Report

CO2 capture from humid flue gases and humid atmosphere using a microporous coppersilicate

See allHide authors and affiliations

Science  16 Oct 2015:
Vol. 350, Issue 6258, pp. 302-306
DOI: 10.1126/science.aab1680

Grabbing CO2 from wet gas streams

It is a challenge to extract CO2 from typical gas streams, such as the flue gas from a power plant. This is because any water in the stream tends to prevent CO2 absorption and may also degrade the absorbing material. Datta et al. developed a microporous copper silicate that avoids these problems. Most other materials have sites that absorb both water and CO2 at the same sites, and in that fight, the water tends to win. Although their material still absorbs water, it has separate sites for the CO2 absorption. It also shows good stability despite the absorbed water and can be reused.

Science, this issue p. 302

Abstract

Capturing CO2 from humid flue gases and atmosphere with porous materials remains costly because prior dehydration of the gases is required. A large number of microporous materials with physical adsorption capacity have been developed as CO2-capturing materials. However, most of them suffer from CO2 sorption capacity reduction or structure decomposition that is caused by co-adsorbed H2O when exposed to humid flue gases and atmosphere. We report a highly stable microporous coppersilicate. It has H2O-specific and CO2-specific adsorption sites but does not have H2O/CO2-sharing sites. Therefore, it readily adsorbs both H2O and CO2 from the humid flue gases and atmosphere, but the adsorbing H2O does not interfere with the adsorption of CO2. It is also highly stable after adsorption of H2O and CO2 because it was synthesized hydrothermally.

Efforts to curtail the increase in atmospheric CO2 concentrations rely on the development of economical methods of capturing CO2 from flue gas and the atmosphere (15). One possible approach involves capture of CO2 by physical adsorption on microporous materials that have high surface areas. To date, various materials that have high CO2 sorption capabilities at 298 K have been developed. They include zeolites (610), metal-organic frameworks (MOFs) (1116), and zeolitic imidazolate frameworks (17, 18). However, these materials require the incoming gas stream to be completely dehydrated, as water causes a drastic reduction in the CO2 sorption capabilities (19, 20) or may even promote their decomposition (13, 14). Although such moisture-sensitive CO2 sorbents can still be used to capture CO2 directly from nonpretreated humid flue gases by charging the column with a water-sorbing layer before the CO2-sorbing layer, the use of a single moisture-insensitive layer would be preferable (4, 5, 21).

Only a limited number of materials meeting this requirement have been discovered, and these substances can only adsorb small to moderate amounts of CO2 from humid flue gas (13, 14, 1820). Furthermore, most flue gases are hot, with temperatures ranging between 373 and 403 K at the point they are released into the atmosphere. Therefore, thermal stability under humid conditions is another key property.

Using a gel consisting of sodium silicate and copper sulfate, we synthesized microporous coppersilicate crystals of uniform size and shape (see supplementary materials). The crystals, which we call SGU-29, have a square bipyramid crystal morphology, which suggests that each crystal has pseudo–four-fold symmetry along the axis (assigned later as the c axis) from the center of the square to the top of the pyramid. Two typical scanning electron microscopy (SEM) images with different crystal sizes are shown in Fig. 1A and the inset. The determined chemical formula was Na2CuSi5O12. Almost all reflections observed in the x-ray powder diffraction pattern matched well with those of ETS-10 (22, 23) and AM-6 (fig. S1) (24, 25). The crystals are stable in air up to 550°C (fig. S2). The effective magnetic moment of Cu (fig. S3A) confirmed that the oxidation state of Cu is 2+. The electron spin resonance spectrum of SGU-29 showed that the electron spins on Cu2+ ions are strongly coupled (fig. S3B). Characteristic features observed by transmission electron microscopy (TEM) can be summarized as follows: (i) In the high-resolution TEM (HRTEM) image taken along the 110 direction (parallel to the channel direction; Fig. 1B), large bright ellipses are arranged horizontally with a period of 14.7 Å with a dark contrast observed between any two neighboring ellipses. Simultaneously, a horizontal row of small white dots arranged in a zigzag manner can also be noticed between the successive arrays of ellipses. The bright contrast of ellipses and small dots corresponds to large and small pores, respectively. The large pores resemble channels formed by 12-membered rings judging from their sizes. These pores belong to single layers that are marked as A, B, C, and D. (ii) There is a horizontal shift by one-quarter of the ellipses between successive layers either to the right in the upper part of the image forming ABCD stacking sequence or to the left in the lower part with that of DCBA. (iii) A stacking sequence of BABAB can be observed at the boundary between the two parts.

Fig. 1 Crystal structure of SGU-29.

(A) SEM image of crystals with uniform size and regular square bipyramidal morphology. Inset: An individual single crystal. (B) HRTEM image with marked unit cells corresponding to monoclinic (M, yellow), monoclinic mirrored (Mm, pink), and tetragonal (T, green) structures. (C) Electron diffraction pattern taken along the [1-10]M zone axis (alternatively [010]T). The arrows mark the principal directions for corresponding unit cells drawn in (B). (D) Electron diffraction pattern along the [001]T zone axis. (E) Structure view along the c axis. (F) Structure view along the channel direction. (G) Perspective projection of several unit cells along the channel direction. (H) [CuO4] square planar geometry, with the average bond length and two Cu–O bond angles that form the fundamental secondary building unit shown in (I). (I) [CuO4] column side view. (J) Five different Na+ positions shown in the channel. (K) Positions of Na+ ions in the channel system.

The monoclinic stacking sequences ABCD and DCBA form crystalline regions on either side, which are mirror images of each other due to the horizontal mirror plane. The boundary BABAB stacking represents the tetragonal structure. The corresponding configurations were confirmed by single-crystal x-ray diffraction analysis. Electron diffraction pattern taken along the channels direction can be interpreted using the above description (Fig. 1C). Diffuse streaks induced by the layer-stacking disorder can be observed. The diffraction pattern taken along the [001]T zone axis (Fig. 1D) confirms the similarity of SGU-29 and the monoclinic polymorph type B reported for the ETS-10 structure by Anderson et al. (22). From these observations we conclude that SGU-29 has the same basic framework structure with stacking defects as they were observed in ETS-10 and AM-6 systems (2225).

Using a synchrotron radiation source (λ = 0.700 Å), the crystal structure solution of SGU-29 was retrieved from a single crystal. The most probable solution was obtained in the centrosymmetric monoclinic lattice system (space group C2/c) with unit cell parameters of a = 20.820 Å, b = 20.819 Å, c = 14.697 Å, and β = 110.73° with Rsym = 0.080 (see table S1 for detailed crystal structure information). The bulk density (final column density) of the SGU-29 powder shown in Fig. 1A is 0.84 g/cm3.

The projections of the monoclinic structure along the [001] and Embedded Image directions are shown in Fig. 1, E and F, respectively, and the perspective view along the Embedded Image direction is shown in Fig. 1G. Relative to the crystal structures of ETS-10 (22, 23) and AM-6 (24, 25), the most critical difference in the crystal structure of SGU-29 is that copper(II) ions adopt a square planar geometry (Fig. 1, H and I), whereas titanium(IV) in ETS-10 and vanadium(IV) in AM-6 have octahedral geometries. Unlike ETS-10 and AM-6, which respectively have one-dimensional -O-Ti(O)4-O- and -O-V(O)4-O- wires, there is no bridging oxygen atom between the neighboring Cu(II) centers. Each CuO4 square plane is supported by four surrounding SiO4 tetrahedra (Fig. 1H) and is arranged almost parallel to each other with the distances of 3.629 and 3.732 Å (Fig. 1I). The positions of five crystallographically distinct Na+ ion positions are shown in Fig. 1, J and K. Na2 exists at the cross section of two channels, and Na3 binds to four oxygen atoms bound to Cu(II) ions from the channel side. Accordingly, Na2 and Na3 can interact with molecules incorporated into the silica channels. The effective window size of the silica channel in SGU-29 is 4.5 × 7.3 Å (fig. S11), which is slightly smaller than the reported values of ETS-10 (4.7 × 7.7 Å, 4.5 × 7.6 Å) (22, 23).

For the purpose of obtaining CO2 adsorption isotherms and dynamic CO2 breakthrough profiles at various temperatures for SGU-29 and the related materials, we used pure CO2 gas and four simulated flue gases denoted F0, F29, F92, and F201, and three simulated air samples denoted A70, A80, and A90 (Table 1). They differ in their compositions of H2O, CO2, O2, and N2 which are represented by their partial pressures (mbar) of the component gases (pPgas). F0 is a dry flue gas and F29, F92, and F201 are humid flue gases with Embedded Image of 29, 92, and 201 mbar, respectively. A70, A80, and A90 are humid air samples with different relative humidities of 70, 80, and 90%, respectively. The Embedded Image values of F0, F29, F92, and F201 are ~100 mbar, which corresponds to the average Embedded Image in flue gas, and that of A70, A80, and A90 is 0.4 mbar or 400 ppm, which corresponds to the average Embedded Image in the atmosphere. The temperatures of the simulated gases are 298 K, except for F92 (318 K) and F201 (338 K).

Table 1 Compositions of the simulated flue gases used in this study.

View this table:

The CO2 adsorption isotherms of SGU-29 in the 0 ≤ Embedded Image ≤ 1000 mbar and 0 ≤ Embedded Image ≤ 100 mbar regions at temperatures between 273 and 373 K are shown in Fig. 2, A and B, respectively. The CO2 adsorption isotherms for SGU-29 in the 0 ≤ Embedded Image ≤ 100 and 0 ≤ Embedded Image ≤ 0.4 mbar regions at 298 K are compared with the corresponding isotherms (Fig. 2, C and D) of AM-6, Mg-DOBDC, ETS-4, ETS-10, and SIFSIX-3-Cu. NaX is also included in the comparisons because it has been considered as a material for industrial applications (4, 9, 21). The Embedded Image values of SGU-29 at Embedded Image = 1000, 100, and 0.4 mbar and 298 K are 156, 126, and 26 cm3/cm3, respectively. Although Embedded Image of SGU-29 at Embedded Image = 1000 mbar ranks third among the top seven materials (Fig. 2E, fig. S12A, and table S4), it ranks first at Embedded Image = 100 mbar (Fig. 2F, fig. S12B, and table S4). We further found that the Embedded Image values of AM-6, ETS-4, and ETS-10 at Embedded Image = 100 mbar rank second, fourth, and fifth, respectively, indicating that the four materials with the same structural origin (SGU-29, AM-6, ETS-10, and ETS-4) are among the top five materials. This indicates that the very high CO2-capturing ability of SGU-29 at Embedded Image = 100 mbar arises primarily from its structure. The increase in Embedded Image ongoing from ETS-10 to AM-6 and to SGU-29 seems to arise from the fine structural variation of the TiO62–, VO62–, and CuO42– centers within the structures and the corresponding change of the electronic properties of the framework oxygen and the charge-balancing cations along the silica channels.

Fig. 2 CO2 adsorption isotherms and isosteric heats of adsorption.

(A and B) CO2 adsorption isotherms of SGU-29 in two different Embedded Image regions. (C and D) Comparison of the CO2 adsorption isotherms of SGU-29, AM-6, ETS-4, ETS-10, SIFSIX-3-Cu, Mg-DOBDC, and NaX in the Embedded Image region between 0 and 100 mbar (C) and between 0 and 0.4 mbar (D) at 298 K. (E to G) Comparison of CO2 uptakes of SGU-29 at Embedded Image = 1000 mbar (E), 100 mbar (F), and 0.4 mbar (G) relative to previously reported CO2 sorbent. (H) Plots of isosteric heat versus uptake of SGU-29 and NaX for H2O, CO2, N2, and O2.

At Embedded Image = 0.4 mbar, Embedded Image of SGU-29 ranks second after that of SIFSIX-3-Cu (43 cm3/cm3) (Fig. 2G and fig. S13); the Embedded Image values of AM-6 and ETS-10 rank third and fourth, confirming the intrinsically high affinity of the ETS-10–type framework structure for CO2.

The isosteric heats of adsorption of CO2, H2O, N2, and O2 [(Qst)gas] onto SGU-29 and NaX are compared in Fig. 2H. For both sorbents, the order of (Qst)gas values is Embedded Image > Embedded Image >> Embedded Image >Embedded Image in the region of U between 0 and 200 cm3/cm3. The initial Embedded Image of SGU-29 is higher than those of most of the materials listed in table S5, including NaX (48.2 kJ/mol). Note that Embedded Image of SGU-29 remains nearly constant regardless of Embedded Image, whereas that of NaX sharply decreases when Embedded Image reaches about two-thirds of its maximum (Fig. 2H). This shows that Embedded Image of SGU-29 is not only higher but also more constant with respect to Embedded Image than that of NaX.

These observations suggest that the attractive interaction between CO2 and the adsorption site in SGU-29 is not only stronger but also more homogeneous than that of NaX, regardless of Embedded Image. They also suggest that the adsorption sites for CO2 in this substance do not cooperate in a negative manner to prevent adsorption of this gas. More important, Embedded Image of SGU-29 is lower than that of NaX by 8 to 14 kJ/mol, regardless of Embedded Image, which indicates that SGU-29 has much lower affinity to H2O than does NaX. Furthermore, the amount of H2O adsorbed onto SGU-29 is much smaller than the amount adsorbed onto NaX and most of the materials shown in figs. S14 and S15. In general, SGU-29 and the materials that belong to the same structural family have much lower Embedded Image values than that of NaX. Even among the group of low-Embedded Image materials, SGU-29 shows the lowest Embedded Image (fig. S14). Such a high affinity and capacity for CO2, together with a very low affinity and capacity for H2O, make SGU-29 a much more suitable CO2 remover from the humid flue gas than NaX and the other materials tested.

The dynamic breakthrough (dynamic column CO2 separation) profiles of seven CO2 adsorbents were obtained at 298 K using a column CO2 separation setup (fig. S10), with F0 (Fig. 3A and table S5) and F29 (Fig. 3B and table S5) as flue gases. The Embedded Image values of SGU-29 from F0 (dry) and F29 (humid) are 117 cm3/cm3 and 115 cm3/cm3, respectively. The small decrease in Embedded Image arises from the decrease of Embedded Image by 3 mbar upon changing F0 to F29. Thus, moisture does not affect the CO2 capture ability of SGU-29, which is a very important feature of SGU-29. The ratio of the Embedded Image obtained from the humid flue gas (F29) to the Embedded Image obtained from dry flue gas (F0) [Embedded Image] is plotted for various materials in Fig. 3C. The ratios of SGU-29 (1.0), AM-6 (1.0), ETS-10 (0.96), SIFSIX-3-Zn (0.96), ETS-4 (0.95), SIFSIX-3-Cu (0.93), and UTSA-16 (0.90) fall between 1.0 and 0.9. They are categorized as moisture-insensitive CO2 sorbents. The ratios of NaX (0.81), Ni-DOBDC (0.78), and Mg-DOBDC (0.35) fall between 0.8 and 0.35; they are categorized as moisture-sensitive CO2 sorbents. Interestingly, SGU-29, AM-6, and ETS-10 are the top three materials in the group of moisture-insensitive sorbents, indicating that the very high Embedded Image ratio of SGU-29 arises from its structure.

Fig. 3 Dynamic CO2 breakthrough profiles, linear Embedded Image-Tad relationship, and recyclability.

(A and B) Dynamic CO2 breakthrough profiles of SGU-29, AM-6, and the other reported best-performing materials with F0 (A) and F29 (B). (C) Plot of Embedded Image with respect to various materials as indicated. (D) Linear relationships between Embedded Image and Tad for SGU-29 and NaX. (E and F) Comparisons of Embedded Image and the recyclability of SGU-29 with those of the other indicated materials, with humid F29 flue gas (E) and humid A70, A80, and A90 air (F), at 298 K.

We measured Fourier transform infrared (FTIR) spectra of SGU-29 and NaX while passing F0 or F29 (flow rate 5 ml/min) onto each dried sorbent placed in an environmental chamber. With F0 as the flue gas and dry SGU-29 as the sorbent, the absorption peaks due to the adsorbed CO2 gradually increased in the 2250 to 2450 cm−1 region and the absorbance reached 2.75 during the period of 40 min (Fig. 4A). When the flow rate of F0 was increased to 50 ml/min, it took only 2 min for the absorbance to reach the maximum (2.75), indicating that the CO2 uptake is a very rapid process (fig. S16). Even with F29 as the flue gas, the growth rate, peak shapes, and intensities of the adsorbed CO2 peaks were identical to those with F0 as the flue gas, as if F29 were a dry flue gas. However, the broad peaks due to adsorbed H2O simultaneously appeared in the 3700 to 2800 cm−1 region, and their intensities gradually grew with time, with the absorbance reaching a maximum at 0.75. These results show that both CO2 and H2O simultaneously adsorb on SGU-29 but the H2O adsorption does not interfere with the CO2 adsorption, indicating that there are different H2O-specific and CO2-specific sites in SGU-29 and they do not interfere with each other (fig. S17A). A similar phenomenon was observed for AM-6 (fig. S18). Recently, various amine-functionalized porous materials have been introduced (2630) and it has been shown that H2O does not interfere with the chemical adsorption of CO2 by amine groups (29), indicating that the chemical reaction between amine groups and CO2 is highly specific, such that water cannot interfere with the reaction. In that respect, SGU-29 can be regarded to have chemical reaction–like, highly specific CO2 sorption selectivity.

Fig. 4 Effect of adsorbed water on adsorbed CO2.

Progressive change of the Fourier transform infrared spectra of (A and B) dried SGU-29 and (C and D) dried NaX, with time (as indicated) under the flow of dry flue gas F0 [(A) and (C)] and humid flue gas F29 [(B) and (D)] at 298 K. Flow rate = 5 ml/min.

With F0 as the flue gas and dry NaX as the sorbent, the IR absorption peaks due to the adsorbed CO2 gradually increased and the absorbance reached 2.75 during the period of 40 min (Fig. 4C). However, with F29 as the flue gas and dry NaX as the sorbent, a very broad (3750 to 2000 cm−1)absorption peak due to the adsorbed H2O appeared, with a maximum absorbance of 1.25 at 3400 cm−1 (Fig. 4D). The integrated area of H2O peak is much larger than that of SGU-29, consistent with the fact that a much larger amount of water is adsorbed into NaX than into SGU-29 (figs. S14 and S15). In NaX, the intensity of the H2O peak rapidly increased to 1.1 during the first 20 min and then slowly increased to 1.25 during the next 20 min. Simultaneously, the intensity of the CO2 peak reached 2.25 during the initial 20 min and subsequently decreased to 2.0 during the next 20 min. This shows that a portion of the adsorbed CO2 is replaced by H2O molecules. We therefore conclude that, in NaX, there are H2O/CO2-sharing sites as well as H2O-specific and CO2-specific sites (fig. S17B). From the observation that Embedded Image = 0.82 (Fig. 3C), the percentage of the H2O/CO2-sharing sites in the total CO2-sorbing sites is 18% in NaX. By the same analysis, the percentage of the H2O/CO2-sharing sites in Ni-DOBDC and Mg-DOBDC are 22 and 65%, respectively.

Both SGU-29 and NaX display negative linear relationships between Embedded Image and temperature (Tad), as is demonstrated by Embedded Image = –aTad + b plots, where a is the slope and b is the intercept at T = 298 K (Fig. 3D). Plot I in Fig. 3D corresponds to a CO2 isotherm for SGU-29 determined by using pure CO2 gas at temperatures between 298 and 373 K. Plots II, III, and IV were obtained by CO2 breakthrough experiments at different Tad using F0, F92, and F201 as flue gases, respectively. The respective slopes of the lines in plots I to IV are –0.97, –0.92, –0.89, and –0.83. The downward shift of the plot ongoing from plot I obtained by pure CO2 gas to plot II obtained by F0 is ascribed to interference of Embedded Image by coexisting N2 and O2, which are the major components of the flue gas. Despite the fact that the increase in Embedded Image on going from F0 to F92 to F201 is drastic, plots II, III, and IV essentially overlap on each other. Accordingly, the Embedded Image values at 378 K are similar (40 to 45 cm3/cm3) for F0, F92, and F201, respectively. The above results again show that Embedded Image does not affect the Embedded Image = –aTad + b plots.

Plot V, generated from NaX using F0, has a slope of –0.82 and an intercept of 73 cm3/cm3; plot VI, generated from NaX using F92, has a slope of –0.80 and an intercept of 58 cm3/cm3. The slopes are similar to those of plots I to IV, but the intercepts are much lower than the corresponding values of SGU-29, consistent with the fact the intrinsic affinity of NaX for CO2 is lower than that of SGU-29. The downward shift from plot V to plot VI is much larger than that from plot II to plot III, which again demonstrates that whereas Embedded Image of NaX is sensitively affected by Embedded Image, that of SGU-29 is not. At 368 K, the Embedded Image values of plot III (by SGU-29 with F92) and plot VI (by NaX with F92) are 51 cm3/cm3 and 0 cm3/cm3, respectively; hence, SGU-29 can remove a substantial amount of CO2 directly from a hot and humid flue gas, whereas NaX cannot.

SGU-29 is also repeatedly reusable (Fig. 3E) after desorption of the adsorbed CO2 by evacuation at room temperature or at elevated temperatures such as 523 K. This indicates that its structure and CO2 sorption sites remain intact during repeated exposure to hot and humid flue gases. Specifically, the Embedded Image of SGU-29 obtained from F29 (relative humidity 90%) at 298 K does not change over the course of 50 CO2 breakthrough experiments. Also, the Embedded Image (40 cm3/cm3) of SGU-29 for flue gas F201 at 378 K remains constant during 10 CO2 breakthrough experiments, confirming the recyclability of this material for CO2 capture from the hot and humid flue gases. Although NaX and some MOFs such as Ni-DOBDC and UTSA-16 (fig. S19) were also found to be recyclable, their Embedded Image values (53, 50, and 37 cm3/cm3) are half that of SGU-29. In contrast, the Embedded Image values of SIFSIX-3-Cu, Mg-DOBDC, and SIFSIX-3-Zn (fig. S19) were found to decrease progressively during repeated use. The estimated power consumption by vacuum swing adsorption at room temperature was 2.04 MJ per kg of CO2 (4, 5).

The initial Embedded Image values for air samples A70, A80, and A90 were 5.3, 4.9, and 4.7 cm3/cm3, respectively; again, these values did not change over repeated experiments. Although the Embedded Image value of SIFSIX-3-Cu was higher than that of SGU-29 at Embedded Image = 0.4 mbar under the dry condition (Fig. 2, D and G), its initial Embedded Image values for CO2 capture from the simulated humid air samples A70, A80, and A90 were lower than those of SGU-29. Furthermore, as the data in Fig. 3F show, SIFSIX-3-Cu is not recyclable under humid conditions.

SGU-29 is a microporous coppersilicate that can be immediately applied in the field as a material for capturing CO2 from humid flue gases without involving the costly prior dehydration step. SGU-29, AM-6, and ETS-10 form an important class of materials having an isostructural platform that gives rise to very high UCO2 (humid)/UCO2 (dry) ratios (>0.96).

Supplementary Materials

www.sciencemag.org/content/350/6258/302/suppl/DC1

Materials and Methods

Figs. S1 to S19

Tables S1 to S5

References (3148)

References and Notes

  1. Acknowledgments: Supported by the Korea Center for Artificial Photosynthesis, located at Sogang University and funded by the Ministry of Science, ICT and Future Planning through the National Research Foundation of Korea, grant 2009-0093886. We thank C.-H. Shin at Chungbuk National University for helpful discussions regarding the isosteric heats of adsorption, and J. Y. Lee for help in drawing the figures.
View Abstract

Navigate This Article