013 2 4 16

20 1,1 + R1 f HCl + R4 1.2 × 1013 2 6 16
21 1,1,2 + R1 f HCl + R6 1.7 × 1013 2 15 17
22 1,1,1,2 + R1 f HCl + R7 1.7 × 1013 2 17 17
23 CHCl3 + R1 f HCl + R8 1.6 × 1013 2 14 18

24 CCl4 + R5 f Di + R8 5.0 × 1011 2 33 est

25 CCl4 + R4 f 112 + R8 1.0 × 1012 2 33 4
26 CCl4 + R6 T 1112 + R8 5.0 × 1011 2 33 est

27 R2 + R1 f VCM + HCl 1.0 × 1013 2 13 4

28 R3 + R1 f Di + HCl 1.0 × 1013 2 12 est
29 R6 + R8 f Di + CCl4 1.0 × 1013 2 13 est

30 2C2H2 + R5 f C6H6 + R1 1.0 × 1014 2 20 est

31 C2H2 +2 R1 f 2C + 2 HCl 1.6 × 1014 2 70 4

tions 15-18 and final step of reaction 30); (c) by addition of relevant below 500 °C if the initial CCl4 feed concentration is
Cl• (back-reaction of radical decay) or other radicals to double >100 ppm or if in situ formation of CCl4 (for example, from
bonds (reactions 10 and 12-14 and initial step of reaction 30); CHCl3) occurs. The role of CCl4 in the EDC feed as a promoter
and (d) by Cl• abstraction from CCl4 (reactions 24-26). (3) has been described in the literature.4,19 The role of CCl4 in our
Chain terminations and decreases of the total radical concentra- reaction model (chain reactions 24-26) is similar to that of Cl2
tion are considered to occur (a) by H abstraction from a radical, in chlorination models. The bond dissociation energies for the
forming two stable compounds (reactions 27-29) and (b) by two molecules are nearly the same as well.
carbon formation (soot/coke) from acetylene (reaction 31). The main reaction pathways from EDC to the desired VCM
Not considered in our model are (1) HCl split-off reaction are H-abstraction reactions 3-9 followed by decay of radical
from chlorinated hydrocarbon in a monomolecular reaction R3 in reaction 15. With decreasing EDC and increasing VCM
[which is unlikely to occur concurrently with fast chain reactions concentration the formation of radical R5 becomes relevant,
because of its high activation energy (>200 kJ/mol)], (2) catalytic which can yield the desired VCM (reaction 4) or unwanted
wall reactions, and (3) complex synthetic pathways from chloroprene and acetylene (reactions 12 and 16, respectively).
possible carbon precursors such as chloroprene and benzene via For the byproducts acetylene, chloroprene, and 1,1-dichloroet-
polycyclic aromatic hydrocarbons (PAHs). hane, a high net formation is observed. The 1,1,2-trichloroethane
In Table 2, the partial reactions and applied kinetic parameters in the EDC feed is found to be mostly consumed.
of the reaction rate constants are given. In the cases of benzene In the present model, a decrease of the total radical
formation (reaction 30) and complex carbon formation (reaction concentration (“chain termination”) is considered only by
31), the parameters of the rate-controlling steps are used. formation of carbon (reaction 30) and H abstraction between
Figure 2 visualizes these reaction pathways to all relevant radicals as in reactions 27-29. Normally, chain termination by
“stable” components found in the feed or product streams after radical recombination needs a solid wall or a third molecule
cracking. Byproducts showing a high change in concentration (as a third collision partner) to absorb the high energy released
are shown in bold. The radicals are represented in ellipses. by recombination. The surface-to-volume ratio in laboratory
Reaction 1 (Cl dissociation from EDC) and reaction 2 (CCl4 studies5,19 is normally much higher than that in technical plants,
dissociation) are the initiation reactions. At 500 °C, the rate or the pressure is low. In our case, with higher pressure, we
constant for reaction 2 is approximately 104 times higher than assume that those wall collisions are negligible relative to
the rate constant for reaction 1. However, with CCl4 from bimolecular collisions in the gas and that terminations on the
reprocessing, the initial CCl4 concentration in the EDC feed wall can be neglected. This assumption is also based on the
stream can give a rate of radical formation that is of the same observation that no significant amounts of C3 compounds from
order of magnitude as that of reaction 1. Reaction 2 becomes recombination of C1 and C2 compounds are found in the
2804 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009

Figure 2. EDC cracking reaction model (measured compounds from large-scale plant with a significant increase or decrease shown in bold).

cracked gas of our technical plant. This is in agreement with the overall process in technical plants. Apart from confidential-
exit gas composition data from Borsa.4 ity, one reason for this dearth of information might be that the
2.2. Coke Formation. The formation of solid carbon par- energy input varies among VCM producers as a result of the
ticles transported by the fluid (so-called “soft coke”) and coke various construction details of the cracking and convection
layers (so-called “hard coke”) and the possible role of tube zones. Moreover, the different numbers or arrangements of
material or iron chloride are discussed in detail elsewhere.9-11,20 heating gas burners can be crucial.
In agreement with other studies,4 we observed that, in technical Both of our EDC crackers consist of a cracking furnace with
plants, deposition of coke is not uniform. two coils with a length of approximately 330 m, each with
One important factor for local coke formation is the net flux horizontal straight segments of approximately 15 m. In the
of local radiation energy and the resulting temperature profile. furnace, there are five horizontal burner lines mounted on each
Another factor is the local concentration of coke precursors in vertical furnace wall plane to the staggered coils (cf. Figure 3).
the reaction mixture. To date, local hot spots have not been A constant ratio of air to fuel in the feed to the burner lines
observed in our plant. However, the heat input through the ensures a uniform adiabatic combustion temperature. The heat
burners needs to be adjusted to optimize the yield and minimize produced by the burner lines decreases stepwise from the EDC
the coke formation. inlet to the cracked-gas outlet.
In both cracking units, we found coke deposits on the wall Our model calculation starts with an overall heat balance of
of each coil outlet close to the radiation zone after 1 year in the complete furnace considering heat recovery from flue gas
operation. Typical deposits in the coil outlet showed more or and stack losses. The total heat input to the coils by radiation
less uniform layers of about 1-5 mm. Visible coke deposits of (subscript rad) and convection (subscript conv) is equal to the
1-5-mm thickness were also on the walls of the insulated heat produced by combustion of the fuel minus the heat
transfer pipes connecting each coil with its heat recovery (two (Qfurnace gas) leaving the cracking zone (compartments 1 and 2
heat exchangers in line) from cracked gas. The strongest in Figure 3) with the flue gas
deposition (>10-mm thickness) was observed at the inlet of the
heat recovery of cracking unit 2. Coke particles were found in Qrad total+Qconv total ) ṁfuelHcomb - Qfurnace gas (2)
the condensed raw product as well.
Deposits on the coils and especially in the heat recovery
∫
Tfurnace gas
Qfurnace gas ) (ṁfuel + ṁair) Cp,flue gas(T) dT (3)
sections cause an increasing pressure loss over time and higher Tgas
external wall temperatures. This pressure loss finally limits the
EDC throughput. where ṁfuel and ṁair are the total mass flows (kg/s) of heating
2.3. Reactor and Furnace Model. Similar design principles gas and air, respectively; Hcomb is the combustion heat of the
of noncatalytic crackers are found in the patent literature (e.g. fuel (kJ/kg); Cp,flue gas is the specific heat of the flue gas [kJ/(kg
ref 343 in our ref 1) and undisclosed company reports. Apart K)]; Tfurnace gas is the temperature of the flue gas leaving the
from the more recent, very complex models of the heating cracking zone (°C); and Tair is the temperature of the air feed
side,22,23 only empirical information is available with regard to to the burners (°C).
 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009 2805
where εF and εP are the dimensionless emissivities of the furnace
wall and pipe wall, respectively, and Ap is the cross-sectional area
of the pipe wall (m2).
The temperature, TF of the emitting furnace wall was about
780 °C. The outer wall temperatures, TPo j, in the coil sections
were calculated in the simulations.
As shown in Figure 4, at steady state, the local heat input on
the coil is equal to the enthalpy consumed for both heating of
the reaction fluid and conversion in the endothermic crack
reaction. Tgas represents the mean gas temperature in the cross
section of the pipe instead of a more realistic radial temperature
profile at turbulent flow. The local heat balance for a coil section
j with volume Vj gives

Qrad j+Qconv j)Qcond j)Qreac j+Qheating j (5)

Qcond j ) λp /spAp(TPo - TPi)j

) λcoke /scoke jAcoke(TPi - TWi)j

) ∑ (-∆H ) r V R k k j + ṁEDC feedCp,gas j(Tgas j - Tgas j-1)

k
(6)

where Qcond j is the heat transferred through the metallic wall

and coke layer (kW); Qreac j is the heat consumed in chemical
reactions (kW); Qheating j is the heat consumed for heating the
Figure 3. Vertical cut through the furnace and the coils. reaction mixture (kW); TPo and TPi are the temperatures (°C) of
the outer and inner walls, respectively, of the metallic pipe; TWi
The total heat input into the coils was about 60% of the heat is the inner wall temperature (°C); λp is the heat conductivity
produced by combustion. A typical value of Tfurnace gas was about of the Incoloy wall [16.3-24.7 W/(m K) at 300-800 °C]; λcoke
700 °C. is the heat conductivity of the coke [i.e., amorphous carbon,
To calculate the local radiation heat transferred on the coils, a 1.6 W/(m K)]; Sp and Scoke are the thicknesses (mm) of the pipe
heat flux profile in the furnace is required that considers the given wall and coke layer, respectively; Ap and Acoke are the areas
geometric conditions, e.g., similar to calculations of the heat flux (m2) of heat transfer surfaces in the coil section; rk is the rate
profiles in ethylene crackers.23 Then, the net local radiation energy, of reaction k [mol/(m3/s)]; ∆HR is the reaction enthalpy (kJ/
Qrad j, absorbed on a coil section can be calculated according to ref mol); mEDC feed is the mass flow of EDC feed (kg/h); Cp,gas j is
24, Chapter K and with the normalized heat flux profile ψj and σS the mean specific heat of the reaction mixture in section j [kJ/
) 5.67 × 10-8 W/(m2 K4) by (kg/K)]; and Tgas j is the mean temperature of the reaction mixture
in section j (°C).
Qrad j ) ψjεFεpσSAp(TF4 - TPo j4) (4) The local heat transfer by convection on the heating side
(typical <5% of the heat input) can be calculated from Qconv j )

Figure 4. Local heat transfer with reaction inside the pipe.

2806 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009

RWoAp∆Tj, where the wall heat-transfer coefficient, RWo, can be VCM ) FEDC × (%EDC/100) × (SP/100) - HCl -
determined by an empirical heat-transfer correlation.24 The mean ∆LE - ∆HE (10)
temperature difference, ∆Tj, is calculated with TPo j and Tfurnace gas.
TWi j and the other temperatures in eq 6 can be calculated where SP is the cracking rate (%), S is the selectivity (%), FEDC
from the mean temperature difference between the inner wall is the EDC feed amount to be cracked (t/h), %EDC is the EDC
and the fluid, ∆Tj ) f(TWi j,Tgas j) ) Qcond j/(RWiAcoke). The local concentration in the feed (wt %), REDC is the EDC amount in
heat-transfer coefficient, RWi, was determined from an empirical the cracked product (t/h), VCM is the VCM amount in the
heat-transfer correlation (ref 24, eq 48) for turbulent flow in cracked product (t/h), HCl is the HCl amount in the cracked
pipes. product (t/h), ∆EDC is the effective EDC amount cracked (t/
The calculated temperature profiles converge strongly, inde- h), ∆VCM is the change in the amount of VCM (t/h), ∆LE is
pendent of the assumed starting temperature profiles of the outer the net change in the amount of low-boiling compounds (t/h),
wall temperature. A coke layer developed in the course of ∆HE is the net change in the amount of high-boiling compounds
production will reduce the local heat flux. At otherwise constant (t/h), and MGi is the molecular weight of compound i (i ) EDC
conditions, this will lead to an increase in the outer wall or VCM) (g/mol).
temperature, TWo. If our model calculation shows an increasing A change in the VCM production can thus be effected by
deviation between calculated and measured TWo values, it is changing either the cracking rate or the EDC feed or EDC
corrected by additional heat resistance of coke. Calculated coke concentration of the feed, whereby the resulting amount of the
layers of about 1-mm thickness at the coil outlet were found to byproduct HCl and the net changes ∆LE and ∆HE are
be in agreement with inspections after 1 year of uptime of the determined by the kinetics and the conditions of the reaction.
furnace. The influence of a potential appearance of local hot HCl is reconverted to EDC in the subsequent oxychlorination
spots, which can result from local deposits of coke, is discussed process. Hence, ∆LE and ∆HE give, together with the EDC
in section 4. losses occurring in the reprocessing section, the raw-material
This model enables simultaneously calculating the noniso- loss of the cracking process resulting from production change
thermal and nonisobaric reactor system for steady state in heat ∆VCM. The cost increase is mainly determined by the losses
transfer using our simplified reaction model. The integration in raw material EDC. Consequently, the marginal costs (GK)
of the mass balances from reaction model was done by using of the next “additional” tonne of VCM can be approximated if
the Runge-Kutta-method25 and a calculation program for stiff the change ∆VCM is significant.
differential equation systems26 with the above-described actual
GK ) (∆RS × EDCK × f + ∆REDC × RK)/∆VCM
temperature profile. Kinetic constants were adjusted to the (11)
composition of the reaction mixture at reactor inlet and outlet
(within ( 5% for the main components and about ( 30% of ∆RS ) ∆LE - ∆HE × fo (12)
the main byproduct).
In eqs 11 and 12, GK is the marginal costs per additional tonne
3. Conversion and Selectivity of VCM (Euro/t of VCM); ∆LE and ∆HE are the net changes
Modification of several essential parameters enables a predic- (t/h) in the heavy ends and light ends, respectively, in the
tive calculation of the change in selectivity (change in the cracking process; ∆RS is the change in the amount of residues
amounts of byproduct). This allows calculation of the VCM (t/h); ∆REDC is the change in the amount of EDC that has to
production, as well as the resulting additional costs for raw be reprocessed (t/h); EDCK is the EDC market price (Euro/t);
materials and energy. These parameters are the cracking rate RK is the specific EDC reprocessing energy costs (at constant
(influenced by the EDC throughput and local energy input), the reflux ratio in distillation, Euro/t of REDC); fo is the loss factor,
EDC quality (mainly influenced by distillation in EDC repro- accounting for EDC losses during HE distillation; and f is the
cessing), the operating time of the plant and/or the differential operating time factor or function.
pressure of the cracker coil (determined by the cracking In complex technical plants, there is typically no steady state.
conditions discussed above), and the net formation of high- For example, changes in the EDC feed mix depend on the EDC
boiling and low-boiling substances (determined by the reaction production in other units. Changes in other cracking parameters
kinetics and the EDC feed quality). result in changes in the raw product and can influence the
The EDC conversion and the cracking rate are related to the reprocessed EDC quality with a time lag of several hours.
effectively cracked EDC. The selectivity, S, is defined as the Therefore, short-term predictions always need an actual balance
VCM amount obtained per unit of cracked EDC. The resulting and check of the data basis for estimating costs from ∆RS
raw-material losses were calculated from the net formation of calculations.
the byproducts (collectively as the changes in high-boiling and
low-boiling compounds, ∆HE and ∆LE, respectively, with 4. Results and Discussion
respect to EDC). The cost effect of the uncracked EDC amount With our simulation program of the heat transfer in an EDC
was taken into account by calculating the EDC losses during cracking furnace together with process data, the reactor profiles
reprocessing together with energy consumption for reprocessing. of temperature, pressure, and conversion were calculated. A
simplified reaction model with all relevant compounds was used
∆EDC ) FEDC × (%EDC/100) - REDC ) VCM +
HCl + ∆LE + ∆HE (7) to calculate the byproduct spectrum found under practical
conditions. With these tools, a prediction of the change in
SP ) (∆EDC × 100)/[FEDC × (%EDC/100)] (8) cracking severity by changing parameters such as conversion
or EDC feed concentration is possible and can be used to provide
S ) (VCM/∆EDC) × (MGEDC/MGVCM) × 100 (9) a forecast of the resulting marginal costs, which are relevant
for additional VCM production based on an actual situation in
The VCM production in EDC cracking was obtained from the production plant. With data transfer from the process control
eqs 7 and 8 as system and analytics, a quasi-online balancing and calculation
 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009 2807

Figure 5. Example profiles of the stepwise simulation along the coils.

is possible (hourly if needed). This tool is useful in investigating thicker than 18 mm. This calculation can be used to define set
the process parameters in view of changes in the fuel, feed points for safety loops.
concentration, or conversion. Our reaction model also shows good agreement with empirical
Figure 5 shows example profiles of the stepwise simulation experience and other reported studies. In agreement with Borsa,4
along the coils. The energy input shows a typical shape with a we found no relevant C3 compounds in the cracked gas, which
maximum around the sixth reactor section. This is a result of otherwise should be there if radical recombination of C1 and
the burner optimization to achieve a high yield and low coke C2 radicals occur. Molecular hydrogen from a spontaneous
formation. Apart from yield and coke formation, it is important decay of acetylene into C and H2, which could explain carbon
to watch the uniformity of burner intensities in order to avoid formation, was not found in the cracked gas products. In contrast
the occurrence of hot spots. These criteria have to be matched to Invaco,19 who found higher amounts of ethylene even when
for each furnace geometry and are an important part of the CCl4 was added, we detected no ethylene, which is in agreement
producer’s acquired know-how. with Borsa.4 Invaco noted that the reason for the difference was
The EDC conversion line in Figure 5 shows that significant the high surface-to-volume-ratio in his microreactor. These
conversion starts above 350 °C. Subsequently, the EDC effects are in agreement with our assumptions for the chain
conversion per coil section rises from about 1% to 4% (in termination mechanism.
the ninth coil section). The low conversion at temperatures As discussed in section 2.1, the effect of small amounts of
below 400 °C agrees with the experimental study of Song et CCl4 (on the order of 100 ppm) cannot be neglected. CCl4 or
al.7 They found less than 1-2% EDC conversion at 400 °C Cl2 both have a promoting effect on EDC conversion but also
and a space time of 11 s. Song et al. also found a high increase the formation of acetylene, dichloroethylene, chloro-
conversion of about 68% and strongly increasing byproduct prene, and other unwanted compounds (cf. reaction 15).
at 500 °C and an 11-s space time. Therefore, in a technical The main pathway to VCM is through formation of radical
plant with typical space time of g20 s, the local heat flux R3, followed by R3 decay to yield VCM and radical R1. The
and resulting heat input profile have to be optimized in view formation of acetylene, chloroprene, and benzene (reactions 16,
of both heating to achieve the needed reaction rate and using 12, and 29, respectively), which are known as soot/coke
the remaining space time for desired conversion at the lowest precursors, depends strongly on the temperature and the
temperature level. In our case, after heating the reaction concentration of radical R5. Generally, in view of these
mixture and starting the conversion, the optimized local heat components, lower reaction temperatures are beneficial for
input limits the heat-consuming conversion in the following the selectivity. However, without additional amounts of promot-
sections. This optimization avoids excessive local tempera- ers, the formation of primary radicals for initiation (reactions 1
tures and minimizes local coke formation. or 2) requires high reaction temperatures up to 500 °C at the
The measured differential pressure at the coils (for a reactor outlet in order to achieve the desired conversion. Yet,
comparable EDC feed) is not sensitive enough to detect local this can lead to above-discussed enhanced byproduct formation
coke deposits. Here, the simulation of the measured wall and reduced severity. Therefore, a compromise between high
temperatures gives an indication of the coke layer. VCM operating rates on one hand and byproduct and coke
Our simulation results show that, at a furnace wall temper- formation on the other hand needs to be found.
ature of 780 °C, a local coke layer of nearly 30 mm is needed In contrast, from an economic point of view, Figures 6 and
to increase the local outer wall temperature beyond the 7 show the principle trends of the marginal costs in relation to
maximum working temperature of this high-alloy steel (ap- EDC quality and cracking rate. In both cases, an increasing
proximately 700 °C). If this happens, the furnace wall temper- operating time results in higher costs due to the wall-layer
ature (i.e., heat input) must be increased to keep the conversion problems. These additional costs for the next tonne of VCM
constant. For example, at a furnace wall temperature of 880 production can reduce the earnings from that tonne dramatically
°C, the temperature limit of the steel is exceeded at coke layers in view of the market price for the raw material EDC.
2808 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009

coking is done after at most 2 years of uptime or when coil

wall temperatures rise over 650 °C (corresponding to coke
layers thicker than 10 mm).

5. Conclusions
Modeling of a cracking unit depends on the individual
construction of the furnace and burners. A simulation model
was set up based on real plant conditions. The yield of the
cracking reaction is determined by the design of the furnace
with its local heat transfer to the coils. Careful adjustment of
this heat input is required to achieve a high yield and to avoid
extensive coke formation. If the local heat flux is too high, then
coking and even local hot spots can occur. In our plant, we
found that EDC conversion above 60% leads to an excessive
Figure 6. Marginal costs as a function of EDC quality. increase in byproducts and costs.
With this simulation model of the cracking process, the
reactor temperature profiles can be calculated on the basis of
actual process data and energy balances. This simulation also
provides an estimate of the thickness of coke layers depositing
during operation time.
Furthermore, the resulting temperature profiles can be used
to avoid unsafe operating conditions in terms of maximum
allowable coil temperatures.
Especially in the case of running high in capacity or at the
end of a typical operating time, marginal costs for the next tonne
of product become more relevant from an economic point of
view. For example, marginal cost increases of some 50% of
standard production costs might arise, which, in fact, would not
justify marginal downstream poly(vinyl chloride) production.
Carefully monitoring this effect on production costs would also
Figure 7. Marginal costs as a function of cracking rate. allow the optimization of multiple EDC cracking furnaces that
are in parallel operation.
Depending on the capacity and current plant performance,
increases in the VCM operating rates can have a strong influence Literature Cited
on the marginal costs. However, the influence of EDC quality (1) Ullmann’s Encyclopedia of Industrial Chemistry, 5th ed.; VCH
is also important, as shown in Figure 6. Decreasing the EDC Verlagsgesellschaft mbH: Weinheim, Germany, 1986; Vol. A6.
purity below 99.5 wt % at constant EDC conversion (i.e., VCM (2) Barton, D. H. R. Kinetics of the Dehydrochlorination of Substituted
production) leads to the shown strong increase in marginal costs. Hydrocarbons. Part I. Induced Dehydrochlorination. J. Chem. Soc. 1949,
148–155.
Under these conditions, it might be beneficial to put more
(3) Barton, D. H. R.; Howlett, K. E. Kinetics of the Dehydrochlorination
emphasis on the downstream distillation even if this is achiev- of Substituted Hydrocarbons. Part II. The Mechanism of the Thermal
able only at higher costs. All our model calculations are based Decomposition of 1:2-Dichloroethane. J. Chem. Soc. 1949, 156–164.
on a price of EDC of 300 Euro/t. (4) Borsa, A. G. Industrial Plant/Laboratory Investigation and Computer
There are two different influences on the cost curve for Modeling of 1,2-Dichloroethane Pyrolysis. Ph.D. Dissertation, Colorado
School of Mines, Golden, CO, 1999.
constant VCM production in Figure 7. One is the nonlinear (5) Huybrechts, G.; Wouters, G. Mechanism of the Pyrolysis of 1,2-
increase of LE/HE residues (material losses, EDC costs) with Dichloroethane in the Absence and Presence of Added Chlorine. Int.
increasing EDC conversion, and the other depends on the costs J. Chem. Kinet. 2002, 34 (5), 316–321.
of energy for EDC recovery by several distillations. The (6) Ashmore, G. P.; Owen, A. J.; Robinson, P. J. Chlorine-Catalysed
Pyrolysis of 1,2-Dichloroethane. Part 1. Experimental Results and Proposed
minimum shown in Figure 7 could even disappear with varying
Mechanism. J. Chem. Soc., Faraday Trans. 1 1982, 78, 657–676.
EDC purity and cost levels of EDC or energy respectively. In (7) Song, B.-H.; Choi, B.-S.; Kim, S. Y.; Yi, J. Kinetics and Mechanism
other words, if energy becomes cheaper or EDC purity is free, of Pyrolysis of the 1,2-Dichloroethane. J. Korean Inst. Chem. Eng. 2001,
then cracking severities of higher than 60% can be economically 99, 481–487.
viable. Under normal market conditions, however, 60-61% (8) Vlasov, P. A.; Warnatz, J. Detailed Kinetic Modeling of Soot
Formation in Hydrocarbon Pyrolysis behind Shock Waves. Proc. Combust.
cracking severity is the upper limit given by the strong increase Inst. 2003, 29, 2335–2341.
in LE/HE. For example, compared to 60% cracking severity, (9) Borsa, A. G.; Herring, A. M.; McKinnon, J. T.; McCormick, R. L.;
the increase in LE/HE is a factor of 1.2 at 61%, a factor of 1.6 Ko, G. H. Coke and Byproduct Formation during 1,2-Dichloroethane
at 62%, and a factor of 2.1 at 63%. Our target was that we use Pyrolysis in a Laboratory Tubular Reactor. Ind. Eng. Chem. Res. 2001, 40,
2428–2436.
this tool to calculate actual resulting marginal costs in advance
(10) Korai, Y.; Ishibashi, M.; Yamamoto, K.; Mochida, I.; Higuchi, K.
to have a decision basis for additional sales. The other economic Mechanism of Carbonaceous Deposition in the Ethylene Dichloride Pyrolytic
objective was to optimize the total selectivity of both cracking Process. J. Jpn. Petrol. Inst. 2003, 46, 308–314.
units. (11) Mochida, I.; Tsunawaki, T.; Sotowa, C.; Korai, Y.; Higuchi, K.
In practice, optimized operating conditions cannot be Coke Produced in the Commercial Pyrolysis of Ethylene Dichloride into
Vinyl Chloride. Ind. Eng. Chem. Res. 1996, 35, 3803–3807.
retained for a long period. Coking is one effect that gradually (12) Shilov, A. E.; Sabirova, R. D. Mechanism of the primary act of
changes cracking conditions. Therefore, decoking of the coils the thermal decomposition of methane derivatives. Zh. Fiz. Khim. 1959, 6,
is necessary after some operation period. In our case, de- 1365.
 Ind. Eng. Chem. Res., Vol. 48, No. 6, 2009 2809
(13) Schneider, M.; Wolfrum, J. Mechanisms of by-product formation (20) Borsa, A. G.; Herring, A. M.; McKinnon, J. T.; McCormick, R. L.;
in the dehydrochlorination of 1,2-dichloroethane. Ber. Bunsen–Ges. 1986, Yamamoto, S.; Teraoka, Y.; Natori, Y. Characterization of Coke Formed
90, 1058–1062. in Vinyl Chloride Manufacture. Ind. Eng. Chem. Res. 1999, 38, 4259–4267.
(14) Ashmore, P. G.; Owen, A. J.; Robinson, P. J. Chlorine-catalysed (21) Stefandis, G.; Heynderickx, G. J.; Merci, B.; Marin, G. B. CFD Modeling
pyrolysis of 1,2-dichloroethane. Part 2. Unimolecular decomposition of the of Thermal Cracking Furnaces Using Detailed Combustion Mechanisms. Comput.
1,2-dichloroethyl radical and its reverse reaction. J. Chem. Soc., Perkin Chem. Eng. 2006, 30, 635–649.
Trans. 1982, 78, 677–693. (22) Park, Y.; Choi, B. S.; Yi, J. Simulation of imbalance reduction
(15) Benson, S. W.; Weissman, M. Mechanism of the pyrolysis of between two reactors in an ethylene dichloride cracker. Chem. Eng. Sci.
C2HCl5, molecular and radical steps. Int. J. Chem. Kinet. 1982, 14, 1287– 2005, 60, 1237–1249.
1304. (23) Colannino, J. Ethylene furnace heat flux correlations. Petrol.
(16) Bryukov, M. G.; Slagle, I. R.; Knyazev, V. D. Kinetics of Reactions Technol. Q. 2008, 13, 93–97.
of Cl Atoms with Ethane, Chloroethane, and 1,1-Dichloroethane. J. Phys. (24) VDI-Wärmeatlas, 8th ed.; Springer-Verlag: Berlin, 1997; pp A7-
Chem. A 2003, 107, 6565–6573. A10 and Chapter K.
(17) Senkan, S. M.; Quam, D. Correlation of reaction rate coefficients (25) Carnahan, B.; Luther, H. A.; Wilkes, J. O. Applied Numerical
for the abstraction of hydrogen atoms from organic compounds by chlorine Methods; John Wiley & Sons: New York, 1969.
radical attack. J. Phys. Chem. 1992, 96, 10837–10842. (26) Villadsen, J.; Michelsen, M. L. Solution of Differential Equation
Models by Polynominal Approximation; Prentice Hall: Upper Saddle River,
(18) Goldfinger, P.; Huybrechts, G.; Martens, G. Elementary rate
NJ, 1978.
constants in atomic chlorination reactions. Part 2. Experiments in competitive ReceiVed for reView April 29, 2008
systems. Trans. Faraday Soc. 1961, 57, 2210–2219. ReVised manuscript receiVed November 13, 2008
(19) Invaco, J. A. A Detailed Quantitative Study of 1,2-Dichloroethane Accepted November 20, 2008
Cracking to Vinyl Chloride by Gas Chromatographic Pyrolysis Device. Ind.
Eng. Chem. Res. 1996, 35, 931–937. IE8006903