UC Merced

Formulating and designing energy efficient polymer additives and base oil molecules is one way to reduce energy losses in mechanical systems. However, the first step to designing novel polymers and base oils with improved performance under a wider range of operating conditions is to decipher the effect of their structure and chemistry on the bulk properties of lubricants. Therefore, this research aimed to investigate the factors and mechanisms that influence the key bulk properties (density, viscosity, traction coefficient) of liquid lubricants at the molecular level using molecular dynamics (MD) simulations, machine learning (ML), and quantitative-structure-property-relationship (QSPR) modeling.

In the first study, the prospect of improving the mechanical efficiency (ME) of hydraulic systems by formulating fluids with viscosity modifiers (VMs) was tested in a pump dynamometer. Lower viscosity fluids provided better ME but decreasing the viscosity of base oil by adding VM did not have the same effect. Simulations showed that viscosity was directly correlated to the elongation of the polymers under shear, which, together with calculations of the key shear rate range in a pump, suggested ways of designing VMs to achieve a specific viscosity profile that maximizes ME. The second study presented a model for predicting the critical shear rate, intending to identify a fluid that shear thins in the key shear rate range in a pump. The model was applied to predict the properties of fluids formulated with VMs and validated by comparison to viscosities obtained from experimental measurements and MD simulations across many decades of shear rates. Results demonstrated that polymer molecular weight plays an important role in determining the critical shear rate, whereas polymer concentration primarily affects the Newtonian viscosity. The simulations showed the molecular origins of shear thinning and critical shear rate.

In the third study, MD simulations were used to identify the QSPR of polymer-enhanced lubricants having commercial grades chemistries. The molecular origins of differences in the viscosity index, thickening efficiency, and traction coefficient between the fluids were investigated by calculating multiple structural properties of the polymers in the simulations. In the fourth study, a python package called PyL3dMD was developed, which enables users to compute nearly 2000 dynamic molecular descriptors by post-processing MD simulation trajectories. This was then used in our fifth study to relate 3D conformations of 305 complex hydrocarbons to their temperature-dependent density and viscosity and, as a result, developed advanced ML-based QSPR models. The models predicted density and dynamic viscosity with the coefficient of determination values of 99.6% and 97.7%, respectively, for all data sets, including a test data set of 45 molecules.

Overall, this dissertation demonstrated the viability of various techniques in understanding molecular interactions and facilitating novel and innovative designs of polymeric additives and base oils for various applications with improved performance.