Design rationale of thermally responsive microgel particle films that reversibly absorb large amounts of CO2: fine tuning the pKa of ammonium ions in the particles† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c5sc01978h

Fine-tuning of pKa value of ammonium ions at both CO2 capture and release temperature is found to be crucial for the design of the thermally responsive gel particle films that reversibly capture large amounts of CO2.


Introduction
The accumulation of CO 2 in the atmosphere due to burning of fossil fuels is considered the main cause of global climate change. 1 Meanwhile, CO 2 is expected to be a carbon source for the production of liquid fuels, such as methanol and dimethyl ether, through the use of regenerative energy. 2 Thus, the development of energy-efficient CO 2 separation and recovery processes for point sources, such as fossil fuel power plants, is essential not only for minimizing climate change but also for the future use of carbon-based energy.
Conventional processes that recover CO 2 from the high humidity exhaust gases of power plants use aqueous solutions of ethanolamine as a CO 2 absorbent. CO 2 in the exhaust gases is selectively captured by the absorbent at a low temperature ($40 C) and then recovered by heating the solution, typically above 140 C. [3][4][5] Although the amine solution exhibits a high CO 2 absorption capacity, the high-energy consumption of the heating process limits its use in the environmentally friendly process. [4][5][6][7][8] Porous solid adsorbents that can be regenerated under relatively mild conditions, such as metal-organic frameworks (MOFs), 9,10 zeolites, 9,10 and silica with physically or chemically supported amine polymers, [9][10][11][12][13] have recently been developed as alternatives to the aqueous amine solution. However, solid CO 2 sorbents are rarely capable of efficiently capturing CO 2 from highly humid gases because the adsorption of water molecules competes with that of CO 2 on the pore surfaces. Furthermore, the capillaries and pores of the sorbents are easily blocked by the liqueed water, preventing CO 2 from diffusing into the pores. [9][10][11][12] Thus, the development of sorbents that absorb/adsorb large amount of CO 2 , even in humid environments, and desorb CO 2 at low temperatures (<100 C) is of great importance.
It has been reported that poly-N-isopropylacrylamide (pNIPAm) undergoes a reversible volume phase transition from a hydrophilic swollen state to a hydrophobic collapsed state at temperatures below and above, respectively, the volume phase transition temperature (VPTT) of $32 C. 14 The phase transition is induced by the entropy-driven dissociation of water from pNIPAm chains aer heating above the VPTT. 14 The temperature responsive pNIPAm-based functional hydrogels have been widely used as materials to reversibly capture targets, such as dyes, 15,16 drugs, 17,18 peptides, 19 proteins, 20,21 nucleotides, 22 cells, 23 and protons, 24 in aqueous media via the volume phase transition. The multipoint interactions are reversibly switched on/off by varying the volume density of hydrophobic functional groups, 17,25,26 the number of charged functional groups, 22,27 and/or the rigidity of the polymer chains 28 at temperatures above and below the VPTT.
Recently, we reported that hydrogel lms composed of temperature-responsive microgel particles (GPs) consisting of NIPAm and N- [3-(dimethylamino)propyl]methacrylamide (DMAPM) reversibly absorbed and released CO 2 via a volume phase transition during cooling (30 C) and heating (75 C) cycles, respectively. 29 Below the VPTT, amines in the swollen GPs were capable of forming ion pairs with absorbed bicarbonate ions. Above the VPTT, shrinkage of the GPs triggered CO 2 desorption. 27 The GP lms showed faster CO 2 capture and release rate 29 than the conventional bulk hydrogel lms due to the fast thermal responsibility of the GPs. 30 However, guidelines to design GP lms that reversibly absorb CO 2 with high capacity and stoichiometry have not been revealed.
In the case of reversible CO 2 absorption by the blood of animals, the pK a value of the Brønsted acid and base in the hemoglobin plays a crucial role for the control of CO 2 solution equilibrium: The pK a value of the ammonium and imidazolium cations decreases due to the drastic conformational change of the hemoglobin caused by oxygen binding to the heme. The shi of the pK a triggers the release of H + into the blood, leading to the release of CO 2 efficiently from lungs (Bohr effect). 31,32 Inspired by the Bohr effect of hemoglobin, we hypothesized that the pK a value of the protonated amines (ammonium ions) within GPs (pK a value of GPs) at the CO 2 capture temperature (30 C) and release temperature (75 C) would contribute to the reversible capture efficiency of CO 2 .
In this study, in order to clarify the design rationale, we prepared a series of GPs with a variety of compositions using different polymerization conditions. The effects of the physical and chemical properties of GPs, such as VPTT, size, swelling ratio, on pK a values and the reversible CO 2 capture stoichiometry against amines were systematically investigated. pK a values of the ammonium ions in the GPs were also tuned by the "microenvironment-imprinting" strategy as we described in the recent report. 24 The reversible CO 2 capture capacity was maximized based on the design rational revealed in this study. Humidied gas (60 C) consisting of 10% CO 2 and 90% N 2 , which is comparable to the post-combustion gas of re power plants aer wet desulfurization process, was used as feed gas. 6,33 The CO 2 was captured at 30 C and released at 75 C under the same atmosphere (10% CO 2 , 90% N 2 , 60 C water moisture).

Preparation of GPs
A series of GPs containing NIPAm, a functional tertiary amine monomer DMPAM, and a crosslinker N,N 0 -methylenebisacrylamide (Bis) were synthesized by precipitation polymerization as reported (Scheme 1). 29 The amount of Bis was varied to prepare GPs with different degrees of crosslinking and swelling ratios. GPs with the same composition but different pK a values were synthesized via a "microenvironment-imprinting" strategy by adding HCl or NaOH into the monomer solution. 24 Larger GPs were obtained by decreasing the concentration of surfactant, while smaller GPs were prepared from solutions with a lower total monomer concentration. To decrease the VPTT of GPs, a more hydrophobic monomer, N-tert-butyl acrylamide (TBAm), was incorporated into the GPs. Details of the polymerization conditions of the GPs are summarized in Table 1. Polymerization process is described in ESI. †

Quantication of hydrodynamic diameters, VPTTs, and swelling ratios of GPs
The hydrodynamic diameters and VPTT of GPs were measured by dynamic light scattering (DLS) as described. 27,29 The swelling ratio is dened as the ratio of the diameters at 30 C (D 30 ) and 75 C (D 75 ).

Quantication of amine content and pK a of ammonium ions in GPs
The amine content and pK a of amines in GPs was quantied by acid-base titration under N 2 purging as described. 27,29 adding 0.5 eq. of HCl to the GP solution. Then, under a N 2 atmosphere with stirring the pH value and temperature of the solution were recorded by the pH meter every 3 s during the heating process.

Preparation of GP lms
The lyophilized GPs were dissolved in methanol. The hydrogel lms were then prepared by casting the methanol solutions containing 120 mg of GPs on the inner bottom surface of a stainless steel container with a surface area of 120 cm 2 . Aer completely evaporating the methanol, 4 mL of water was added per gram of GPs.

Measurement of CO 2 absorption capacity of GP lms
The CO 2 absorption capacities of the GP lms are quantied as illustrated in Scheme S1. † 10% CO 2 (90% N 2 , 10 mL min À1 ) gas was passed through 60 C water to generate saturated water vapor. The resulting 60 C water-saturated gas mixture was channeled into a stainless steel container with the GP lm on the inner surface, and then to a gas chromatograph aer condensing the water moisture at 5 C.
More detailed information about materials and experimental process are shown in ESI. †

Results and discussion
Effect of pK a values of GPs on the reversible CO 2 capture stoichiometry of GP lms The dynamic and reversible transition of the pK a values of Brønsted acid and base groups in proteins plays an important role in their functions, such as molecule/ion transfer, enzymatic reactions, and molecular recognition. 32,34 The pK a values are signicantly inuenced by the microenvironment around the functional groups, such as hydrophobicity, hydrogen bonding, and distance to the neighboring charges. 32,[34][35][36][37][38] Thus, the pK a value can be further dynamically shied by the conformational change of polypeptides around the functional groups, which induces a microenvironment change.
In the case of CO 2 transfer through the bloodstream of animals, the pK a value of the ammonium and imidazolium cations in/on the hemoglobin decreases due to the drastic conformational change of the polypeptides caused by oxygen binding to the heme. The shi of the pK a value triggers the release of H + into the blood, lowers the pH of the blood, and drives the CO 2 efficiently from the lungs (Bohr effect). 32,38 Inspired by the function of hemoglobin, we expected that the pK a value of the protonated amines (ammonium ions) within GPs (pK a value of GPs) at the CO 2 capture temperature (30 C) and release temperature (75 C) would contribute to the reversible capture efficiency of CO 2 .
To clarify the effect of the pK a value of GPs on the reversible CO 2 capture stoichiometry of GP lms in the temperature swing process from 30 C to 75 C, a series of GPs with the same composition but different pK a values were prepared by the "microenvironment-imprinting" strategy. 24,39 Imprinting is a method to create polymers with specic microstructures that show strong affinity to target molecules and ions by crosslinking the polymer networks in the presence of targets. 40,41 It has been reported that pNIPAm-based acrylic acids (AAc) containing GPs with different pK a values can be prepared by the "proton-imprinted" strategy. 24 The GPs polymerized at a pH below the pK a of AAc showed a much higher pK a value at the collapsed state than those prepared at a high pH because stronger proton-affinity sites were incorporated into the relatively hydrophobic microenvironment around the protonated AAc. Note that the GPs were polymerized at the collapsed state in the temperature above the VPTT.
We anticipated that amine-containing GPs with different pK a values could also be prepared via the "microenvironment- imprinting" strategy. Thus, the GPs were polymerized in the presence of acid (1 or 1/2 eq. of HCl against the amine monomer DMAPM) or base (1/2 eq. of NaOH). All GPs were polymerized in the aqueous monomer solution consisting of 5 mol% DMAPM, 2 mol% Bis and 93 mol% NIPAm at the temperature above the VPTT of the growing GPs (70 C). The VPTTs of the GPs were determined by monitoring the scattering intensity of the GP solutions during the heating process ( Fig. S1a in ESI †). Despite the different polymerization conditions, all four GPs, D5B2-1/1HCl (1 eq. HCl), D5B2-1/2HCl (1/2 eq. HCl), D5B2-1/2NaOH, (1 eq. NaOH), and D5B2 (without HCl or NaOH), show similar VPTTs at about 38 C. Similar swelling ratios in the range of 2.3-2.7 are also observed ( Fig. S1b in ESI †), indicating the VPTTs and swelling ratios of GPs are independent of the polymerization conditions.
The apparent pK a values of the GPs at 30 C and 75 C were determined by acid-base titration of the GP solutions and are plotted as the top and bottom, respectively, of the gray bars in Fig. 1a. 27 All GPs show lower pK a values than the amine monomer DMAPM, at both temperatures, indicating the relatively low dielectric constant and high steric hindrance around the amine groups in the GPs, as well as the high electrostatic repulsion between neighboring charges. 42,43 Moreover, in accordance with our expectation, the pK a values of D5B2-1/1HCl and D5B2-1/2HCl at 75 C (7.4 and 7.0, respectively) were signicantly higher than those of D5B2 and D5B2-1/2NaOH (5.3 and 4.9, respectively).
The large difference in pK a values can be explained by the "microenvironment-imprinting" effect. When the GPs are polymerized in a proton-rich solution (in the presence of HCl), the protonated DMAPM is imprinted in the polar and hydrophilic microenvironment of the GPs (D5B2-1/1HCl and D5B2-1/ 2HCl), resulting in higher pK a values than those of the GPs polymerized in the proton-poor solution (D5B2 and D5B2-1/ 2NaOH). Although there is variation in the GP diameters ( Fig. S1b in ESI †), the large difference in the pK a values (>2) at 75 C cannot be a result of the size differences because the smallest GP, D5B2 (74 nm), and the largest GP, D5B2-1/2NaOH (136 nm), show a pK a difference of only 0.4 (5.3-4.9). Fig. 1a also shows that the pK a difference between 30 C and 75 C (DpK a ) of all GPs (1.5-2.5) is larger than that of the monomer (0.9). For better understanding of the temperature dependent pK a difference, the apparent pK a values of D5B2 and the monomer DMAPM, were recorded during the heating process. Fig. 1b shows that a sharp pK a transition of D5B2 occurs at a temperature around its VPTT, and a steep pK a decrease continues over a wide temperature range above the VPTT. However, this sharp pK a transition is not observed in the case of the monomer, indicating that the sharp pK a decrease was induced by a volume phase transition of GPs, during which the increased hydrophobicity and steric hindrance, and the decreased dielectric constant at high temperature make the amine groups more difficult to be protonated. 42 In Fig. 1a, the DpK a values between 75 C and 30 C of the GPs polymerized with HCl, D5B2-1/1HCl (1.5) and D5B2-1/2HCl (1.6), are signicantly lower than those of D5B2 (2.5) and D5B2-1/2NaOH (2.3), although the swelling ratios are similar (2.3-2.7). We hypothesize that in comparison with D5B2 and D5B2-1/2NaOH, which were polymerized with 0 and 0.5 eq. of NaOH, respectively, the protonated DMAPM groups (ammoniums) within D5B2-1/1HCl and D5B2-1/2HCl are relatively distant from the hydrophobic pNIPAm moieties and backbones in the collapsed GPs. This is because the protonated amine groups are relatively polar and hydrophilic. In other words, the amine groups of the GPs prepared with HCl are located in a "pseudo-swollen" microenvironment analogous to the swollen state even at the collapsed state. Aer the volume phase transition from the collapsed to the swollen state, swelling has less effect on the amines within the "pseudo-swollen" GPs. As a result, the DpK a values of the GPs prepared with HCl are less than those of the GPs prepared without HCl.
The reversible CO 2 capture stoichiometries of the GP lms are presented as red plots in Fig. 1a. D5B2-1/1HCl and D5B2-1/2HCl show very low reversible CO 2 capture stoichiometries (0.18 and 0.27 mol CO 2 per mol amine, respectively), while the values for D5B2 and D5B2-1/2NaOH are both $0.97 mol CO 2 per mol amine. The difference in the stoichiometry can be interpreted by the acid-base theory as follows. It has been reported that the tertiary amines in GPs and CO 2 form ammonium ions (RN 3 H + ) and bicarbonate ions (HCO 3 À ) via base-catalyzed reactions. 44,45 In aqueous media containing RN 3 H + and HCO 3 À , there is always an equilibrium between their corresponding conjugated base-acid, as described in Scheme 2.
When the pK a of NR 3 H + is above the pK a1 of H 2 CO 3 , which indicates that NR 3 H + holds a proton more tightly than H 2 CO 3 does, the proton will move from the stronger acid, H 2 CO 3 , to the stronger base, NR 3 . As a result, more R 3 N will be consumed and more CO 2 will be captured, as presented by the blue arrows in Scheme 2. In contrast, when the pK a of NR 3 H + is below the pK a1 of H 2 CO 3 , more CO 2 will be released, as indicated by the red arrows in Scheme 2. 46 The pK a1 value of carbonic acid (H 2 CO 3 ) is in the range of 6.35 AE 0.05 at both 30 C and 75 C, as determined by Harned and Davis, 47 however, the pK a values of the amine-containing thermal responsive GPs dramatically depend on the temperature, as presented in Fig. 1. The high pK a values of D5B2-1/1HCl and D5B2-1/2HCl at 75 C (7.4 and 7.0, respectively) are higher than the pK a1 of H 2 CO 3 (6.35) and inhibit the release of CO 2 . As a result, the reversible CO 2 capture stoichiometries of D5B2-1/ 1HCl and D5B2-1/2HCl are low (0.18 and 0.27 mol CO 2 per mol amine, respectively), although the pK a values at 30 C (8.9 and 8.6, respectively) are high enough to capture CO 2 efficiently. For D5B2 and D5B2-1/2NaOH, the high reversible CO 2 capture stoichiometries (0.97 mol CO 2 per mol amine for both) are achieved due to the low pK a values at 75 C (5.3 and 4.9, respectively), which are below the pK a1 of H 2 CO 3 , and meanwhile, the high pK a values at 30 C (7.8 and 7.2, respectively), which are above the pK a1 of H 2 CO 3 .
Effect of DpK a values of GPs on the reversible CO 2 capture stoichiometry of GP lms Despite the low pK a values of D5B2 and D5B2-1/2NaOH at 75 C (Fig. 1a), their larger DpK a values between 30 C and 75 C (2.5 and 2.3, respectively) than those of D5B2-1/1HCl and D5B2-1/2HCl (1.5 and 1.6, respectively) may also be responsible for their high reversible CO 2 capture stoichiometries.
In order to distinguish the effects of the pK a value and the DpK a on the reversible CO 2 capture stoichiometry of the GP lms, a GP with a smaller DpK a than D5B2 was prepared by increasing the Bis cross-linker to 10 mol% into the GP that contains 5 mol% DMAPM (D5B10), since it has been reported that the crosslink degree inuences the DpK a of the GPs. 24 The pK a values of D5B2 and D5B10 at 30 C and 75 C are plotted as the top and bottom of the gray bars in Fig. 2. The DpK a of D5B10 (1.4) is apparently less than that of D5B2 (2.5). However the pK a value of D5B10 (5.9) at 30 C is below the pK a1 of H 2 CO 3 . Thus the possible reason for the lower reversible CO 2 capture stoichiometry of D5B10 than D5B2 (0.74 and 0.97 mol CO 2 per mol amine, respectively), the smaller DpK a of D5B10 than D5B2 and/or the low pK a value of D5B10 at 30 C which is below the pK a1 of H 2 CO 3 , still cannot be distinguished.
To identify the main factor that governs the reversible CO 2 capture stoichiometry, we designed another GP with a greater pK a value than D5B10 and a lower DpK a than D5B2, using the "microenvironment-imprinting" strategy described above. The GPs were prepared by polymerizing D5B10 in the presence of an appropriate amount of HCl (D5B10-1/4HCl). Fig. 2 shows the pK a values and DpK a of D5B10-1/4HCl. The pK a values at 30 C and 75 C of D5B10-1/4HCl (6.8 and 5.3, respectively) are higher than those of D5B10 (5.9 and 4.5, respectively), but the DpK a of D5B10-1/4HCl (1.5) is comparable to that of D5B10 (1.4). D5B10-1/4HCl shows a much higher CO 2 capture stoichiometry (0.97 mol CO 2 per mol amine) than D5B10 (0.74 mol CO 2 per mol amine), despite the similar DpK a values.
On the other hand, as displayed in Fig. 2, the pK a value of D5B10-1/4HCl at 75 C (5.3) is similar to that of D5B2 (5.3), while the DpK a of D5B10-1/4HCl (1.5) is much less than that of D5B2 (2.5). However, both GPs exhibit comparably high reversible CO 2 capture stoichiometries (0.97 mol CO 2 per mol amine), even though the DpK a values are much different.
We conclude from these results that the DpK a has a minor inuence on the reversible CO 2 capture stoichiometry, and a high reversible CO 2 capture stoichiometry is possible for GPs that exhibit reduced DpK a values as long as the pK a values lie within the appropriate range. The pK a value of the GPs must be tuned above the pK a1 of H 2 CO 3 (6.35) at the CO 2 capture temperature (30 C) and below the pK a1 of H 2 CO 3 (6.35) at the CO 2 release temperature (75 C) in order to achieve a high reversible CO 2 capture stoichiometry.
Effect of VPTT of GPs on the reversible CO 2 capture stoichiometry of GP lms As shown in Fig. 1b, the thermal responsive volume phase transition of the GPs leads to a sharp pK a transition. In Scheme 2 Equilibrium of the tertiary amine-CO 2 reaction. addition, the pK a value of the GPs has been revealed essential for the reversible CO 2 capture stoichiometry. Therefore, it is interesting to observe the effect of the VPTT of GPs on the reversible CO 2 capture stoichiometry of the GP lms.
A GP with a lower VPTT than D5B2 was synthesized by polymerizing 40 mol% of TBAm, which is more hydrophobic than NIPAm, together with 5 mol% DMAPM and 2 mol% Bis (D5B2T40). The relative scattering intensity of the GP solutions is shown in Fig. 3a. Compared with D5B2 (VPTT ¼ 38 C), the VPTT of D5B2T40 is only 12 C. The higher intra-particle hydrophobic interaction of the GPs containing 40 mol% TBAm causes the entropy-driven collapse of D5B2T40 to occur at a lower temperature. 48,49 Fig. 3b shows the temperature dependent pK a values of D5B2 and D5B2T40. Similar to D5B2, D5B2T40 shows a sharp pK a transition at a temperature around its VPTT (12 C) and a steep pK a decrease continues over a wide temperature range above the VPTT. As a result, the pK a values of D5B2T40 at 30 C and 75 C (6.0 and 4.8, respectively) are both low. The lower reversible CO 2 capture stoichiometry of D5B2T40 (0.60 mol CO 2 per mol amine) than that of D5B2 (0.97 mol CO 2 per mol amine), can be explained by the low pK a value of D5B2T40 at 30 C (6.0), which is below the pK a1 of H 2 CO 3 .
Besides D5B2 and D5B2T40, another pair of GPs with different VPTTs were prepared by increasing the feed ratio of DMAPM to 30 mol% (D30B2 and D30B2T40). 48,49 Fig. 3c shows that the VPTT of D30B2 is above 60 C. To lower the VPTT, 40 mol% of TBAm was incorporated into the 30 mol% DMAPM-containing GP (D30B2T40). In comparison, the VPTT of D30B2T40 is only 40 C (Fig. 3c). In Fig. 3d, the pK a value of D30B2 is linearly dependent on temperature, and no obvious sharp transition is observed in the temperature range due to the high VPTT. However, the pK a value of D30B2T40 exhibits a transition from the temperature around its VPTT (40 C), and a steeper pK a decrease than that of D30B2 is observed at the temperatures above the VPTT. As a result, D30B2T40 has lower pK a values at higher temperatures (>40 C) and a larger DpK a between 30 C and 75 C than D30B2 (Fig. 3e).
The degree of pK a transition of D30B2T40 (Fig. 3d) induced by the phase transition is less signicant than D5B2 and D5B2T40 (Fig. 3b). This may be a result of the increased intra-GP charge repulsion within D30B2T40 caused by the smaller amine-amine distance.
The pK a values at 30 C and 75 C of D30B2 (7.8 and 6.8, respectively) and D30B2T40 (7.7 and 6.3, respectively) are shown in Fig. 3e. The reversible CO 2 capture stoichiometry (0.72 mol CO 2 per mol amine) of D30B2 is quite low; however, that of D30B2T40 is much higher at 0.97 mol CO 2 per mol amine. The improved stoichiometric efficiency can be explained by the relatively low pK a value of D30B2T40 at 75 C (6.3), which is below the pK a1 of H 2 CO 3 . The VPTT of D30B2T40, which is close to 30 C, induces a sharp pK a transition and a steep pK a decrease over a wide temperature range above the VPTT, lowers the pK a value at 75 C, and consequently improves the reversible CO 2 capture stoichiometry.
In summary, the results in Fig. 3 lead to a comparable conclusion to those of Fig. 1 and 2: the pK a values of the GPs at 30 C and 75 C govern the reversible CO 2 capture stoichiometry. However, the VPTT of the GPs also plays a crucial role. The GPs with a much lower VPTT than the CO 2 capture temperature (30 C) show low reversible CO 2 capture stoichiometries due to the low pK a value. Meanwhile, to achieve a high reversible CO 2 capture stoichiometry, the VPTT should be above and as close to 30 C as possible to generate a large pK a transition over a wide temperature range above the VPTT.
However, a VPTT above and close to 30 C is not enough for a high CO 2 capture stoichiometry if the pK a values at 30 C and 75 C lie at an improper level. For example, in Fig. 1, though the VPTTs of D5B2-1/1HCl and D5B2-1/2HCl are both 38 C, the high pK a values at 75 C (7.4 and 7.0, respectively) considerably restrict efficient CO 2 release (0.18 and 0.27 mol CO 2 per mol amine, respectively).

Effect of GP size on the reversible CO 2 capture stoichiometry of GP lms
We have reported that the GP lms exhibited larger CO 2 capture capacities than conventional bulk hydrogel lms due to the small dimensions of GPs, which lead to a fast response and fast ion diffusion. 29 Herein, the effect of GP size on the reversible CO 2 capture stoichiometry is discussed.
Larger GPs can be prepared by increasing the concentration of monomer or by decreasing the concentration of surfactant. 50 In this study, GPs with the same composition (5 mol% DMAPM and 2 mol% Bis) but different size were designed. A GP with a larger hydrodynamic diameter (D5B2-L) than D5B2 was prepared by decreasing the concentration of surfactant, while a smaller GP (D5B2-S) was synthesized using a monomer solution with a lower total concentration.
The diameters of the GPs at 30 C and 75 C are shown in Fig. 4a. As expected, D5B2-L (638 nm at 30 C and 322 nm at 75 C) has a larger diameter than D5B2 (196 nm at 30 C and 74 nm at 75 C), and D5B2-S has the smallest diameter (115 nm at 30 C and 49 nm at 75 C). The pK a values of the GPs at 30 C and 75 C are presented as the top and bottom, respectively, of the bars in Fig. 4b. It can be seen clearly that the pK a values of the GPs at both temperatures increase with the decreasing diameter.
The amine groups located at the exterior of the GPs are expected to show higher pK a values than those of the interior groups because of the high dielectric constant of water, as well as the reduced steric hindrance. In the meantime, the amine groups of the exterior of the smaller GP account for a much larger percentage than those of the larger GPs. Therefore, smaller GPs show higher pK a values than the larger GPs. Fig. 4b also shows that the three GPs exhibit similar, high reversible CO 2 capture stoichiometries (0.96-1.0 mol CO 2 per mol amine), despite the difference in GP diameters and pK a values.
Another pair of GPs showing different pK a ranges from those of the GPs in Fig. 4a and b was polymerized using the "microenvironment-imprinting" strategy by adding 0.5 eq. of NaOH to the monomer solution (D5B2-1/2NaOH and D5B2-1/2NaOH-S). D5B2-1/2NaOH-S, having a smaller diameter than D5B2-1/ 2NaOH, was prepared by lowering the total monomer concentration. Fig. 4c and d show the diameters and pK a values, respectively, of the GPs at 30 C and 75 C. In good agreement with the results in Fig. 4a, the diameters of D5B2-1/2NaOH-S (83 nm at 30 C and 57 nm at 75 C) are much smaller than the diameters of D5B2-1/2NaOH (715 nm at 30 C and 349 nm at 75 C). The pK a values of the smaller GP are also higher than those of the larger GP. However, the reversible CO 2 capture stoichiometry of D5B2-1/2NaOH-S (0.74 mol CO 2 per mol amine) is lower than that of D5B2-1/2NaOH (0.98 mol CO 2 per mol amine), as shown in Fig. 4d.
The different results from Fig. 4b and d could also be ascribed to the different pK a values of the GPs. The high pK a values of D5B2-L, D5B2, D5B2-S, and D5B2-1/2NaOH at 30 C (7.0, 7.8, 8.5, and 7.2, respectively), which are above the pK a1 of H 2 CO 3 (6.35), and the low pK a values at 75 C (4.8, 5.3, 6.1, and 4.9, respectively), which are below the pK a1 of H 2 CO 3 , result in high reversible CO 2 capture stoichiometries. However, the pK a values of D5B2-1/2NaOH-S at 30 C and 75 C are 8.4 and 6.6, respectively. Its high pK a value at 75 C (6.6), which is above the pK a1 of H 2 CO 3 , restricts the efficient release of CO 2 , resulting in a low reversible CO 2 capture stoichiometry (0.74 mol CO 2 per mol amine).
In general, the results of Fig. 4 lead to the conclusion that the reversible CO 2 capture stoichiometry of GPs can be improved by regulating their pK a values at both 30 C and 75 C through varying their diameter. The smaller diameter leads to a higher pK a value.
Effect of swelling ratio of GP on the reversible CO 2 capture stoichiometry of GP lms The swelling ratio of thermal responsive GPs plays an important role in their function such as the reversible target binding. 26,51,52 The swelling ratio of GPs is typically controlled by the crosslink degree. 53 Herein, we prepared a series of GPs containing 5 mol% of DMAPM with 0, 2, 5, and 10 mol% Bis crosslinker (D5B0, D5B2, D5B5, and D5B10, respectively) to investigate the effect of the swelling ratio on the reversible CO 2 capture stoichiometry of GPs.
The diameters at 30 C and 75 C, and the swelling ratios of the GPs are plotted in Fig. 5a. It is noteworthy that GP can be obtained without Bis crosslinker (D5B0), possibly due to the slight self-crosslinking. 54 As expected, the swelling ratios of D5B2 (2.7), D5B5 (2.0), and D5B10 (1.6) decrease with the increasing crosslink degree. The diameters of the GPs at 75 C increase slightly with the increasing amount of Bis crosslinker, indicating that Bis inuences the nucleation process during polymerization, as observed by Pelton's group. 50 Fig . 5b shows the pK a values of the GPs at 30 C and 75 C, and the DpK a values. The DpK a values of D5B2 (2.5), D5B5 (2.1), and D5B10 (1.3) between 30 C and 75 C show a positive correlation with the swelling ratios of the GPs (2.7, 2.0, and 1.6, respectively). This suggests that from 30 C to 75 C the microenvironmental changes, such as the change of dielectric constant and polymer density, around the less cross-linked amines are greater than those of the highly cross-linked amines in the GPs. At 30 C, the pK a values of the swollen GPs, D5B2 (7.8), D5B5 (6.7), and D5B10 (5.9), decrease with the reduction in the swelling ratio, indicating the GP with a smaller swelling ratio is in a relatively collapsed state at 30 C compared to those with higher swelling ratios, because swelling of the GP is inhibited by the high crosslink degree. The pK a value of D5B2 at 30 C (7.8) is similar to that of D5B0 (7.7), although D5B0 exhibits a much larger swelling ratio, indicating that D5B2 is also fully swollen at 30 C.
In Fig. 5b, D5B0, D5B2, and D5B5 exhibit similar reversible CO 2 capture stoichiometries (0.97-1.0 mol CO 2 per mol amine), while that of D5B10 (0.74 mol CO 2 per mol amine) is much lower. The difference in the reversible CO 2 capture stoichiometries can also be explained by their pK a values. The higher pK a values of D5B0, D5B2, and D5B5 at 30 C (7.7, 7.8, and 6.7, respectively), which are above the pK a1 of H 2 CO 3 , and the lower pK a values at 75 C (5.2, 5.3, and 4.4, respectively), which are below the pK a1 of H 2 CO 3 , lead to high reversible CO 2 capture stoichiometries. However, the pK a value of D5B10 at 30 C (5.9) is too low and the basicity is too weak to capture CO 2 efficiently.
Conclusively, the swelling ratio (D 30 /D 75 ), which can be controlled by crosslink degree, is an important factor to tune the DpK a and consequently the pK a values of GPs to improve the reversible CO 2 capture stoichiometry.
Design rationale of GP lm with large reversible CO 2 capture stoichiometry The above discussion has revealed the design rationale of GPs that show large reversible CO 2 capture stoichiometry within a narrow temperature range of 30 C to 75 C. The pK a values of the GPs at 30 C and 75 C are the principal factors that govern the reversible CO 2 capture stoichiometry. Higher pK a values at 30 C, above the pK a1 of H 2 CO 3 (6.35), and lower pK a values at 75 C, below the pK a1 of H 2 CO 3 , enable high reversible CO 2 capture stoichiometries of GP lms. This is because the GPs are capable of capturing CO 2 efficiently at 30 C due to the stronger basicity of the amine than HCO 3 À , and then releasing CO 2 sufficiently at 75 C because of the weaker basicity. High reversible CO 2 capture stoichiometry can be achieved for GPs that exhibit a smaller DpK a , as long as the pK a values lie within the appropriate range.
The pK a value of the GPs can be readily adjusted to the desired level by varying the VPTT of the GPs above and close to the CO 2 capture temperature (30 C in this study), because the volume phase transition of GPs always brings out a large pK a transition throughout a wide temperature range above the VPTT. The pK a value of the GPs can also be tuned by controlling the size of the GPs since smaller GPs show higher pK a values. Another method is to regulate the GP swelling ratio, which inuences DpK a and consequently the pK a value of GPs. Finally, the imprinted microenvironment around the amine groups in the GPs can also inuence the pK a values of the GPs because the GPs synthesized in the presence of a large amount of protons exhibit higher pK a values. The inter-relationship between these factors is illustrated by Scheme 3.

Optimization of GPs to maximize reversible CO 2 capture capacity
In order to improve the CO 2 capture capacity by GP lms, the ratio of functional amine monomer, DMAPM, was maximized to 55 mol% (GPs with a higher DMAPM concentration precipitated during polymerization process). However, D55B2 exhibited a low CO 2 capture stoichiometry (0.73 mol CO 2 per mol amine) due to its high pK a value at 75 C (6.8) (Fig. 6a). According to the design rationale described above, highly efficient CO 2 capture cycles could be achieved if the pK a value of the 55 mol% DMAPM-containing GP at 75 C is below the pK a1 of H 2 CO 3 (6.35).
Since the VPTT of D55B2 is greater than 75 C, one approach to reducing the pK a value of the 55 mol% DMAPM-containing GP at 75 C is to lower its VPTT. Thus, 43 mol% TBAm was incorporated into the 55 mol% DMAPM-containing GP (D55B2T43).
In accordance with the design rationale, the pK a value of D55B2T43 at 75 C decreases to about 6.35 (Fig. 6a) due to the reduced VPTT via the incorporation of TBAm. Consequently, a high reversible CO 2 capture stoichiometry (0.93 mol CO 2 per mol amine) is obtained. As a result of the high amine content and the high stoichiometric efficiency, as shown in Fig. 6b, D55B2T43 shows the largest reversible CO 2 capture capacity (68 mL CO 2 per g dry GPs, 3.0 mmol CO 2 per g dry GPs) in this study.
Though the GP D55B2T43 showed high CO 2 capture capacity (68 mL CO 2 per g dry GPs, 3.0 mmol CO 2 per g dry GPs), taking into account of water and support that are necessary for the GP lms, the CO 2 capture capacity of the GP lm is lower than the best adsorbents. 55 However the GP lms can be used directly to capture CO 2 from desulfurized post-combustion exhausted gas, without pre-treatment of the exhausted gas to remove water vapor inside, nor to decrease the temperature of the water vapor. Furthermore, the low regeneration temperature of the GP lms (75 C) enables the utilization of abundant and low cost waste heat (<100 C) of factories as an energy source. Consequently the energy consumption of the GP lms might be lowered. Fig. 6a summarizes the pK a values and the reversible CO 2 capture stoichiometries of the GPs studied in this paper. Overall, it can be concluded that the GPs with pK a values at 30 C above the pK a1 of H 2 CO 3 , and pK a values at 75 C below the pK a1 of H 2 CO 3 (red) generally show larger reversible CO 2 capture stoichiometries than other GPs (blue).

Reversibility of the optimized GPs as CO 2 absorbent in wet environment
In order to show the cycle stability of the GP lm, the reversible CO 2 capture-release by D55B2T43 lm was carried out for 10 cycles. The result is shown in Fig. 7 red plots. The reversible CO 2 capture capacity decreased gradually from 0.93 mol CO 2 per mol amine of the 1 st cycle to 0.71 mol CO 2 per mol amine of the 10 th cycle. The reason is that the GP lm partly dried out, as can be Fig. 6 (a) pK a values (left Y axis) at 75 C (bottom of bar) and 30 C (top of bar) of the GPs studied in this paper, and the reversible CO 2 capture stoichiometries (right Y axis, red dot) of the GP films in the temperature range 30-75 C. (b) Reversible CO 2 capture capacities of the GP films in the temperature range 30-75 C. Red represents GPs with high reversible CO 2 capture stoichiometries (>0.9 mol CO 2 per mol amine) and blue represents those with low stoichiometries (<0.75 mol CO 2 per mol amine). Fig. 7 CO 2 capture/release stoichiometry (right Y axis) and CO 2 capture/release capacity (left Y axis) of GP D55B2T43 film with 4 mL water per g GPs. Red: first 10 cycles of CO 2 release/capture by the GP film. After the first 10 cycles, 4 mL water per g GPs was supplied to the dried part of the GP film. Blue: the CO 2 release-capture by the GP film after supplying 4 mL water per g GPs to the dried part of the GP film.