www.afm-journal.

de
www.MaterialsViews.com

FULL PAPER
Synthesis–Structure–Property Relationships for
Hyperbranched Aminosilica CO2 Adsorbents
By Jeffrey H. Drese, Sunho Choi, Ryan P. Lively, William J. Koros,
Daniel J. Fauth, McMahan L. Gray, and Christopher W. Jones*

cycle (IGCC), oxycombustion, chemical

Hyperbranched aminosilica (HAS) adsorbents are prepared via the looping, and membrane-aided solid oxide
fuel cell (SOFC) schemes.[3–5]
ring-opening polymerization of aziridine in the presence of mesoporous silica
Post-combustion capture technologies,
SBA-15 support. The aminopolymers are covalently bound to the silica such as absorption with liquid amines or
support and capture CO2 reversibly in a temperature swing process. Here, a adsorption with solid adsorbents, have been
range of HAS materials are prepared with different organic loadings. The proposed as possible retrofits to existing PC
effects of organic loading on the structural properties and CO2 adsorption plants that are integral to modern power
properties of the resultant hybrid materials are examined. The residual generation infrastructure. Liquid absorp-
tion technologies are mostly based on the
porosity in the HAS adsorbents after organic loading, as well as the molecular
chemical fixation of CO2 with aqueous
weights and degrees of branching for the separated aminopolymers, are solutions of amines. The benchmark for
determined to draw a relationship between adsorbent structure and this strategy is the use of amines, such
performance. Humid adsorption working capacities and apparent adsorption as monoethanolamine (MEA), although
kinetics are determined from experiments in a packed-bed flow system numerous amines have been investigated
over the last several decades.[6–10] However,
monitored by mass spectrometry. Dry adsorption isotherms are presented for
these liquid amine-based absorption tech-
one HAS adsorbent with a high amine loading at 35 and 75 -C. These nologies are very energy intensive and are
combined results establish the relationships between adsorbent synthesis, widely viewed to be potentially cost prohi-
structure, and CO2 adsorption properties of the family of HAS materials. bitive, although several commercial tech-
nologies exist.[3]
As alternatives, several classes of solid
1. Introduction CO2 adsorbents have been investigated, and each class has its own
advantages and disadvantages.[11] Zeolites and activated carbons
Recent consensus in the international scientific community capture CO2 via physical adsorption, and consequently are typically
suggests that anthropogenic CO2 emissions have caused a rise in hindered in the presence of water.[12,13] Metal oxides (CaO, MgO,
atmospheric CO2 concentration and, in turn, a measurable change etc.), lithium zirconates, hydrotalcites, and solid-supported
in the global climate.[1] The largest point sources for these amines capture CO2 via chemical adsorption. Metal oxides and
emissions are pulverized coal-fired power plants (PC plants) that hydrotalcites frequently degrade upon regeneration.[14,15] Many
account for about a third of all CO2 emissions in the USA.[2] The lithium zirconates suffer from slow adsorption kinetics.[16] Solid-
amount of energy required to operate under the status-quo dictates supported amines are susceptible to poisoning by flue gas
that, at least for the short- to mid-term future, fossil-fuel-based contaminants, such as SO2 and NOx.[17,18] Metal-organic frame-
electricity generation will be required until alternative energy works (MOFs) are an emerging class of solid adsorbents that
generation technologies can be adequately refined/scaled to meet capture CO2 by physical adsorption except in cases where amines
the ever-increasing societal demand. Numerous alternative fossil- are incorporated into the structure or, possibly, when open metal
fuel-based power generation schemes that include CO2 capture coordination sites are generated.[19,20] At this time, most MOFs
have been proposed, such as integrated gasification combined have relatively low adsorption capacities at low CO2 partial
pressures.[21]
A key objective in the design of a solid adsorbent is to increase
[*] Prof. C. W. Jones, J. H. Drese, Dr. S. Choi, R. P. Lively, Prof. W. J. Koros the available adsorption sites per mass of material to reduce the
School of Chemical & Biomolecular Engineering
Georgia Institute of Technology size of the required process equipment and the associated
311 Ferst Drive, Atlanta, GA 30332 (USA) operating cost. In a previous report, we described the synthesis of a
E-mail: [email protected] silica-supported amine adsorbent with a very high loading of
D. J. Fauth, M. L. Gray amine adsorption sites compared to previous covalently attached
U.S. Department of Energy amine adsorbents.[22] This adsorbent, called a hyperbranched
National Energy Technology Laboratory
626 Cochrans Mill Road, Pittsburgh, PA 15236 (USA)
aminosilica (HAS), was prepared by the acid-catalyzed, ring-
opening polymerization of aziridine off of a mesoporous silica
DOI: 10.1002/adfm.200901461 (SBA-15) support. The HAS adsorbent displayed superior CO2

Adv. Funct. Mater. 2009, 19, 3821–3832 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3821
 www.afm-journal.de
www.MaterialsViews.com

adsorption capacities from humid simulated flue gas compared to The functionalization of ordered mesoporous silicas through
FULL PAPER

other types of silica-supported amine adsorbents. The adsorption reaction with surface silanols has been extensively studied.[34,35]
capacity was fully regenerable over 11 adsorption/desorption Typically, these grafting reactions involve the use of silane
cycles. chemistry where alkoxy groups from a silane condense with the
Here, we describe the range of material characteristics and surface silanols to form covalent tethers to the silica surface
adsorptive properties attainable with HAS adsorbents by modify- through Si
O
Si bonds, incorporating organic functionalities
ing certain synthesis conditions, such as reaction solvent and through substituents on the silane molecule.[34] In some cases,
aziridine-to-silica weight ratio. An analysis of the aminopolymers these silane-functionalized materials can be further modified to
formed in these reactions and the behavior of the HAS materials create large amine-containing dendron structures that nearly fill
in dry and humid CO2 adsorption follows with the intent of the pore space.[36–38] The HAS synthesis directly incorporates
developing synthesis–structure–property relationships for this organic functionalities to the silica surface by forming Si
O
C
class of adsorbents. bonds through ring-opening polymerization with surface OH
groups, if the polymerization is carried out on a bare silica surface.
Alternately, aziridine can be polymerized off of silane-functional-
ized surfaces that contain nucleophilic reactive centers, such as
amines.
2. Results and Discussion HAS adsorbents with varied amine loadings (hereafter referred
to as the HAS varied loading adsorbent set, or HAS-VLAS) were
2.1. Synthesis and Characterization of HAS Adsorbents with prepared similar to reported procedures[2,22] by varying the weight
Varied Loadings ratio of aziridine to SBA-15 used in the synthesis. Table 1 displays
the amine loadings, pore volumes, pore diameters, and surface
The equilibrium CO2 capacities of solid amine-based adsorbents areas of the HAS adsorbents prepared as well as the aziridine-to-
are often used as the primary metric to evaluate their (SBA-15) weight ratios used in their syntheses and the resulting
performance.[11,23,24] This is because the higher the capacity of yields. The yields were reported as the percent of aziridine that
a given adsorbent, the lower the quantity of adsorbent that is remained attached to the silica support after the synthesis relative
required to capture a given amount of CO2 from a flue gas stream. to the amount added to the solution. The organic loadings were
Intuition suggests that the higher the equilibrium adsorption assessed by thermogravimetric analysis (TGA), and the amine
capacity the better; however, the importance of the other loadings were stoichiometrically estimated from this value. The
characteristics of an adsorbent should not be overlooked. For Brunauer–Emmett–Teller (BET) surface areas were determined at
instance, an adsorbent with very high capacity, but very slow 77 K by N2 physisorption (isotherms shown in Supplementary
adsorption kinetics would likely be useless for capturing CO2 from Information, SI). The spatial characteristics of the porosity were
flue gas in a practical way, where gas flow rates are high and fast determined from the N2 physisorption data interpreted by the
adsorption cycle times are required. For this reason, it is Frenkel–Halsey–Hill-modified Broekhof–de Boer method (BdB-
sometimes more useful to discuss capture capacities in terms FHH), which has been shown to be more accurate than the
of working capacities that are achievable in a fixed amount of time. Barrett–Joyner–Halenda method (BJH) for practical application to
For the silica-supported amine class of adsorbents, it has been mesoporous materials.[39–44] Confirmation of organic functional-
shown that there is indeed a tradeoff between capacity and kinetics, ization with aliphatic amines was determined for the HAS-VLAS
especially in cases where the silica supports are so highly loaded by Fourier Transform IR (FTIR) spectroscopy (see SI).
that there is little or no remaining pore volume.[11,25,26] Diffusion HAS adsorbents in the VLAS were prepared with amine
of CO2 into the interior of the adsorbent can be blocked or retarded loadings as low as 2 mmol N g
1 adsorbent and as high as
by organic functionalities that have drastically reduced the open 10 mmol N g
1 adsorbent, by adjusting the aziridine-to-silica
pore space. A set of HAS adsorbents possessing a range of organic weight ratio in the synthesis (Table 1). Thus, the amine loading of
loadings were synthesized to investigate these issues. These the HAS adsorbent is tunable and controllable within these limits.
adsorbents consist of a mesoporous silica support (SBA-15) However, as with all silica-supported amine adsorbents, there is a
covalently functionalized with a hyperbranched aminopolymer physical upper-bound to the amount of amines that can be loaded
with ethylenimine repeat units, achieved by the acid-catalyzed into the pores of the support that is equal to its total pore volume.
ring-opening polymerization of aziridine from silanol groups on The SBA-15 used for much of this study had a pore volume of
the silica surface. It should be noted that a silicon-based material 0.78 cm3 g
1; thus the theoretical maximum amount of
was first functionalized with aminopolymer in this manner by aminopolymer that could be loaded into the pore space is
Park and co-workers, although they worked on flat substrates.[2] 0.84 g organic g
1 silica, which translates to 10.6 mmol N g
1
Subsequently, SBA-15 was also functionalized with aminopoly- adsorbent.[45] To put this into perspective, pure PEI, an
mers by aziridine polymerization in the work of Linden for a aminopolymer widely used in constructing effective adsorbents,
different application,[27–31] and by our group for CO2 capture.[22] has an amine loading of 23 mmol N g
1 polymer.[24,32,33,46–49]
Many researchers have studied a similar material that is prepared However, pure PEI is a viscous liquid and thus, a poor candidate as
by the physical impregnation of already-synthesized poly(ethyle- a single-component liquid absorbent. Nevertheless, it is important
neimine) (PEI) into mesoporous silica.[24,32,33] However, the to note that, unless another polymer with higher amine content is
amine in these materials are not covalently bound to the silica used, the maximum amine loading for an amine-based solid
support and thus may be susceptible to leaching under certain adsorbent is just under 23 mmol N g
1, regardless of the type of
conditions.[22] solid support.

3822 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 3821–3832
 www.afm-journal.de
www.MaterialsViews.com

Table 1. Compositions and physical characteristics of HAS adsorbents.

FULL PAPER
Sample Amine Aziridine-to- Yield [a] Pore BET Pore Pore Micropore Aminopolymer Remaining
loading (SBA-15) ratio [mol N mol
1 aziridine] diameter surface volume volume volume volume [b] pore volume [c]
[mmol N g
1] [w w
1] [nm] area [cm3 g
1] [cm3 g
1 SiO2] [cm3 g
1] [cm3 polymer g
1 SiO2] [cm3 g
1 SiO2]
2
1
[m g ]

HAS1 [d] 9.78 2.65 0.29 4.9 45 0.11 0.20 0.00 0.72 -0.11
HAS2 [d] 8.30 2.03 0.30 5.0 71 0.15 0.25 0.00 0.56 -0.01
HAS3 [d] 7.25 1.33 0.37 4.9 119 0.25 0.38 0.00 0.46 -0.03
HAS11 [e] 5.27 5.01 0.06 5.4 234 0.39 0.55 0.01 0.30 -0.07
HAS13 [f ] 3.65 0.53 0.38 5.9 278 0.49 0.64 0.02 0.19 -0.02
HAS10 [e] 2.84 0.33 0.43 6.4 314 0.51 0.61 0.03 0.14 0.04
HAS14 [f ] 2.27 0.13 0.86 6.3 579 0.69 0.81 0.08 0.11 -0.11
SBA1 0 6.2 865 0.81 0.81 0.08
SBA2 0 5.9 840 0.78 0.78 0.06
SBA3 0 6.3 870 0.81 0.81 0.09

[a] Calculated as mol N g
1 SiO2 in the prepared HAS divided by the reactant ratio [mol N g
1 SiO2]. [b] Calculated as aminopolymer loading obtained from
TGA divided by the density of commercial PEI (1.07 g cm
3). [c] Theoretical unaccounted-for pore volume, calculated by subtracting the adsorbent pore
volume and calculated aminopolymer volume from the pore volume of unfunctionalized SBA-15. [d] SBA1 used as silica support in synthesis. [e] SBA2 used
as silica support in synthesis. [f ] SBA3 used as silica support in synthesis.

As the aziridine-to-(SBA-15) reactant ratio was increased from 0.09 cm3 g
1. Linden and co-workers claimed that aminopolymer-
values below 1 to about 3, the HAS amine loading obtained functionalized SBA-15 first filled the silica micropores and
increased in a roughly linear fashion, as shown in Figure 1. When a subsequently the mesopores.[29,31] They also suggested that the
very large reactant ratio of 5 was used, the amine loading decreased nitrogen adsorption isotherms indicated that no partial pore
to a value below the maximum, suggesting that a large amount of blocking occurred after functionalization, as the adsorption and
aziridine may have polymerized in solution before it could diffuse desorption branches were essentially parallel and exhibited
into the pores of the silica particles. Aminopolymer formed in a narrow hysteresis.[29] Unfortunately, the authors did not report
solution in this manner was washed away from the silica particles the aziridine-to-silica ratios used in their syntheses, so a direct
during workup. comparison to the work here is not possible. The results from
The BET surface areas, BdB-FHH pore volumes, and pore Table 1 indicate that micropore volume does not disappear in
diameters were calculated from N2 physisorption data to the HAS materials described here until relatively high amine
determine the HAS-VLAS’s pore characteristics (Table 1). When loadings are obtained, higher than those reported for
the organic loadings of the HAS materials increased, the BET aminopolymer-functionalized, calcined SBA-15 prepared by
surface areas and pore volumes decreased. A plot of surface Linden and co-workers.[29] It should be noted that SBA-15
area and pore volume versus amine loading is shown in Figure 1. materials prepared under slightly different conditions may
There was a sharp decrease in both characteristics as amine have significantly different amounts of micropores and micropore
loading increased at low values. Surface area and pore volume size distributions, and this may account for some of the differences
linearly decreased at increasing amine loadings, eventually seen in these studies.
reaching plateaus with the highest loaded materials. The three The average pore diameters found for the HAS-VLAS also
unfunctionalized SBA-15 silicas used for these syntheses had decreased as the amine loading of the HAS material increased
BET surface areas of 840 to 870 m2 g
1 and total pore volumes of from low to moderate loadings (Fig. 2). However, at higher
0.78 to 0.81 cm3 g
1, including micropore volumes from 0.06 to loadings, the pore diameters remained at 5 nm, and there was

Figure 1. BET surface area and pore volume versus amine loading of the Figure 2. Pore diameter distributions from N2 physisorption at 77 K for the
parent HAS adsorbent. HAS-VLAS.

Adv. Funct. Mater. 2009, 19, 3821–3832 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3823
 www.afm-journal.de
www.MaterialsViews.com

little or no adsorption from pores below 4.5 nm in size. This trend reached, which may coincide with a nearly complete coverage of
FULL PAPER

was also reported for the polymerization of isopropylacrylamide the silica surface.
monomers on SBA-15, where substantial inclusion of polymer Indeed, it is possible to selectively functionalize the exterior
(27.5 wt%) resulted in only a 1 nm reduction in pore diameter.[50] If silica surface with aminopolymer. Linden and co-workers achieved
increasing organic inclusion into the silica pores occurred evenly a loading of 1.7 mmol N g
1 material for aminopolymer-function-
via layer-by-layer growth of aminopolymer on the pore walls, the alized nonporous silica nanoparticles.[27] In a separate work,
pore diameters of highly loaded materials where the pores are Linden and co-workers claimed to preferentially functionalize the
almost completely filled, but not blocked, should continue to exterior of SBA-15 particles with 6.55 mmol N g
1 material by
decrease in approximately a linear fashion approaching zero. performing aziridine polymerization on as-prepared SBA-15 and
Instead, it appears that as pores were filled with aminopolymer, later removing the pore-forming organic template via solvent-
functionalization occurred in the majority of the pores until a aided extraction.[28] After extraction of the template, this material
critical diameter of 5 nm was reached. Past this point, we suggest had a pore volume of 0.99 cm3 g
1 SiO2, compared to a pore
additional polymer inclusion occurred as pores were closed or volume of 1.16 cm3 g
1 SiO2 for unfunctionalized, template-
blocked by the aziridine polymerization, leaving the average pore extracted SBA-15. The authors claimed that because the two
diameter unchanged at the critical diameter. This hypothesis is materials possessed the same average pore diameters, at least a
consistent with the observation of there being a critical pore portion of the added polymer was located outside of the pores. It is
diameter, after which the pore size does not decrease further, but also possible that some of the organic template of the as-prepared
the number of pores of this size decreases as loading increases (see silica was displaced during aziridine polymerization and that the
Fig. 2). The location where a pore closes due to polymerization is difference between pore volumes of the two materials represents
unknown, but likely occurs at or near the pore mouth because of the incorporation of some aminopolymer inside the pores. In fact,
the high reactivity of aziridine.[51] Partially blocked pores have been the volume of aminopolymer present in the functionalized
shown to have an inkbottle effect on the desorption branch of the material was 0.37 cm3 g
1 SiO2, whereas the difference in pore
N2 adsorption isotherm,[50] deviating from the parallel adsorption volume was 0.17 cm3 g
1 SiO2, meaning that potentially 46% of the
and desorption branches observed for open cylindrical pores with added polymer was actually located in the pores. These examples
small deviations in pore diameter.[31,52] It is likely that the blocked illustrate that under certain conditions, large amounts of
pores of the HAS-VLAS are completely or nearly completely aminopolymer can be grown off of the silica particle exterior,
blocked because the adsorption and desorption branches of the N2 but that the aziridine molecule may be small enough to access and
adsorption isotherms are relatively parallel. polymerize in the pores even in situations where the pores are
A possible consequence of this pore blocking phenomenon is partially filled with organic material.
the increased likelihood of polymerization occurring outside of the The reaction yield over the series of HAS-VLAS syntheses
pore spaces. To further investigate this, a simple calculation was decreased as the reactant ratio increased (Table 1, Fig. 3). At low
performed to reconcile the N2 adsorption results with the TGA reactant ratios the yield decreased sharply as the ratio increased,
results for these materials (Table 1, columns 10 and 11). In the ideal but then followed a roughly linear behavior thereafter. This
case of an HAS material with all of the aminopolymer located suggests that the functionalization of SBA-15 with aziridine is
inside the pores, the sum of the volume of added aminopolymer diffusion-controlled under these conditions, since materials at low
per weight of silica and pore volume of the HAS material per reactant ratios had low amine loadings and high pore volumes,
weight of silica, subtracted from the original pore volume of the making them more accessible to entering reactants. Yet for
unfunctionalized SBA-15 should equal zero. A negative value of materials with higher reactant ratios and smaller resultant pore
this parameter (Remaining pore volume) suggests that amino- volumes, increasing quantities of aziridine remained unreacted or
polymer is located outside of the pores, while a positive value
suggests that there exists some pore volume that is inaccessible to
N2 adsorption. Similarly, it is possible to have both inaccessible
pore volume and aminopolymer outside of the pores that could
result in positive or negative values, depending on the relative
magnitudes. Table 1 shows that for all the HAS-VLAS except for
HAS10, the sums of the pore volumes and aminopolymer volumes
(assuming a density of 1.07 g cm
3 for our polymer, if it is similar
to commercial PEI) are greater than the pore volume of the bare
support and thus, some aminopolymer is likely located on the
particle exterior. Although there is not a trend in the remaining
pore volume versus amine loading or any of the pore
characteristics, the two most negative values were observed for
the highest- and lowest-loaded materials. The pore volume and
pore diameter of the lowest-loaded HAS material (HAS14) are very
near those of the silica support; this negative value for the
remaining pore volume suggests that aziridine preferentially
reacts with the silica surface (including the particle exterior). This Figure 3. Reaction yield for the HAS-VLAS (calculated as moles N in
supports the hypothesis that the majority of the silica pores are prepared HAS material per moles aziridine used in reaction) versus the
functionalized with aminopolymer until a critical radius is aziridine-to-(SBA-15) weight ratio used in the synthesis.

3824 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 3821–3832
 www.afm-journal.de
www.MaterialsViews.com

polymerized in solution to form aminopolymer that was removed differently loaded HAS adsorbents, it was hypothesized that the

FULL PAPER
during workup. aminopolymer structure could vary in its degree of branching and
polymer chain length as the available pore space in the silica
support changed. To investigate this possibility, the aminopoly-
mers of the HAS-VLAS had to be separated from their silica
2.2. Aminopolymer Characterization: Degree of Branching and supports so that traditional polymer characterization techniques
Molecular Weight could be used. This was accomplished by dissolving the
mesoporous silica in strong base, effectively cleaving the
In addition to the amine loading and pore characteristics of a given aminopolymer from its support. The aminopolymer was then
HAS adsorbent, it is also important to know the distribution of recovered by phase separation after partial solvent removal and a
amine types (primary, secondary, and tertiary). In the HAS subsequent physical separation. To characterize these materials,
synthesis, aziridine polymerizes off of the silica support, forming a aqueous size exclusion chromatography (ASEC) was used to assess
hyperbranched aminopolymer that contains a mixture of primary, the aminopolymer molecular weight and inversely-gated solution
13
secondary, and tertiary amines (hereafter referred to as 18, 28, and C NMR was used to measure the 18:28:38 amine ratio.
38, respectively). The 18:28:38 ratio of PEI synthesized in solution The inversely-gated 13C NMR experiments were conducted on
can vary over a range depending on synthesis conditions and all adsorbents from the HAS-VLAS. This proton-decoupled
molecular weight. For example, Kissel determined that the 18:28:38 technique produced quantitative 13C spectra. For commercial
amine ratios for eight commercial PEI’s fell inside of a range of PEI, the spectrum consisted of eight shifts between 40 and 60 ppm
(31–42): (33–39): (25–31).[53] The amine type is important to the that corresponded to carbons in slightly different environments
efficacy of CO2 adsorbents because the reaction of CO2 with 18, 28, within the polymer chain. The 18:28:38 amine ratios were then
and sterically hindered amines is usually described by a found by peak integration and are shown in Table 2. Additional
zwitterionic mechanism, whereas the reaction with 38 amines details regarding the carbonaceous environments and the 18:28:38
is governed by the base-catalyzed hydration of CO2.[11,54] In the amine ratio calculation are described in the Supplementary
former mechanism, the zwitterion is formed through the attack of Information. The cleaved aminopolymers (Table 2, rows 2 through
the lone pair of electrons on the amine on the carbon in CO2. The 7) do not significantly differ from each other in terms of their
stable carbamate is then formed from the deprotonation of the degree of branching. They have an average 18:28:38 amine ratio of
zwitterion by free base, such as another amine, or water under 28:46:26. The error in the analysis was estimated from three
humid conditions. Therefore, under dry conditions, two amines repeated runs with sample 6. Indeed, the ratios of the cleaved
are required to capture one CO2 molecule, but under humid samples do not change very much outside of this error. However,
conditions only one amine and one water are required to capture the cleaved aminopolymers are distinctly different from the two
one CO2 molecule, theoretically.[11] The first step in the reaction commercial PEI’s (Table 2, rows 1 and 9). These data suggest that
involving 38 amines involves the amine acting as catalyst to the degree of branching of the HAS aminopolymer is affected by
dissociate water to form a quaternary cationic species and a the reaction conditions, and perhaps by the pore diameter of the
hydroxide ion. The hydroxide ion then attacks CO2 to form a silica support (roughly the same in all cases) and not as dependent
bicarbonate anion. The last step is then the ionic association of the on the resulting polymer loading or the concentration of aziridine
protonated amine and bicarbonate. Since the different amine types used in the adsorbent’s synthesis.
react with CO2 in different ways, especially in the presence of The cleaved aminopolymers were also analyzed by ASEC to
water, it is important to determine the relative amounts of each determine their average molecular weights. In this technique,
contained in the HAS adsorbent, if adsorption capacities and molecules are eluted from the column in decreasing order of
kinetics are to be optimized. their molecular weight, usually with the small-molecular-weight
Our first report on the HAS adsorbent identified an amine ratio mobile-phase components eluting last. Figure 4 shows the
of 28:47:25 for a moderately loaded adsorbent. However, for molecular weights of the cleaved aminopolymers as a function

Table 2. Retention times of cleaved aminopolymers and calculated molecular weights relative to commercial PEI standards.

Sample Amine loading [mmol N g
1] Amine Ratio [a] [%] Avg. Mw [Daltons] Avg. repeat units [number]

18 28 38

Low Mw PEI N/A 44 33 23 800 19

HAS1 9.78 29 43 28 5695 132
HAS2 8.30 24 49 27 5695 132
HAS8 7.89 31 42 27 6577 153
HAS11 5.27 29 50 22 5383 125
HAS13 3.65 26 47 27 4814 112
HAS10 2.84 31 44 25 2140 50
HAS14 2.27 N/M N/M N/M 661 15
High Mw PEI N/A 25 41 34 60000 1395

[a] Calculated amine ratio error was estimated by triplicate analysis of HAS13. This error applied to the calculated average over the HAS-VLAS is given by the
range, 18:28:38 ¼ (27.9–29.2): (45.1–46): (25.7–26.3).

Adv. Funct. Mater. 2009, 19, 3821–3832 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3825
 www.afm-journal.de
www.MaterialsViews.com

grown aminopolymer is distinctly different from HAS aminopo-

FULL PAPER

lymer, with a larger fraction of 28 amines, or a more linear-like

conformation, and larger molecular weight. The collected data are
consistent with the hypothesis that the aminopolymers grow in the
silica pores and eventually block the pore mouths once a critical
radius is reached. As an increasing number of pores in the SBA-15
particle become blocked, the average molecular weight approaches
the chain length required to block a pore.
Aminopolymer grown in solution (PEI) should also differ from
HAS aminopolymer. The molecular weight of PEI grown in
solution is likely dependent on a number of factors including
solvent, concentration, catalyst concentration, and temperature.
Thus, it is important to know the structure of PEI grown in solution
under the conditions used in the HAS synthesis to compare it to
HAS aminopolymer. To investigate this, aziridine was added to
toluene at approximately the same concentration as in the HAS
synthesis, and reacted at room temperature in the presence of a
small amount of acetic acid (4 drops). The toluene and unreacted
aziridine were removed under vacuum at 50 8C, leaving the viscous
Figure 4. ASEC retention times for aminopolymers cleaved from HAS liquid PEI product. The molecular weight was found to be 946
adsorbents with different amine loadings. Daltons by ASEC. The 18:28:38 amine ratio was found to be
25:52:23 by inversely-gated 13C NMR. This degree of branching is
similar to HAS aminopolymer, albeit with a slightly larger amount
of the amine loading of the parent HAS material from which the of secondary amines, and is distinctly different from commercial
polymers were obtained. At low HAS amine loadings, the retention low-molecular-weight PEI. These results indicate that PEI grown
times decreased as amine loading increased, reflecting increasing in solution is also distinctly different from the HAS aminopolymer
aminopolymer molecular weight. However, at higher amine grown in the constrained environment of a silica support.
loadings the retention times leveled off, suggesting that the
molecular weights were all similar among the materials with high
loadings. The relative molecular weights of the cleaved amino-
polymers (Table 2) were determined by comparing their retention
times to a calibration curve constructed from commercial PEI 2.3. Modification of the HAS Adsorbent Synthesis
samples (see SI).
One might hypothesize that at high amine loadings, polymer- Small changes to the HAS synthesis were made to determine their
ization occurred on the particle exterior instead of inside the pores, effect on the resulting material’s aminopolymer loading. The first
resulting in comparably high-molecular-weight surface-bound experiment was to examine the catalytic effect of the acetic acid in
polymers. These ASEC results seem to contradict that hypothesis the HAS synthesis. The first mechanistic step of the ring-opening
because molecular weight should increase with amine loading polymerization reaction is suggested to be the protonation of
throughout the data series, not just at low amine loadings, and aziridine by acid in solution when acetic acid is present. However,
there would likely be a bimodal distribution to the data, although the silica surface silanols are mildly Brønsted acidic, and thus
admittedly, bimodal distributions at very low molecular weights acetic acid may not be needed in this synthesis. HAS9 was
may be difficult to resolve with typical ASEC techniques. synthesized without the addition of acetic acid, and it was
Furthermore, particle-exterior-only polymerization should not determined that added acid had little or no effect on the final
be constrained by the silica support, and one might expect the aminopolymer loading, resulting in 7.37 mmol N g
1 at an
18:28:38 amine ratio to be different from the HAS aminopolymers. aziridine-to-silica ratio of 2. Therefore, the acidity of the SBA-
To confirm this, aziridine was reacted with Cab-O-Sil, a nonporous, 15’s surface silanol groups is sufficiently strong to catalyze the
amorphous silica, under typical HAS reactions conditions. The aziridine ring-opening reaction, as expected.
amine loading achieved was 2.25 mmol N g
1, as determined by The ring-opening polymerization is a fast reaction and thus can
TGA. The aminopolymer was then cleaved from the Cab-O-Sil create aminopolymers uncontrollably, sometimes to the detriment
silica by the same procedure used to cleave aminopolymer from of material properties such as gas diffusion when the polymeriza-
HAS materials. The 18:28:38 amine ratio of the cleaved polymer tion occurs rapidly enough that it may block pores. In an effort to
was found to be 19:59:22 by inversely-gated 13C NMR using the affect the reaction rate, HAS12 was synthesized by the dropwise
same experimental parameters as for the HAS aminopolymer addition of aziridine to the reaction environment over 90 min
experiments. The average molecular weight of the cleaved without the addition of a catalytic amount of acetic acid. The
aminopolymer was found to be 9080 Daltons by ASEC, obtained organic loading is lower than expected for the aziridine-
approximately 3000 Daltons longer than the average size of the to-silica ratio used (3). This unexpected result suggests that a
aminopolymers cleaved from the highly-loaded HAS materials. concentration gradient from the bulk solution to the interior pore
With the caveat that factors, such as particle size and surface silanol space may be crucial for aziridine diffusion into the pore space and
density, may differ from those of SBA-15, the particle-exterior- subsequent productive reaction in the pore system.

3826 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 3821–3832
 www.afm-journal.de
www.MaterialsViews.com

Table 3. Compositions and physical characteristics of HAS adsorbents prepared through various synthesis modifications.

FULL PAPER
Sample Amine Aziridine-to- Yield [a] Pore diameter BET Pore vol. Pore vol. Micropore vol.
loading (SBA-15) [mol N mol
1 aziridine] [nm] surface area [cm3 g
1] [cm3 g
1 SiO2] [cm3 g
1]
[mmol N g
1] ratio [w w
1] [m2 g
1]

HAS7 [b] 7.58 2.00 0.29 4.5 47 0.09 0.16 0.00

HAS8 [b] 7.89 2.00 0.30 4.7 60 0.12 0.20 0.00
HAS9 [b][c] 7.37 2.01 0.27 4.9 68 0.12 0.21 0.00
HAS12 [d][e] 4.46 3.10 0.08 5.1 245 0.38 0.50 0.01
SBA2 0 6.2 865 0.81 0.81 0.06
SBA4 0 5.9 774 0.74 0.74 0.05

[a] Calculated as mol N g
1 SiO2 in the prepared HAS divided by the reactant ratio [mol N g
1 SiO2]. [b] SBA2 used as silica support in synthesis. [c]
Synthesis performed without the addition of acetic acid. [d] SBA4 used as silica support in synthesis. [e] Synthesis performed without the addition of acetic
acid and with aziridine added dropwise to the reaction over a period of 90 min.

Another important aspect in the preparation of any complex reactant ratio of 2 in diethyl ether, tetrahydrofuran (THF),
material is synthetic reproducibility. To investigate whether the dichloromethane (DCM), and acetonitrile (ACN). A small amount
uncontrolled but constrained polymerization was reproducible, of glacial acetic acid was added to catalyze the reactions, and the
two HAS materials were synthesized side by side with the same reaction mixtures were left overnight at room temperature under
quantities of aziridine, mesoporous silica, and acetic acid from vigorous stirring in glass pressure vessels. In a typical synthesis
the same batches. They were reacted for the same length of time using toluene, considerable amounts of product fell out of
in glass pressure vessels of the same size and shape and were dispersion and built up in the form of white gel on the wall of the
stirred at the same rate with stir bars of the same size and shape. pressure vessel during reaction. This is likely due to the highly
The nitrogen loadings of the two materials were 7.58 and hydrophilic character of the aminopolymer, in that once a
7.89 mmol N g
1 and the surface areas were 47 and 60 m2 g
1, sufficient quantity of the polymer forms on the silica support,
respectively, indicating that the two materials were very similar the particles agglomerated due to repulsion from the surrounding
(Table 3, HAS7 and HAS8). However, it should be noted that organic solvent. However, this effect was not observed in the
these materials do not fit exactly into the trends in composition reactions performed in THF, DCM, and ACN, while the reaction
and pore characteristics described for the HAS-VLAS. Instead, it with diethyl ether showed a small amount of gel-like agglomera-
appears that while the HAS synthesis is reproducible with the tion, albeit much less than was observed in toluene. The organic
same batch of materials, the use of different batches of aziridine, loadings of the resulting HAS materials were all substantially
SBA-15, toluene, as well as small variations in laboratory lower than those obtained from reaction in toluene (Table 4). These
conditions such as humidity and temperature, may make it observations are possibly related to the polarity or hydrogen-
difficult to exactly reproduce HAS adsorbents with specific bonding nature of the various organic solvents disrupting the
loadings and pore characteristics. polymerization of amines or preventing the fixation of organics on
The role of solvent in the HAS synthesis is complex. First, the the silica surface. It may be that low-dielectric-constant solvents
solvent must be capable of dispersing the mesoporous silica provide a driving force for aziridine to diffuse to the relatively
adequately under stirring, so that the reaction environment is well- hydrophilic environment of the silica pore surface.
mixed and the mass transfer resistance of aziridine to the particle
surface is minimized. Second, aziridine must be soluble in the
reaction solvent, again for the reaction environment to be well-
mixed. Third, the solvent must be nonreactive to the mesoporous 2.4. Adsorption Capacities of the HAS-VLAS
silica and the aziridine molecule, which excludes all protic
solvents. However, several common organic solvents meet these As discussed in Section 2.1, the adsorption capacity is one of
criteria while still displaying different physical characteristics, several important parameters that characterize adsorbent perfor-
such as dipole moment and dielectric constant. mance. Adsorption capacities are sometimes reported as working
To understand the effect of reaction solvent on the resulting capacities rather than equilibrium capacities, when adsorption is
HAS material, reactions were run with an aziridine-to-(SBA-15) measured for a shorter time than required to reach equilibrium.

Table 4. Organic contents and pore properties of various samples synthesized in different reaction solvents.

Sample Solvent Dipole moment Dielectric constant Amine loading Pore diameter BET surface area Pore vol.
[debye] [mmol N g
1] [nm] [m2 g
1] [cm3 g
1]

HAS8 Toluene 0.37 2.4 7.98 4.6 61 0.09

SCHAS1 Diethyl ether 1.15 4.3 5.19 5.4 284 0.44
SCHAS2 THF 1.75 7.5 5.49 5.3 302 0.48
SCHAS3 DCM 1.6 9.1 5.60 5.4 233 0.38
SCHAS4 ACN 3.92 37 5.07 5.5 270 0.43

Adv. Funct. Mater. 2009, 19, 3821–3832 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3827
 www.afm-journal.de
www.MaterialsViews.com

Working capacities may also be reported when an incomplete because a sufficiently long adsorption time was chosen (200 min)
FULL PAPER

desorption step is performed, leaving part of an adsorbent’s rendering kinetic effects less important, leaving the measured
capacity inaccessible for subsequent adsorption steps. In either capacity mainly a function of the accessible adsorption sites.
instance, working capacities are perhaps more relevant to an Several duplicated runs using the same material and measure-
adsorbent’s performance in a real CO2 capture process where cycle ment conditions were found to be within 5% of each other. The
times will likely be short and temperature swings performed as increasing linear trend was observed for both operating
narrowly as possible to reduce energy costs. Typically, cycle times temperatures that were investigated, where the capacity at 25 8C
in a practical adsorption process will be significantly shorter than was 1.1 mmol g
1 higher than the capacity at 75 8C, averaged over
those used here, where relatively long cycle times were used to the data set, a trend that is in good agreement with the previously
generate working capacities. Again, all these factors must be reported results.[22] The highest capacity measured was
evaluated, and the tradeoff between amount of adsorbent and 5.55 mmol g
1 for an HAS adsorbent with an amine loading of
process costs for that adsorbent considered, for a specific local 9.78 mmol N g
1. It should be noted that since these working
application, to determine an adsorbent’s efficacy. capacities do not represent full equilibrium capacities and were
The first report on HAS adsorbents described a sample with a measured over a temperature range of 50 8C, they should not be
CO2 capacity of 3.11 mmol CO2 g
1 adsorbent (hereafter shor- used for the calculation of fundamental constants such as heats of
tened to mmol g
1) for a moderately loaded adsorbent with a adsorption. The heats of adsorption reported in the literature for
humidified 10% CO2 test gas to simulate flue gas capture.[22] covalently-tethered silica-supported amines fall into the range of
However, the capacities reported here are slightly higher than
48[55] to –90[56] kJ mol
1, depending on the adsorbents and
those reported previously for similar materials. The experiments analyses used. It seems reasonable to expect that the HAS
discussed in this paper were conducted with similar parameters adsorbent falls somewhere in this admittedly large range. Future
and equipment (see Experimental Section), but the step size detailed studies will be directed at quantitative assessments of the
between scans in the mass spectrometer (MS) was decreased from heat released during adsorption.
7 to 0.7 s. This increases the number of data points by a factor of ten Another important measure of an amine-based adsorbent’s
and substantially reduces the error obtained from numerical performance is its amine efficiency, or the ratio of the moles of
integration. Additionally, the data was analyzed in a slightly CO2 captured to the moles of nitrogen present in the material.
different manner; adsorption was measured from the time CO2 The maximum amine efficiency of an adsorbent depends on
was first introduced to the adsorbent bed, not from the time the the amine type and the presence or absence of water. Under
first CO2 signal was detected with the MS (breakthrough time). For dry conditions, 18 and 28 amines react with CO2 to form
more information on the data analysis, please see the SI. In most carbamates through a zwitterionic mechanism.[6,7,10,54,57,58] This
cases, this distinction did not have a great effect on the final mechanism requires free base to deprotonate the zwitterion to
determined capacity because of the relatively small size of the form the carbamate. Thus, under dry conditions, 2 mol of amines
adsorbent bed and the minimal pressure drop experienced across are required to capture 1 mol of CO2, reducing the maximum
it. Nevertheless, the effect was noticeable over the range of amine efficiency to 0.5 mol CO2 mol
1 N. In humid conditions,
adsorbents prepared. H2O or OH
can act as free base, raising the maximum amine
The CO2 working capacities of the HAS-VLAS were measured at efficiency to 1.0 mol CO2 mol
1 N as long as stoichiometric
25 and 75 8C under humid conditions via methodology described quantities of H2O are present. Tertiary amines are not suggested
in the Experimental Section and the SI and are summarized in to react directly with CO2; rather, they catalyze the formation of
Figure 5. The working capacities increased as amine loading bicarbonate in the presence of water.[54,59,60] The first step in this
increased in a roughly linear manner. This result was expected mechanism is the dissociation of H2O to form a cationic
quarternary amine and OH
, which then reacts with CO2 to
form the bicarbonate anion. Therefore, the maximum amine
efficiency for 38 amines is 1.0 in the presence of water, but
the expected value is close to 0 under dry conditions. Primary
and 28 amines may also react via this mechanism, however, the
reaction rate for the formation of carbonates in solution by
this mechanism is generally much lower than for the zwitter-
ionic mechanism.[6,54] Taking both of these reaction
mechanisms into consideration, the maximum amine efficiency
for the HAS adsorbent is 1.0 under humid conditions and 0.37
under dry conditions, assuming an average amine ratio of
18:28:38 ¼ 28:46:26. Clearly, the performance of the HAS
adsorbent is fundamentally aided by the presence of water, unlike
other types of solid adsorbents, such as zeolites, activated carbons,
and MOFs.[11–13]
Several additional factors can cause a decrease in amine
efficiency. Amines can be inaccessible to entering CO2 due to pore
blockage, location within micropores, or steric hindrance from
Figure 5. CO2 capture working capacities at 25 and 75 8C and associated neighboring species, especially after long adsorption periods
amine efficiency versus HAS amine loading for the HAS-VLAS. where equilibrium is nearly reached and almost every amine has

3828 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 3821–3832
 www.afm-journal.de
www.MaterialsViews.com

captured a CO2 molecule. Amine efficiency can be reduced if, in the isotherm. In some cases, single expansions took several weeks

FULL PAPER
spite of the presence of water, CO2 reacts with the amines to form to reach equilibrium, so it is likely that heat transfer and reaction
carbamates. There is some evidence in the literature that suggests rate effects, which have timescales on the order of minutes or less,
that at early adsorption times, the formation of carbamates is were not relevant to the measured equilibrium. However, it is
kinetically favored, although it is unclear whether this is an interesting to note that the results indicate a higher dry capacity at
inherent characteristic of supported amines or due to other factors, 75 8C than at 35 8C. This goes against intuition since the adsorption
such as accessibility and amine type.[23,61–64] Amine efficiency may of CO2 with amines is an exothermic process with a negative
also be reduced when CO2 adsorbs in the form of bidentate species. change in entropy. Fundamentally, thermodynamics suggests that
Chuang and co-workers observed the formation of bidentate capacity should decrease with increasing temperature. Hence, this
carbonate (1575, 1390 cm
1) and bidentate bicarbonate observation is probably associated with severely hindered
(1634 cm
1) for CO2 adsorption with primary amine-functional- diffusion at the 35 8C adsorption condition. Similar results have
ized SBA-15 under dry conditions.[63] been observed for solid amine-impregnated silica adsorbents
Figure 5 shows the amine efficiencies for the HAS-VLAS at possessing large amine loadings with little or no remaining pore
25 8C. At low loadings, the amine efficiency decreased as amine volume.[24,32,65] This result again fits with the hypothesis that
loading increased and reached a plateau at higher loadings. nearly complete pore blockage occurs at or near the pore mouths of
Since these measurements were made after 200 min of adsorption, the HAS adsorbent, effectively blocking off access to amines
these amine efficiencies are somewhat lower than they would be at already in the pores.
full equilibrium. Nevertheless, they exhibit the fact that HAS
adsorbents with high loadings do not take full advantage of their
high amine density. Since the highly loaded adsorbents displayed 2.5. Adsorption Kinetics
low pore volumes and likely possessed blocked pores, the reduced
amine efficiencies were probably due to limited mass transfer or The kinetics of adsorption and desorption are also important in
steric constraints. assessing an adsorbent’s performance. The application of
The above discussion described experiments designed to capturing CO2 from flue gas requires that adsorbents capture
investigate CO2 capture from a simulated flue gas consisting of CO2 as quickly as possible to shorten adsorption/desorption cycle
humidified 10% CO2 in Ar. An alternative way to describe an times and to reduce the total quantity of adsorbent required to
adsorbent’s performance is to measure its adsorption isotherm reduce CO2 concentration to a required level. A general theme in
over a broad range of CO2 partial pressures, since flue gases vary the silica-supported amine literature is a fast initial adsorption step
slightly in CO2 concentration and other alternative power- followed by a much slower approach to equilibrium,[66] which may
generation strategies may produce process streams in need of be so gradual as to cause most researchers to simply end the
CO2 separation that vary over a wide concentration range. experiment before true equilibrium is reached.[11] In addition to
Additionally, isotherms are often used to extract information on obvious material characteristics, such as the density of adsorption
key physical constants of an adsorbent, such as heats of adsorption sites a material possesses, there is also a strong link between the
and Henry’s Law constants. open pore volume of the adsorbent and adsorption kinetics.[26,32]
Single-component isotherms for a highly loaded HAS This is related to the mass transfer of CO2 through the pores to
adsorbent (HAS2) were measured at 35 and 75 8C (Fig. 6) using the adsorption sites and illustrates the importance of amine
pressure–volume–temperature (PVT) techniques. In these mea- accessibility.
surements, a small amount of sample (0.8 g) was exposed to a The adsorption kinetics of the HAS-VLAS measured in the
known volume of CO2 and allowed to approach equilibrium. The packed bed flow system and reported below are not true
moles of CO2 adsorbed at a given pressure were then calculated, adsorption kinetics, but rather, a combination of process related
and the experiment repeated at an increased pressure, assembling effects (adsorbent bed bypass, gas mixing; see SI for description
of application of backgrounds, etc.), internal mass transfer
effects, and intrinsic chemical reaction rates. They should be
interpreted as being experiment-dependent and not necessarily
scalable, and should not be used to derive mass transfer
coefficients. Rather they should be viewed as a kinetic trend and
evidence that, at least in this simulated packed-bed experiment,
the adsorbents exhibit Fickian transport behavior and the
diffusion might be described by a partial immobilization
model.[67] However, quantitative kinetic evaluation will require
additional experimentation.
The MS at the outlet of the packed bed flow system collects
transient concentration data as adsorption occurs. With the
assumption that deviation from the value at 200 min is attributed to
active adsorption, the amount adsorbed at a certain time is the
integral of the concentration differences from the start of the
experiment to that given time. This translates to an uptake curve of
Figure 6. Single-component CO2 adsorption isotherm for a highly-loaded the mass adsorbed (Mt, here normalized by weight of adsorbent
HAS adsorbent at 35 and 75 8C. used) versus square root time (to highlight the large changes in

Adv. Funct. Mater. 2009, 19, 3821–3832 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3829
 www.afm-journal.de
www.MaterialsViews.com

reached a critical density and then caused pore blocking, likely at or

FULL PAPER

near the pore mouths. This inferred physical characteristic was

likely the cause of the HAS-VLAS’s observed decline in amine
efficiencies and adsorption kinetics at high amine loadings. It is
important to note that the pore characteristics of the parent SBA-15
support were the physical boundaries that limited the amount of
amines that were incorporated into the adsorbent and also the
mass transfer to those amines. Future research will investigate the
variation of the silica support to further increase the performance
of HAS adsorbents.
Despite the tunability of the HAS amine loading and the
Figure 7. Adsorption halftimes of the HAS-VLAS at operating tempera- following predictability of the adsorption behavior, the aminopo-
tures of 25 and 75 8C. The halftime at 25 8C for an amine loading of lymer growth was still the result of an uncontrolled poly-
10 mmol N g
1 was omitted because it was off the scale of the plot, with a merization. However, it appears that the polymerization may
value of 2300 s. have been constrained by the silica support, as evident by in the
similar degree of branching of the aminopolymer (18:28:38
early times). Uptake plots for the HAS-VLAS are given in the SI for amines ¼ 28:46:26) and an upper bound in the molecular weight
adsorption temperatures of 25 and 75 8C, respectively. (approximately 6500 Daltons). Efforts to further constrain the
Given the long times required for some of these adsorbents to polymerization by changing reaction conditions or adding
approach equilibrium, it is perhaps more reasonable to discuss inhibitors may be successful, but it is unlikely that the degree
their kinetics in terms of working capacities that are reported of branching or molecular weight will ever be fully controlled using
for some time less than the time required to reach full capacity. It this monomer. Nevertheless, the outcome of the HAS synthesis is
is common practice to use the measure of adsorption halftime fairly predictable and, to that end, makes the HAS adsorbent a
(the time where the adsorbent reaches half of its capacity at the end promising candidate for CO2 capture process technologies where
of the experiment) for this purpose.[52] Figure 7 shows the the balance between desired adsorption capacities, kinetics, and
adsorption halftimes for the HAS-VLAS at 25 and 75 8C. At room operating temperature (among others) will likely be decided on a
temperature, the adsorption halftimes of the HAS materials with local, application-specific basis.
high loadings were significantly longer than those of adsorbents
with lower loadings. Since process and measurement variables
were all similar for these measurements, these data show that mass
transfer limitations of the highly loaded adsorbents had a 4. Experimental
significant detrimental effect on the adsorption kinetics, despite All chemicals were purchased from Sigma-Aldrich and used without
the presence of a larger number of potential adsorption sites. modification. Single-component gases were purchased from Airgas, Inc.
Again, this fits our hypothesis that highly loaded HAS materials Custom gas mixtures were purchased from Matheson Tri-Gas, Inc.
likely have a large number of their pores blocked at or near the pore SBA-15 Synthesis: Mesoporous SBA-15 was synthesized similar to
mouths, dramatically reducing mass transfer. The data set previously reported methods [69–71]. In a typical synthesis, 17.9 g of
Pluronic 123 (EO-PO-EO block copolymer), 99.4 g of HCl, and 561 g of H2O
obtained at 75 8C also displayed this trend, albeit to a lesser
were stirred overnight in a 1 L Erlenmeyer flask. To the micellular solution,
degree. This is expected since diffusional resistances will be less 39.6 g of tetraethylorthosilicate (TEOS) was added and stirred for 5 min.
significant at higher temperatures. The variation in adsorption The solution was transferred to an oven and stirred at 35 8C for 20 h.
capacity and kinetics show that these HAS materials can be tuned Stirring was then terminated and the solution was post-treated at 80 8C for
to have different adsorption rates and capacities which, coupled 24 h. The white solid was filtered and washed with copious amounts of
with process variations (fixed-bed, fluidized-bed, fibers[68]) can deionized (DI) H2O and dried overnight at 75 8C. The organic template was
enable the scientist or engineer significant leeway in designing an removed by calcination using the following temperature program: 1)
ramping the temperature to 200 8C at a rate of 1.2 8C min
1, 2) holding at
optimized CO2 capture process. 200 8C for 1 h, 3) ramping the temperature to 550 8C at a rate of
1.2 8C min
1, and 4) holding at 550 8C for 6 h. The calcined SBA-15 was
then dried at 200 8C under vacuum overnight.
Aziridine Synthesis: Aziridine was synthesized and purified similar to
3. Conclusion reported methods [2,22]. In a typical synthesis, 34.5 g of 2-chloroethyla-
We have investigated the degree of control that can be exerted on mine hydrochloride was added to a solution of 34.1 g of KOH in 200 g of DI
H2O. The solution was stirred at 50 8C for 2 h. Aziridine was then distilled
the HAS synthesis to tune the adsorbent’s composition, adsorption from solution under a partial static vacuum of 530 mm Hg and collected in
capacity, and adsorption kinetics. Specifically, it was found that by a cold trap. The collection flask was sealed and then stored at 0 8C. 1H NMR
simply adjusting the aziridine-to-silica reactant ratio, materials (400 MHz, D2O, d): 1.56 (s); 13C NMR (400 MHz, CD2Cl2, d): 18.1. Caution:
with a wide range of characteristics could be prepared. Amine Aziridine is a carcinogen and reproductive hazard. It is toxic if swallowed,
loadings, and therefore, potential active adsorption sites ranged inhaled, or absorbed through the skin. Use extreme caution when handling
from 2 to 10 mmol N g
1 adsorbent, resulting in working aziridine. Only handle aziridine in a ventilated fume hood and always wear
proper personal protection equipment.
adsorption capacities ranging from 2.0 to 5.6 mmol g
1 at 25 8C. HAS synthesis: The HAS synthesis has been reported elsewhere [22].
As adsorbent organic content increased, pore volume and surface Variations of the syntheses with different aminopolymer loadings are
area decreased. At high loading levels, N2 physisorption data and reported here (Table 1). In a generalized synthesis, a predetermined
ASEC data indicated that polymer growing inside the silica pores amount of aziridine was added to a suspension of a measured amount of

3830 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 3821–3832
 www.afm-journal.de
www.MaterialsViews.com

SBA-15 in toluene. Approximately 4 drops of glacial acetic acid were added to introduce the CO2 to the sample. The pressure loss in the sample cell

FULL PAPER
to catalyze the ring-opening polymerization. The mixture was stirred over time was recorded for each expansion until pressure stabilization
overnight at room temperature in a glass pressure vessel, sealed with a occurred, allowing for the generation of adsorption isotherms.
polytetrafluoroethylene (PTFE) cap. The solid was filtered and washed with
copious amounts of toluene and methanol and then dried at 50 8C under
vacuum.
Removal of Aminopolymer from Silica Support: For a given reaction Acknowledgements
targeted at cleaving the aminopolymer from the inorganic support,
This work was supported by the U.S. Department of Energy, National
approximately 500 mg of HAS adsorbent was stirred in a basic solution of
Energy Technology Laboratory, and the Camille and Henry Dreyfus
33 g of KOH in 100 g of DI H2O at 50 8C overnight, after which the silica
Postdoctoral Program in Environmental Chemistry. Supporting Informa-
support was degraded into soluble species. About half of the H2O was
tion is available online from Wiley InterScience or from the author.
removed by vacuum distillation at 75 8C and the remaining solution was
stored in a freezer overnight. The aminopolymer phase separated and was
recovered by decanting/physical separation. Received: August 4, 2009
Characterization: Pore characteristics of the bare silica support and the Published online: November 9, 2009
HAS adsorbents were assessed via N2 physisorption analysis at 77 K using
a Micromeritics ASAP 2010. Amine loadings were determined by TGA
using a Netzsch STA409. Total organic loading was estimated as the weight
[1] S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt,
loss from 160 to 760 8C, followed by inference of a stoichiometric ratio of
M. Tignor, H. L. Miller, IPCC 2007, Summary for Policymakers in Climate
ethyleneimine units (molecular weight of 43 Daltons) after accounting
for silica surface silanol condensation (1 wt%). FTIR spectroscopy using Change 2007: The Physical Science Basis. Contribution of Working Group 1 to
KBr pellets and FT-Raman spectroscopy were obtained on a Bruker Vertex the Fourth Assessment Report of the Intergovernmental Panel on Climate
80v optical bench with a RAMII Raman module. Solution 1H NMR Change, Cambridge, UK 2007.
experiments were conducted on a Varian Mercury Vx 400. Inversely-gated [2] H. J. Kim, J. H. Moon, J. W. Park, J. Colloid Interface Sci. 2000, 227, 247.
solution 13C NMR experiments were conducted on a Bruker DRX 500. [3] R. Steeneveldt, B. Berger, T. A. Torp, Chem. Eng. Res. Des. 2006, 84, 739.
Samples were prepared with D2O and a single drop of Cr(acac)2 (where [4] J. Davison, Energy 2007, 32, 1163.
acac is acetylacetonate) in deuterated dimethylsulfoxide (DMSO) to aid [5] J. W. Dijkstra, D. Jansen, Energy 2004, 29, 1249.
relaxation and minimize interference from possible residual silica or base [6] P. V. Danckwerts, Chem. Eng. Sci. 1979, 34, 443.
salts. Typically, 6000 to 10 000 scans were gathered with a relaxation time of [7] P. V. Danckwerts, M. M. Sharma, The Chemical Engineer 1966, CE, 244.
4 s in the experiments using an inversely-gated pulse sequence. This [8] S. Bishnoi, G. T. Rochelle, Chem. Eng. Sci. 2000, 55, 5531.
relaxation time is in excess of the 3.3 s required for quantitative analysis [9] D. A. Glasscock, J. E. Critchfield, G. T. Rochelle, Chem. Eng. Sci. 1991, 46,
determined in a full T1 study by Pierre and Geckle for branched PEI [72]. 2829.
Molecular weights of PEI and cleaved aminopolymer were determined by [10] G. F. Versteeg, L. A. J. van Dijck, W. P. M. van Swaaij, Chem. Eng. Commun.
ASEC at 30 8C. The ASEC system comprised a Shimadzu LC-20AD pump, a 1996, 144, 113.
Shimadzu RID-10A RI detector, a Shimadzu SPD-20A UV detector, a [11] S. Choi, J. Drese, C. Jones, ChemSusChem 2009, 2, 796.
Shimadzu CTO-20A column oven, and Viscotek TSK Viscogel PWXL Guard, [12] F. Brandani, D. M. Ruthven, Ind. Eng. Chem. Res. 2004, 43, 8339.
G3000, G4000 and G6000 columns mounted in series. The mobile phase
[13] Y. X. Wang, Y. P. Zhou, C. M. Liu, L. Zhou, Colloids Surf, A 2008, 322, 14.
consisted of 0.5 M acetic acid and 0.3 M Na2SO4, and the flow rate was
[14] R. Barker, J. Appl. Chem. Biotechnol. 1973, 23, 733.
maintained at 0.5 mL min
1. Commercial PEI’s of a range of molecular
[15] J. R. Hufton, S. Mayorga, S. Sircar, AIChE J. 1999, 45, 248.
weights from 800 to 10 000 Daltons were used to generate a calibration
curve. [16] J. Ida, R. Xiong, Y. S. Lin, Sep. Purif. Technol. 2004, 36, 41.
Adsorption: Humid adsorption capacities and apparent adsorption [17] R. A. Khatri, S. S. C. Chuang, Y. Soong, M. Gray, Energy Fuels 2006, 20, 1514.
kinetics were measured with a packed bed flow reactor with an online mass [18] X. C. Xu, C. S. Song, B. G. Miller, A. W. Scaroni, Fuel Process. Technol. 2005,
spectrometer as described in a previous report [22]. Briefly, approximately 86, 1457.
60 to 80 mg of adsorbent was dispersed in 300 mg of sieved sand (250 [19] B. Arstad, H. Fjellvag, K. O. Kongshaug, O. Swang, R. Blom, Adsorption
to 425 mm) and loaded into a Pyrex tubular reactor (1/4 in. outer diameter; 2008, 14, 755.
1 in. ¼ 2.54 cm). The assembly was then placed in a temperature-controlled [20] L. Pan, K. M. Adams, H. E. Hernandez, X. T. Wang, C. Zheng, Y. Hattori,
tube furnace. Before an experiment was started, desorption was performed K. Kaneko, J. Am. Chem. Soc. 2003, 125, 3062.
for at least 3 h at 110 8C under flowing Ar to ensure complete desorption. [21] A. R. Millward, O. M. Yaghi, J. Am. Chem. Soc. 2005, 127, 17998.
Also, before an adsorption experiment, test gas (10% CO2 balance Ar) was [22] J. C. Hicks, J. H. Drese, D. J. Fauth, M. L. Gray, G. G. Qi, C. W. Jones, J. Am.
pre-humidified to a steady state by flowing through two glass water Chem. Soc. 2008, 130, 2902.
bubblers at room temperature. To begin an experiment, the inlet to the [23] R. Serna-Guerrero, E. Da’na, A. Sayari, Ind. Eng. Chem. Res. 2008.
reactor was switched from desorption gas to test gas and the time was [24] X. C. Xu, C. S. Song, J. M. Andresen, B. G. Miller, A. W. Scaroni, Energy Fuels
noted. Adsorption was then performed at a test gas flow rate of 2002, 16, 1463.
20 mL min
1 for 200 min at the desired temperature. The CO2 concentra- [25] R. S. Franchi, P. J. E. Harlick, A. Sayari, Ind. Eng. Chem. Res. 2005, 44, 8007.
tion at the reactor outlet was measured with a Pfeiffer Vacuum QMS 200 [26] N. Hiyoshi, K. Yogo, T. Yashima, Chem. Lett. 2004, 33, 510.
Prisma Quadrupole Mass Spectrometer. The mass of CO2 adsorbed was
[27] L. Bergman, J. Rosenholm, A. B. Ost, A. Duchanoy, P. Kankaanpaa,
calculated from the MS output as described in the SI. To determine total
J. Heino, M. Linden, J. Nanomater. 2008.
CO2 uptake in the HAS-VLAS as a function of CO2 pressure at constant
[28] J. M. Rosenholm, A. Duchanoy, M. Linden, Chem. Mater. 2008, 20, 1126.
temperature, a simple pressure decay method, originally developed for
polymers, was used [73]. The HAS sample was held in a 0.5 mm stainless [29] J. M. Rosenholm, M. Linden, Chem. Mater. 2007, 19, 5023.
steel filter element and capped with aluminum foil, and then loaded into [30] J. M. Rosenholm, M. Linden, J. Controlled Release 2008, 128, 157.
the adsorption cells. The adsorption cells were submerged in constant [31] J. M. Rosenholm, A. Penninkangas, M. Linden, Chem. Commun. 2006,
temperature oil baths and the oil baths were set to 115 8C and vacuum was 3909.
pulled on the sample cells for one day to completely evacuate the cell and [32] W. J. Son, J. S. Choi, W. S. Ahn, Micropor. Mesopor. Mater. 2008, 113,
desorb the adsorbent. After the sample drying step the oil bath was set to 31.
35 or 75 8C and CO2 was introduced to the reservoir and allowed to [33] T. C. Drage, A. Arenillas, K. M. Smith, C. E. Snape, Micropor. Mesopor.
equilibrate. After thermal equilibrium, the sample valve was opened briefly Mater. 2008, 116, 504.

Adv. Funct. Mater. 2009, 19, 3821–3832 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3831
 www.afm-journal.de
www.MaterialsViews.com

[34] J. S. Beck, J. C. Vartuli, W. J. Roth, M. E. Leonowicz, C. T. Kresge, K. D. Schmitt, [54] P. D. Vaidya, E. Y. Kenig, Chem. Eng. Technol. 2007, 30, 1467.
FULL PAPER

C. T. W. Chu, D. H. Olson, E. W. Sheppard, S. B. McCullen, J. B. Higgins, [55] R. A. Khatri, S. S. C. Chuang, Y. Soong, M. Gray, Ind. Eng. Chem. Res. 2005,
J. L. Schlenker, J. Am. Chem. Soc. 1992, 114, 10834. 44, 3702.
[35] D. J. Macquarrie, Chem. Commun. 1996, 1961. [56] C. Knofel, J. Descarpentries, A. Benzaouia, V. Zelenak, S. Maornet,
[36] E. J. Acosta, C. S. Carr, E. E. Simanek, D. F. Shantz, Adv. Mater. 2004, 16, P. L. Llewellyn, V. Hornebecq, Micropor. Mesopor. Mater. 2007, 99, 79.
985. [57] M. Caplow, J. Am. Chem. Soc. 1968, 90, 6795.
[37] D. M. Ford, E. E. Simanek, D. F. Shantz, NANO 2005, 16, S458. [58] V. V. Mahajani, J. B. Joshi, Gas Sep. Purif. 1988, 2, 50.
[38] S. Yoo, J. D. Lunn, S. Gonzalez, J. A. Ristich, E. E. Simanek, D. F. Shantz, [59] T. L. Donaldson, Y. N. Nguyen, Ind. Eng. Chem. Res. Fund. 1980, 19,
Chem. Mater. 2006, 18, 2935. 260.
[39] W. W. Lukens, P. Schmidt-Winkel, D. Y. Zhao, J. L. Feng, G. D. Stucky, [60] T. L. Donaldson, J. A. Quinn, Proc. Natl. Acad. Sci. USA 1974, 71, 4995.
Langmuir 1999, 15, 5403. [61] N. Hiyoshi, K. Yogo, T. Yashima, Micropor. Mesopor. Mater. 2005, 84, 357.
[40] E. P. Barrett, L. G. Joyner, P. P. Halenda, J. Am. Chem. Soc. 1951, 73, 373. [62] N. Hiyoshi, K. Yogo, T. Yashima, J. Jpn. Pet. Inst. 2005, 48, 29.
[41] J. C. Broekhof, J. H. Deboer, J. Catal. 1967, 9, 8. [63] A. C. C. Chang, S. S. C. Chuang, M. Gray, Y. Soong, Energy Fuels 2003, 17,
[42] T. L. Hill, Adv. Catal. 1952, 4, 211. 468.
[43] G. Halsey, J. Chem. Phys. 1948, 16, 931. [64] X. C. Xu, C. S. Song, B. G. Miller, A. W. Scaroni, Ind. Eng. Chem. Res. 2005,
[44] G. D. Halsey, Adv. Catal. 1952, 4, 259. 44, 8113.
[45] The theoretical maximum aminopolymer loading is calculated here by [65] M. B. Yue, Y. Chun, Y. Cao, X. Dong, J. H. Zhu, Adv. Funct. Mater. 2006, 16,
multiplying the pore volume of the unfunctionalized SBA-15 by the density 1717.
of PEI (1.07 g cc
1) giving units of [g organic g
1 SiO2]. [66] M. B. Yue, L. B. Sun, Y. Cao, Y. Wang, Z. J. Wang, J. H. Zhu, Chem. Eur. J.
[46] T. Tsuda, T. Fujiwara, J. Chem. Soc. Chem. Commun. 1992, 1659. 2008, 14, 3442.
[47] M. G. Plaza, C. Pevida, A. Arenillas, F. Rubiera, J. J. Pis, Fuel 2007, 86, [67] D. R. Paul, W. J. Koros, J. Polym. Sci, Part B: Polym. Phys. 1976, 14, 675.
2204. [68] R. P. Lively, R. R. Chance, J. H. Drese, C. W. Jones, W. J. Koros, Ind. Eng.
[48] S. Kim, J. Ida, V. V. Guliants, J. Y. S. Lin, J. Phys. Chem. B 2005, 109, 6287. Chem. Res. 2009, 48, 7314.
[49] E. P. Dillon, C. A. Crouse, A. R. Barron, NANO 2008, 2, 156. [69] J. C. Hicks, R. Dabestani, A. C. Buchanan, C. W. Jones, Chem. Mater. 2006,
[50] B. S. Tian, C. Yang, J. Phys. Chem. C 2009, 113, 4925. 18, 5022.
[51] W. G. Barb, J. Chem. Soc. 1955, 2564. [70] J. C. Hicks, C. W. Jones, Langmuir 2006, 22, 2676.
[52] D. D. Do, Adsorption Analysis: Equilibria and Kinetics, Vol. 2, Imperial College [71] D. Y. Zhao, Q. S. Huo, J. L. Feng, B. F. Chmelka, G. D. Stucky, J. Am. Chem.
Press, London 1998. Soc. 1998, 120, 6024.
[53] A. von Harpe, H. Petersen, Y. X. Li, T. Kissel, J. Controlled Release 2000, 69, [72] T. S. Pierre, M. Geckle, J. Macromol. Sci.: Chem. 1985, A22, 877.
309. [73] W. J. Koros, D. R. Paul, J. Polym. Sci, Part B: Polym. Phys. 1976, 14, 1903.

3832 ß 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2009, 19, 3821–3832