Below is a very abridged list of high-level research topics carried out in the MSE group. If you are interested in discussing any of these in more detail, please contact any of the team investigators and we will be happy to discuss them.
Integrated Molecular & Process Design
This work focuses on developing an integrated molecular and process synthesis methodology which involves the simultaneous solution of optimal molecular and process variables for a given system. We have investigated the design of fluid-fluid separation systems at steady state, where the molecules (materials) to be designed are pure component solvents that act as mass separating agents in the process.
We have proposed a deterministic outer approximation (OA) algorithm for the solution of the integrated molecular and process design flowsheet, which is a challenging mixed integer nonlinear programming (MINLP) problem. Novel tests are embedded within the MINLP solution framework which eliminate infeasible regions of the molecular and process domain (Fig. 3). The algorithm is applied to separation processes, where the separations of carbon dioxide from methane and of butanol from water are explored. Overall, the proposed framework avert evaluations of infeasible primal problems and enhance convergence to locally optimal solutions of challenging integrated material and process design and synthesis problems. This work has been published in Burger et al. (2015) and Gopinath et al. (2016).
This work focuses on developing an integrated molecular and process synthesis methodology which involves the simultaneous solution of optimal molecular and process variables for a given system. We have investigated the design of fluid-fluid separation systems at steady state, where the molecules (materials) to be designed are pure component solvents that act as mass separating agents in the process.
We have proposed a deterministic outer approximation (OA) algorithm for the solution of the integrated molecular and process design flowsheet, which is a challenging mixed integer nonlinear programming (MINLP) problem. Novel tests are embedded within the MINLP solution framework which eliminate infeasible regions of the molecular and process domain (Fig. 3). The algorithm is applied to separation processes, where the separations of carbon dioxide from methane and of butanol from water are explored. Overall, the proposed framework avert evaluations of infeasible primal problems and enhance convergence to locally optimal solutions of challenging integrated material and process design and synthesis problems. This work has been published in Burger et al. (2015) and Gopinath et al. (2016).
|
|
Molecular dynamics of transport of fluids through porous media
This research looks into the study of molecular-level flow in nano-confined spaces. Applications include the design of membranes for desalinization of water, the flow of gases and crude oils through reservoirs and the efficient separation of gases. |
Accelerating chemical reactions by computer-aided molecular design
What is the best solvent for a given chemical reaction? Given that the rate and selectivity of chemical reactions can vary by several orders of magnitude in different solvents, this question has important ramifications for the exploration of novel reaction routes and the development of industrial processes. When investigating new liquid phase reactions, it is essential to find a “good” solvent to avoid dismissing a valuable chemistry because of a solvent which suppresses the desired reaction. At the process development level, the problem of solvent choice is further compounded by the numerous safety, environmental and process constraints that the solvent must satisfy. Despite the complexity inherent in solvent selection for reactions, few tools exist to support this decision and researchers are often left to choose on the basis of their intuition and/or extensive and costly experimental investigations. As a result, the improved understanding of liquid phase reactions and the development of solvent selection techniques have recently been highlighted as key priority areas by the ACS Green Chemistry roundtable.
In this series of projects, we have been developing a methodology for optimal solvent design for enhanced reaction kinetics, QM-CAMD, which relies on the integration of continuum solvation quantum mechanical calculations into a computer aided molecular design (CAMD) framework. This approach allows the exploration of a solvent design space consisting of thousands of potential molecules and leads to a shortlist of promising solvents that can then be assessed experimentally. To manage the computational cost, a surrogate model for the quantum mechanical (QM) calculations is built and improved iteratively. As a result, only a small number of QM calculations need to be performed during the course of the QM-CAMD algorithm. This approach has been successfully applied to the SN2 reaction of phenacyl bromide and pyridine, leading to a 40% increase in the reaction rate. The results have been verified experimentally, using in-situ kinetic monitoring techniques.
This Nature Chemistry paper gives more details on the topic.
What is the best solvent for a given chemical reaction? Given that the rate and selectivity of chemical reactions can vary by several orders of magnitude in different solvents, this question has important ramifications for the exploration of novel reaction routes and the development of industrial processes. When investigating new liquid phase reactions, it is essential to find a “good” solvent to avoid dismissing a valuable chemistry because of a solvent which suppresses the desired reaction. At the process development level, the problem of solvent choice is further compounded by the numerous safety, environmental and process constraints that the solvent must satisfy. Despite the complexity inherent in solvent selection for reactions, few tools exist to support this decision and researchers are often left to choose on the basis of their intuition and/or extensive and costly experimental investigations. As a result, the improved understanding of liquid phase reactions and the development of solvent selection techniques have recently been highlighted as key priority areas by the ACS Green Chemistry roundtable.
In this series of projects, we have been developing a methodology for optimal solvent design for enhanced reaction kinetics, QM-CAMD, which relies on the integration of continuum solvation quantum mechanical calculations into a computer aided molecular design (CAMD) framework. This approach allows the exploration of a solvent design space consisting of thousands of potential molecules and leads to a shortlist of promising solvents that can then be assessed experimentally. To manage the computational cost, a surrogate model for the quantum mechanical (QM) calculations is built and improved iteratively. As a result, only a small number of QM calculations need to be performed during the course of the QM-CAMD algorithm. This approach has been successfully applied to the SN2 reaction of phenacyl bromide and pyridine, leading to a 40% increase in the reaction rate. The results have been verified experimentally, using in-situ kinetic monitoring techniques.
This Nature Chemistry paper gives more details on the topic.
SAFT-g force fields for simulation of molecular fluids
Detachment of a micelle of an anionic surfactant (C10E4) from a water-air interface. Coarse grain modelling using SAFT force fields
Many interesting and important phenomena in soft matter, such as self-assembly of large complex molecules, protein folding or colloidal aggregation, are typically observed in the mesoscale regime. The spatial and time scales involved in these processes are very large, making atomistic simulations very challenging. The growing area of coarse-graining (CG) methods has made possible the use of conventional molecular simulation techniques to study large systems in a reasonable computing time. In generic CG methodologies molecules are described as being formed by segments of bundles of matter (super-atoms) interacting via effective CG potentials. This approach clearly involves a loss in resolution of the description of the system, hence it is important to have an appropriate methodology that preserves a good overall description of the key target properties that one wants to represent after the coarse-graining. We have introduced a new coarse-graining methodology, where the statistical associating fluid theory (SAFT) is used as a link between the experimental fluid phase equilibria data and novel CG force fields based on the Mie potential, thereby facilitating parametrization. For more details and publications see here.
Systematic methodologies for crystal structure predictions
Methodologies for the systematic prediction of the polymorphs of organic molecules solely from the knowledge of the molecular connectivity diagram have undergone rapid improvements in the last few years. Progress has been achieved by combining better models of the different forces at play (e.g., electrostatics, dispersion) with efficient and reliable numerical techniques. In our most recent works, we have significantly extended the range of flexibility that can be handled, by developing local approximate models that allow us to achieve quantum mechanical accuracy at a fraction of the cost. Using an approach based on a global search stage, with our CrystalPredictor code, followed by a more accurate calculation of the lattice energy with our CrystalOptimizer code, we have successfully found experimental structures for large and flexible molecules such as molecule XX of the latest Blind Test (Fig. 1) and a pharmaceutical compound provided by BMS (Fig. 2). We have also been able to identify all seven known polymorphs of ROY, the molecule with the largest number known (anhydrate) polymorphs.
Methodologies for the systematic prediction of the polymorphs of organic molecules solely from the knowledge of the molecular connectivity diagram have undergone rapid improvements in the last few years. Progress has been achieved by combining better models of the different forces at play (e.g., electrostatics, dispersion) with efficient and reliable numerical techniques. In our most recent works, we have significantly extended the range of flexibility that can be handled, by developing local approximate models that allow us to achieve quantum mechanical accuracy at a fraction of the cost. Using an approach based on a global search stage, with our CrystalPredictor code, followed by a more accurate calculation of the lattice energy with our CrystalOptimizer code, we have successfully found experimental structures for large and flexible molecules such as molecule XX of the latest Blind Test (Fig. 1) and a pharmaceutical compound provided by BMS (Fig. 2). We have also been able to identify all seven known polymorphs of ROY, the molecule with the largest number known (anhydrate) polymorphs.
Molecular simulations of asphaltene aggregation
1. Model development. Asphaltenes are a class of molecules in crude oil linked to aggregation and deposition phenomena, which cause serious operational problems in the oil industry. Asphaltenes are defined in terms of solubility as soluble in aromatics (like toluene) and insoluble in alkanes (like heptane). We have developed coarse-grained (CG) versions of model asphaltenes to study the aggregation problem at a molecular level using molecular dynamics simulations. The CG molecule is built in a group-contribution fashion with interaction parameters obtained through the SAFT-g methodology. To validate the proposed CG model, we used fully atomistic simulations of asphaltenes in good (toluene) and bad (heptane) solvents as a benchmark. In the figure, we show the results for simulation boxes with 27 asphaltenes in explicit solvent. The plot corresponds to the cluster-size distributions for every system. In toluene, clusters of no more than 20 molecules appear, while in heptane, the aggregate sizes are limited by the simulation box.
2. Asphaltene aggregation. The validated asphaltene model was further studied using simulation boxes with up to 2000 asphaltene molecules in explicit solvent (toluene + heptane mixtures). In the figure, we show a schematic representation of the trends observed in the cluster-size distributions as the temperature is lowered, the asphaltene concentration is incremented, or the proportion of heptane is increased in the solvent. At good solubility conditions, asphaltenes are predominantly organised as monomers, dimers, and small clusters. Unstable asphaltene systems, however, present a behaviour consistent with a phase separation case, where an asphaltene-rich region coexists with a distribution of small clusters. In some intermediate cases, we observed distributions characterised by a shoulder, which can be the indication of phase-separated systems frustrated by too small simulation boxes. The second reason for such frustration is the slow kinetics of the aggregates that affects not only simulations but also experimental results in real systems and might be the cause of the different interpretations in aggregation observations.
More details and results from asphaltene atomistic simulations in:
Headen TF, Boek ES, Jackson G, Totton TS, Müller EA.”Simulation of Asphaltene Aggregation through Molecular Dynamics: Insights and Limitations.” Energy & Fuels, 2017, 31, pp 1108-1125. doi: 10.1021/acs.energyfuels.6b02161
CG model development, simulation details and more results in:
Jiménez-Serratos G, Totton TS, Jackson G, Müller EA. “Aggregation Behaviour of Model Asphaltenes Revealed from Large-Scale Coarse-Grained Molecular Simulations.” Journal Of Physical Chemistry B, 2019, 123, pp 2380-2396. doi: 10.1021/acs.jpcb.8b12295
D. M. Kaimaki, et al. “Multiscale Approach Linking Self-Aggregation and Surface Interactions of Synthesized Foulants to Fouling Mitigation Strategies,” Energy Fuels, 33, 7216–7224, 2019.
1. Model development. Asphaltenes are a class of molecules in crude oil linked to aggregation and deposition phenomena, which cause serious operational problems in the oil industry. Asphaltenes are defined in terms of solubility as soluble in aromatics (like toluene) and insoluble in alkanes (like heptane). We have developed coarse-grained (CG) versions of model asphaltenes to study the aggregation problem at a molecular level using molecular dynamics simulations. The CG molecule is built in a group-contribution fashion with interaction parameters obtained through the SAFT-g methodology. To validate the proposed CG model, we used fully atomistic simulations of asphaltenes in good (toluene) and bad (heptane) solvents as a benchmark. In the figure, we show the results for simulation boxes with 27 asphaltenes in explicit solvent. The plot corresponds to the cluster-size distributions for every system. In toluene, clusters of no more than 20 molecules appear, while in heptane, the aggregate sizes are limited by the simulation box.
2. Asphaltene aggregation. The validated asphaltene model was further studied using simulation boxes with up to 2000 asphaltene molecules in explicit solvent (toluene + heptane mixtures). In the figure, we show a schematic representation of the trends observed in the cluster-size distributions as the temperature is lowered, the asphaltene concentration is incremented, or the proportion of heptane is increased in the solvent. At good solubility conditions, asphaltenes are predominantly organised as monomers, dimers, and small clusters. Unstable asphaltene systems, however, present a behaviour consistent with a phase separation case, where an asphaltene-rich region coexists with a distribution of small clusters. In some intermediate cases, we observed distributions characterised by a shoulder, which can be the indication of phase-separated systems frustrated by too small simulation boxes. The second reason for such frustration is the slow kinetics of the aggregates that affects not only simulations but also experimental results in real systems and might be the cause of the different interpretations in aggregation observations.
More details and results from asphaltene atomistic simulations in:
Headen TF, Boek ES, Jackson G, Totton TS, Müller EA.”Simulation of Asphaltene Aggregation through Molecular Dynamics: Insights and Limitations.” Energy & Fuels, 2017, 31, pp 1108-1125. doi: 10.1021/acs.energyfuels.6b02161
CG model development, simulation details and more results in:
Jiménez-Serratos G, Totton TS, Jackson G, Müller EA. “Aggregation Behaviour of Model Asphaltenes Revealed from Large-Scale Coarse-Grained Molecular Simulations.” Journal Of Physical Chemistry B, 2019, 123, pp 2380-2396. doi: 10.1021/acs.jpcb.8b12295
D. M. Kaimaki, et al. “Multiscale Approach Linking Self-Aggregation and Surface Interactions of Synthesized Foulants to Fouling Mitigation Strategies,” Energy Fuels, 33, 7216–7224, 2019.