Options
Modelling and simulation of the production process of thin film composite membranes for gas separation
Citation Link: https://doi.org/10.15480/882.17313
Publikationstyp
Doctoral Thesis
Date Issued
2026
Sprache
English
Author(s)
Advisor
Referee
Brinkmann, Thorsten
Title Granting Institution
Technische Universität Hamburg
Place of Title Granting Institution
Hamburg
Examination Date
2026-04-23
Institute
TORE-DOI
Citation
Technische Universität Hamburg (2026)
Despite the growing attention for membrane-based processes, especially those targeted at the reduction of greenhouse gas emissions, the transfer of a membrane production process from laboratory scale to the large scale is still a challenge. Many production steps involve manual fine-tuning and a large number of experiments to find the appropriate process parameters for the desired coating result. Adding to the complexity of the problem, increasing awareness is raised to sustainability of the fabrication process and materials utilized for membranes. The anticipated increase in restrictions on the use of toxic or environmentally harmful solvents fuels the search for greener alternatives. This implies a growing demand for a better understanding of how chemical and physical properties of the used polymers and solvents interact in the fabrication process of membranes, and how this can be modelled and predicted to save time and resources.
In this thesis, the roll-to-roll coating process in the large-scale fabrication of thin film composite membranes is modelled and simulated by means of computational fluid dynamics. The interface of the free-surface flow is captured using a moving mesh approach and the evaporation of solvent and its effect on the polymer concentration, and thus viscosity and density of the solution, is incorporated. To reduce the computational effort, the model was set up in two dimensions only.
For the model validation, several measurement techniques were employed. The final polymer thickness on the membrane is determined via electron microscopy images and also derived from mass-loss measurements of the polymer solution feed during coating experiments. Furthermore, the liquid film thickness on the applicator roll of the coating device is measured via a confocal laser displacement sensor and compared to the model predictions. As a first step, the predictive quality of the model was evaluated for the polydimethylsiloxane (PDMS) based coating of the protective top layer of the membranes and demonstrated the principal validity of the approach. The final polymer thickness could be predicted with an error of approximately 20% compared to the empirical mean value for several coating conditions.
Subsequently, the same model was used to predict coating thicknesses of PolyActive™ solution, a multiblock copolymer that is used for the selective layer of membranes that are designated for CO2 separation tasks. The polymer is dissolved in a mixture of three solvents, which are concealed for confidential reasons and will be abbreviated as solvent mixture component 1-3 (SMC 1-3). This marks another difference to the PDMS coating which is done in one solvent (isooctane).
In contrast to the PDMS homopolymer, the coating thickness on the membrane was found to deviate strongly from model predictions, showing an approximately four-times thicker polymer layer than expected. Two main changes in the model were done that might account for this, namely the inclusion of viscoelastic effects of the polymer and the Marangoni effects caused by the inhomogeneous distribution of the polymer at the interface, which is thought to act like a surfactant that induces surface tension gradients and therefore secondary flow patterns. Shear rate and oscillatory viscosity measurements of PolyActive™ solution were performed and, together with simulation results and data from literature, suggest that viscoelasticity is not likely to be the cause of the observed thickening effect. In contrast, the implemented Marangoni effects were able to reproduce the observed thickening effect in large parts. As literature data on similar block copolymers demonstrate the amphiphilic character and surface activity of such polymers, this explanation seems more plausible, although several uncertainties concerning numerical robustness and model parameters remain to be addressed in future research. In addition, the coating of PolyActive™ was repeated with pure SMC 3 as the solvent. Here, the thickening effect was even stronger and could not be explained via simulations so far, even though some results hint at thermal effects that might play a role due to the high volatility of SMC 3. Nevertheless, the implemented computational fluid dynamics (CFD) model delivers valuable insights into the complexities of this particular coating process and highlights the sensitivity of the system to fluid-mechanical force gradients on the micro-scale.
In this thesis, the roll-to-roll coating process in the large-scale fabrication of thin film composite membranes is modelled and simulated by means of computational fluid dynamics. The interface of the free-surface flow is captured using a moving mesh approach and the evaporation of solvent and its effect on the polymer concentration, and thus viscosity and density of the solution, is incorporated. To reduce the computational effort, the model was set up in two dimensions only.
For the model validation, several measurement techniques were employed. The final polymer thickness on the membrane is determined via electron microscopy images and also derived from mass-loss measurements of the polymer solution feed during coating experiments. Furthermore, the liquid film thickness on the applicator roll of the coating device is measured via a confocal laser displacement sensor and compared to the model predictions. As a first step, the predictive quality of the model was evaluated for the polydimethylsiloxane (PDMS) based coating of the protective top layer of the membranes and demonstrated the principal validity of the approach. The final polymer thickness could be predicted with an error of approximately 20% compared to the empirical mean value for several coating conditions.
Subsequently, the same model was used to predict coating thicknesses of PolyActive™ solution, a multiblock copolymer that is used for the selective layer of membranes that are designated for CO2 separation tasks. The polymer is dissolved in a mixture of three solvents, which are concealed for confidential reasons and will be abbreviated as solvent mixture component 1-3 (SMC 1-3). This marks another difference to the PDMS coating which is done in one solvent (isooctane).
In contrast to the PDMS homopolymer, the coating thickness on the membrane was found to deviate strongly from model predictions, showing an approximately four-times thicker polymer layer than expected. Two main changes in the model were done that might account for this, namely the inclusion of viscoelastic effects of the polymer and the Marangoni effects caused by the inhomogeneous distribution of the polymer at the interface, which is thought to act like a surfactant that induces surface tension gradients and therefore secondary flow patterns. Shear rate and oscillatory viscosity measurements of PolyActive™ solution were performed and, together with simulation results and data from literature, suggest that viscoelasticity is not likely to be the cause of the observed thickening effect. In contrast, the implemented Marangoni effects were able to reproduce the observed thickening effect in large parts. As literature data on similar block copolymers demonstrate the amphiphilic character and surface activity of such polymers, this explanation seems more plausible, although several uncertainties concerning numerical robustness and model parameters remain to be addressed in future research. In addition, the coating of PolyActive™ was repeated with pure SMC 3 as the solvent. Here, the thickening effect was even stronger and could not be explained via simulations so far, even though some results hint at thermal effects that might play a role due to the high volatility of SMC 3. Nevertheless, the implemented computational fluid dynamics (CFD) model delivers valuable insights into the complexities of this particular coating process and highlights the sensitivity of the system to fluid-mechanical force gradients on the micro-scale.
Subjects
TFC membranes
Polyactive
Computational fluid dynamics
COMSOL
Roll-to-roll Coating
DDC Class
660.2: Chemical Engineering
Loading...
Name
Dissertation_Florian_Brennecke.pdf
Size
9.54 MB
Format
Adobe PDF