Experimental studies and modeling of solute-solvent interactions

Abstract

Enamik praktilist tähtsust omavatest keemilistest protsessidest toimub vedelikes – mitte ainult tööstuslik süntees ja laborikeemia, vaid ka bioloogilised protsessid nagu rakkude hingamine toimuvad molekulaarsel tasemel keerulise koostisega lahustes. Neist protsessidest arusaamine ja molekulide käitumise ennustamine lahustes on tähtis arvukate uurimisvaldkondade jaoks, meditsiinist ja farmakoloogiast naftakeemiani. Kahjuks on ainete omaduste ennustamine vedelikes arvutuskeemia jaoks üks keerulisemaid ülesandeid. Käesolevas töös hinnati olemasolevate arvutusmetoodikate sobivust vesiniksideme tekke kirjeldamiseks orgaanilistes lahustites ning molekulide jaotuse kirjeldamiseks kahe vedeliku vahel (sisuliselt vedelik-vedelik ekstraktsiooni modelleerimiseks). Peamine kasutatud arvutusmeetod oli COSMO-RS (Conductor-like Screening Model for Real Solvents), valitud oma erakordse sobivuse tõttu kontsentreeritud ja mitmekomponendiliste lahuste omaduste ennustamiseks ja molekulaardisainiks. Töö käigus leiti, et vesiniksidemed neutraalsete molekulide vahel on kirjeldatavad suhteliselt hästi, kuid vaadeldud arvutusmetoodikad pole piisavalt täpsed negatiivselt laetud vesiniksidemega komplekside modelleerimiseks. Vedelik-vedelik ekstraktsiooni tulemuste ennustamine COSMO-RS meetodiga oli üldjuhul edukas. Saadud tulemustele (nii lõpp- kui vahepealsetele parameetritele) saadi täpsuse hinanngud. Peale selle arendati uus metodoloogia tundmatute ühendite jaotuse ennustamiseks kahe mitteseguneva vedeliku vahel ilma vajaduseta ühendeid identifitseerida. See lihtsustab parima lahusti valikut ainete isoleerimiseks või puhastamiseks, vähendades töö- ja kemikaalide kulu ning jäätmete kogust.

The majority of practically relevant chemical processes occur in liquids. Those are not limited to industrial synthesis and laboratory chemistry – biological processes such as cellular respiration on molecular level take place in complex solutions. Understanding and being able to predict the behaviour of molecules in solutions is essential for numerous branches of science, ranging from medicine and pharmacology to petroleum chemistry. However, predicting the behavior of chemical compounds in liquids, especially in many-component solutions, is one of the most challenging tasks for computational chemistry. In this work existing computational methodologies were tested for suitability for describing hydrogen bond formation in organic solvents and distribution of organic compounds between liquids (essentially modeling of liquid-liquid extraction). The main computational method in this work is COSMO-RS (Conductor-like Screening Model for Real Solvents), chosen for its unequalled ability to predict properties of concentrated and multicomponent solutions and usability in molecular design. It was found that properties of hydrogen bonds between uncharged molecules can be predicted relatively well, but the tested computational approaches were not accurate enough for description of hydrogen bonds involving negatively charged ions. Modeling of liquid-liquid extraction using COSMO-RS was generally successful. Accuracy of the predictions and intermediate parameters was evaluated and problematic cases identified and discussed. Also, a new methodology was developed for predicting the distribution of unknown compounds between immiscible solutions without need for compound identification. It allows simplifying the solvent selection for compound isolation or purification, reducing the workload, expenses and waste amount.