Competitive and Cooperative CO2–H2O Adsorption through Humidity Control in a Polyimide Covalent Organic Framework

In order to capture and separate CO2 from the air or flue gas streams through nanoporous adsorbents, the influence of the humidity in these streams has to be taken into account as it hampers the capture process in two main ways: (1) water preferentially binds to CO2 adsorption sites and lowers the overall capacity, and (2) water causes hydrolytic degradation and pore collapse of the porous framework. Here, we have used a water-stable polyimide covalent organic framework (COF) in N2/CO2/H2O breakthrough studies and assessed its performance under varying levels of relative humidity (RH). We discovered that at limited relative humidity, the competitive binding of H2O over CO2 is replaced by cooperative adsorption. For some conditions, the CO2 capacity was significantly higher under humid versus dry conditions (e.g., a 25% capacity increase at 343 K and 10% RH). These results in combination with FT-IR studies on equilibrated COFs at controlled RH values allowed us to assign the effect of cooperative adsorption to CO2 being adsorbed on single-site adsorbed water. Additionally, once water cluster formation sets in, loss of CO2 capacity is inevitable. Finally, the polyimide COF used in this research retained performance after a total exposure time of >75 h and temperatures up to 403 K. This research provides insight in how cooperative CO2–H2O can be achieved and as such provides directions for the development of CO2 physisorbents that can function in humid streams.

S-2 Cumulative pore volume S-11 Isosteric enthalpy of adsorption S-12 Supplementary breakthrough data S-13 Supplementary TGA data on COFs with pre-adsorbed water S-17 Supplementary FT-IR data on water-COF binding studies S-19 S-3

Characterization techniques
FT-IR spectra were recorded on a PerkinElmer Spectrum 100 FT-IR Spectrometer with an universal ATR accessory over a range of 4000 to 650 cm -1 . TGA analyses were performed from 30 to 860 °C, under a nitrogen atmosphere at a heating rate of 10 °C⋅min -1 using a Perkin Elmer TGA 4000.

Synthesis of TAPB-NDA COF
TAPB (351 mg, 1.0 mmol) and NDA (402 mg, 1.5 mmol) were added to a glass cylindrical reactor and subsequently o-DCB (10 mL), NMP (10 mL) and isoquinoline (0.1 mL) were added. Thereafter, the reactor was sealed, vacuum degassed, and placed in an oil bath. Under gently shaking the reactor vessel, it was heated to a temperature of 150 °C. Upon complete dissolution of the monomers, the temperature was stepwise (10 °C per 15 min) increased to 190 °C and kept without agitation at that temperature for 3 days. The reaction mixture was cooled to room temperature, suspended in 60 mL methanol, and mixed thoroughly.
The solid was separated from the liquid through centrifugation (10 min at 4400 rpm), after which the solid was washed with methanol (3 x 30 mL) and acetone (1 x 30 mL). After drying the solid in a vacuum oven at 60 °C for one hour, it was subjected to Soxhlet extraction with THF for 20 hours. After that, the COF

S-4
was allowed to dry in a vacuum oven at 60 °C for 20 hours. The TAPB-NDA-COF was isolated as an ochrebrown, fluffy powder (610 mg, 87 %).

Post-synthetic processing
The COF powder was pelletized in batches of ~ 100 mg, using a pellet die with a diameter of 20 mm. For each batch, a hydraulic press was used to apply a pressure of 31 MPa for 60 seconds. After that, the pellets were crushed and then, through sieving, we collected COF pellets in the particle size range between 300 and 425 μm. These pellets were used without further treatment for characterization and breakthrough studies.

Isosteric enthalpy of adsorption,
The common Freundlich-Langmuir fit/Clausius-Clapeyron approach for the calculation of ΔH ads was executed according to the procedure described by Alexander Nuhnen and Christoph Janiak (DOI: 10.1039/D0DT01784A) and applied to all CO 2 adsorption isotherms. The corresponding Freundlich-Langmuir fitting parameters are presented in Table S1.

Water-COF FT-IR binding studies
Around 5 mg of pelletized COF was transferred to an open vial. This vial was placed in a 250 mL glass jar that also contained a vial of ~ 10 mL of salt solutions / suspensions. A relative humidity sensor was added to the system, which was subsequently sealed with Parafilm. The COFs were allowed to equilibrate for 16 hours at room temperature. Thereafter, they were immediately subjected to standard FT-IR analysis. The following RH values were monitored for the specific salt solutions: calcium chloride for 33 %, magnesium chloride for 38 %, sodium iodide for 42 %, sodium bromide for 55 %, magnesium bromide for 60 %, sodium chloride for 70 % and distilled water for 80 % relative humidity. The latter value is relatively low compared to other systems with distilled water, but this is likely a kinetic effect (i.e. insufficient equilibration time).
On the other hand, the lower RH values agree with literature values. RH values of ~ 0 % are obtained by directly retrieving the COF vial from a vacuum oven set at 130 °C. Those vials were directly sealed when retrieved from the oven and utilized directly for FT-IR analysis.

S-5
primarily consists of He, N 2 and CO 2 feed flows (regulated by mass flow controllers, Brooks 5850 series), water saturators and a packed bed column within which temperature and pressure can be regulated. The bed height of the COF (95 mg) in the stainless steel column (length 79 mm, inner diameter 4 mm) is 39 mm, and the bed is sandwiched between two layers of quartz wool. A four-way switching valve allows the operator to switch between N 2 (at 8 mL⋅min -1 ) / CO 2 (at 2 mL⋅min -1 ) mixture (feed 1) and He (feed 2) gas streams and back-pressure regulators ensure a constant pressure when switching. The second He flow (at 10 mL⋅min -1 ) is added after the column to prevent flow disruptions and securing a constant flow to the analysis instruments. The breakthrough response is monitored by a mass spectrometer (MS, LPM T100 Gas Analyzer) and a relative humidity sensor (PosiTector Dew Point Meter, DeFelsko). A complete overview of the setup is displayed in Figure S7.

Experiments
Prior to every experiment (set of 4 cycles per experimental parameter set), the system is flushed overnight (~ 16 h) with 10 mL⋅min -1 He and the oven is set to 130 °C. Then, the column is allowed to reach the specific temperature necessary for the experiment. An additional pre-treatment step is used for humid experiments, where the column is exposed to a humid He stream with a controlled RH value and temperature, and equilibrated overnight (~ 16 h). Typically, 60 minutes are used for the adsorption and desorption cycles unless stated otherwise. Dead volume measurements were performed by replacing the COF with non-adsorbent silicon carbide (350 μm) using a similar bed height.

Analysis of breakthrough data
MS signal to flow rate The conversion of MS signal to estimated flow rates from the exit are based on the following equations.

S-6
But, to have an expression for we have to use a correction factor: there is a constant flow going into ( ) the MS, but the total flow rate that exits the adsorption column changes over time (due to adsorption / desorption of different components). So, the time-dependent correction factor should correct for the ( ) fact that the total flow rate at steady state can be different than the total flow rate during adsorption: While eq. 4 can be written as three separate equations for all components, these three equations are not independent and thus cannot be solved in order to get an expression for that is defined by only known ( ) values. We also cannot assume to be constant (which would help solving the equations), as it is used eq. 5 The normalized corrected flow for component x is defined as , otherwise written as in the main ( ) ( F( ) 0 ( ) ) manuscript.

CO 2 breakthrough time and capacity
The CO 2 breakthrough time is classified as the breakthrough time difference between N 2 and CO 2 breakthrough. In this research, we identify the exact point of breakthrough when there is a significant difference in the slope of the curve (see Figure S11). CO 2 capacities were calculated from their breakthrough curves (time vs F(CO 2 )), by integrating the area above this curve. Areas were always integrated using:  the x-axis range: time t = 0 (defined as N 2 breakthrough) to t = 1200 seconds (chosen as here typically all curves have plateaued, see Figure S11), and  the y-axis range: F(CO 2 ) = 0 to F(CO 2 ) = 2.
Then, this area is subtracted by the integrated area above the CO 2 breakthrough curve from the silicon carbide experiments (dead volume). The dry-silicon carbide area was used for the dry COF experiments, S-7 and the humid-silicon carbide area for the humid COF experiments. The value of the final integrated area is divided by the weight of the COF in the column (95 mg) to obtain the CO 2 capacity in cm 3 ⋅g -1 (at STP) or mmol⋅g -1 (Table 1).

TGA measurements of COFs with pre-adsorbed water
The protocol for TGA measurements of COFs with pre-adsorbed water was adopted from Llewellyn and co-workers with minor changes (https://doi.org/10.1002/cssc.201601816). The pressure during these experiment is kept constant at atmospheric values. First, COFs were equilibrated in a closed vessel in the presence of water at 80 % RH. Then, ~ 7 -10 mg of this powder was transferred to the TGA crucible and submitted to the following program: 1. 100 mL⋅min -1 N 2 flow, 298 K, 60 minutes (equilibration, desorption of weakly bound water) 2. 100 mL⋅min -1 CO 2 flow, 298 K, 30 minutes (adsorption) 3. 100 mL⋅min -1 N 2 flow, 298 K, 30 minutes (desorption) 4. 100 mL⋅min -1 N 2 flow, heating to desired temperature at 10 K⋅min -1 (controlled water desorption) 5. 100 mL⋅min -1 N 2 flow, cooling to 298 K at 10 K⋅min -1 6. 100 mL⋅min -1 N 2 flow, 298 K, 15 minutes (ensure temperature stabilization) 7. 100 mL⋅min -1 CO 2 flow, 298 K, 30 minutes (adsorption) 8. 100 mL⋅min -1 N 2 flow, 298 K, 30 minutes (desorption) We used water desorption temperatures of 333, 363, 393 and 423 K at step 4. All weights from the TGA measurements were corrected with the weight of the empty crucible following the same protocol. After the four measurement cycles of step 1.-8., the weight of the COF with 0 mmol⋅g -1 H 2 O adsorbed was obtained by first drying the COF powder under 100 mL⋅min -1 N 2 flow at 423 K for three hours, cooling to 298 K and stabilizing at that temperature for 60 minutes, after which the monitored weight was taken as the dry-COF weight. Then, the dry COF was again subjected to a final CO 2 ad-and desorption cycle (replicate step 7 and 8). The supplementary data to the main figure (Figure 3) are presented in supplementary Figure S13.
The data were analysed as is listed in the supplementary protocol of Llewellyn and co-workers. Similar to the issues with their data analysis, we calculated 2 CO 2 and 2 H 2 O capacities (including and excluding the unaccounted weight increase, step 4 in Figure S13A) which were averaged for the final data.

S-8
Supplementary experimental data Figure S1. FT-IR spectra of TAPB-NDA-COF and its originating monomers TAPB and NDA. The right graph is a zoom-in of the left graph.   Table S1. Fitting parameters for the Freundlich-Langmuir fits applied to CO 2 adsorption isotherms. The equation used for the fit is B = (a*b*p^c)/(1+b*p^c), with p being the absolute pressure (kPa) and B the CO 2 uptake (mmol g -1 ).

S-16
Figure S12. Example CO 2 breakthrough curve (dry experiment, cycle 1) for illustration of how analysis was conducted. Left: light-red area indicates integrated area to determine capacity. Right: the last data point before the slope of the curve starts to significantly increase is taken as the point at breakthrough.

S-17
Figure S13. Top: Breakthrough conditions to which a single batch of TAPB-NDA-COF was exposed over time. Bottom: Normalized N 2 (8 mL/min) and CO 2 (2 mL/min) TAPB-NDA-COF breakthrough curves under dry conditions at 3.1 bar and 298 K, comparing t1 and t2 as indicated by the top figure. TGA measurements of COFs with pre-adsorbed water Temperatures of 333, 363, 393 and 423 K were chosen to step-wise desorb water and observe how the CO 2 capacity of the material changed as a result of that ( Figure S13A). This trend has been visualised in Figure   S13B. After all 4 runs had been executed, the COF powder was dried completely by flowing N 2 at 423 K for three hours. Then the powder was cooled to 298 K and stabilized, and its weight was taken as the weight at 0 mmol⋅g -1 H 2 O adsorbed. A final CO 2 ad-and desorption cycle was performed to get the CO 2 capacity of the dry powder. Figure S14. A) TGA curves of water-equilibrated TAPB-NDA-COF, using for every consecutive run the same steps: N 2 flow at 298 K (1), CO 2 flow at 298 K (2), N 2 flow at 298 K (3), heating to indicated temperature and cooling to 298 K under N 2 flow and both rates 10 K min -1 (4), CO 2 flow at 298 K (5), N 2 flow at 298 K (6). Flow rates of N 2 and CO 2 were 100 mL⋅min -1 . The pressure is kept constant at atmospheric pressure. B) Evolution of CO 2 uptake as a function of the amount of pre-adsorbed water of TAPB-NDA-COF. The line between the data points functions to guide the eye. S-19