Project Leader: Joe McGonigle

Partners: Michelle Darjean, Tammy Lai



The proposed distillation design for ethylene oxide production consists of four stages: an absorber and desorber for a crude separation of ethylene oxide from the gaseous reactor effluent, followed by a stripper for light ends removal and a final distillation column for fine separation. The process optimization focuses mainly on minimizing the number of stages and the overall dimension of the columns coupled with determining the most efficient reflux ratio to minimize utilities' costs. This design exceeds the 99.5% purity requirement, producing 134,000 metric tons per year of 99.8% pure ethylene oxide with a fractional recovery of 98.7%. The process has a total capital cost of $9,100,000 and a yearly utility cost of $13,800,000 resulting in a cumulative net present value of $360,000,000 over a 15-year lifespan.


JMT Engineering was contracted to design a separation process to produce 99.5% pure ethylene oxide (EO) from gaseous reactor effluent produced by jEO & Associates' reaction process. The process was modeled in Aspen using Radfrac columns. When it was necessary to make a specification, the distillate to feed ratio and reflux ratio were used. The distillate to feed ratio was utilized to induce the desired separation. All columns were assumed to have a negligible pressure drop throughout. An NRTL property package with regressed binary parameters was used for physical property calculations.

Separation Description

The reactor effluent stream fed into the separations process was as given by jEO & Associates, Inc. with compositions modified to include the downstream side reactions to acetaldehyde and formaldehyde (Table I). This stream also contains several other contaminants. Carbon dioxide and water are present from the combustion of ethylene in the reactor as well as inert gases that are used as ballast in the reactor. In addition the reactor effluent contains unreacted ethane and oxygen, which should be recycled back to the reactor to increase overall reactor conversion. A proper separation sequence must generate a product stream of 99.5% purity in order to be commercially viable.

Product Description

Ethylene oxide is a colorless liquid or gas with a faint ether-like odor, which is reactive due to the highly strained, three-member ring structure. Although non-corrosive, ethylene oxide ignites in air at 804°F and decomposes explosively at 1040°F. The decomposition reaction can occur in anaerobic conditions, however dilution of ethylene oxide with an inert gas can prevent this reaction. The boiling point of ethylene oxide is 51°F with lower and upper flammability limits of 2.6% and 100% respectively.

Safety and Environmental Issues

Due to the reactive nature of ethylene oxide, the most important safety aspect involves determining appropriate materials of construction for the distillation process. All metallic equipment should be of carbon steel or 300 series stainless steel; materials such as copper, silver, magnesium, mercury, and cast iron should be avoided. Any packing or gaskets should be composed of or reinforced with Teflon or Grafoil. To prevent water or ethylene oxide adsorption and exothermic reactions, porous insulation should be avoided; only closed-cell insulation materials should be used with stainless steel sheathing and banding. Valves should not allow any ethylene oxide to stagnate within the gate; therefore ball and plug valves need to be avoided. Overall, the main safety concern of any ethylene oxide separations process involves avoiding buildups of unreacted ethylene oxide throughout the distillation procedure. The proposed design has columns made entirely out of stainless steel for an extra degree of safety. This reduces the chance that any stagnant EO in the liquid or vapor phase will encounter rust, which can catalyze a highly exothermic polymerization reaction.

Proposed Design

The proposed design features a 98.7% recovery of EO from the feed and a product stream of 99.8% EO on a molar basis, which exceeds the required purity (see Table 1 and Appendix A).

Table 1. Feed and Product Stream Composition







Pressure (bar)



Flow (kmol/hr)




































Illustrated below is the basic flowsheet for the separations process (Fig. 1). The design was focused on the purification of the liquid absorber stream EORICH; the refinement of the compounds in the gas-phase stream CO2LOOP was beyond the scope of the project.


Figure 1: Optimized Ethylene Oxide Distillation Process



The reactor effluent column is first sent to a 45-stage absorber, which removes a majority of the light gases. A cold process water stream is fed to the top of the absorber and the separation takes advantage of the fact that, at the high column pressure of 13 bar, EO is much more soluble in water than any of the other components. Increasing pressure and lowering temperature in the column serves to make all the gases more soluble, so a balance had to be found that allowed the EO to dissolve while the other gases were still vented out the top. The bottoms of the absorber, containing mainly EO and water, are sent to a desorber and the gas-phase top stream is sent to a CO2 recovery loop.

CO2 Loop

As the CO2 loop is not critical to obtaining a pure stream of EO, it was not explicitly modeled for this project. Basically in this loop CO2 and the other light gases are vented while ethylene and oxygen are sent back to the reactor. The first step in the loop is an absorber, which has a feed of hot potassium carbonate solution. The bottoms stream of the column consists of a carbon dioxide-rich solution, which is sent to an intermediate stripper that removes any remaining ethylene. Finally a desorber is used to remove the rest of the carbon dioxide. The stream leaving the bottom of the desorber is then recycled back to the absorber, creating a loop.


The desorber, a 24-tray column with a reboiler, is used to remove the large amount of water added in the absorber. The feed to the desorber is preheated by condensing 175 psi steam in a heat exchanger. The bottom stream contains water and formaldehyde while the top contains the EO and the rest of the impurities. In this column it is favorable to have a lower pressure and higher temperature, thus it is operated at 3.5 bar. The reboiler has a very high duty in this column due to the large amounts of material that must be vaporized.


The stripper removes all of the remaining light gas impurities. It is a 24-tray column with a condenser and reboiler. The distillate contains all of the gas contaminants while the bottoms contains only water, EO and acetaldehyde. One major problem with this column is that the condenser must be operated at -64.5°C in order to provide the necessary reflux for the separation. This means that refrigerated water cannot be used. Due to time and project scope limitations, only a preliminary investigation into coolants was conducted. Lewin et. al. suggest that ethylene is a suitable refrigerant, but it would unfortunately increase costs.


The final column has 42 trays with a condenser and reboiler. The EO, along with remaining impurities, goes out the top as distillate and the bottoms contains water, and acetaldehyde and small quantities of EO.

Optimization Techniques

Property Package Regression

In order to properly model the interactions of polar molecules in the separations process, a property package had to be chosen that would give reliable vapor-liquid equilibrium calculations. Experimental data was obtained for several binary systems:

· ethylene oxide and water

· acetaldehyde and water

· formaldehyde and water

· ethylene oxide and acetaldehyde.

The data was used to generate y vs. x diagrams for the various systems and these diagrams were compared to those generated in Aspen using various property packages. Wilson, UNIQUAC and NRTL packages were all used and it was found that none of these gave accurate representations for the formaldehyde and water, EO and acetaldehyde, and acetaldehyde and water systems. As NRTL provided the closest representation of the real data, it was the property package chosen. The binary parameters for these systems were modified using the regression tool in Aspen. For the EO and water system, the default NRTL parameters provided a very good representation of the real data. Since using the regressed values for this system caused errors in the simulation, the default parameters were used. For the formaldehyde and water system, even the regressed package provided a very poor representation of the data, but the results were reasonable regardless of the choice of parameters used. It would be desirable to have more data in order to obtain a better regression. The data available was generally at very low pressures and at a range of temperatures outside of that in the design. Figure 2 illustrates the significant improvements of the regressed parameters as compared to the original Aspen values. Although the Aspen parameters (red) result in a decent representation of the experimental data (blue) at low ranges, the Aspen values become inaccurate at higher purity separations. The regressed data (green) closely mimics the experimental data. The same improvements were obtained by performing regression analyses on the remaining binary systems.



Figure 2: Y vs X for Acetaldehyde and Water

Purification Column

In optimizing the design the focus was on the final purification column. The costs associated with this column depend on several factors. The first is the capital cost of the column, which increases with diameter, number of stages, and condenser and reboiler duty. The second is the utilities costs associated with steam in the reboiler and cooling water in the condenser. The one-time capital cost is very small in comparison to utilities costs for the thirteen years of plant operation. Thus optimizing the column focuses mainly on the reduction of utilities, which correspond directly to the specified reflux ratio in the column. The reflux ratio is a measure of how much of the distillate is condensed and sent back to the column. A high ratio results in large condenser and reboiler duties and thus the column should be run at as close to minimum reflux as possible. A low reflux ratio lowers the capital cost by decreasing the diameter of the column, but it increases capital cost, as it requires more stages to achieve the separation. The number of trays cannot be increased indefinitely because columns should be kept to a reasonable height to prevent wind or other severe weather from causing equipment damage. Sensitivity runs were conducted in Aspen to find the number of purification column stages and reflux ratio required to obtain the desired product stream composition (Fig. 3). It was decided that 16 ideal stages and a reflux ratio of 3 would not only produce a 99.7% pure product, but would also be a good balance between column height and reflux ratio.


Figure 3: Optimal Number of Purification Column Stages Based on Reflux Ratio

Optimization of Other Columns

The specifications for the absorber and desorber were not an intensive part of the design process. The absorber did not have a condenser or reboiler so the only specification was the number of stages and operating pressure. The goal for the absorber was to have all the EO come out the bottom and as much of the gases out of the top. The major factor determining this split was the flow rate of process water to the column. We found this flow by reducing the amount of water to the column until EO started to come out the top. Increasing the number of ideal stages beyond 15 provided no further benefits for separation. The pressure of 13.5 bar was chosen from the results of sensitivity runs over varying pressures. The pressure is high enough to force the EO to dissolve in water, but not so high as to cause the other gases to become soluble.

For the desorber, the specified duty was the minimum necessary to get all of the EO to come out in the distillate. The desorber and pre-heater duties are the largest expense in the process, and there is little which can be done to dramatically reduce their value. The number of stages was chosen by a trial and error process to be the smallest amount that would still give the necessary separation.

The stripper was first simulated using a shortcut column with the Winn-Underwood-Gilliland method. This calculated a minimum reflux ratio of 4.4, an optimal feed stage of 5 and a table of reflux versus number of stages (Figure 4). The design heuristic of using 1.2 times the minimum reflux as the operating reflux was used and the shortcut column was replaced with a Radfrac column with 13 stages and a reflux ratio of 5.3. However, the results were dramatically different in the Radfrac as this type of column no longer assumes ideality. In order to get the column to converge the reflux ratio and number of stages had to be tweaked resulting in the final values of 10 stages and a reflux ratio of 9 with the feed to stage 3.

Absorber Process Water Reduction

Process water fed to the absorber accounts for $540,000 of the yearly utilities costs. This is the next highest utility cost after high-pressure steam. The process is being run using the minimum amount of water required to dissolve all of the EO. Recycling the almost pure water from the desorber back into the absorber column could dramatically reduce the cost for process water. A preliminary investigation showed several problems with this strategy. The first is that the cost of chilling the desorber bottoms back to 40 C is almost equivalent to the yearly cost of the process water. Also the desorber bottoms stream contains formaldehyde, which would build up to high levels within the recycle loop. This could be prevented through the use of a purge stream, but it was felt that these modifications would add considerable complexity to the design for relatively little gain.

Economic Analysis

Equipment Cost

Capital costs were estimated using the CAPCOST utility. For columns, tray sizing runs were conducted in Aspen once the optimization process was finished. The height between trays was the Aspen default of 2 ft. or 0.6096 m. The number of trays used for cost estimation was 3 times the number used in Aspen because of an assumed 33% tray efficiency. Stainless steel was chosen for both the vessel and the trays because of the greatly decreased risk of rust, a catalyst for polymerization. Although it is only recommended that stainless steel be used for places in which EO may be stagnant, JMT Engineering feels that the added capital cost associated with using all stainless steel is far outweighed by increased safety.

The condensers and the heater (block B1 in Fig. 1) were all assumed to be floating head with the maximum 900m2 heat transfer area allowed in CAPCOST and the minimum operating pressure of 10 barg. The materials of construction were stainless steel for the process stream side and carbon steel for the coolant side, which helped to minimize capital costs. (See Appendix C)

The reboilers were all kettle reboilers, and like the condensers and heater, the maximum heat transfer area (100 m2) and minimum operating pressure or 10 barg.


Table 2: Equipment Cost


Bare Module Cost ($)









2 Condensers (Floating Head)


3 Kettle Reboilers


Heater (Floating Head)




Operating Cost

To find the needed heating steam flowrates m for the various heat exchangers, the relationship


was used, where the duty was calculated in Aspen and Hout - Hin was the enthalpy change of condensing saturated steam at 175 psi, the pressure given in Lewin, et. al. (See Appendix D)

Similarly, to estimate the needed cooling water molar flowrate m for the purifier, the equation


was used, where

1. the duty Q was calculated by Aspen

2. Tout - Tin was the change in cooling water temperature, assumed to be from 32.2°C to 48.9°C

3. the heat capacity cp was found using Antoine's equation and the average temperature of the cooling water.

The temperature of the stripper condenser was found to be -64.1°C, which requires refrigeration. Since detailed heat transfer analysis was beyond the scope of the assignment, a very rough cost analysis was done using refrigerated water at -30°C as the coolant. As mentioned in the Proposed Design section, ethylene would be a more suitable coolant and merits further study. It was assumed that the water was saturated liquid, vaporizing during cooling; the relationship given above to find heating steam flowrates was also used in this case.

All prices for utilities were taken from Lewin, et. al.

Cash Flow Analysis

Cash flow analysis (see Table 3, Fig. 5, Appendix B) was conducted according to the method outlined by Lewin, et. al. The following conditions were used in the analysis:

· 15-year useful life of plant

· 2 years of construction

· 10% discount rate for calculation of net present value

· 37% tax rate

· 8400 operating hours each year

· operation at 50% and 75% of capacity for first two years after construction, respectively

The proposed plant would become profitable after the first year of operation with estimated net earnings of $57,750,000 per year after the first 6 years of operation.


Table 3: Economic Analysis Summary



The following topics need further investigation and research before implementation to maximize efficiency and profit:

· A wider range of experimental binary interaction data needs to be obtained to improve the simulation at pressures and temperatures closer to current and maximum operating parameters.

· Investigate the feasibility of recycling wastewater from the desorber and/or the purification column to greatly reduce process water costs.

· Investigate the use of the steam generated in the reactor section of the plant to supplement the steam fed to the heat exchangers in the process. This would most definitely increase profitability, considering the amount of steam required.

· Further study of refrigerant for use in condenser on stripper.



We recommend the implementation of the optimized ethylene oxide separation design for several reasons:

· The use of regressed binary interaction data with the NRTL property package greatly improves the accuracy and precision of the distillation model.

· By optimizing the number of stages and the reflux ratio, capital costs for the columns and heat exchangers are kept to a minimum.

· The optimized process conserves utility costs by significantly reducing the amount of process water fed to the absorber.

· The 99.8% concentrated ethylene oxide solution exceeds purity specifications of 99.5% and maintains a fractional recovery of 98.7%

· The proposed design will become profitable after only one year of operation, with estimated net earnings of $57,750,000 per year after the first 6 years of operation.



2. Gerhartz, Wolfgang. Ullman's Encyclopedia of Industrial Chemistry. 5th Edition.

Vol. A10. New York: VCH Publishers, 1987.

3. Gmehling, Onken, Arlt. Vapor-Liquid Equilibrium Data Collection. Vol I, 1a, Supplement 1. Frankfurt, Germany: Schon & Wetzel GmBH, 1981.

4. Gmehling, Onken, Arlt. Vapor-Liquid Equilibrium Data Collection. Vol I, Parts 3 and 4. Frankfurt, Germany: Schon & Wetzel GmBH, 1981.

5. Kister, Henry Z. Distillation Design. Boston, MA: McGraw Hill, 1992

6. Kirk-Othmer Encyclopedia of Chemical Technology. Vol. 9. New York: John Wiley & Sons, Inc., 1994.

7. Lewin, Daniel R., Seider, Warren D., and Seader, J. D. Process Design Principles. New York: John Wiley & Sons, Inc., 1999.