Understanding How Synthetic Impurities Affect Glyphosate Solubility and Crystal Growth Using Free Energy Calculations and Molecular Dynamics Simulations

TL;DR

This study addresses how synthesis-derived glycine impurities modulate glyphosate crystallization by combining Direct Coexistence MD simulations, free energy calculations, and targeted experiments. The authors show that glycine adsorbs at crystal surfaces, transiently coating interfaces and hindering glyphosate incorporation, while simultaneously lowering glyphosate’s solvation free energy and increasing its solution solubility, yielding a reduced thermodynamic driving force for crystallization. The results are robust across crystal faces and supported by experimental data showing faster dissolution and slower nucleation/growth in glycine-containing systems. Together, the work provides a mechanistic, molecular-level framework for impurity effects in agrochemical crystallization and demonstrates a transferable computational–experimental workflow for optimizing industrial crystallization processes.

Abstract

Glyphosate, the most widely used herbicide worldwide, crystallizes through complex intermolecular interactions that are strongly influenced by synthesis-derived impurities. Understanding this process at the molecular scale is critical for optimizing production, ensuring product quality, and assessing environmental impact. Here, we employ direct coexistence molecular dynamics simulations and free energy calculations to elucidate how glycine-a prevalent synthesis byproduct-modulates glyphosate solubility and crystal growth in aqueous solutions. Our simulations identify two major mechanisms by which glycine hinders crystallization. First, direct coexistence simulations show that glycine preferentially adsorbs at crystal surfaces, hindering glyphosate attachment and slowing growth. Second, free energy calculations demonstrate that glycine enhances glyphosate solubility, reducing the supersaturation driving force to incorporate into the crystal phase. Experimental measurements corroborate our predictions, confirming both enhanced solubility and reduced crystallization kinetics in glycine-bearing systems. These findings establish that glycine-typically considered an inert impurity-actively disrupts glyphosate crystallization by promoting its dissolution. More broadly, this integrated computational-experimental approach highlights the power of molecular simulations to disentangle impurity effects, interfacial phenomena, and solution thermodynamics in crystallization, providing molecular-level insights for optimizing industrial protocols and predicting agrochemical behavior under relevant environmental conditions.

Figures (11)

• Figure 1: a) Left: Snapshot of the glyphosate bulk crystal structure knuuttila1979crystal from our simulations. Right: Time-evolution of the crystal density from $NpT$ simulations at 300K compared to the experimentally determined value from Ref. wilson2023discerning (dashed red horizontal line). b) Top: Representative snapshot from a DC simulation where water molecules have been rendered semi-transparent to better visualize the glyphosate molecules in the bulk. Bottom: Density profile along the longest box direction of one of our systems ($C_0 = 0.069 \text{m}$) without vacancies upon reaching equilibrium. The density profiles have been separated into different species as indicated in the legend. The region that we define as the bulk solution is in between the two dashed orange vertical lines. c) Time-evolution of the glyphosate average concentration in the solution for different systems, each with a different initial glyphosate concentration in the bulk liquid phase. The values were obtained by calculating the converged average value of the concentration at all times $<n$ time. The average value among all the different trajectories (i.e., solubility limit, $m$) is indicated with a dashed horizontal line. The inset shows a zoomed-in view of the Y-axis.
• Figure 2: a) Snapshots of the different crystallographic planes studied in a 4 x 4 x 4 unit cell. b) Convergence of the molality of two independent DC simulations with different crystal orientations and different initial glyphosate concentrations as a function of time. The values were obtained by calculating the arithmetic accumulated average value of the solubility at all times $<n$ time. The average value of both trajectories is depicted by an horizontal dashed line. c) Bar plot of the obtained solubility of the two different crystallographic faces. The whisker represents the uncertainty bounds.
• Figure 3: a) Top: Snapshot from a DC simulation where water and glyphosate molecules have been rendered semi-transparent to better visualize the glycine molecules in the bulk and at the crystal surface. Bottom: Density profile of a DC simulation along the longest box direction, separated into its different components as indicated in the legend. The inset presents a zoom-in insight of the density profile showing the glycine density peaks at the crystal surface. b) Glyphosate concentration vs. time for two different systems, one containing 2%wt glycine (shown in red), and another in absence of glycine (shown in blue). The values were obtained by calculating the accumulated average value of the glyphosate concentration at times $<n$ time. The convergence value (solubility limit, $m$) is indicated with a dashed horizontal line (shown in violet and black in presence vs. absence of glycine, respectively).
• Figure 4: a) Bar plot of the obtained solubility in the different studied systems. The experimental value is shown in red gly_sol. Simulations without impurities for different crystal planes are shown in blue, whereas simulations where glycine is present as an impurity at different concentrations are shown in green. b) Solvation free energy as a function of glycine weight percentage from free energy calculations. We observe that increasing glycine content leads to a further decrease in solvation free energy.
• Figure 5: a) Solubility temperatures (measured as clear points) as a function of glycine percentage in weight. Experiments are performed at constant concentration of glyphosate in acidic water for increasing concentrations of glycine. Our data refer to the mean values of triplicates, with the error bars indicating the associated standard deviation. b) Empirical cumulative distribution functions (ECDFs) of crystallization times obtained from the measured detection times at constant glyphosate supersaturation ($S=2$), and varying concentrations of glycine as indicated in the legend. The experiments are performed at constant temperature, $T=293K$. c) Growth times obtained from the detection time analysis against the glycine weight percentage in the solution. A second-order fit, shown as a dashed red line, is included as a guide for the eye. The experimental uncertainty of these measurements is of the order of the symbol size.