Reliable prediction of pore size distribution for nano-sized adsorbents with minimum information requirements

Direct estimation of pore size distribution (PSD) for nano-structured adsorbents suffers from a number of heavy shortcomings. On the other hand, conventional regularization techniques also require profoundly detailed and accurate information about the proper local adsorption isotherm (kernel) to provide reliable PSDs. Selection of improper kernel or use of inexact values for isotherm parameters can lead to serious errors in PSD profiles.Two new PSD determination techniques are presented in this article for efficient PSD recovery from mere condensation data or condensation isotherms in the presence of adsorption. The second method extends the first approach by considering a pre-adsorbed layer prior to condensation. Both procedures use entirely different linear regularization technique compared to conventional regularization methods. The new proposed techniques do not require any a priori information about the shape of local adsorption isotherm. They recruit minimum number of physical data or assumptions to produce reliable performances for successful PSD recoveries.Several comparative synthetic and real case studies are employed to illustrate the promising performance of the newly proposed Techniques. The leave one out cross validation (LOOCV) criterion and the generalized singular value decomposition (GSVD) technique are crucial for fast, efficient and reliable estimation of optimal regularization level (λ*). Simultaneous manual verification is also recommended to ensure best possible performance. Although various heuristics are considered for estimation of adsorbed film thickness, however, the proposed method provides almost same result for all different heuristics.