Adsorption of CO 2 on Amine-Modiﬁed Silica Particles in a Conﬁned-Fluidized Bed

: To reduce the anthropogenic CO 2 emissions produced from fossil fuel burning plants, the application of carbon capture and storage (CCS) is necessary and development of a more e ﬃ cient and economically feasible CO 2 capture process is essential as an alternative to the conventional amine scrubbing process which uses aqueous amine solutions. CO 2 capture can be enhanced by improving both the gas–solid contact e ﬃ ciency and by tuning a speciﬁc high-performance sorbent. The aim of this research is to investigate the adsorption of CO 2 using impregnated mesoporous silica in a “conﬁned-ﬂuidized bed”. This non-conventional ﬂuidized bed (sometimes also termed the “packed-ﬂuidized bed”) seems suitable for improving the e ﬃ ciency of gas–solid processes for which the bypass e ﬀ ect of the gas–solid contact caused by bubbling represents a major drawback. Results, expressed as grams of CO 2 adsorbed per kilogram of material, are discussed in terms of amine load in the sorbent, breakthrough time and fraction of bed utilized. The stability of the materials after regeneration cycles is also discussed. The results obtained conﬁrm that the conﬁnement of the bed allows exploiting ﬂuidization technology in adsorption operations. The operating velocity can be ﬁxed at a value at which the thermal e ﬀ ects also connected to the operation are kept under control. R.G. and F.T.; investigation, R.B. and G.L.; resourses, R.G. and F.T; data curation, R.G. and F.T.; writing—original draft preparation, R.G.; writing—review and editing, R.G., F.T. and B.F.; visualization, R.G.; supervision, B.F.; project administration, F.T.; funding acquisition, F.T. and B.F.


Introduction
The estimates released in a new report from the Global Carbon Project, an international research consortium dedicated to tracking the world's greenhouse gas emissions, represents an increasing long-term trend of the total carbon emissions from all human activities, including industrial activities and burning of fossil fuels, agriculture and land use [1]. International climate goals outlined in the Paris Agreement will be harder to reach as long as burning coal will be the cheapest way of generating energy and coal is relatively abundant in large energy-consuming countries. So far, Carbon dioxide Capture and Storage (CCS) is considered the most promising post-combustion technology to sequestrate CO 2 from large stationary sources of CO 2 emissions [2].
The presence of other contaminants in the exhaust gas, such as NO x , SO x , and particulate matter, are not crucial because they are usually separated before the CO 2 capture process, but the low partial pressure of CO 2 in the flue gas at atmospheric pressure, derived by the CO 2 concentration within 13-15%, is the limiting factor, as it is a very low driving force for its separation. For this reason, monoethanolamine-based chemical absorption or a scrubbing process is currently the most suitable In this sketch, the shaded zone represents the bed of fine adsorbent fluidized in the voids of the glass beads forming a much taller fixed bed (the clearer zone in the figure).
Glass beads have a diameter of 11 mm and a density of 2.48 g/cm 3 . Their packing, never subjected to fluidization, was always 450 mm high, so that its mass was of 1230 g. In it, a smaller amount of fine adsorbent was loaded by percolating through the voids of the glass beads, this material formed a bed of height Hfc, which gradually increases during the fluidization test. The gas utilized for fluidization of the fine adsorbent bed was air, whose flow rates were regulated by a set of two mass flow controllers covering the range 0 ÷ 6000 Nl/h. For the adsorption tests, the CO2 concentration in the gas feed stream was set by adding CO2 from a cylinder to the air supplied by a compressor. A column filled with zeolites and activated carbon was located before the CO2 mixing point to remove the humidity of the air. The total pressure drop across the solid bed was measured by a U-tube water manometer. The CO2 concentration in the gas stream entering or leaving the adsorption unit was measured by a gas analyzer, GA-21 plus (Madur Polska Sp. z o.o., Zgierz , Poland) and acquired on a personal computer. A thermocouple vertically immersed in the bed was used to determine its temperature.

Adsorbent Preparation and Characterization
Commercial silica gel and the chemical (3-Aminopropyl)triethoxysilane (APTES, >98%) were provided by Sigma-Aldrich (St. Louis, MO, USA). Ethanol (analysis grade) was provided by J.T. Baker-Avantor (Radnor, PA, USA). Before the functionalization procedure, the 400-500 μm sample was prepared by sieving to obtain a sample with a short granulometric dispersion and its particle size distribution, reported in Figure 2, was determined by a laser diffractometer, Helos (Sympatec GmbH, Clausthal-Zellerfeld, Germany). Particle shape was characterized thanks to an image analysis system, Qicpic (Sympatec GmbH, Clausthal-Zellerfeld, Germany) and Figure 3 reports some examples of the images obtained. Characteristic parameters of the distribution obtained are reported in Table 1. In this sketch, the shaded zone represents the bed of fine adsorbent fluidized in the voids of the glass beads forming a much taller fixed bed (the clearer zone in the figure).
Glass beads have a diameter of 11 mm and a density of 2.48 g/cm 3 . Their packing, never subjected to fluidization, was always 450 mm high, so that its mass was of 1230 g. In it, a smaller amount of fine adsorbent was loaded by percolating through the voids of the glass beads, this material formed a bed of height H fc , which gradually increases during the fluidization test. The gas utilized for fluidization of the fine adsorbent bed was air, whose flow rates were regulated by a set of two mass flow controllers covering the range 0 ÷ 6000 Nl/h. For the adsorption tests, the CO 2 concentration in the gas feed stream was set by adding CO 2 from a cylinder to the air supplied by a compressor. A column filled with zeolites and activated carbon was located before the CO 2 mixing point to remove the humidity of the air. The total pressure drop across the solid bed was measured by a U-tube water manometer. The CO 2 concentration in the gas stream entering or leaving the adsorption unit was measured by a gas analyzer, GA-21 plus (Madur Polska Sp. z o.o., Zgierz, Poland) and acquired on a personal computer. A thermocouple vertically immersed in the bed was used to determine its temperature.

Adsorbent Preparation and Characterization
Commercial silica gel and the chemical (3-Aminopropyl)triethoxysilane (APTES, >98%) were provided by Sigma-Aldrich (St. Louis, MO, USA). Ethanol (analysis grade) was provided by J.T. Baker-Avantor (Radnor, PA, USA). Before the functionalization procedure, the 400-500 µm sample was prepared by sieving to obtain a sample with a short granulometric dispersion and its particle size distribution, reported in Figure 2, was determined by a laser diffractometer, Helos (Sympatec GmbH, Clausthal-Zellerfeld, Germany). Particle shape was characterized thanks to an image analysis system, Qicpic (Sympatec GmbH, Clausthal-Zellerfeld, Germany) and Figure 3 reports some examples of the images obtained. Characteristic parameters of the distribution obtained are reported in Table 1.   XRD (X-Ray Diffraction) data were collected using an XPert3 diffractometer (Malvern Panalytical Ltd, Malvern, UK) (Cu Kα radiation λ = 1.5406 Å). Small angles were scanned between 0.7° and 10° 2θ for identifying any possible mesostructure of the material. High angles were scanned between 5° and 60° 2θ in order to verify the presence of order at the atomic scale.
In Figure 4, XRD diffraction at small angles of the silica particles is reported showing the absence of the mesostructure. The XRD diffraction at high angles gives evidence of the amorphous character of the material (figure not reported).    XRD (X-Ray Diffraction) data were collected using an XPert3 diffractometer (Malvern Panalytical Ltd, Malvern, UK) (Cu Kα radiation λ = 1.5406 Å). Small angles were scanned between 0.7° and 10° 2θ for identifying any possible mesostructure of the material. High angles were scanned between 5° and 60° 2θ in order to verify the presence of order at the atomic scale.
In Figure 4, XRD diffraction at small angles of the silica particles is reported showing the absence of the mesostructure. The XRD diffraction at high angles gives evidence of the amorphous character of the material (figure not reported).   XRD (X-Ray Diffraction) data were collected using an XPert3 diffractometer (Malvern Panalytical Ltd, Malvern, UK) (Cu Kα radiation λ = 1.5406 Å). Small angles were scanned between 0.7 • and 10 • 2θ for identifying any possible mesostructure of the material. High angles were scanned between 5 • and 60 • 2θ in order to verify the presence of order at the atomic scale.
In Figure 4, XRD diffraction at small angles of the silica particles is reported showing the absence of the mesostructure. The XRD diffraction at high angles gives evidence of the amorphous character of the material (figure not reported).
Two functionalized mesoporous silica samples were prepared using a modified procedure reported by Quang et al. (2013) [8]. In total, 100 g of silica gel material was soaked with 400 g of an APTES/ethanol impregnating solution 20/80 (APTES20) and 40/60 (APTES40) w/w in a laboratory flask for 72 h at room temperature and, successively, for 48 h at 40 • C under mechanical stirring. Finally, the slurry was dried in an oven at 60 • C for 48 h. After a CO 2 adsorption test, regeneration cycles were performed heating the powder in an oven at 120 • C for 3 h.
According to the method proposed by Abrahamsen and Geldart (1980) [15], the particle apparent density of the porous adsorbent was determined by comparison with a non-porous solid having the same particle size distribution, similar shape and known density. This method is useful for porous material and is based on the observation that the minimum packed voidage is virtually the same for solids of similar size and particle shape.
Panalytical Ltd, Malvern, UK) (Cu Kα radiation λ = 1.5406 Å). Small angles were scanned between 0.7° and 10° 2θ for identifying any possible mesostructure of the material. High angles were scanned between 5° and 60° 2θ in order to verify the presence of order at the atomic scale.
In Figure 4, XRD diffraction at small angles of the silica particles is reported showing the absence of the mesostructure. The XRD diffraction at high angles gives evidence of the amorphous character of the material (figure not reported).  In this work, 140 g of a siliceous sand of density ρ = 2.66 g/cm 3 , exactly weighed and used as a reference, was loaded into a graduated cylinder and compacted until the minimum volume was reached. A value of voidage of 0.457 was obtained from the relation: Subsequently, the same amount of silica gel was weighed, packed and its occupied volume was determined. Assuming that the voidage of the silica gel particles was equal to that of the reference material when the minimum volume was reached [15]-i.e., that ε sand = ε mesoporous-silica -application of Equation (1) to silica gel data made it possible to calculate its unknown density, which was equal to 0.947 g/cm 3 . The same procedure, applied to two functionalized samples, yielded higher values of particle density, as indicated in Table 2. Nitrogen adsorption-desorption isotherms were acquired using a ASAP 2020 Plus analyzer, (Micromeritics, Norcross, GA, USA). Samples were degassed at 120 • C in nitrogen atmosphere at a pressure of 10 µmHg for 3 h. Surface area was calculated by the Brunauer-Emmett-Teller (BET) method [16] and pore volume by the Barrett-Joyner-Halenda (BJH) method [17] using desorption isothermal data.
Isotherms at 77 K for the three samples studied are reported in Figure 5. The isotherms of as-made silica are of type IV according to the IUPAC (International Union of Pure and Applied Chemistry) classification [18]. Hysteresis loops between 0.5 and 0.9 p/p 0 evidence the presence of mesopores. The surface area calculated with the BET model is 584 m 2 /g and the pore volume is 0.89 cm 3 /g (see Table 2). These properties make the sample of the as-made silica well suitable for grafting through an impregnation process with APTES. In fact, the APTES20 and APTES40 samples adsorbed a considerably lower amount of N2 with respect to the precursor (SBET of 180 and 73 m 2 /g, respectively).
The pore size distribution of the as-made silica ( Figure 6) is centred at 5 nm and ranges from 4 to c.a. 13 nm. Pore volume of APTES20 sample is 0.30 cm 3 /g and the pore size distribution is similar to that of the as-made material giving evidence that mesopores are not completely occluded. In APTES40, the sample pore volume decreases to 0.10 cm 3 /g and only porosity up to 5 nm is present. Micrographs were obtained using a Field Emission Scanning Electron Microscopy technique by means a Merlin (Zeiss Instruments, Oberkochen, Germany). All the samples were coated with a 5 nm layer of platinum under a tension of 3 kV. In Figure 7, the micrographs of the as-made silica and of APTES20 and APTES40 are compared-the surface of the silica particles is essentially rough but it becomes smooth for the functionalized samples. In addition, small aggregates (up to 200 nm) of the organic molecules are present with a higher population density for the APTES40. In fact, the APTES20 and APTES40 samples adsorbed a considerably lower amount of N 2 with respect to the precursor (S BET of 180 and 73 m 2 /g, respectively).
The pore size distribution of the as-made silica ( Figure 6) is centred at 5 nm and ranges from 4 to c.a. 13 nm. Pore volume of APTES20 sample is 0.30 cm 3 /g and the pore size distribution is similar to that of the as-made material giving evidence that mesopores are not completely occluded. In APTES40, the sample pore volume decreases to 0.10 cm 3 /g and only porosity up to 5 nm is present. In fact, the APTES20 and APTES40 samples adsorbed a considerably lower amount of N2 with respect to the precursor (SBET of 180 and 73 m 2 /g, respectively).
The pore size distribution of the as-made silica ( Figure 6) is centred at 5 nm and ranges from 4 to c.a. 13 nm. Pore volume of APTES20 sample is 0.30 cm 3 /g and the pore size distribution is similar to that of the as-made material giving evidence that mesopores are not completely occluded. In APTES40, the sample pore volume decreases to 0.10 cm 3 /g and only porosity up to 5 nm is present. Micrographs were obtained using a Field Emission Scanning Electron Microscopy technique by means a Merlin (Zeiss Instruments, Oberkochen, Germany). All the samples were coated with a 5 nm layer of platinum under a tension of 3 kV. In Figure 7, the micrographs of the as-made silica and of APTES20 and APTES40 are compared-the surface of the silica particles is essentially rough but it becomes smooth for the functionalized samples. In addition, small aggregates (up to 200 nm) of the organic molecules are present with a higher population density for the APTES40. Micrographs were obtained using a Field Emission Scanning Electron Microscopy technique by means a Merlin (Zeiss Instruments, Oberkochen, Germany). All the samples were coated with a 5 nm layer of platinum under a tension of 3 kV. In Figure 7, the micrographs of the as-made silica and of APTES20 and APTES40 are compared-the surface of the silica particles is essentially rough but it becomes smooth for the functionalized samples. In addition, small aggregates (up to 200 nm) of the organic molecules are present with a higher population density for the APTES40. Thermogravimetric analysis (TGA) was performed using a DSC-TGA STA409C analyzer (Netzsch-Gerätebau GmbH, Selb, Germany) from 20 to 800 °C at a heating rate of 10 °C in static air. Figure 8 reports the TGA thermograms for the as-made silica, APTES20 and APTES40 samples. Mass loss from room temperature up to ca. 150 °C is attributed to the water physiosorbed on the particles' surfaces. From 120 to 800 °C, vicinal Si-OH groups react with production of H2O molecules and the formation of -Si-O-Si-groups. When APTES molecules impregnate the silica particles, an important mass loss is present between ca 300 and 600 °C due to degradation and/or combustion of the organic moieties adsorbed or chemically linked on the silica surface. In this case, water is also adsorbed on the organic sites. These data can be utilized for an evaluation of the amount of the APTES molecules. If we consider APTES molecules (MW 221.37 g/moles) as a reference, the amount of the organic moieties is about 55.5 mmol per mol of silica for the APTES20 sample and 89.7 mmol per mol of silica in APTES40.

Fluid-Dynamic Characterization
Adsorption tests on the amino-functionalized solids were carried out in a confined-fluidized bed, where the gas-solid contact was verified as being improved [13,14]. In a conventional column, i.e., without packing, the thermal effects of the adsorption process do not recommend a fixed bed configuration for the adsorbent; at the same time, the bubbling regime of the fluidized bed reduces the mass exchange efficiency. These limitations were overcome adopting a non-conventional fluidization technique, by which the fine adsorbent undergoes homogeneous expansion in the voids of the packing of coarse spheres once the state of incipient fluidization is reached. Following the guidelines indicated in a previous study [11], glass beads with a diameter of 11 mm and particle density of 2.48 g/cm 3 were selected as packing particles allowing an interstitial voidage εp of 0.44. In a typical experiment of confined fluidization, 1230 g of coarse spheres were poured onto the column. Thermogravimetric analysis (TGA) was performed using a DSC-TGA STA409C analyzer (Netzsch-Gerätebau GmbH, Selb, Germany) from 20 to 800 • C at a heating rate of 10 • C in static air. Figure 8 reports the TGA thermograms for the as-made silica, APTES20 and APTES40 samples. Mass loss from room temperature up to ca. 150 • C is attributed to the water physiosorbed on the particles' surfaces. From 120 to 800 • C, vicinal Si-OH groups react with production of H 2 O molecules and the formation of -Si-O-Si-groups. When APTES molecules impregnate the silica particles, an important mass loss is present between ca 300 and 600 • C due to degradation and/or combustion of the organic moieties adsorbed or chemically linked on the silica surface. In this case, water is also adsorbed on the organic sites. These data can be utilized for an evaluation of the amount of the APTES molecules. If we consider APTES molecules (MW 221.37 g/moles) as a reference, the amount of the organic moieties is about 55.5 mmol per mol of silica for the APTES20 sample and 89.7 mmol per mol of silica in APTES40. Thermogravimetric analysis (TGA) was performed using a DSC-TGA STA409C analyzer (Netzsch-Gerätebau GmbH, Selb, Germany) from 20 to 800 °C at a heating rate of 10 °C in static air. Figure 8 reports the TGA thermograms for the as-made silica, APTES20 and APTES40 samples. Mass loss from room temperature up to ca. 150 °C is attributed to the water physiosorbed on the particles' surfaces. From 120 to 800 °C, vicinal Si-OH groups react with production of H2O molecules and the formation of -Si-O-Si-groups. When APTES molecules impregnate the silica particles, an important mass loss is present between ca 300 and 600 °C due to degradation and/or combustion of the organic moieties adsorbed or chemically linked on the silica surface. In this case, water is also adsorbed on the organic sites. These data can be utilized for an evaluation of the amount of the APTES molecules. If we consider APTES molecules (MW 221.37 g/moles) as a reference, the amount of the organic moieties is about 55.5 mmol per mol of silica for the APTES20 sample and 89.7 mmol per mol of silica in APTES40.

Fluid-Dynamic Characterization
Adsorption tests on the amino-functionalized solids were carried out in a confined-fluidized bed, where the gas-solid contact was verified as being improved [13,14]. In a conventional column, i.e., without packing, the thermal effects of the adsorption process do not recommend a fixed bed configuration for the adsorbent; at the same time, the bubbling regime of the fluidized bed reduces the mass exchange efficiency. These limitations were overcome adopting a non-conventional fluidization technique, by which the fine adsorbent undergoes homogeneous expansion in the voids of the packing of coarse spheres once the state of incipient fluidization is reached. Following the guidelines indicated in a previous study [11], glass beads with a diameter of 11 mm and particle density of 2.48 g/cm 3 were selected as packing particles allowing an interstitial voidage εp of 0.44. In a typical experiment of confined fluidization, 1230 g of coarse spheres were poured onto the column.

Fluid-Dynamic Characterization
Adsorption tests on the amino-functionalized solids were carried out in a confined-fluidized bed, where the gas-solid contact was verified as being improved [13,14]. In a conventional column, i.e., without packing, the thermal effects of the adsorption process do not recommend a fixed bed configuration for the adsorbent; at the same time, the bubbling regime of the fluidized bed reduces the mass exchange efficiency. These limitations were overcome adopting a non-conventional fluidization technique, by which the fine adsorbent undergoes homogeneous expansion in the voids of the packing of coarse spheres once the state of incipient fluidization is reached. Following the guidelines indicated in a previous study [11], glass beads with a diameter of 11 mm and particle density of 2.48 g/cm 3 were selected as packing particles allowing an interstitial voidage ε p of 0.44. In a typical experiment of confined fluidization, 1230 g of coarse spheres were poured onto the column. Then, 50 g of the finer solid (the adsorbent) was loaded in the column and, after a complete fluidization-defluidization cycle, their heights were measured (for an aspect ratio H fc /D = 1.6-2.6). The bed voidage in the packed-fluidized bed, i.e., the fraction of volume available to the gas flow, is calculated as: where m p and ρ p , m f and ρ f , are the mass and density of the packing and of the fine sorbent, respectively. These tests were performed at ambient temperature and pressure. The pressure drop and bed expansion curves obtained by measuring pressure and bed height in experiments performed both at increasing and decreasing superficial gas velocity were found to coincide. Figures 9 and 10 show these diagrams for the three samples investigated-as-made silica and the two solids after functionalization. Then, 50 g of the finer solid (the adsorbent) was loaded in the column and, after a complete fluidization-defluidization cycle, their heights were measured (for an aspect ratio Hfc/D = 1.6-2.6). The bed voidage in the packed-fluidized bed, i.e., the fraction of volume available to the gas flow, is calculated as: where mp and ρp, mf and ρf, are the mass and density of the packing and of the fine sorbent, respectively. These tests were performed at ambient temperature and pressure. The pressure drop and bed expansion curves obtained by measuring pressure and bed height in experiments performed both at increasing and decreasing superficial gas velocity were found to coincide. Figures 9 and 10 show these diagrams for the three samples investigated-as-made silica and the two solids after functionalization. As previously shown, the silica gel functionalization changes particle density, a fact that results in a proportional increase in umfc values (see Figure 9 and data in Table 3).  On the contrary, the ability of the sorbent to expand in the confined bed configuration remains unchanged. Actually, the homogeneous fluidization parameters εfc, u0 and n, as determined from Then, 50 g of the finer solid (the adsorbent) was loaded in the column and, after a complete fluidization-defluidization cycle, their heights were measured (for an aspect ratio Hfc/D = 1.6-2.6). The bed voidage in the packed-fluidized bed, i.e., the fraction of volume available to the gas flow, is calculated as: where mp and ρp, mf and ρf, are the mass and density of the packing and of the fine sorbent, respectively. These tests were performed at ambient temperature and pressure. The pressure drop and bed expansion curves obtained by measuring pressure and bed height in experiments performed both at increasing and decreasing superficial gas velocity were found to coincide. Figures 9 and 10 show these diagrams for the three samples investigated-as-made silica and the two solids after functionalization. As previously shown, the silica gel functionalization changes particle density, a fact that results in a proportional increase in umfc values (see Figure 9 and data in Table 3).  On the contrary, the ability of the sorbent to expand in the confined bed configuration remains unchanged. Actually, the homogeneous fluidization parameters εfc, u0 and n, as determined from As previously shown, the silica gel functionalization changes particle density, a fact that results in a proportional increase in u mfc values (see Figure 9 and data in Table 3). On the contrary, the ability of the sorbent to expand in the confined bed configuration remains unchanged. Actually, the homogeneous fluidization parameters ε fc , u 0 and n, as determined from Figure 10, remain almost constant with a little decrease in the exponent n for the sample APTES40, denoting an increased difficulty of expansion through the interstitial voids of the packing (see Table 3).
The regular homogeneous expansion observed was described by a modified form of the Richardson and Zaki's equation [10], in which the exponents n and u 0 are calculated according to a method reported in a previous work [11].

Test on As-Made Silica
The sample of silica gel was tested as a blank potential adsorbent; a test in a conventional fixed bed was carried out with a mixture containing 12.5% of CO 2 and a gas velocity of 3.7 cm/s. A total of 120 g of silica gel was pulled onto the column and, since the minimum fluidization velocity (u mf ) is 4.5 cm/s at ε mf = 0.43, the system was in the fixed bed condition, with a constant height equal to 131 mm. The adsorbed mass of CO 2 up to the equilibrium was only 6.15 g/kg of silica, as determined by the breakthrough curve shown in Figure 11. This value is very low and is justified by weak interactions of molecules with a silica surface, characterized only by the presence of silanol groups (SiOH) [8,19,20]. For this reason, a test in the packed-fluidized bed configuration was not carried out, as it was neither useful nor interesting.
Processes 2020, 8, x FOR PEER REVIEW 9 of 16 Figure 10, remain almost constant with a little decrease in the exponent n for the sample APTES40, denoting an increased difficulty of expansion through the interstitial voids of the packing (see Table  3). The regular homogeneous expansion observed was described by a modified form of the Richardson and Zaki's equation [10], in which the exponents n and u0 are calculated according to a method reported in a previous work [11].

Test on As-Made Silica
The sample of silica gel was tested as a blank potential adsorbent; a test in a conventional fixed bed was carried out with a mixture containing 12.5% of CO2 and a gas velocity of 3.7 cm/s. A total of 120 g of silica gel was pulled onto the column and, since the minimum fluidization velocity (umf) is 4.5 cm/s at εmf = 0.43, the system was in the fixed bed condition, with a constant height equal to 131 mm. The adsorbed mass of CO2 up to the equilibrium was only 6.15 g/kg of silica, as determined by the breakthrough curve shown in Figure 11. This value is very low and is justified by weak interactions of molecules with a silica surface, characterized only by the presence of silanol groups (SiOH) [8,19,20]. For this reason, a test in the packed-fluidized bed configuration was not carried out, as it was neither useful nor interesting.

Test on Amino-Functionalized Silica
After functionalization and new dynamic characterization, the solids were used for an adsorption test operated in the confined-fluidized bed configuration. For each sample, two values of the fluidization velocity were chosen in order to analyze the effect of the flow rate on the level of CO2 capture: 5.7 and 8.6 cm/s for APTES20 and 6.9 and 10.9 cm/s for APTES40. Both values correspond to a condition expansion of the adsorbent over its incipient fluidization velocity threshold. It has to be borne in mind that all the gas fed to the fluidized bed is in contact with the adsorbent; in a packedfluidized bed the bypass effect in the gas-solid contact typical of the bubbling regime is avoided, so that the value of the gas velocity directly influences the contact time between gas and particles as well as the mass and heat transfer rates. Figures 12-15 show the breakthrough curves (the ratio between the concentration downstream and upstream of the adsorption column versus time) and the temperature trends for both amino-functionalized samples at various values of the superficial velocity of the gas mixture. In order to compare the performance of adsorbents with a different concentration of APTES, a mixture with 12.5% of CO2 was used in all the adsorption tests and a mass of sorbent of 100 g.

Test on Amino-Functionalized Silica
After functionalization and new dynamic characterization, the solids were used for an adsorption test operated in the confined-fluidized bed configuration. For each sample, two values of the fluidization velocity were chosen in order to analyze the effect of the flow rate on the level of CO 2 capture: 5.7 and 8.6 cm/s for APTES20 and 6.9 and 10.9 cm/s for APTES40. Both values correspond to a condition expansion of the adsorbent over its incipient fluidization velocity threshold. It has to be borne in mind that all the gas fed to the fluidized bed is in contact with the adsorbent; in a packed-fluidized bed the bypass effect in the gas-solid contact typical of the bubbling regime is avoided, so that the value of the gas velocity directly influences the contact time between gas and particles as well as the mass and heat transfer rates. Figures 12-15 show the breakthrough curves (the ratio between the concentration downstream and upstream of the adsorption column versus time) and the temperature trends for both amino-functionalized samples at various values of the superficial velocity of the gas mixture. In order to compare the performance of adsorbents with a different concentration of APTES, a mixture with 12.5% of CO 2 was used in all the adsorption tests and a mass of sorbent of 100 g.
To determine the flow rates of both gases used in the experiments, the CO2 fraction, xCO2, and the gas velocity of the mixture, u, have to be set as an input for the following relationships: From each breakthrough curve the main parameters were determined: the breakthrough time, the saturation time, the fraction of bed utilized at the breakpoint and the amount of CO2 adsorbed.
The breakthrough time tb, or breakpoint, is the time at which CO2 at the outlet of the adsorption column reaches 5% of the inlet concentration. Usually, in a continuous process, it represents the time at which the gas feed must be switched to a stand-by column while the saturated sorbent of the first column is regenerated.
The saturation time, ts, is the time at which CO2 at the outlet of the adsorption column reaches 95% of the inlet concentration.
The fraction of bed utilized at the breakpoint, W, represents the ratio between the amount of CO2 adsorbed up to the breakpoint and that adsorbed at saturation. This parameter is strictly related to the performance of the sorbent; for a given contact time, the higher W the lower the amount of sorbent required. Conversely, for a fixed mass of solid in the column, a higher value of W indicates that sorbent saturation occurs after a longer time.
Finally, the mass of CO2 adsorbed per unit mass of adsorbent, mCO2 (reported as mg of CO2 adsorbed per gram of sorbent) is calculated by integration of the breakthrough curve. This is the main parameter to analyze when different adsorbents are compared.
The maximum temperature Tmax reached during the adsorption test (measured by the thermocouple immersed in the bed) allowed us to check also the thermal effects associated with the process. All the value of these parameters are shown in Table 4. The characteristic contact time for this continuous process is defined as where and the inlet flow rate of the gas mixture fed to the column is To determine the flow rates of both gases used in the experiments, the CO 2 fraction, x CO2 , and the gas velocity of the mixture, u, have to be set as an input for the following relationships: From each breakthrough curve the main parameters were determined: the breakthrough time, the saturation time, the fraction of bed utilized at the breakpoint and the amount of CO 2 adsorbed.
The breakthrough time t b , or breakpoint, is the time at which CO 2 at the outlet of the adsorption column reaches 5% of the inlet concentration. Usually, in a continuous process, it represents the time at which the gas feed must be switched to a stand-by column while the saturated sorbent of the first column is regenerated.
The saturation time, t s , is the time at which CO 2 at the outlet of the adsorption column reaches 95% of the inlet concentration.
The fraction of bed utilized at the breakpoint, W, represents the ratio between the amount of CO 2 adsorbed up to the breakpoint and that adsorbed at saturation. This parameter is strictly related to the performance of the sorbent; for a given contact time, the higher W the lower the amount of sorbent required. Conversely, for a fixed mass of solid in the column, a higher value of W indicates that sorbent saturation occurs after a longer time.
Finally, the mass of CO 2 adsorbed per unit mass of adsorbent, m CO2 (reported as mg of CO 2 adsorbed per gram of sorbent) is calculated by integration of the breakthrough curve. This is the main parameter to analyze when different adsorbents are compared.
The maximum temperature T max reached during the adsorption test (measured by the thermocouple immersed in the bed) allowed us to check also the thermal effects associated with the process. All the value of these parameters are shown in Table 4. Even if the gas velocities in the expansion regime are different, their values are comparable as they represent conditions in which the distance from the onset of fluidization is constant; in the first case the value of u is c.a. 25% higher than u mfc of the solid, thus corresponding to a low level of expansion of the bed; in the second case, with a value of u c.a. 93% higher than u mfc , the bed of sorbent reached an extended homogeneous expansion. This circumstance is confirmed by contact time values, reported in the Table 4, which are very close at similar expansion conditions.

Thermal Stability
Usually, CO 2 adsorption involves a reaction between the acidic CO 2 molecules and the modified basic active sites and for this reason amine-based adsorbents are extensively studied with regard to enhancing CO 2 adsorption capacity and selectivity [7]. Amine impregnation is advantageous because it increases tolerance to water vapor in the gas feed of mesoporous silicas. On the other hand, several disadvantages are also present including high regeneration cost, equipment corrosion and oxidative degradation of some amines [21]. To investigate this last aspect, special attention has been devoted to the thermal stability of sorbents investigated here. Figure 16 reports the TG thermograms of APTES20 (a) and APTES40 (b), respectively, after different cycles of regeneration. In the APTES20 sample, the initial amount of organic was ca. 16%, with a decrease to 14% after the first cycle of regeneration, a value that remains constant during all the other cycles of CO 2 adsorption. On the other hand, the APTES40 sample (24.56% total weight loss before the adsorption tests) shows a weight loss of ca. 24% after the first cycle of adsorption; such a value decreases to 20% and then remains constant after the ninth cycle. Even if the gas velocities in the expansion regime are different, their values are comparable as they represent conditions in which the distance from the onset of fluidization is constant; in the first case the value of u is c.a. 25% higher than umfc of the solid, thus corresponding to a low level of expansion of the bed; in the second case, with a value of u c.a. 93% higher than umfc, the bed of sorbent reached an extended homogeneous expansion. This circumstance is confirmed by contact time values, reported in the Table 4, which are very close at similar expansion conditions.

Thermal Stability
Usually, CO2 adsorption involves a reaction between the acidic CO2 molecules and the modified basic active sites and for this reason amine-based adsorbents are extensively studied with regard to enhancing CO2 adsorption capacity and selectivity [7]. Amine impregnation is advantageous because it increases tolerance to water vapor in the gas feed of mesoporous silicas. On the other hand, several disadvantages are also present including high regeneration cost, equipment corrosion and oxidative degradation of some amines [21]. To investigate this last aspect, special attention has been devoted to the thermal stability of sorbents investigated here. Figure 16 reports the TG thermograms of APTES20 (a) and APTES40 (b), respectively, after different cycles of regeneration. In the APTES20 sample, the initial amount of organic was ca. 16%, with a decrease to 14% after the first cycle of regeneration, a value that remains constant during all the other cycles of CO2 adsorption. On the other hand, the APTES40 sample (24.56% total weight loss before the adsorption tests) shows a weight loss of ca. 24% after the first cycle of adsorption; such a value decreases to 20% and then remains constant after the ninth cycle. The nitrogen adsorption/desorption isotherms of APTES20 and APTES40 samples after nine cycles of regeneration, reported in Figure 17, support the results presented above. First of all, the shapes of the two isotherms are very similar to those obtained from the as-prepared samples (see The nitrogen adsorption/desorption isotherms of APTES20 and APTES40 samples after nine cycles of regeneration, reported in Figure 17, support the results presented above. First of all, the shapes of the two isotherms are very similar to those obtained from the as-prepared samples (see Figure 5). On the other hand, S BET and pore volume are slightly higher with respect to the values reported in Table 2. Specific surfaces are 226 and 98 m 2 /g for APTES20 and APTES40, respectively. In the same way the pore volume increases to 0.30 and 0.14 g/cm 3 . The combination of all these results allows us to affirm that during the exhaustion/regeneration cycles only the organic molecules weakly linked to the silica surface are removed. It is likely that the amount of easily desorbed molecules is lower in APTES20.
Processes 2020, 8, x FOR PEER REVIEW 13 of 16 Figure 5). On the other hand, SBET and pore volume are slightly higher with respect to the values reported in Table 2. Specific surfaces are 226 and 98 m 2 /g for APTES20 and APTES40, respectively. In the same way the pore volume increases to 0.30 and 0.14 g/cm 3 . The combination of all these results allows us to affirm that during the exhaustion/regeneration cycles only the organic molecules weakly linked to the silica surface are removed. It is likely that the amount of easily desorbed molecules is lower in APTES20. These results are associated with high thermal stability and are in good agreement with those obtained by Quang et al. [19], indicating that the amino groups are linked to the silica surface by chemical bonds.

Discussion and Conclusions
From the analysis of Figures 12a-15a and of data of Table 4, it can be noticed that the breakthrough curve for the sample APTES20 has a higher breakpoint time, indicating that this sorbent can capture CO2 for a longer time during an industrial process; moreover, the saturation time is lower than that of the sample APTES40, suggesting that faster kinetics of adsorption are related to a larger availability of active surface.
The amount of CO2 adsorbed, calculated by integration of the breakthrough curve, is slightly higher for APTES40. Nevertheless, this sample exhibits remarkably lower values of W, indicating a slow saturation kinetic and consequently an insufficient use of the sorbent up to the breakpoint. This behaviour is incompatible with industrial applications when switching to regeneration of the sorbent at the breakpoint is required. On the whole, the values of CO2 adsorbed measured, over 30 mg/g, are better than the result obtained by Leal et al. [22], equal to 26.4 mg/g at 23 °C.
Regarding the effect of the operating gas velocity, its increase reduces the contact time but does not modify the process performance in terms of the mass of CO2 adsorbed. In a previous work [14], it has been demonstrated that the increase in gas velocity can improve the control of the thermal effects of the adsorption process. As Figures 12b-15b show, for the amino-functionalized silica used in this experimental investigation the thermal effects are moderate, resulting in a variation of 10 °C for APTES20 and of only 5 °C for APTES40. Though small in all cases, the increase in the adsorption temperature depends on the operating gas velocity of the process (see Tmax in Table 4); this is consistent with previous findings [14] indicating that lower temperatures are measured at higher gas velocities.
The experimental results of Quang et al. [19] have shown that the nitrogen content in the adsorbent increases with the increase in APTES concentration in the solid, but also that a maximum These results are associated with high thermal stability and are in good agreement with those obtained by Quang et al. [19], indicating that the amino groups are linked to the silica surface by chemical bonds.

Discussion and Conclusions
From the analysis of Figures 12a, 13a, 14a and 15a and of data of Table 4, it can be noticed that the breakthrough curve for the sample APTES20 has a higher breakpoint time, indicating that this sorbent can capture CO 2 for a longer time during an industrial process; moreover, the saturation time is lower than that of the sample APTES40, suggesting that faster kinetics of adsorption are related to a larger availability of active surface.
The amount of CO 2 adsorbed, calculated by integration of the breakthrough curve, is slightly higher for APTES40. Nevertheless, this sample exhibits remarkably lower values of W, indicating a slow saturation kinetic and consequently an insufficient use of the sorbent up to the breakpoint. This behaviour is incompatible with industrial applications when switching to regeneration of the sorbent at the breakpoint is required. On the whole, the values of CO 2 adsorbed measured, over 30 mg/g, are better than the result obtained by Leal et al. [22], equal to 26.4 mg/g at 23 • C.
Regarding the effect of the operating gas velocity, its increase reduces the contact time but does not modify the process performance in terms of the mass of CO 2 adsorbed. In a previous work [14], it has been demonstrated that the increase in gas velocity can improve the control of the thermal effects of the adsorption process. As Figures 12b, 13b, 14b and 15b show, for the amino-functionalized silica used in this experimental investigation the thermal effects are moderate, resulting in a variation of 10 • C for APTES20 and of only 5 • C for APTES40. Though small in all cases, the increase in the adsorption temperature depends on the operating gas velocity of the process (see T max in Table 4); this is consistent with previous findings [14] indicating that lower temperatures are measured at higher gas velocities.
The experimental results of Quang et al. [19] have shown that the nitrogen content in the adsorbent increases with the increase in APTES concentration in the solid, but also that a maximum CO 2 load exists. At a high concentration, APTES soaked in the pores produces a pore blockage so that CO 2 gas cannot access hidden amine sites, a circumstance that causes a reduction in CO 2 adsorption. For this reason, in our study lower APTES concentrations have been explored, as the new gas-solid contact configuration provided by the packed-fluidized bed mass transfer is improved thanks to the extension of the expansion regime. In fact, APTES20 offered the best performance for a potential industrial application of CCS: high CO 2 capture ability, fast kinetics of adsorption, good efficiency of use of the sorbent in a continuous process, small thermal effects at long contact times and high stability for long-term operation with the optimum adsorption and regeneration temperature of 120 • C. Finally, APTES20 sample preparation consumes a lower amount of an expensive organic reactant.
In conclusion, as-made silica did not show any remarkable CO 2 adsorption capacity, consistently with what was expected. In any case, the material confirmed the role of heterogenous support able to allow an efficient dispersion of the adsorption sites. On the contrary, APTES-functionalized samples showed a significantly higher CO 2 adsorption capacity. Such a capacity is of the same order of magnitude for the two materials, although the process kinetics are different. When all the aspects of the process are considered, it can be concluded that APTES20 offers a better performance in a continuous process.