Machine learning guided design of covalent organic frameworks for CO2
capture in wet flue gas
Abstract
Discovery of remarkable porous materials for CO2 capture from wet flue
gas is of great significance to reduce the CO2 emissions, but
elucidating the most critical structure features for boosting CO2
capture capabilities remains a great challenge. Here, machine learning
assisted computational screening on 516 experimental covalent organic
frameworks (COFs) identify the superior secondary building units (SBUs)
for wet flue gas separation, which are tetraphenylporphyrin unit in
sql-type COFs and functional groups. Accordingly, 1233 COFs are
assembled using the superior SBUs. Combined with DFT calculations, the
“electron-donating induced vdW interaction” effect is discovered to
design better-performing COFs with superior CO2 uptake, which can
achieve 4.4 mmol·g-1 with CO2/N2 selectivity of 104.8; while the
“electron-withdrawing induced vdW+electrostatic coupling interaction”
effect is proposed to design better-performing COFs with superior CO2/N2
selectivity, which can arrive 277.6 with CO2 uptake of 2.2 mmol·g-1, in
this case, H2O contributes to improving the CO2/N2 selectivity.