Design, synthesis, and physicochemical study of a biomass-derived CO2 sorbent 2,5-furan-bis(iminoguanidine)

Summary In this study, the concept of biomass-based direct air capture is proposed, and the aminoguanidine CO2 chemical sorbent 2,5-furan-bis(iminoguanidine) (FuBIG) was designed, synthesized, and elucidated for the physicochemical properties in the process of CO2 capture and release. Results showed that the aqueous solution of FuBIG could readily capture CO2 from ambient air and provided an insoluble tetrahydrated carbonate salt FuBIGH2(CO3) (H2O)4 with a second order kinetics. Hydrogen binding modes of iminoguanidine cations with carbonate ions and water were identified by single-crystal X-ray diffraction analysis. Equilibrium constant (K) and the enthalpies (ΔH) for CO2 absorption/release were obtained by thermodynamic and kinetic analysis (K7 = 5.97 × 104, ΔH7 = −116.1 kJ/mol, ΔH8 = 209.31 kJ/mol), and the CO2-release process was conformed to the geometrical phase-boundary model (1-(1-α)1/3 = kt). It was found that the FuBIGH2(CO3) (H2O)4 can release CO2 spontaneously in DMSO without heating. Zebrafish models revealed a favorable biocompatibility of FuBIG.


INTRODUCTION
The heavy reliance and massive consumption of fossil resources in modern society has caused continuous rising of atmospheric CO 2 concentration and resulted in an alarming change of global climate (Cox et al., 2020;Davis et al., 2018;Anderson et al., 2018;Jin et al., 2020). Carbon capture and storage (CCS) has been proposed and implemented as a feasible strategy to reduce the point-source CO 2 emissions (Lackner, 2003;Reiner, 2016;Burant et al., 2017;Zhai et al., 2015;Sha et al., 2016;Bui et al., 2018;Wang, 2016). However, as the dispersed CO 2 emissions account for 50% of total greenhouse emissions (Seipp et al., 2017), the application of point-source CCS technologies is unlikely to stabilize the atmospheric CO 2 concentration at a desirable level (Sanz-Pé rez et al., 2016). Accordingly, the concept of direct air capture (DAC) has been put forward which aims at capturing CO 2 from ambient air (National Research Council, 2015;Keith, 2009;Sutherland, 2019;Jacobson, 2019;Bajamundia et al., 2019;Shi et al., 2020). In the past two decades, many efforts have been devoted for the development of various DAC sorbents, such as porous organic polymers (POPs), metal-organic frameworks (MOFs), solid-supported amine-based sorbents, and small molecular organic sorbents. Among them, the POPs are polymeric materials constructed by organic covalent bonds which are considered to be promising materials for CO 2 storage due to the hyper-crosslinked structures and high stability (Zou et al., 2017;Li et al., 2017;Wang et al., 2017;Yang et al., 2019). The MOFs are highly porous materials constructed by metal cations and organic ligands which have large surface areas and can effectively trap carbon dioxide in the cages (Boyd et al., Trickett et al., 2017;Gonzá lez-Zamora and Ibarra, 2017). Solid-supported amine-based sorbents are solid adsorbents which incorporate amine moieties into solid supports including zeolites, carbons and organic resins, etc (Thakkar et al., 2017;Pang et al., 2017;Holewinski et al., 2017;Sarazen and Jones, 2017). These adsorbents offer advantages such as high selectivity for CO 2 and relatively low cost, but they displayed lower CO 2 capacities compared to other sorbents. In recent years, some specially designed organic compounds have shown great potential for DAC. For example, in 2014, Hossain group developed an organic compound with six urea groups which absorbed atmospheric CO 2 as CO 3 2À via 12 strong N-H•••O bonds under mild conditions (Pramanik et al., 2014). In 2017, Custelcean et al. reported an innovative iminoguanidine type sorbent, namely 2,6pyridine-bis(iminoguanidine) (PyBIG) which could capture CO 2 from ambient air, crystallize as an insoluble carbonate and regenerate PyBIG by mild heating with concomitant CO 2 releasing (Seipp et al., From the viewpoints of green chemistry and climate change, the chemical conversion and utilization of biomass represent a kind of carbon neutral processes which would be more eco-friendly compare with the utilization of fossil resources (Chheda et al., 2007;Ji et al., 2018;Strube et al., 2011). As an indirect source of solar energy, biomass is abundant and renewable in nature. It is generally accepted that largescale and efficient utilization of biomass would cause no net increase of carbon and would exert beneficial effects against global warming (Chheda et al., 2007;He et al., 2019;Ma et al., 2011).
Platform molecules are recognized as vital hallmarks of biomass conversion and utilization which can be utilized as starting materials or building blocks for the production of a great number of downstream chemicals (Serrano-Ruiz et al., 2011;Arias et al., 2020;Luterbacher et al., 2014;Zhu et al., 2017). Among them, the 5-hydroxymethylfurfural (5-HMF) is one of the most versatile and highly transformable bio-based platform molecules originated from lignocellulose (Ma et al., 2011;Zhu et al., 2017;Saha and Omar, 2014;Solanki and Rode, 2019;Zhao et al., 2007). The partial oxidized derivative of 5-HMF, namely, 2,5-diformylfuran (DFF) has also been widely used as a starting material for the synthesis of pharmaceuticals, macrocyclic ligands, and functional polymeric materials (Ghatta et al., Zhang and Huber, 2018;Zhao et al., 2018). Inspired by the concept of bioenergy with carbon capture and storage (BECCS) (Davis et al., 2018), and as part of our continuing efforts on biomass conversion and utilization (Wu et al., 2012;Luo et al., 2019;Huang et al., 2018;Sheng et al., 2016;Zhang et al., 2014aZhang et al., , 2014bZhang et al., , 2014cZhang et al., , 2015, we envisioned that the concept of biomass conversion and DAC could be combined to the description of biomass-based direct air capture (BBDAC, Figure 1.). More specifically, we envisaged that the development of an iminoguanidine type of CO 2 sorbents based on biomass-derived platform molecules would be rational and more favorable to achieve the goal of negative emissions. Herein, we report the design, synthesis, physicochemical study, and swing property of an efficient CO 2 sorbent, the 2,5-furan-bis(iminoguanidine) (FuBIG), starting from the biomass-derived DFF.

Synthesis, physical properties and crystallization analysis of FuBIG
The FuBIG was readily obtained by the imine condensation of the biomass-derived DFF with aminoguanidinium chloride, followed by neutralization with aqueous NaOH. Gratifyingly, the FuBIG showed an improved aqueous solubility (0.4029 M, 25 C) than PyBIG (0.0012 M, 25 C), and when the aqueous solution of FuBIG was left open to ambient air for a few days, the formation of prism shaped, yellowish-brown single crystals was found, which was consistent with the composition of a tetrahydrated carbonate FuBIGH 2 (CO 3 ) (H 2 O) 4 by Fourier transform infrared spectroscopy (FTIR), elemental analysis (EA), and single-crystal X-ray diffraction analysis (CCDC: 2038310, iScience Article was significant and would be the main driving force for the capture of CO 2 in ambient air. Moreover, the energy of the hydrogen bonds has been analyzed by DFT study. The energy of the hydrogen bonds between FuBIGH 2 2+ and CO 3 2is calculated to be 16.4 kcal/mol, and the energy of the hydrogen bonds between FuBIGH 2 (CO 3 ) and H 2 O is calculated to be 6.5 kcal/mol (see Table S12: The cartesian coordinates (xyz) for hydrogen bonds on DFT calculation for details). These results suggest that hydrogen bonds would provide a delicate balance between the stability of FuBIGH 2 (CO 3 )(H 2 O) 4 and the property of FuBIG regeneration and CO 2 release.
Thermodynamic and kinetic analysis of CO 2 absorption and heat release The thermodynamic and kinetic study of CO 2 absorption and heat release are crucial tasks in the search for CO 2 sorbents, which could provide accurate physicochemical parameters and minimum energy requirements for the reactions in stepwise and overall manner, thereby paving the way for further optimization and application. The reactions involved in the CO 2 absorption and heat release regarding FuBIG are shown in Scheme 1, and the corresponding thermodynamic parameters are listed in Table 1.
The overall equilibrium constant for CO 2 absorption is 5.97310 4 (K 7 = K 1 3K 2 3K 3 3K 4 3K 5 3K 6 , Table 1). It is worth-mentioning that the values of K 3 , K 4 and K 5 are invariable under ideal conditions, so the extent of the overall reaction of any specially designed aminoguanidine sorbent could be improved with the increase of basicity (K 2 ) and solubility of the free base (K 1 ) and the decrease of solubility of its carbonate salt (K 6 ). Specifically, strong basicity of the sorbent would facilitate the transformation of CO 2 to CO 3 2À , high solubility of the free base would make the aqueous solution of the sorbent to interact with the gaseous CO 2 more efficiently, whereas the low solubility of the corresponding carbonate salt would lead to the precipitation more thorough and the separation of which by filtration more convenient. However, as the differences of basicity between iminoguanidine sorbents are generally indistinctive, therefore, the solubility of the free base and its carbonate salt would be the simple and determining factors for this kind of sorbents. Accordingly, The value [R s ] was introduced in this study which was defined as the solubility ratio between FuBIG and its carbonate FuBIGH 2 (CO 3 ) (H 2 O) 4 at 25 C ([R s ] = 43.12, Table 1, Tables S4 and S5). Moreover, according to solubility data reported by Custelcean et al. (Brethomé et al., 2018;Williams et al., 2019), the [R s ] value for PyBIG and GBIG were determined to be 7.97 and 1.60, respectively. Thus, the results suggested that the FuBIG would be more favorable as a CO 2 sorbent than PyBIG and GBIG in terms of [R s ] value.
When the crystals of FuBIGH 2 (CO 3 ) (H 2 O) 4 were heated in an oven at 120 C for 1h, the crystals transformed to FuBIG and changed their appearance from transparent to opaque, while maintaining the original yellow color ( Figure S15). Thermogravimetric analysis (TGA) provided a quantitative measurement of the decomposition process ( Figure 3A). FuBIGH 2 (CO 3 ) (H 2 O) 4 showed a 35.71% mass loss between 72 C and 148 C, which was consistent with the loss of 1 equiv. of carbonic acid (as CO 2 and H 2 O) and 4 equiv. of H 2 O (36.22% theoretical mass loss). After the complete regeneration of FuBIG at 148 C, the TGA curve became flat with the increase of temperature to about 220 C, this result indicated that there was no thermal decomposition occurred in this temperature range, demonstrating a high thermostability of FuBIG. Subsequently, it was found that the FuBIG began to decompose at around 240 C with further increment of temperature. In addition, the very similar TGA pattern between FuBIGH 2 (CO 3 ) (H 2 O) 4 and FuBIG at above 140 C, fully confirmed the regeneration process  Figure S10 2 FuBIG protonation a DH 2 = 46.08 e K 2 = 1.91 3 10 À12 j - Figure S12 3 crystallization DH 6 = À101.57 f K 6 = 1.63310 7 k S 6 = 0.009344 Figure S11 7 overall CO 2 absorption DH 7 = À116.1 g K 7 = 5.97 3 10 4l -  Figure S16). Before the temperature reached the point for CO 2 release, it could be calculated from the DSC measurements that an enthalpy of 48.40 kJ/mol was needed for the specific heat capacity of FuBIGH 2 (CO 3 ) (H 2 O) 4 ( Figure S17). Hence, the total enthalpy requirement for the regeneration of FuBIG would be 257.71 kJ/mol.   Figure S18). Among them, it was found that the practical and effective data was obtained from temperature range of 80, 90, 100, and 110 C ( Figure 3B). After plotting the fractional conversion (a) as a function of time ( Figure S19), the most common solid-state reaction kinetics, including Avrami-Erofeev, Prout-Tompkins, Ginstling-Brounstein, Jander, and geometrical phase-boundary models were screened ( Figure S20). It was found that the most fitted model would be assigned to the geometrical phase-boundary model characterized by Equation (9): Where m 0 represents the initial sample weight of FuBIG carbonate salt, m f represents the final weight after isothermal heating, and m represents the sample weight at a certain time ( Figure S20). This result revealed that the reaction initiated on the surface of crystals, followed by inward advance to the center, and resulting in a decelerator a-t curve with the decrease of interface. Moreover, activation barrier (E a = 92.24 kJ/mol) for the decomposition of FuBIGH 2 (CO 3 ) (H 2 O) 4 was obtained from Arrhenius analysis of the rate constants (k) under different temperatures ( Figure S21). ), respectively ( Figure 3C). By monitoring the intensity change of these absorption peaks, the corresponding concentration changes of FuBIG and FuBIG carbonate salt could be obtained. As shown in Figure S23, when CO 2 was put into the aqueous solution of FuBIG, the concentration of FuBIG decreased with the formation of FuBIG carbonate salt (the concentration change of carbonate anion was relatively complicated owing to the multifactorial influence toward the labile FuBIG carbonate salt under thermal conditions in solution state). By analyzing the concentration-time curve of FuBIG with integral method, it could be found that the process of CO 2 absorption was in accordance with the second order reaction kinetics with a rate constant (k) of 4.8102 3 10 À4 L/mol$s at 25 C ( Figure 3D).

Applicability evaluation of FuBIG in practical process
It is widely accepted that the recycling of both CO 2 and the sorbent is indispensable in practical CCUS process (Gao et al., 2020;Flores-Granobles and Saeys, 2020;Leclaire and Heldebrant, 2018;Gonzalez-Diaz et al., 2020). Therefore, in this study, the crystals of FuBIGH 2 (CO 3 ) (H 2 O) 4 were heated at 110 C in oven for one week for the investigation of the robustness of FuBIG in CO 2 capturing and releasing process. The weight was measured every 6 hr. After the release of CO 2 and H 2 O, FuBIG showed no sign of decomposition ( Figure 3E). The weight of the solid was fluctuating within a narrow range.
Next, we ran a full CO 2 separation cycle using CO 2 balloon to assess the recyclability of this sorbent. CO 2 gas was injected into the saturated aqueous solution of FuBIG, leading to the formation of yellow precipitate within minutes. The solid was collected by filtration and the filtrate was analyzed by ultravioletvisible spectroscopy to determine the concentration of the free FuBIG left in the solution. At the first cycle, 99.36% of FuBIG was converted to the FuBIG carbonate salt. Then the FuBIG carbonate salt was heated at 110 C for 4h, leading to complete release of CO 2 and H 2 O. The regenerative FuBIG was then dissolved into the filtrate and precipitated again by CO 2 . Overall, ten consecutive CO 2 capture/release cycles were conducted, and the conversion rate could still maintain at 92.49% ( Figure 3F). Although the longterm stability and recyclability of FuBIG remains to be explored over more and more capture/release cycles under practical conditions, our preliminary results indicated that this biomass-derived sorbent is remarkably robust.
The transformation of CO 2 into bulk chemicals or value-added products represents an attractive strategy for CO 2 utilization. In our previous work, a facile and operationally simple method (room temperature, 1 atm of CO 2 balloon) for the synthesis of various O-aryl carbamates via one-pot three-component coupling of aryl carboxamides, CO 2 , and amines has been established (Luo et al., 2019). We reasoned that the CO 2 captured by biomass-derived sorbent FuBIG could be utilized in chemical reactions. Accordingly, in this study, a round bottom flask filled with FuBIGH 2 (CO 3 ) (H 2 O) 4 in place of CO 2 balloon was connected with the reaction system, the CO 2 airflow was stably generated at 80 C, and reacted with aryl carboxamide and amine to afford the desired O-aryl carbamate in 70% yield in the presence of CuI and MnO 2 ( Figure  iScience Article (Luo et al., 2019). This result demonstrated the feasibility for the circulation of CO 2 capture, release and utilization regarding the transformation between FuBIG and FuBIG carbonate salt and represented a practical example for negative emissions under the proposed BBDAC rationale.

Spontaneous CO 2 release of FuBIG carbonate salt in DMSO
The regeneration convenience and long-term robustness of CO 2 sorbents are vital factors in the development of practical and efficient DAC technologies. It is quite obvious that, for an optimal CO 2 sorbent, large energy requirement should be avoided for CO 2 release and sorbent regeneration. However, to the best of our knowledge, current DAC technologies including humidity swing, pressure swing, or temperature swing, are energyintensive, strongly endothermic and complicated in the operating processes. Surprisingly and delightfully, during the recording of the NMR spectra for FuBIG carbonate salt, some interesting results displaying a unique, unprecedented, and spontaneous way for CO 2 release from FuBIG carbonate salt were obtained. It was found that when DMSO-d 6 was added into the FuBIG carbonate salt with oscillation at room temperature, the insoluble solid gradually dissolved with concomitant release of CO 2 in an endothermic manner in a few minutes. The 1 HNMR and 13 CNMR spectra gave identical results compared with that of FuBIG, demonstrating a unique and easy way for CO 2 release and regeneration of FuBIG from its carbonate salt in DMSO-d 6 ( Figures S4 and  S5). Subsequently, the non-deuterated DMSO was tested (showed by React IR in Figure 4A), and the same result was obtained. To gain further insight into the spontaneous process for CO 2 release of FuBIG carbonate salt in DMSO, calculation studies were carried out by density functional theory (Hohenberg and Kohn, 1964;Kohn and Sham, 1965). As illustrated in Figures 4B and 4C, owing to the existence of hydrogen bonds, the binding energy representing by DH and DG between H 2 CO 3 and FuBIG are more negative in value than that of between DMSO and FuBIG (for interaction of H 2 CO 3 with FuBIG: DH/DG = À26.7/-2.5 kcal/mol; for interaction of DMSO with FuBIG: DH/DG = À10.9/10.3 kcal/mol). Therefore, the extrusion of CO 2 from FuBIG carbonate salt in DMSO should not result from stronger interaction between DMSO and FuBIG. Moreover, as shown in Figure 4D, the reaction between H 2 CO 3 and FuBIG in water to form FuBIG carbonate salt is spontaneous (DG = À1.5 kcal/mol), whereas the same reaction taken place in DMSO is nonspontaneous (DG = 35.5 kcal/mol) (see Table S13: The cartesian coordinates (xyz) for all optimized structures on DFT calculation for details). Accordingly, it could be concluded that H 2 CO 3 and FuBIG are more in inclined to form the ion pair (FuBIGH 2 2+ ,CO 3 2À ) in water, while the counterreaction of which is more likely to take place in DMSO.
Interestingly, we have also found that the DMSO solution saturated with FuBIG could still promote the release of CO 2 from FuBIG carbonate, and result in the regeneration of FuBIG. This intriguing result indicated that a dynamic equilibrium might be achieved for CO 2 release and FuBIG regeneration in these conditions, thereby providing a proof of concept for a highly efficient and energy-saving protocol for the cycling of CO 2 capture, release and sorbent regeneration. It is obvious that under these conditions, the recovery of FuBIG (by filtration) from DMSO could be realized spontaneously and continuously with minimum DMSO consumption, and evaporation of DMSO is no longer needed for the sorbent recovery ( Figures S25  and S26). In addition, freeze drying might be another option for DMSO removing and sorbent recovery.

Biocompatibility assay of FuBIG
The future application scenarios of DAC sorbents could be classified into two categories: exceptionally largescale deployments for CO 2 capture either from point-sources or from ambient atmosphere, and small-scale facilities for CO 2 capture in enclosed cabins such as space capsules or submarines. Accordingly, the biocompatibility properties of DAC sorbents should be taken into consideration. In this study, zebrafish were used as the model species, the tests of acute toxicity and embryo toxicity were conducted, respectively, for the biocompatibility evaluation of FuBIG and the previously reported PyBIG. Results showed that the FuBIG was less toxic than PyBIG in acute toxicity test in terms of maximum non-lethal concentration (MNLC) and 10% lethal concentration (LC 10 ), with the corresponding measurements for FuBIG (MNLC = 39.6 mM, LC 10 = 54 mM) and PyBIG (MNLC = 17.1 mM, LC 10 = 24 mM), respectively ( Figures 5A and 5B). In target organ toxicity tests of FuBIG (1.9, 5.7, 17.1, and 24.0 mM), no toxicity was observed in 1.9 and 5.7 mM groups, whereas the delay of yolk sac absorption was found in 16.7% and 23.3% of zebrafish in 17. iScience Article treated with PyBIG (1.9, 5.7, 17.1, and 24.0 mM), no toxicity was observed in 1.9 mM group, whereas the delay of yolk sac absorption was found in 13.3% of zebrafish in 5.7 mM group. Moreover, it was observed that in 17.1 mM group treated with PyBIG, 6.7-30.0% of zebrafish developed renal edema, pericardium edema, delay of yolk sac absorption, lack and slowing down of blood flow, respectively. Further increment of PyBIG concentration to 24.0 mM, more serious toxic responses was observed with 10.0% of deaths occurred. In addition, FuBIG was found to be less toxic than PyBIG in embryo toxicity assay (for FuBIG: MNLC = 26.7 mM, LC 10 = 55.9 mM; for PyBIG: MNLC = 15.8 mM, LC 10 = 26.7 mM. Figures 5C and 5D). Notably, target organ toxicity tests of FuBIG and PyBIG in the same concentrations (1.8, 5.3, 15.8, and 26.7 mM) also revealed that the FuBIG has less embryo toxicity than PyBIG. The above-mentioned results suggested that the biomass-derived CO 2 sorbent might be more eco-friendly and more favorable in biocompatibility.  ), and H 2 O were determined by single-crystal X-ray diffraction analysis. The stepwise and overall thermodynamic and kinetic parameters for CO 2 absorption and heat release have been obtained through van't Hoff analysis, TGA, DSC and in situ reaction analysis. The reaction for CO 2 absorption has an overall enthalpy value (DH 7 ) of À116.10 kJ/mol, and an overall equilibrium constant (K 7 ) of 5.97310 4 , showing that the absorption of CO 2 in aqueous solution of FuBIG is highly advantageous. Moreover, the reaction for CO 2 heat release of FuBIGH 2 (CO 3 ) (H 2 O) 4 displayed a relatively lower energy requirement with an enthalpy value (DH 8 ) of 209.31 kJ/mol. Besides, a simple and intuitive symbol for the evaluation of CO 2 sorbents, namely, the [R s ] value which was defined as the solubility ratio between sorbents and their carbonate salts was proposed in this study ([R s ] = 43.12 for FuBIG, 7.97 for PyBIG, and 1.60 for GBIG, respectively, at 25 C). In addition, React IR analysis showed that the CO 2 absorption process was consistent with the second-order reaction kinetics with a rate constant (k) of 4.8102 3 10 À4 L/mol$s at 25 C, whereas the isothermal TGA analysis demonstrated that the kinetic characters for the release of CO 2 and H 2 O from FuBIG carbonate salt was in line with the geometrical phaseboundary model. Notably, it was found that the spontaneous CO 2 release of FuBIG carbonate salt occurred in DMSO, which might represent a near-zero-energy technique for DAC, this amazing process for spontaneous CO 2 release in DMSO was further elucidated by DFT calculations. Finally, the acute toxicity and embryo toxicity assay in zebrafish model displayed that the biomass-derived CO 2 sorbent of FuBIG was more favorable in terms of biocompatibility. Further investigations to develop more efficient biomass-derived sorbents and researches on structure-property relationships are underway in our group.

Limitations of the study
This study designed and synthesized an iminoguanidine type CO 2 sorbent, the FuBIG, starting from the biomass-derived platform compound DFF as the core structure. However, the side chain of iminoguanidine (transformed from aminoguanidine) was not biomass-sourced at present. Technically speaking, the aminoguanidine could be biomass-sourced when it is needed (see Figure S1 for details). In addition, the capacity of FuBIG decreases by 7% within 10 cycles, showing an insufficient efficiency in CO 2 capture/release and sorbent regeneration process. Considering that the weight loss in sample transfer procedure was iScience Article inevitable, especially at a laboratory scale, it is possible that when the experiment is performed at a largescale, the weight loss in sample transfer process would be greatly reduced. Furthermore, techno-economic analysis and life cycle analysis are warranted for future work.

Resource availability Lead contact
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Yong Zou (zouyong3@mail.sysu.edu.cn).

Materials availability
Full experimental procedures are provided in the Supplemental Information.

METHODS
All methods can be found in the accompanying transparent methods supplemental file.   Shi, X., Xiao, H., Azarabadi, H., Song, J., Wu, X., Chen, X., and Lackner, K.S. (2020). Sorbents for the direct capture of CO 2 from ambient air.

Experimental
Common reagents used in the synthesis were obtained commercially and used without further purification unless otherwise specified. All water used was deionized (18 mΩ). The 1 H NMR and 13 C NMR spectra were recorded using TMS as the internal standard on a Bruker BioSpin GmbH spectrometer at 400, 500 MHz respectively. UV-vis spectra were measured in 10-mm-path-length quartz cuvettes using a UV-2600 (SHIMADZU). pH measurements were conducted with a PHS-3E pH meter (Shanghai INESA Scientific Instrument CO., Itd) using an E-301F pH electrode. PXRD measurements were performed with X-ray single crystal diffractometer (Xcalibur Nova). 12423 diffraction points and 5860 independent diffraction points (R int = 0.0204, R sigma = 0.0213) were collected within the range of 2θ max = 134° at T = 100 K using Cu-Kα ray (λ = 1.54184 Å) was used as a light source. The crystal structure was analyzed by XS (Sheldrick, 2008) and refined by SHELXL (Sheldrick, 2015). TGA was under a nitrogen atmosphere using a STA 409PC (NETZSCH). DSC measurements were conducted under nitrogen with a DSC 3 (METTLER TOLEDO). In situ reaction analysis was conducted with a React IR 15 (METTLER TOLEDO). Acute toxicity and embryo toxicity experiment in zebrafish were observed with stereomicroscope (SZX7, OLYMPUS, Japan), and photos taken by CCD camera (VertA1, Shanghai Tusen Vivion Technology Co., Ltd, China). Sorbents were weighed with a precision electronic balance (CP214, OHAUS, USA).

Methods
Synthesis of FuBIG 2,5-furandialdehyde (12.4 g, 0.1 mol), aminoguanidine hydrochloride (22 g, 0.2 mol) and ethanol (100 mL) were added to a 250 mL round bottom flask fitted with a condenser. The mixture was heated at 70 °C with stirring for 8 h. After the reaction was ended, the reaction mixture was subjected to standing for 4h at 4 °C. The FuBIGH 2 Cl 2 product was collected by vacuum filtration as a light yellow solid, washed with ethanol for three times, then was dried under vacuum. The procedure yielded 34.7 g (98% yield). The chloride salt recrystallized from ethanol to be used in pK a experiment. The above 2,5-furyldiiminoguanidine hydrochloride hydrate was put in a 250 mL round bottom flask, added with 50 ml of 2M sodium hydroxide aqueous solution, stirred for 0.5h at room temperature, then subjected to standing for 12h at 4 °C. Resulting yellow solid was collected by vacuum filtration and dried to obtain 22.66 g (96% yield) of FuBIG.
CO 2 capture from air using aqueous FuBIG FuBIG (2.36 g, 10 mmol) was dissolved into 100mL of water and stirred for 12h at room temperature under the condition of sufficiently contacting with air to separate out a yellow solid. The yellow solid was filtered at reduced pressure and dried to obtain yellow powder (3.52 g, 95%), which was FuBIGH 2 (CO 3 )(H 2 O) 4 .
Single crystal X-ray diffraction X-ray quality single crystals of FuBIGH 2 (CO 3 )(H 2 O) 4 were obtained by preparing an aqueous solution of FuBIG (20 mL, 5 mM) in a 50 mL round bottom flask under ambient air, and let it at room temperature few days. 12423 diffraction points and 5860 independent diffraction points (R int = 0.0204, R sigma = 0.0213) were collected within the range of 2θ max = 134° at T = 100 K using Cu-Kα ray (λ = 1.54184 Å) was used as a light source. The crystal structure was analyzed by XS (Sheldrick, 2008) and refined by SHELXL (Sheldrick, 2015). pK a determination by potentiometric titrations The variable temperature titrations of pK a of FuBIG were done at the desired temperature using a circulating water bath. The electrode was calibrated by potassium hydrogen phthalate buffer (pH 4.00), mixed phosphate buffer (pH 6.86) and sodium tetraborate buffer (pH 9.18) respectively. A 50 mL ultrapure water solution containing FuBIG (5 mM), HCl (6 mM) and NaCl background electrolyte (0.2 mM) was titrated with a standard 0.1 M NaOH solution using a 200 μL pipette. The potential readings were recorded 5 minutes after each NaOH addition to allow the solution to equilibrate. The volume and pH value of each titration were recorded. Table S7 lists the pK a values obtained in the 15-35 °C range.

Solubility measurements
The solubilities of FuBIG and FuBIGH 2 (CO 3 )(H 2 O) 4 under variable temperature were determined by measuring the UV-Vis absorption spectra of the corresponding saturated ultrapure water solutions and comparing with a calibration curve obtained using solutions of FuBIGH 2 Cl 2 of known concentrations. The FuBIGH 2 Cl 2 aqueous solutions having the concentrations of 1.346 × 10 -5 M, 2.690 × 10 -5 M, 3.365 × 10 -5 M, 6.730 × 10 -5 M and 1.350 × 10 -4 M respectively. The absorbance of the FuBIGH 2 Cl 2 samples was recorded under the maximum absorption wavelength (Table S3). Saturated solutions were prepared by suspending an excess of the crystalline solids in 10 mL H 2 O inside reaction tube, and stirred for 24h inside a circulating water bath at different temperatures in the range of 15 to 35 °C. All measurements were run in triplicate. The average solubility values of FuBIG and FuBIGH 2 (CO 3 )(H 2 O) 4 are reported respectively in Table S4 and S6.

Determination of K sp of FuBIGH 2 (CO 3 )(H 2 O) 4
As the calculation of K sp at 25°C for an example: K sp is the reaction equilibrium constant of equation 1. The concentration of FuBIGH 2 2+ and CO 3 2in equilibrium and the activity coefficient (γ ± ) at 25°C need be measured and calculated. The concentration of carbonate anion was calculated by equation 2 and mass balance, it was determined to be 0.0001331 M considering the pK a of HCO 3 is 10.32 and the pH of the saturated carbonate solution is 8.48. The concentration of the FuBIGH 2 2+ cation was determined by taking into account of the measured solubility of FuBIGH 2 (CO 3 )(H 2 O) 4 (0.009344 M), the pK a values of FuBIG in 25 °C (7.57 (equation 3) and 8.71 (equation 4)) and the pH of the saturated solution is 8.48. The concentration of the ligand is: FuBIG 0.003214 M, FuBIGH + 0.005458 M, FuBIGH 2 2+ 0.0006715 M. Ionic strength can be calculated and the value is 0.001609. The activity coefficients (γ ± ) were estimated at 0.828 using the Debye-Huckel limiting law (Peiper and Pitzer, 1982;Stefánsson et al., 2013;Huang, 2010). The Values of K sp at other temperature were calculated as the same measurement. The pK a of HCO 3 and value of A in Debye-Huckel limiting law at 15-35 °C were obtained from previous references (Peiper and Pitzer, 1982;Huang, 2010). Table S5 lists the pH of saturated

TGA measurements
The TGA was under a nitrogen atmosphere. The sample was ramped at 10 °C/min to 600 °C. For the isothermal measurement, FuBIG carbonate salt was ramped 5 °C/min to 50, 60, 70, 80, 90 °C and 10 °C/min to 100, 110 °C, then held the temperature for 180 min.
DSC measurements DSC was conducted under a nitrogen atmosphere. The sample was measured in a temperature range of 30-200 °C and temperature ramp of 10 °C/min.

CO 2 absorption and release of FuBIG monitored by ReactIR
The infrared spectrum of FuBIG and FuBIGH 2 (CO 3 )(H 2 O) 4 was collected through ReactIR. The peak at 1533 cm -1 is the characteristic N-H absorption of FuBIG, while the peak at 1365 cm -1 representing the wavenumber of the carbonate salt. The calibration curve was determined by FuBIG aqueous solutions having the concentrations of 0.15890 M and diluted 5, 10, 25 and 50 times respectively. The absorbance of the above concentration FuBIG solutions was measured in ReactIR. FuBIG (2.83 g, 12 mmol) was added to a 100 mL three-necked flask and dissloved into 40 mL water at 25°C. One neck of flask was inserted into an on-line Infrared Dicomp Probe and fixed with a Teflon adapter. One data is collected every 0.5 min. After the absorbance is stabilized, a CO 2 balloon was inserted into the other mouth of the flask, and continued to stir until FuBIG was completely converted into FuBIGH 2 (CO 3 )(H 2 O) 4 . Then the mixture was heated to 70 °C, FuBIGH 2 (CO 3 )(H 2 O) 4 precipitate released CO 2 and converted to FuBIG aqueous solution again.
CO 2 separation cycles FuBIG (9.5 g, 0.04 mol) was dissolved into 100 mL ultrapure water in a 250 mL round bottom flask and marked liquid level. The absorbance of FuBIG at 368 nm was measured to be substituted into the standard equation to calculate the concentration. Then the CO 2 balloon was bubbled through the solution for 1 hour. A yellow precipitate started to form after 5 min. The yellow solid was collected by vacuum filtration. The filtrate was collected and analyzed by UV-Vis spectroscopy to determine the concentration of FuBIG left in solution. The filtrate was saved for the next cycle. The precipitate was placed in a crystallization dish and heated for 4 hours in an oven at 100 °C. The regenerated FuBIG was redissolved into the filtrate saved from the previous cycle and the ultrapure water was added to the mark. And the resulting FuBIG solution was recycled. Overall, ten consecutive cycles had been run.

DFT computational
All reported structures were optimized by the density functional theory (DFT) (Kohn and Sham, 1965;Hohenberg and Kohn, 1964) with the B3LYP functional (Vosko et al., 1980;Lee et al., 1988;Becke, 1993) with 6-31G (d, p) basis sets (Petersson et al., 1988(Petersson et al., , 1991 in the gas phase. Based on the recent studies by Grimme, the empirical dispersion correction was considered to be important in accurate prediction of the reaction free energy (Chakraborty et al., 2014). Hence, the D3 version of Grimme's dispersion correction with the original D3 damping function was considered in structure optimizations and energy calculations (Grimme et al., 2010). Frequency analysis calculations of optimized structures were performed at the same level of theory to characterize the structures to be minima (no imaginary frequency). Based on the B3LYP-D/6-31G (d, p) optimized geometries, the energy results were further refined by calculating the single point energy at the B3LYP-D/6-311++G (d, p) (Petersson et al., 1988(Petersson et al., , 1991 level of theory. The bulky solvation effects were simulated by SMD (Marenich et al., 2009) continuum solvent mode at the B3LYP-D/6-311++G (d, p) level of theory, with water (ɛ = 78.4) and DMSO (ɛ = 46.8), respectively, according to their corresponding reaction conditions. All the calculations were performed with the Gaussian 09 program (Frisch et al., 2013). The 3D optimized structures were displayed by CYLview visualization program (Legault, 2009).

Zebrafish handling
Adult AB strain zebrafish were fed with live brine shrimp twice daily and dry flake once a day. The culture temperature was controlled by aquaculture facility with a standard 14 h/10 h light/dark photoperiod (Westerfield, 1995). Four to five pairs of zebrafish were set up for nature mating every time. On average, 200-300 embryos were generated. Embryos were maintained at 28 °C in fish water (0.2% Instant Ocean Salt in deionized water, pH 6.9-7.2, conductivity 480-510 μS/cm and hardness 53.7-71.6 mg/L CaCO 3 ). The embryos were washed and staged at 6 h post-fertilization (hpf) and 24 hpf (Kimmel et al., 1995). Zebrafish were housed in Hunter Biotechnology, Inc., which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAA LAC) International.

Determination of maximum non-lethal concentration (MNLC) and LC 10
Zebrafish larvae were treated with FuBIG or PyBIG from 48 to 120 hpf for the acute toxicity, from 4 to 120 hpf for the embryonic toxicity assay (He et al., 2013;Zhu et al., 2014). Mined as those lacking an observable heartbeat under a dissecting stereomicroscope. Seven concentrations were used for each sample. If LC 10 (10% lethal concentration) and MNLC were not reached, additional testing concentrations up to 2000 μM and down to 0.001 μM were tested. Mortality curves were generated using Origin 8.0 (OriginLab, USA). MNLC and LC 10 were estimated from this curve.

Identification of target organs
Four concentrations (1/9 MNLC, 1/3 MNLC, MNLC and LC 10 ) were used to identify the toxicity target organs. Zebrafish larvae were treated with samples from 48 to 120 hpf. At the end of treatment, zebrafish from each group were randomly selected for visua observation and image acquisition. Major zebrafish organs and tissues were visually assessed, and toxic target organs were identified based on morphological abnormalities. After treatment, the heart, brain, eyes, liver, intestine, spine, and behaviours of each fish were observed under the microscope. The occurrence of edema, hemorrhage, and thrombosis were also observed in the animals.

Embryo toxicity
Four concentrations (1/9 MNLC, 1/3 MNLC, MNLC and LC 10 ) were used to identify the embryonic toxicity. Zebrafish larvae were treated with samples from 4 to 120 hpf. At the end of treatment, zebrafish from each group were randomly selected for visual observation and image acquisition. Embryos were daily observed up to 120 h with the dissecting stereomicroscope (SZX7, OLYMPUS, Japan), recording the four apical observations as indicators of lethality: coagulation of fertilized eggs, lack of somite formation, lack of detachment of the tailbud from the yolk sac, and lack of heartbeat. During the exposure period, developmental alterations, teratological parameters, and percentage of hatching were also recorded. Major zebrafish organs and tissues were visually assessed, and toxic target organs were identified based on morphological abnormalities. After treatment, the heart, brain, eyes, liver, intestine, spine, and behaviors of each fish were observed under the microscope. The occurrence of edema, hemorrhage, and thrombosis were also observed in the animals. Figure S1. Synthetic routes of biomass derived aminoguanidine. Related to Figure 1.