New Materials and Process Development for Energy-Efficient Carbon

Capture in the Presence of Water Vapor

Investigators All investigators are at Northwestern University. Note that the experimental part of the

project is described in a separate document, although experiment and modeling are

tightly integrated in the project.

Randall Q. Snurr, Professor, Chemical & Biological Engineering

Fengqi You, Assistant Professor, Chemical & Biological Engineering

Hanyu Gao, Graduate Researcher, Chemical & Biological Engineering

Emmanouil Tyllianakis, Post-doctoral Researcher, Chemical & Biological Engineering

Abstract

Using quantum chemical methods, we have screened a variety of functional groups

for their ability to bind CO2 and water. Most functional groups bind water more strongly

than CO2, making them ineffective for incorporation into materials for removing CO2

from humid flue gas. Preliminary calculations suggest that adding perfluorinated

functional groups strongly suppresses water binding, while having little effect on CO2

binding. Preliminary process-level modeling has shown that if materials are not available

for capturing CO2 from humid flue gas, it may be economically feasible to remove much

of the water from the flue gas prior to CO2 removal. This opens up a wider range of

material and process combinations for cost-effective CO2 capture.

Introduction

Adsorption processes based on porous solids are a promising technology for

removing CO2 from power plant flue gas. The premise of this proposal is that game-

changing improvement of adsorption separation processes for carbon capture and storage

(CCS) will require simultaneous development of new materials and specially designed

processes that take advantage of these new materials. Competitive adsorption of water is

regarded as the single greatest technical hurdle to adsorption-based CCS, but there is very

little research on new water-tolerant CCS adsorbents. One goal of this project is to

develop new metal-organic framework (MOF) materials with extraordinarily high

selectivity for CO2 over nitrogen and water, together with suitably high absolute capacity

for CO2. Several related strategies are being explored, and molecular modeling is being

used to guide the synthetic effort. In addition, we are using state-of-the-art process-level

modeling to optimize adsorption processes around the new class of sorbents and explore

the real efficiency limits of MOF-based adsorption processes that meet desired CCS

technical and economic criteria, while minimizing the life-cycle environmental impact.

Adsorption processes are already used in large-scale applications, such as air separation,

with high reliability. Discovery of new, water-stable and water-tolerant adsorbents for

CO2 capture and new, optimal process configurations for these sorbents would be a step-

out development in CCS.

Background

Research on MOFs continues to explode. Many research groups are investigating

MOFs for separation of CO2/N2 mixtures, motivated by carbon capture. However, very

few papers report results that take water into account. A small number of papers have

appeared recently on the stability of MOFs, including their stability in water or humid gas

streams.

Results

Molecular-level Modeling

We are using molecular-level modeling to 1) screen new candidate materials and 2)

obtain insights into the structures that provide the best performance. One focus this year

was to screen functional groups that could be incorporated into existing MOF materials

for selective adsorption of CO2 over N2 in the presence of water vapor. ZIF-8 was chosen

as a candidate MOF for functionalization due to its stability in the presence of water

(even under boiling water conditions). In addition, this material has been shown to be

very hydrophobic and to adsorb almost no water. This has been demonstrated

experimentally [1-2] and by simulations from our group [3].

A variety of different functional groups were tested for their ability to enhance the

target properties of ZIF-8. The binding energies for water and CO2 were calculated as an

initial measure of the ability to adsorb CO2 in the presence of water. The goal was to find

functional groups that would increase the adsorption of CO2 compared to the

unfunctionalized material but that would not adsorb too much water. Quantum

mechanical calculations were performed to obtain the binding energies. Calculations

were performed on a small and a large cluster extracted from the ZIF-8 structure, as

shown in Figures 1. For the small clusters, we used a high level of theory, namely second

order Møller–Plesset perturbation theory (MP2) with large basis sets such as 6-

311++G(2d,p). For the large clusters, we used less expensive density functional methods

using the wB97xD functional and a 6-31G(d,p) basis set.

Figure 1. Small and large clusters of ZIF-8 used for the quantum chemical

calculations.

For the large cluster, the different functional groups were added not only on the

central ring of the cluster but on all of the six peripheral rings as well. This was done to

observe any possible synergetic effect between the functional groups of adjacent rings,

and indeed we did observe such effects for some functional groups, where the interaction

of guest molecules with the cluster was significantly enhanced when groups were present

on every ring of the cluster rather than only on the central ring.

As expected from our choice of functional groups, we observed an increase in the

interaction energy between the ZIF cluster and the CO2 molecules upon functionalization

with all of the functional groups tested. In some cases, the enhancement was quite

pronounced, for example an increase in the magnitude of the binding energy from -13

kJ/mol to -94 kJ/mol. At the same time, when these groups were tested for their ability to

interact with water, the same trend was observed: in every case, the functional group

increased the binding of water as well. There was no functional group among the ones

tested that could bind CO2more strongly than H2O, supporting our initial idea that simple

functionalization was unlikely to be a successful strategy.

Functionalization with perfluoro group was given special attention. In a similar set of

calculations, the ZIF-8 methyl groups on the imidazole ring (Figure 1) were replaced by

perfluorinated methyl groups. The perfluorination of the methyl groups resulted in a

significant decrease in the interaction with water (from -36 kJ/mol to -24 kJ/mol), making

the material more hydrophobic. At the same time, this change had almost no effect on the

interaction with CO2, with binding energies of-13 kJ/mol for the original cluster and -12

kJ/mol for the cluster with the perfluorinated methyl groups.1 Thus, the material with

perfluoromethyl groups is predicted to be significantly more hydrophobic than the parent

material but to adsorb essentially the same amount of CO2. While the parent ZIF-8

material is not particularly good for CO2 capture, this finding may allow us to find other

materials that can be made hydrophobic while retaining their ability to adsorb CO2.

Process-level Modeling

There are two main objectives of the process-level modeling in this project: first, to

design a system that utilizes materials discovered in the project and, second, to provide

feedback to the materials discovery effort on the desired material properties. While our

primary focus is the development of materials that can adsorb CO2 in the presence of

water, we thought it worthwhile to first investigate how much it would cost to remove

water vapor from the flue gas prior to the CO2/N2 separation. A new graduate student

was recruited for this effort, and during the past several months, we designed several

different processes for water removal and determined the corresponding cost under

different requirements and operating conditions.

The total cost of the water removal system was estimated using a framework

proposed by Hasan et al. [4]. The total annualized cost (TAC) was divided into the

annualized investment cost (AIC) and the annual operating cost (AOC). The detailed

scheme for calculating the AIC is shown in Table 1. The AOC was determined by the

usage of utilities including process utilities (cooling water, steam, refrigerant) and

electricity. All equipment costs and utility costs were estimated using the Aspen Process

Economic Analyzer.

1It should be noted that with the simple cluster models, these energies are not expected to reflect the

adsorption enthalpies measured in the real materials. Rather, we are seeking trends with these calculations.

Table 1. Parameters for the economic analysis

parameter Value

Equipment (mover) installation cost (EIC) 80% of equipment

purchase cost(EPC)

Equipment (column, exchanger, etc.) installation cost (EIC) 4% of EPC

Total installed cost (TIC) All EPCs and EICs

Indirect cost (IDC) 32% of TIC

Balance of plant cost (BPC) 20% of TIC

Total plant cost (TPC) TIC+IDC+BPC

Annual maintenance cost (AMC) 5% of TPC

Capital recovery factor 15.4%

There are many different technologies for water removal. We mainly focused on two

of them, absorption with triethylene glycol (TEG) and cooling & condensation. Other

technologies include compression and condensation, cryogenic separation, membranes,

and solid adsorption. Compression and condensation was not considered because of the

high installation cost and electricity consumption for compressing the entire flue gas

stream. Cryogenic separation requires extensive use of coolant. Membrane technology is

also reported to cost more than TEG absorption by Hasan et al. [4] and was not

considered here. Solid adsorption needs further evaluation.

TEG absorption and cooling & condensation processes were simulated using Aspen

Plus. We assumed feed flue gas available at 1 bar and 55℃ with 15% H2O, 69.5% N2,

5.5% O2 and 10% CO2 (volumetric fractions). The flue gas then goes through a direct

contact cooler where it is saturated with water at 35℃ (corresponding to 5.5% H2O). The

flow rate of this stream was taken to be 1kmol/s (which is about the CO2 emission of a 30

MW power plant) for simplicity.

The process flow diagram for cooling & condensation is shown in Figure 2. Wet flue

gas is cooled down, and the liquid phase is separated from the vapor phase in a flash tank

operating at 1 atm. The temperature of the cooled mixture exiting the heat exchanger

depends on the requirement set for the water content in the dry gas. For example, if we

require less than 1% water in the dry gas, the cooling temperature is 75˚F. If the

requirement is 0.1%, the cooling temperature has to be 41˚F.

Figure 2. Cooling and condensation process flow diagram.

The cost of this process was estimated for different requirements set on the water

content. Refrigerant at 10˚F was assumed to be available at $1.70 per ton-day [5]. With a

1% water requirement, the AIC and AOC are $39,504 and $296,154, respectively. The

total plant cost (TPC) is then $335,658, and the water removal costs $2.4 per ton of CO2

captured. With a 0.5% water requirement, the cost per ton of CO2 captured is $4.2. The

cost of cooling and condensation is low, because it is a simple process. The only utility

requirement is refrigeration, and the cooling temperature is not very low. However, this

simple process leads to significant loss of CO2, especially when requiring low water

content in the dry gas. For example, when a 0.1% water requirement is set, more than

25% of the CO2 is lost from the bottom of the flash tank. Our goal is to capture at least

90% of the CO2. If we instead require no more than 10% loss of CO2, the water content

can only go down to as low as 0.5%, which places significant restriction on this

technology.

The other technology studied for dehydration was TEG absorption. The process flow

diagram is given in Figure 3. Wet flue gas is contacted with 99.5% TEG in the absorption

column and dry gas is obtained from the top. TEG absorbed with water is then

regenerated using a flash tank operating at high temperature and low pressure.

Figure3. TEG absorption process flow diagram.

The cost of TEG absorption is generally higher than cooling & condensation under

the same requirement on water content. Figure 4 shows the cost per ton of CO2 captured

for the TEG absorption process under different requirements when the flash tank is

operating at 120℃ and 0.04 bar and the cooler outlet temperature is 65℃. The general

range is between $10 and $20 per ton of CO2 captured, while the cost goes to $73/ton

CO2 captured when the water requirement is 0.1%. This is because the operating

condition is far from optimal for this case.

Figure 4. Dependence of cost on water requirement in the outlet stream for TEG

absorption.

TEG absorption can achieve good separation of H2O from flue gas without sacrificing

CO2 recovery. The loss of CO2 is less than 3% even when the water percentage in the dry

gas is set to 0.1%. Unlike cooling & condensation, the operating conditions of TEG

absorption are adjustable even when the requirement on water removal is fixed. Changing

the temperature and pressure of the flash tank and the temperature of cooler outlet can

significantly influence the cost of the process, so the operating conditions need to be

optimized in order to achieve lower costs. The three main process variables are the flash

temperature (ranging from 100-140℃), the flash pressure (0.04-0.2 bar), and the cooler

0

20

40

60

80

0.0% 0.2% 0.4% 0.6% 0.8% 1.0% 1.2%

cost

pe

r to

n C

O2

cap

ture

d

(in

$)

water persentage required

outlet temperature (45-65℃). For the 1% water requirement case, 30 sampling points

were taken within these ranges using the Latin Hypercube Sampling method in

MATLAB. The cost was estimated for each sampling point. Among these sampling

points, the cost of CO2 capture ranges from $5.8 to $18.1 per ton of CO2 captured, with

the lowest cost found when the flash temperature is 108℃, flash pressure is 0.09 bar and

cooler outlet temperature is 61℃. For the 0.1% water requirement case, the lowest cost is

found at $10.9 per ton of CO2 captured when the flash temperature, flash pressure and

cooler outlet temperature are 140℃, 0.04 bar, and 40℃, respectively.

Progress

Our preliminary process-level modeling indicates that removing water prior to

capturing CO2 from flue gas may be economically and technically feasible. The question

of how much of the water to remove requires an integration of the process modeling and

the materials development and will be one focus in the coming year.

Development of fluorinated MOFs may provide a solution to the difficult problem of

selectively adsorbing CO2 over N2 in the presence of water vapor.

Future Plans

Computational screening of MOFs will continue in close association with the

experimental effort, as described in the original proposal. One new direction is suggested

by the process-level modeling, namely, an assessment of how MOFs perform in the

presence of a small amount of water vapor (as opposed to a stream that is essentially

saturated with water vapor).

We will finish our assessment of the cost of water removal. We have made good

progress in writing a MATLAB code for modeling adsorption units and will soon begin

modeling carbon capture processes based on MOF materials.

Publications and Patents There are no publications or patents yet.

References 1. V. Colombo, S. Galli, H.-J. Choi, G.-D. Han, A. Maspero, G. Palmisano, N. Masciocchi, J. R.

Long, Chem. Sci., 2, 1311 (2009).

2. P. Küsgens, M. Rose, I. Senkovska, H. Fröde, A. Henschel, S.Siegle, S. Kaskel, Micropor.

Mesopor. Mater.,120, 325 (2009).

3. P. Ghosh, K.C. Kim, R.Q. Snurr, manuscript in preparation.

4. M.M.F. Hasan, R.C. Baliban, J.A. Elia, C.A. Floudas, C. A., Ind. Eng. Chem. Res., 51, 15665

(2012).

5. W.D. Seider, J.D. Seader, D.R. Lewin, S. Widagdo,Product and Process Design Principles:

Synthesis, Analysis, and Evaluation, 3rd ed.; John Wiley: Hoboken, NJ, 2009.

Contacts

Randall Q. Snurr, snurr@northwestern.edu

Fengqi You, you@northwestern.edu

New Materials and Process Development for Energy-Efficient

Carbon Capture in the Presence of Water Vapor Investigators All investigators are at Northwestern University. Note that the modeling part of the project is described in a separate document, although experiment and modeling are tightly integrated in the project. Omar K. Farha, Research Associate Professor, Chemistry Joseph T. Hupp, Professor, Chemistry Pravas Deria, Post-doctoral Researcher, Chemistry Zachary Brown, Graduate Student, Chemistry Olga Karagiaridi, Graduate Student, Chemistry Marianne Lalonde, Graduate Student, Chemistry Abstract The experimental portion of the project is currently focused on designing, synthesizing, characterizing, and functionally evaluating nanoporous materials that feature the following properties: a) excellent stability under conditions encountered in flue-gas environments, b) high capacity for gas uptake, c) ability to capture CO2 at pressures found in flue gas, and d) ability to inhibit competitive adsorption of water. Our initial strategy has focused on metal-organic framework (MOF) materials that present high surface areas and large total pore volumes. We have explored two approaches to incorporation of perfluorinated functional groups that can render MOF interior surfaces both hydrophobic and strongly attractive for carbon dioxide. 1. Decorating the Mesopores of the Zr-Based Metal-Organic Framework Material NU-1000 with Perfluoroalkyl Functional groups for Carbon Capture Under Humid Conditions. Among the requirements for effective sorption-based carbon capture are that the sorbent material: (a) have high capacity for CO2, (b) possess high selectivity for CO2 over N2, (c) be stable, and perform well, under humid condition of flue gas, and (d) display fast adsorption and desorption kinetics. It appears that isosteric heats of adsorption of ~50 kJ/mol (at 0.15 bar) would be required in order for a sorbent material to interact strongly enough with CO2 to capture 90% of the carbon present in flue gas while still allowing easy regeneration. For a sorbent material to function successfully under humid (~5% H2O in flue gas) conditions, it additionally must have a strongly preferential interaction with CO2 over H2O. MOFs that possess open metal coordination sites and exposed primary or secondary alkyl-amines often show high heats of adsorption. But, they are either water intolerant or exhibit excessive large isosteric heats of adsorption. The latter imply excessive energy consumption for subsequent carbon release and sorbent regeneration. To achieve these targeted physical properties within sorbent materials, we sought to functionalize a mesoporous MOF with fluorinated alkanes of various lengths. Notably, perfluoroalkanes are widely used as surfactants in processes involving

supercritical CO2. The strengths of interaction of C-F sites with carbon dioxide are believed to be 10 to 20 kJ/mol greater than strengths of interaction C-H sites with carbon dioxide. Additionally, perfluorinated materials typically are strongly hydrophobic, i.e. they typically exhibit little interaction with water.

We used the recently developed material, NU-1000 (Figure 1A), as a chemically robust and highly porous as a platform for grafting flexible fluoroalkane moieties onto the walls of a MOF. We identified conditions for quantitatively flexible, linear fluoroalkanes of various lengths (CnF2n+1COOH; n=1, 3, 5, 7, 9) to the MOF interior surface (Figure 1B), while keeping the overall structure intact (as evidenced in part by unchanged powder X-ray diffraction pattern data).

N2-gas adsorption results show a systematic decrease of (a) surface area, (b) pore diameter and volume, and (c) total gas uptake with increasing the decorated perfluoroalkane chain length (e.g. comparing CF3- to C7F15- functionalized NU-1000 MOF; the surface area, pore volume and total gas uptake reduced by ~1/2, whereas the pore width reduced by only ~2 Å). Preliminary CO2 adsorption studies suggest that going from CF3- to C7F15- functionalized NU-1000 MOF the isosteric heats of adsorption increase by ~6 kJ/mol while the water uptake is reduced by half. The latter changes are nicely correlated with the pore volumes measured for these two NU-1000 MOF derivatives (Figure 2A, B). Ongoing work is aimed at completing the adsorption study as a function of chain length, this novel perfluoroalkane decorated mesoporous MOF system will be further modified, next, to fine tune the physical adsorption parameters for CO2 and H2O uptake. These modifications would involve incorporation of bi-functional system, e.g. via installing CO2-philic and hydrophobic functionalities in the same Zr6 node.

A

B

Figure 1. (A) Structural units of NU-1000; (B) Cartoon depicting perfluoroalkyl functionalized NU-1000.

2. MOF channel functionalization for carbon capture via solvent assisted linker exchange (SALE). Linker exchange offers a second means of incorporating desired flouroalkane functional groups into porous MOFs (Figure 3). Although MOFs are often regarded as rigid, almost inert materials, we have shown that SALE can be applied to some of the most robust members of the MOF family, such as zeolitic imidazolate frameworks (ZIFs). We have synthesized ZIFs with unprecedented linker-topology combinations by replacing the 2-ethylimidazolate linkers in CdIF-4 (rho topology) with 2-methylimidazolate and the 2-methylimidazolate linkers in ZIF-8 (sod topology) with unsubstituted imidazolate. The latter case led to an increase in the aperture size of the sodalite cage, and a resulting uptake of solvents of larger kinetic diameter, which was confirmed with thermogravimetric analysis experiments.

We were able to further use the SALE strategy to create a series of hydrophobic zeolitic imidazolate frameworks (ZIFs.) ZIF-69 is a gme topology ZIF composed of a 1:1 ratio of 2-nitroimidazole and 5-chlorobenzimidazole.1 Upon soaking in a 2 M solution of 5-trifluoromethyl benzimidazole at 100 °C for 3 days, the 5-trifluoromethyl benzimidazole selectively replaces only the 5-chlorobenzimidazole up to 95%. The resulting material is SALEM-10, the first ZIF

A

B

Figure 2. Plot of (A) isosteric heat of adsorption of CO2 and (B) water adsorption isotherm in CF3- (black) to C7F15- (red) functionalized NU-1000 MOF. Water adsorption isotherm for a Cr3-cluster based mesoporous MIL-101(light purple triangle) and ZIF-8 (light cyan square) MOFs are presented for comparison purpose.

synthesized using selective SALE. (Figure 4.) SALEM-10 contains 14 Å trifluoromethyl-decorated channels, allowing for maximum interaction of potential sorbates with trifluoromethyl functional groups.

Figure 3. Schematic representation of SALE

Figure 4. 2-nitroimidazole is retained, while 5-chlorobenzimidazole is substituted for 5-trifluoromethyl benzimidazole, in the first example of selective SALE in ZIFs.

The procedure was repeated for ZIF-76 (lta, 1:1 imidazole and 5-chlorobenzimidazole), ZIF-78 (gme, 1:1 2-nitroimidazole and 5-nitrobenzimidazole) and ZIF-95 (poz, 5-chlorobenzimidazole.) In the ZIF-76 and ZIF-78 cases, the 5-trifluoromethyl benzimidazole replaces the 5-chlorobenzimidazole selectively, up to 90% and 92% respectively, resulting in SALEM-11 and SALEM-10-B. In the ZIF-95 case, 5-chlorobenzimidazole is the sole linker, and it is replaced up to 81% to form SALEM-12. (Table 1.)

Table 1: Parent ZIFs and Product SALEMs. Percent exchange represents the percentage of each benzimidazole component exchanged for 5-trifluoromethyl benzimidazole.

While fluorinated moieties are considered attractive for CO2 sorption,2 there were no increases in CO2 uptake from the ZIFs to their SALEM analogues. However, there was a significant increase in hydrophobicity from the SALEM to the analogous ZIF as based on contact angle measurements. Likewise, SALEM-10 showed a decreased water vapor uptake from ZIF-69. (Figures 5-7.) As water is a non-negligible component of flue gas, hydrophobicity is a desired trait of potential CO2 sorbents, especially as many MOFs are either unstable in water,3 or competitively adsorb water over carbon dioxide. Further studies are focusing on introducing other functional groups that may simultaneously enhance carbon dioxide sorption and inhibit competitive water sorption.

Figure 5. A summary of contact angle measurements for parent ZIFs (blue) and product SALEMs (green) averaged over 15 trials.

Figure 6. Representative contact angle measurements for ZIF-69 and SALEM-10.

Figure 7. Water vapor sorption measurements for SALEM-10 and ZIF-69.

References 1. Banerjee, R., Hiroyasu, F., Britt, D., Knobler, C., O’Keeffe, M., Yaghi, O. M. J. Am. Chem. Soc., 2009, 131, 3875. 2. Almantariotis, D., Gefflaut, T., Padua, A. A. H., Coxam, J.-Y., Costa Gomes, M. F., J. Phys. Chem. B., 2010, 114, 3608. 3. Küsgens, P., Rose, M., Senkovska, I., Fröde, H., Henschel, A., Siegle, S., Kaskel, S. Microporous Mesoporous Mater., 2009, 120, 325.