Strategies to Maximize Ethane Recovery with High-CO2 Feeds

Kyle Ross, PhD, PE

Bryan Research & Engineering, Inc.
Bryan, TX, U.S.A.

Barry Keys, BSEE, MSETM, PE

Enbridge
Longview, TX, U.S.A

Abstract

Elevated feed-CO2 levels decrease the maximum ethane recovery in GSP and similar demethanizer
configurations. This can be attributed to two causes. First, as CO2 displaces methane in the feed, the feed
tends to behave more like a richer gas, which tend to have decreased ethane recoveries. Secondly, the
choice of column operating conditions is complicated by the tendency of high-CO2 feeds to form dry ice
in the upper sections of the column, or in the feed or residue lines. This often requires a sacrifice of ethane
recovery to avoid dry ice formation. However, by manipulating column operating parameters, namely the
vapor and liquid split to the subcooler, the column operating pressure, and the inlet split to the reboiler,
the CO2 concentration profile in the column can be tailored to avoid dry ice formation, while still
maximizing ethane recovery. This paper explores the effects of adjusting these operating parameters with
both lean and rich feeds.
Background
Enbridge operates a gas processing facility in Longview, TX that includes a demethanizer followed by
amine treatment of the NGL (liquid product) stream from the demethanizer. The feed to the facility is a
fairly lean gas (~90% methane) with a CO2 content that varies between 1.6% and 1.8% on a molar basis.
The facility has repeatedly experienced operational stability problems in the demethanizer when attempts
are made to increase ethane recovery. Some of the problems are: increased column pressure drop,
carryover of liquid in the overhead line, and unsteady liquid production from the tower.
ProMax® [1] simulations of the demethanizer under conditions that correspond with the carryover
mentioned above do show that dry ice formation is likely near the top of the tower. If dry ice were to form
in the top section of the tower, it could accumulate in the packing. This would reduce cross-sectional area
available for flow, which would lead to increased vapor and liquid velocity and possibly column flooding.
Symptoms of flooding at the top of the tower would be increased column pressure drop, carryover of
liquid into the overhead line, and unsteady liquid production from the tower, exactly the problems that
Enbridge has experienced.
A study by Enbridge, Bryan Research & Engineering, and Koch identified other column hardware issues
that appeared to be contributing to column flooding, but it is likely that resolving these other issues will
not prevent dry ice formation. ProMax models showed that dry ice had the potential to form in the top
section of the tower. Since the plant was running so poorly to begin with, due to the lack of heat from a
clogged side reboiler, it was very difficult to determine if dry ice was forming. Previous inspections of the
demethanizer clearly indicated that CO2 freezing had occurred at some point(s) in the past as the mist
eliminator had been “relocated” and lodged in the outlet piping, and the flow distribution trough above
the top bed had been thoroughly mangled. This makes an interesting basis for a case study on how to deal
with elevated CO2 feed in a demethanizer.
Introduction
This paper will focus on a GSP (Gas Subcooled Process) facility as shown in Fig. 1, but many of the
lessons learned can be applied to other demethanizer configurations. In ethane recovery mode, methane is
separated from ethane and heavier components in the tower with methane concentrated into the residue
gas, and the ethane and heavier components recovered in the NGL. In ethane rejection, ethane is directed
to the residue gas. This paper will focus on ethane recovery mode. CO2 volatility is between that of
methane and ethane. CO2 recovery will be lower than ethane recovery, but will typically increase as
ethane recovery increases.
Stepping through the GSP plant,

 The inlet flow is split and follows two possible cooling paths before recombining at the inlet of
the Low Temperature Separator (LTS).
o In the top path the gas is cooled by the exiting residue gas.
o In the bottom path the gas is used as a heat source for the bottom and side reboiler of the
tower.
o A bypass path is typically included as a way to moderate cooling before the LTS.
 The vapor leaving the LTS is split, with approximately 70% directed to the expander and the
remaining 30% [2] sent to the subcooler.
o The subcooler condenses and subcools the gas by heat exchange with the residue gas
exiting the top of the tower. The subcooled liquid is then flashed down to column
pressure, which provides additional cooling before entering the column.
 o The expander extracts energy from the gas to cool it before it enters the column below the
subcooled liquid. The expander is typically combined with the booster compressor to
harness the extracted energy. By lowering the column pressure, more energy can be
extracted, thus lowering the temperature and increasing ethane recovery.
o The fraction split to the subcooler is typically adjusted to maximize ethane recovery
when in recovery mode.
 The liquid leaving the LTS can also be split, with some flashed across a JT valve before entering
the column below the expander outlet connection. A portion of the liquid can also combine with
the LTS vapor entering the subcooler to provide a richer feed mixture at the top of the column.
 Residue gas leaves the top of the tower. Typically it must be compressed to pipeline pressure. The
energy recovered from the expander provides a modest pressure increase in the booster
compressor, but in most cases, not enough to reach pipeline pressure. This leaves the residue
compressor to provide the bulk of the recompression energy. When the tower pressure is lowered
to increase recovery, residue compression power increases. Thus, there is a tradeoff between
increased revenue from higher recovery versus higher residue compression costs.
 NGL leaves the bottom of the tower as a liquid and can be pumped at a relatively low cost
compared to residue gas compression.
22 23 Residue Gas

From expander Q-Res

Booster Residue Compressor

Subcooler

20 19
21
1 10 11 1
SCL Valve
Mixer 2
LTS Vapor 9
3
Splitter
2 3
8 13 4
Gas/Gas HEX 12
Expander 5
Feed Splitter Feed Mixer To booster
1 Feed 18 To reflux
7 LTS
6

14 4 15 7
LTS Liq Valve
8
6
Side Reboiler LTS Liq Splitter
23 9

24 10

11

5 12
Demethanizer
17 13

25 14
4 To Reboiler
Bottom Reboiler

NGL

Figure 1. GSP layout

For this paper, the fractions of the LTS vapor and liquid sent to the subcooler will hereafter be called the
Vapor Split and the Liquid Split. The fraction of the inlet sent through the reboiler will be called the
Reboiler Inlet Split. This is shown in Figure 2.

2
 22 23 Residue Gas

From expander Q-Res

Booster Residue Compressor

Subcooler

20 19
21

Liquid 1
Mixer
10
SCL Valve
11 1

Split LTS Vapor Vapor

2

3
Splitter
2 3
8
Split 13 4
Gas/Gas HEX 12
Expander 5
Feed Splitter Feed Mixer To booster
1 Feed 18 7 LTS
6

14 4 15 7
LTS Liq Valve
8
6
Side Reboiler LTS Liq Splitter
23 9

24 10

11

5 12
Reboiler Demethanizer
17 13
Inlet Split 25 14

Bottom Reboiler

NGL

Figure 2. Reboiler Inlet Split, Liquid Split and Vapor Split Definitions
There are many demethanizer plants where CO2 removal occurs after methane removal, typically in the
NGL stream. There are some benefits to this type of operation. CO2 removal from the NGL is more
efficient since the CO2 concentration will be higher in the NGL than in the feed gas. Oftentimes, the CO2
in the residue gas can bypass the removal steps altogether since the residue gas specifications typically
allow for some CO2 [3].
The drawback to removing CO2 downstream of the demethanizer, as opposed to upstream, is that CO2 is
known [3] to decrease the maximum ethane recovery. There are a few reasons for this. First, as the CO2
content in the feed rises, the gas behaves as a richer gas. But, heat integration in a demethanizer plant is
poorer with a richer gas [4]. This leads to warmer temperatures at the top of the demethanizer, which
allows more ethane to escape into the residue gas.
Furthermore, as the CO2 content of the feed gas rises, the minimum allowable temperature near the top of
the tower increases (becomes less negative) due to the need to avoid dry ice formation. A typical method
to achieve higher recoveries is to increase the cooling across the turboexpander and JT valves by lowering
the column pressure. The tradeoff is increased residue compressor power. Many times this tradeoff is
justified and an economic optimum can be found that balances increased recovery versus compression
power. However, with high-CO2 feed concentrations it may not be possible to operate at the optimal
tower pressure due to the need to avoid low temperatures and dry ice formation in the top of the tower or
in associated feed and product piping.

3
Dry ice formation
Dry ice forms as an essentially pure, solid CO2 phase in natural gas. At a given temperature and pressure,
CO2 can form a stable, solid phase in a vapor (V), liquid (L) or vapor/liquid (VL) mixture when the
𝑆
chemical potential of pure solid CO2, 𝜇𝐶𝑂 2
, is lower than or equal to the chemical potential of CO2 in the
𝑉 𝐿 𝑉 𝐿
other phase(s), 𝜇𝐶𝑂2 and/or 𝜇𝐶𝑂2 [5]. (If both a liquid and vapor are present, then 𝜇𝐶𝑂 2
= 𝜇𝐶𝑂2
at
equilibrium.)
This can be understood by first reviewing the phase envelope for pure CO2 shown in Fig. 3. If we start at
𝑆 𝑉
a point A (290 psig, 60 °F), CO2 exists solely as a vapor because 𝜇𝐶𝑂2
> 𝜇𝐶𝑂2
.

1200
Freezing Condensation Sublimation
1000

800
Pressure, psig

600 Solid Liquid

400
C B A
200
Vapor
0
-150 -100 -50 0 50 100
Temperature, °F

Figure 3. Pure CO2 phase diagram

If the CO2 is then chilled from A at constant pressure, it will remain a vapor until the temperature reaches
𝑉 𝐿 𝑆
0 °F (Point B). At this point, 𝜇𝐶𝑂 2
= 𝜇𝐶𝑂2
< 𝜇𝐶𝑂2
. If heat is further removed, CO2 will condense at constant
temperature (since it is a pure substance) until all the vapor has condensed. With continued heat removal,
the temperature will decrease and the CO2 will remain as a liquid until -69 °F (Point C). At this
𝐿 𝑆
temperature, 𝜇𝐶𝑂 2
= 𝜇𝐶𝑂2
and continued heat removal will isothermally transform liquid CO2 into solid
CO2. After all CO2 solidifies, the temperature will decrease with continued heat removal. Below the
𝑆 𝐿
freezing temperature, 𝜇𝐶𝑂 2
< 𝜇𝐶𝑂2
.

Transitioning from pure CO2 to CO2-containing mixtures, there are some interesting behaviors that
𝑆 𝑉
appear. Dry ice will form in natural gas mixtures as essentially pure CO2 whenever 𝜇𝐶𝑂 2
≤ 𝜇𝐶𝑂 2
and/or
𝐿 𝑉 𝐿
𝜇𝐶𝑂2 . However, in a mixture, 𝜇𝐶𝑂2 and 𝜇𝐶𝑂2 decrease as the CO2 mole fraction decreases, which means
that the dry ice formation temperature will decrease as the CO2 mole fraction decreases. Fig. 4 below
shows the phase envelope and dry ice formation curve for the 2nd-stage composition of a typical
demethanizer. Comparing Fig. 4 to Fig. 3, not only is the shape of the dry ice curve different, but it is also
shifted to colder temperatures in the mixture versus pure CO2.
𝑉 𝑆
Starting in the vapor phase at Point A in Fig. 4, once again 𝜇𝐶𝑂2
< 𝜇𝐶𝑂2
. With isobaric cooling, the
𝑉 𝐿 𝑆
mixture will remain a vapor until Point B, the dewpoint, where 𝜇𝐶𝑂2 = 𝜇𝐶𝑂 2
< 𝜇𝐶𝑂2
. With additional heat

4
removal, the overall liquid fraction will increase and the temperature will decrease (as opposed to the
isothermal phase change of pure CO2) until the mixture arrives at Point C, which lies on the dry ice curve.
𝑉 𝐿 𝑆
At Point C, dry ice can begin to form since 𝜇𝐶𝑂 2
= 𝜇𝐶𝑂2
= 𝜇𝐶𝑂2
.

800

Phase Envelope Dry ice Critical point

700

600

500 Liquid-Solid Vapor

Pressure, psig

Liquid
(L-S) Vapor-Liquid
400 (V-L)
F
300 G
E D C B A
200

Solid-Vapor-Liquid
100 (S-V-L)

0
-200 -180 -160 -140 -120 -100 -80 -60
Temperature, °F

Figure 4. Phase diagram of typical Stage 2 composition

With further cooling, additional vapor will condense and the solid fraction will increase to a maximum.
Then, counterintuitively, the dry ice will begin to “melt” and completely disappear when the mixture is
cooled to Point D. This melting behavior can be explained by the fact that CO2 is more stable in the liquid
phase than in the vapor phase at these conditions. Under the right conditions (such as between points C
and D), when a portion of vapor condenses, it can hold more CO2 without forming dry ice than it
previously could as a vapor. Therefore, as more liquid forms it absorbs CO2 from the solid phase, thereby
dissolving the dry ice. This can be mathematically expressed as follows. At the conditions on Stage 2,
𝐾𝐶𝑂2 < 1, where
𝑦𝐶𝑂2
𝐾𝐶𝑂2 = ⁄𝑥𝐶𝑂
2

𝑦𝐶𝑂2 , 𝑥𝐶𝑂2 = vapor- and liquid-phase CO2 mole fractions

When vapor condenses with 𝐾𝐶𝑂2 < 1, 𝑦𝐶𝑂2 must decrease to maintain vapor-liquid equilibrium. As 𝑦𝐶𝑂2
𝑉 𝑆 𝑉 𝐿 𝑆
decreases, 𝜇𝐶𝑂 2
can become lower than 𝜇𝐶𝑂 2
. If 𝜇𝐶𝑂2
= 𝜇𝐶𝑂2
< 𝜇𝐶𝑂2
, CO2 will move back from the solid
phase into the vapor and liquid phases to reestablish equilibrium.
Between Points D and E, the mixture will transform from a VL mixture to a single liquid phase. At point
𝐿 𝑆
E, 𝜇𝐶𝑂2
= 𝜇𝐶𝑂2
, and dry ice will start to form in the liquid mixture. With further heat removal, the
temperature will decrease and more CO2 will transfer from the liquid phase into the pure CO2 solid phase.

5
There are two interesting corollaries to this behavior. First, by shifting the pressure higher from Point A to
Point F of Fig. 4, the S-V-L region can be avoided and a significant decrease in the dry ice formation
temperature can be achieved. For instance, when isobarically cooling from Point F, dry ice will not form
until Point G, which is roughly 30 °F colder than the dry ice formation temperature when starting at Point
A. This can be explained by the fact that for a fixed composition and temperature with a pressure below
the critical pressure, the overall liquid fraction will increase as the pressure is increased. By shifting to a
higher pressure, which increases the amount of liquid available to absorb CO2 from the vapor,
𝑉 𝑆
𝜇𝐶𝑂 2
remains below 𝜇𝐶𝑂 2
and dry ice does not form in the 2-phase region.

Secondly, if a rich and lean gas both have the same overall CO2 mole fraction, 𝑧𝐶𝑂2 , the S-V-L region will
shift to lower temperatures and pressures for the richer gas. To illustrate this, Figure 5 shows dry ice
formation curves for four gas compositions, all containing 1.8% CO2. The compositions are defined in
Table 1. Observe that a richer gas composition has a dry ice curve that is shifted to colder temperatures
and lower pressures.

500

450

400
C2+ GPM
350 0.05 5.0
Pressure, psig

300 5.0 10.4

250

200

150

100

50

0
-190 -180 -170 -160 -150 -140 -130 -120 -110 -100
Temperature, °F

Figure 5. Impact of methane content on dry ice curve with constant (1.8%) CO2 mole fraction
Finally, CO2 content has a noticeable effect on the location of the dry ice curve. Fig. 6 compares the dry
ice formation curve at 1.5%, 1.8%, and 2.1% CO2 with a constant C2+ GPM = 2.3 gal/MSCF.

6
 500

450
CO2 fraction
400

350 1.5% 1.8% 2.1%

Pressure, psig

300

250

200

150

100

50

0
-190 -180 -170 -160 -150 -140 -130 -120 -110 -100
Temperature, °F

Figure 6. Effect of CO2 content on dry ice curve.

Implications for operating GSP demethanizer with high-CO2 feeds

As mentioned previously, high-CO2 feeds can prevent operation at what would be considered
economically optimum ethane recoveries due to the need to avoid dry ice formation. As every plant offers
a different combination of feed conditions and compositions, there is not a single solution to maximize
profit while avoiding freezing in the column. What is offered here are levers that can be pulled to move
the demethanizer operating envelope out of the region where dry ice can form.
These levers are,

 Column pressure
 Reboiler Inlet Split
 Liquid Split
 Vapor Split
In this study it will be helpful to refer to a few different feed gas compositions for comparison purposes.
These are listed below in Table 1. All simulations were performed in ProMax with the GSP layout shown
in Fig. 1. The feed flowrate, temperature and pressure were 50 MMSCFD, 100 °F, and 800 psig for all
cases. Constant UA values were maintained in the bottom reboiler, side reboiler, subcooler, and gas/gas
exchanger, as 55,000, 120,000, 480,000 and 760,000 BTU/(hr∙°F), respectively. Unless otherwise
indicated, the NGL C1/C2 liquid volume ratio was controlled at 0.015, and the tower was modeled with 14
ideal stages, to correspond with 28 real trays [6]. The turboexpander isentropic efficiency was 85%.

7
 Table 1. Gas compositions used in study
Gas compositions (mole %)
Component A B C D E F G
CO2 1.8 1.5 2.1 1.8 1.8 3.8 1.8
Methane 90 90.3 89.7 98 80 78 60
Ethane 5 5 5 0.15 10 10 25
Propane 1.2 1.2 1.2 0.05 6.2 6.2 11.2
i-Butane 0.4 0.4 0.4 0 0.4 0.4 0.4
n-Butane 1 1 1 0 1 1 1
i-Pentane 0.2 0.2 0.2 0 0.2 0.2 0.2
n-Pentane 0.4 0.4 0.4 0 0.4 0.4 0.4
Richness,
2.3 2.3 2.3 0.05 5.0 5.0 10.4
C2+ GPM

Column Pressure
In some cases one might be able to select a column pressure above the S-V-L region to significantly
decrease the dry ice formation temperature. All else being equal, increasing the column pressure will
lower ethane recovery by decreasing the cooling through the expander and J-T valves. However, the
required pressure increase could be small, such that it doesn’t significantly decrease ethane recovery.
Dry ice can form in a number of places in and around the top section of the tower. It is possible for dry ice
to form in the feed, overhead product, or in the top section of the tower above the expander outlet.
Although the temperature rises moving down the column, the CO2 can concentrate below the top of the
tower under certain conditions [7] as it is carried up by the methane leaving the expander and then
recondensed by the colder temperatures at the top of the tower. The increased CO2 concentration will
expand the S-V-L region, which can lead to dry ice formation on stages below the top of the tower even
though the temperature increases. An example of this is shown in Table 2.
Table 2. Top of the tower conditions for Feed A (0% LTS liquid and 27% LTS vapor to top).
Highlighted cells show potential dry ice formation.
Top Stage Vapor CO2, Liquid CO2, Temperature, Dry Ice Formation
Stage
Pressure, psig mol% mol% °F Temperature, °F
1 1.07 5.00 -144.9 -174.4
2 1.81 8.10 -137.2 -134.9
315
3 2.28 8.74 -128.3 -159.8
4 2.13 6.61 -118.6 -164.7
1 1.16 5.21 -142.4 -173.2
2 1.93 8.13 -133.9 -161.9
325
3 2.33 8.36 -124.7 -160.3
4 2.12 6.24 -115.5 -165.6

By increasing the pressure by 10 psi, the column is able to operate in a region that avoids dry ice
formation. Fig. 7 compares the operating point for Stage 2 in the column at the two pressures with the
ProMax-generated phase envelope overlaid with the dry ice formation curve. This shows the operating
point for this tray to be within the S-V-L region at the lower pressure while, with a 10-psi pressure
increase, the operating conditions in the tower avoid the dry ice region. Higher pressure operation did two

8
things to move into the dry ice-free region. First, higher pressure increased the temperature and pressure
on the tray, moving the operating point up and to the right. Secondly, higher pressure increased the liquid
content, moving the S-V-L region down and to the left. By moving to a higher-pressure, the ethane
recovery decreased from 86.9% to 84.8%, while the residue compressor power decreased from 2,288 hp
to 2,191 hp. So, this is the first tradeoff. Raising the pressure allows one to avoid the S-V-L region and
lowers the residue compression power requirement at the expense of reduced ethane recovery.

900
Phase Envelope
800
Dry ice 315 psig
700
600
Pressure, psig

500
400
(-137.2 °F, 315.3 psig)
300
200
100
0
-200 -180 -160 -140 -120 -100 -80 -60
Temperature, °F

900
800
325 psig
700
600
Pressure, psig

500
400
(-133.9 °F, 325.3 psig)
300
200
100
0
-200 -180 -160 -140 -120 -100 -80 -60
Temperature, °F

Figure 7. Stage 2 operating point at 315 and 325 psig (Feed A)

Reboiler Inlet Split
A typical GSP plant has the ability to change the bottom and side reboiler duty by adjusting the fraction
(Reboiler Inlet Split) of the warm inlet gas that passes through these exchangers. A common method for
decreasing the methane fraction in the NGL product is to increase the bottoms temperature by increasing
the Reboiler Inlet Split. This will also decrease the amount of CO2 in the NGL, boiling the CO2 back up

9
the column. Conversely, lowering the Reboiler Inlet Split will increase the CH4 and CO2 fractions in the
NGL product, allowing the CO2 to leave in the column bottoms.
As the column bottoms temperature is lowered, the CO2 concentrations at the top of the tower will
decrease as shown in Figure 8. This shifts the dry ice formation curve to lower pressures and colder
temperatures, as detailed in Figure 6. Therefore, if one is trying to avoid dry ice formation then operate
the column bottom as cold as NGL specifications will allow. Typically, the NGL specifications will limit
the amount of CH4 and/or CO2 in the NGL. These limits effectively set a lower bound on the bottoms
temperature.

45

40

35
(Temp., °F)/(C1/C2 Vol. Ratio)
Vapor CO2 mole fraction

30 50/0.00013
40/0.0004
25 30/0.006
25/0.017
20

15

10

0
1 2 3 4 5 6 7 8 9 10 11 12 13
Stage

Figure 8. Effect of NGL temperature on CO2 concentrations in the tower

Liquid Split and Vapor Split
In the discussion about tower pressure, the dry ice region could be avoided with just a small change in the
operating pressure. In some cases, the pressure change needed to get above the S-V-L region might be so
large that it causes an unacceptable drop in ethane recovery. Another alternative is to increase the fraction
of heavy components in the top of the tower, which, as shown in Figure 5, will move the dry ice
formation curve to lower pressures and colder temperatures. This can be done by increasing the LTS
liquid split to overhead (Liquid Split). The effect can be magnified by simultaneously decreasing the LTS
vapor split to overhead (Vapor Split). Fig. 9 shows how the vapor CO2 concentration in the top of the
tower decreases when increasing Liquid Split and decreasing Vapor Split. This occurs even though the
LTS liquid has a higher CO2 concentration than the LTS vapor.

10
 2.5

2
Vapor CO2 mole fraction, %

1.5

0.5 0% Liq/27% Vap

100% Liq/27% Vap
100% Liq/23% Vap
0
1 2 3 4 5
Stage

Figure 9. Effect of Liquid Split and Vapor Split on CO2 concentrations near top of tower
For the set of conditions used in this example, dry ice formation is expected on stages 2 and 3 with 0%
Liquid Split. Fig. 10 compares the dry ice formation curve on Stage 2 for 0% versus 100% Liquid Split.
Other parameters are the same in these scenarios (290-psig tower pressure, 27% Vapor Split). As can be
seen, the additional liquid to the top moves the dry ice curve to lower pressures and colder temperatures,
giving a larger dry-ice free operating envelope, which allows the operating point for the 100% Liquid
Split to lie outside of the S-V-L region

11
 600

Stage 2
500
Dry ice lines &
operating points
400
"0%" "100%"
Pressure, psig

300

200

100

0
-170 -165 -160 -155 -150 -145 -140 -135 -130 -125 -120
Temperature, °F

Figure 10. Comparison of Stage 2 operating point and dry ice curve for 0% and 100% Liquid Split
(Feed A, Pressure = 290 psig, 27% Vapor Split)
Intuition would lead one to think that sending a richer composition to the top of the tower would decrease
ethane and propane recovery. Table 3 shows the maximum ethane and propane recoveries for for 0%
versus 100% Liquid Split (if one neglects the potential to form dry ice). The maximum ethane and
propane recovery is found by varying the Vapor Split with a fixed Liquid Split. The results are shown for
a top-stage pressure of 290 psig, but similar results are found at other pressures. The results confirm our
intuition that sending richer liquid to the top of the tower results in loss of the richer components into the
residue gas. However, simulation models can predict dry ice formation and then ignore dry ice
accumulation in the column calculations. In real equipment dry ice accumulation can lead to flooding,
poor separation, or complete flow blockage. Therefore, simulation results that indicate dry ice formation
should be treated with caution. In the present case of Table 3, the 100% Liquid Split is the only viable
option of the two.
Table 3. Effect of Liquid Split on Maximum Ethane & Propane Recovery
(Feed composition A, Pressure = 290 psig)
Liquid Split, % Vapor Split, % C2 Recovery, % C3 Recovery, %
0* 29 91.0 99.2
100 27 90.9 98.7
*Dry ice formation predicted at this condition
Also of note is that the Vapor Split that yields the maximum ethane recovery, and coldest overhead
product temperature, is lower when the Liquid Split increases. This is a trend that will continue to appear
later in this paper. Therefore, if the Liquid Split increases, reduce Vapor Split to maximize recovery.

12
Optimization
The operational levers mentioned above are not exclusive of each other. As is often the case, a
combination of lever adjustments will yield optimal performance. The definition of optimal is not fixed.
The relative value of increased ethane recovery versus increased residue compressor power must be
determined for each facility regardless of whether dry ice formation is an issue. The possibility of dry ice
formation simply complicates the analysis by constraining the available process setpoints.
Presented below is a comparison of the operating points that yield the maximum dry ice-free ethane
recovery for 0%, 50%, and 100% Liquid Split at pressures from 270 to 320 psig. (Dry ice-free operation
is defined as not having a predicted dry ice formation temperation within 5 °F of the operating
temperature.) For each pressure and Liquid Split combination, the Vapor Split is adjusted to find the
maximum dry ice-free ethane recovery for that combination. Tables 4, 5 and 6 present the results for feed
composition A (1.5% CO2), B (1.8% CO2), and C (2.1% CO2). Moving from A to C, CO2 replaces
methane, while the rest of the gas composition is fixed to give a C2+ GPM of 2.3 gal/MSCF.
Table 4 shows that with 1.5% CO2 in the mixture, dry ice formation cannot be avoided at or below 290
psig with 0% Liquid Split. At 300 psig, the column is able to avoid dry ice formation if the Vapor Split is
reduced to 23%. This is lower than the typical GSP Vapor Split of 30% [4]. By operating at a reduced
Vapor Split, the heat integration is suboptimal. The resultant warmer tower allows more CO2 to escape in
the residue gas, reduces the concentration in the tower, and shrinks the dry ice-formation region of the
operating envelope. By shifting the pressure 10 psi higher, the Vapor Split can move to 30% without dry
ice formation, achieve better heat integration, and actually increase ethane recovery. (This is opposite of
the typical recovery decrease with higher column pressure.) For 50% and 100% Liquid Split, the tower is
able to operate with a Vapor Split that yields maximum heat integration. The recoveries and residue
compressor power go down as the pressure is increased, except as noted above for the suboptimal case.
Propane recovery seems to decrease more with increased Liquid Split than does ethane recovery.
In Table 5 with higher CO2 feed concentration (1.8%) there are fewer options for dry ice-free operation.
Below 300 psig, the 100% Liquid Split is the only one that avoids dry ice. Below 290 psig, even with
100% Liquid Split, the Vapor Split has to be set lower than what achieves the lowest residue gas
temperature. Due to this, as the tower pressure is changed from 270 to 290 psig the achievable ethane
recovery only decreased 0.3%, while the residue power decreased by 9%. At 300 psig, 50% Liquid Split
becomes an option but requires a reduced Vapor Split, making the ethane recovery lower than at 100%
Liquid Split. The same trend holds that propane recovery decreases more than ethane recovery with
higher Liquid Split.

13
 Table 4. Recoveries and residue compression power at conditions producing maximum ethane
recovery. (Feed composition A, 1.5% CO2)

Recovery, %
Top-stage
Liquid Split, % Vapor Split,% C2 C3 Res. Comp Power, hp
pressure, psig
0 * -- -- --
270 50 28 93.1 99.1 2778
100 27 92.7 98.9 2771
0 * -- -- --
280 50 28 92.5 99.0 2655
100 27 92.1 98.8 2648
0 * -- -- --
290 50 29 91.7 99.0 2541
100 27 91.4 98.7 2530
0 23 88.8 99.2 2463
300 50 28 90.8 98.9 2431
100 28 90.5 98.6 2425
0 30 89.4 99.0 2345
310 50 29 89.6 98.7 2332
100 27 89.4 98.4 2318
0 29 87.6 98.9 2242
320 50 29 87.9 98.6 2229
100 27 87.9 98.3 2215
*At these conditions, no Vapor Split was found that maintained a 5 °F buffer above dry ice formation temperature.

Table 5. Recoveries and residue compression power at conditions producing maximum ethane
recovery. (Feed composition B, 1.8% CO2)

Recovery, %
Top-stage
Liquid Split, % Vapor Split,% C2 C3 Res. Comp Power, hp
pressure, psig
0 * -- -- --
270 50 *
100 21 91.2 98.6 2761
0 * -- -- --
280 50 * -- -- --
100 24 91.1 98.7 2632
0 * -- -- --
290 50 * -- -- --
100 27 90.9 98.7 2516
0 * -- -- --
300 50 23 88.3 98.6 2433
100 28 89.8 98.6 2415
0 * -- -- --
310 50 28 88.7 98.7 2321
100 28 88.5 98.4 2313
0 29 86.7 98.9 2238
320 50 29 87.1 98.5 2227
100 28 87.1 98.2 2220

14
With 2.1% CO2, Table 6 shows a more limited range of dry ice-free conditions. Even with 100% Liquid
Split the column must be operated with a suboptimal Vapor Split.

Table 6. Recoveries and residue compression power at conditions producing maximum ethane
recovery. (Feed composition C, 2.1% CO2)

Recovery, %
Top-stage
Liquid Split, % Vapor Split,% C2 C3 Res. Comp Power, hp
pressure, psig
0 * -- -- --
310 50 * -- -- --
100 25 87.2 98.2 2301
0 * -- -- --
320 50 * -- -- --
100 27 86.1 98.2 2206
*At these conditions, no Vapor Split was found that maintained a 5 °F buffer above dry ice formation temperature.

Tables 4-6 compare the effect of CO2 concentration with a relatively lean gas (2.3 C2+ GPM). Table 7
shows how a richer gas (Composition E, 5.0 C2+ GPM) behaves. GSP processes with rich gas feeds
typically need external mechanical refrigeration [5]. In this case mechanical refrigeration is used to cool
the low-temperature separator to -30 °F. (The refrigeration power listed in Table 7 assumes a power to
duty ratio of 250 hp/MMBtu.)
It can be seen in Table 7 that ethane recoveries are lower than with the leaner gas. As ethane recovery
decreases, so do CO2 recovery and concentrations in the tower. This, along with more liquids in the
column, moves the dry ice formation region farther from the operating envelope. Consequently, except
for the 0% Liquid Split at 270 psig, the need to avoid dry ice formation does not affect the maximum
recovery Vapor Split in Table 7. The richer feed gas tends to suppress dry ice formation.
Some interesting observations can be made about Table 7. Some important parameters move inversely
with Liquid Split. As Liquid Split increases these items decrease,

 propane recovery (more so than with leaner feed gas)

 residue compression power (more so than with leaner feed gas)
 refrigeration power
However, there appears to be a maximum ethane recovery between 0% and 100% Liquid Split.

15
 Table 7. Recoveries and residue compression power at conditions producing maximum ethane
recovery. (Feed composition E, 1.8% CO2, 5.0 C2+ GPM)

Recovery, % Power, hp
Top-stage
Liquid Split, % Vapor Split, % C2 C3 Residue Refrigeration
pressure, psig
0 28 84.1 99.4 2518 616
270 50 25 88.4 98.7 2472 666
100 18 85.4 97.9 2418 613
0 35 87.0 99.4 2446 709
280 50 25 87.3 98.6 2358 683
100 18 84.5 97.7 2304 634
0 36 85.3 99.3 2333 744
290 50 25 86.2 98.4 2241 713
100 18 83.5 97.5 2185 673
0 37 83.7 99.2 2210 808
300 50 25 85.1 98.2 2120 763
100 18 82.6 97.2 2064 728
0 38 81.9 99 2060 933
310 50 25 83.9 98.1 1981 855
100 18 81.6 97.0 1925 828
0 38 80.0 98.9 1824 1216
320 50 25 82.6 97.9 1797 1053
100 18 80.5 96.7 1741 1030

If the feed gas CO2 content rises, most of the trends seen in Tables 4-7 continue. This is shown in Table 8,
where the feed CO2 content increases to 3.8%, while maintaining the same richness (5.0 C2+ GPM) as in
the previous example. At the higher CO2 content, a higher Liquid Split and/or lower Vapor Split is
necessary to avoid dry ice formation at lower pressures, similar to the leaner gas. Also, with the richer
gas, more pronounced difference in recoveries, compression power, and refrigeration power versus Liquid
Split are still present.

16
 Table 8. Recoveries and residue compression power at conditions producing maximum ethane
recovery. (Feed composition F, 3.8% CO2, 5.0 C2+ GPM)

Recovery, % Power, hp
Top-stage
Liquid Split, % Vapor Split,% C2 C3 Residue Refrigeration
pressure, psig
0 * -- -- -- --
270 50 * -- -- -- --
100 16 83.3 97.5 2403 678
0 * -- -- -- --
280 50 20 82.6 98.0 2341 672
100 16 82.4 97.3 2301 680
0 * -- -- -- --
290 50 25 84.1 98.1 2266 742
100 16 81.6 97.1 2200 688
0 * -- -- -- --
300 50 25 83 97.9 2163 755
100 16 80.7 96.8 2098 705
0 * -- -- -- --
310 50 25 82 97.7 2060 778
100 16 79.8 96.6 1994 738
0 * -- -- -- --
320 50 25 80.9 97.5 1953 819
100 16 78.9 96.3 1889 778
*At these conditions, no Vapor Split was found that maintained a 5 °F buffer above dry ice formation temperature.

Conclusion

Dry ice formation in a demethanizer tower is a problem that must be managed as feed CO2 levels
increase. There are multiple operational levers in a GSP facility that can be adjusted to avoid dry ice,
including tower pressure, Inlet Reboiler Split, Vapor Split and Liquid Split. The effect of these levers is
summarized below.

 Tower pressure—Higher tower pressures may allow operation above the S-V-L region of the
phase envelope. All else being equal, higher pressure will warm the tower and decrease
recoveries.
 Inlet Reboiler Split—Increased flow to the bottom reboiler will increase the bottom temperature,
which will force CO2 back up the column and lead to higher CO2 concentration in the top of the
tower. This could cause dry ice formation. Decreased flow to the reboiler will lower the bottom
temperature, which will increase NGL CO2 and CH4 concentrations.
 Vapor Split—There is a vapor split that will produce the coldest overhead product temperature,
which typically corresponds to maximum ethane and propane recovery. To avoid dry ice
formation, it may be necessary to adjust the vapor split to warm the column.
 Liquid Split—Increasing the Liquid Split tends to suppress dry ice formation, but does have an
adverse effect on propane recovery for both rich and lean feed gas, and ethane recovery for lean
gas. With richer gas, there appears to be an optimal Liquid Split between 0% and 100% that
maximizes ethane recovery.

This summary is captured in Table 9.

17
 Table 9. Effect of operating levers on plant performance.
Lever direction set to reduce likelihood of dry ice formation.

Compressor
Direction to Likelihood C2 Recovery/ Bottoms
Lever Power/Fuel
move of Dry Ice NGL production Temp
Usage

Column Pressure

Reboiler Inlet Split

*
Liquid Split

**
Vapor Split

* For richer gases, there appears to be a Liquid Split between 0% and 100% that maximizes C2 recovery.
** There is a vapor split that maximizes recovery and NGL production.

Equipment limitations and process economics will dictate the available range of adjustment for these
levers. A thorough analysis can help find the optimal combination.

The steps below are suggested as a methodology to determine the right set of conditions for a facility to
maximize profitability while avoiding dry ice formation.
1. Work with the Operations team to determine the operating limits of the existing equipment, and
the product specifications that must be maintained when changing process settings. For instance,
at the Enbridge Longview facility the column pressure is controlled by the speed of the residue
compressors. Lower compressor speed causes the column pressure to increase. However, the
residue compressors have a minimum flow limit. This flow limit can prevent higher tower
pressures when the feed flowrate to the demethanizer is low.
2. Develop a working simulation model that can reproduce plant operating data. Keep in mind that
the plant might not be in “as new” condition, which might make it more difficult to match
operating data. As an example,
a. In the Longview plant, the Operations group was deeply concerned as to the actual flow
through the side reboiler. All passes of the gas-gas heat exchanger had been previously
cleaned and hydrostatically tested with no indications of any leaks and/or excessive
fouling. Tests indicated that the side reboiler flow was essentially zero (based on
temperature profiles around the reboiler as the plant was manipulated). Radiography of
the reboiler piping revealed no obvious metallic obstructions. The plant was shutdown
and the piping was probed with remote video cameras where it was discovered that the
inlet piping to the reboiler, from the tower, was completely plugged with packing. This
led to a longer duration shutdown where the tower was unpacked, copious amounts of
18
 packing removed from in between the flow distributors between the packing beds, and
the tower was repacked. Further investigation of the flow distributors revealed that they
had not been manufactured in accordance with the manufacturer’s drawings. When
pressure surges had occurred over time, packing had worked its way through the flow
distributors and ultimately in between the packing beds and into the reboiler side draw
piping.
b. The first iteration of the simulation model assumed that the reboiler feed was a total draw
from the middle of the column. To fit the data better, this had to be modeled as a partial
draw instead.
3. Obtain management support to manipulate the plant and encourage the operators to move away
from the thinking of “that is not the way we do it”.
4. With a validated model in hand, work with Operations to develop an overall test plan. The model
can be used to produce expected results for a set of operating setpoints for the tower pressure,
Inlet Reboiler Split, Vapor Split and Liquid Split. Confirm with Operations the desirability of
these results. For instance, setpoints to maximize recovery may cause excessive C1/C2 ratios or
unacceptable NGL CO2 concentration.
a. Establish a series of scenarios to test how well the plant responds in accordance with the
predictions made by the process model. For example, at the Longview plant, several
scenarios were developed to test the plant response to raising the tower pressure in 10 psi
increments. The first goal of the test was to determine the adverse effects of operating the
plant at higher tower pressures. In this case, due to customer requirements, the tower
pressure could not be raised by more than about 35 psig over normal operating
conditions. This constraint was then carried forward into the development of other
scenarios.
b. Include in the plan a pathway for moving from current setpoints to new setpoints. The
simulation model can be used to determine if intermediate steps between current and
desired setpoints might be susceptible to dry ice formation.
c. Expect the unexpected, since the simulation model might not have captured everything.
Think in terms of a long-distance race, not a short-distance sprint. Make changes slowly
and maintain the rigor and discipline to not initiate a new change until all of the
observations from the existing change have been completed and the resulting behavior
understood.
d. If unexpected behavior does occur, determine whether the model should be refined to
incorporate this behavior while still maintaining fidelity to previous operating data. If
necessary, revise the test plan before moving forward.

19
 References

[1] ProMax, 4.0.16308, Process Simulation Software, Bryan Research & Engineering, Inc., 2017

[2] Poe, W. A., and Mokhatab, S., Modeling, Control, and Optimization of Natural Gas Processing
Plants, Elsevier, 2017

[3] Wilkinson, J. D. and Hudson, H. M., “Turboexpander Plant Designs Can Provide High Ethane
Recovery Without Inlet CO2 Removal”, Oil & Gas Journal, May 3, 1982.

[4] Mehra, Y. R., Gaskin, T., “Guidelines offered for choosing cryogenics or absorption for gas
processing”, Oil & Gas Journal, May 1, 1999.

[5] Hlavinka, M. W., Hernandez, V. N., and McCartney, D., “Proper Interpretation of Freezing and
Hydrate Prediction Results from Process Simulation”, Proceedings of the 85th Annual GPA
Convention.

[6] GPSA Engineering Data, Thirteenth Edition, Gas Processors Suppliers Association, 2012.

[7] Campbell, R.E. and Wilkinson, J.D., U.S. Patent No. 4,157,904.

20