A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Real solutions deviate from Raoult's law due to molecular interactions; activity coefficients γᵢ correct chemical potentials and equilibrium constants. Models like Regular Solution theory (symmetric) and NRTL/UNIQUAC (asymmetric, molecular-scale) predict γ from intermolecular forces, without quantum calculation. Accurate γ values are essential for phase equilibrium calculations, solubility predictions, and industrial separation processes.
From solution thermodynamics, you know that an ideal solution obeys Raoult's law: the vapor pressure of each component is proportional to its mole fraction, P_i = x_i P_i*. This works when the molecules of different components interact with each other in exactly the same way they interact with themselves. Real molecules are not so accommodating. Ethanol and water, for instance, form strong hydrogen bonds with each other that differ from the ethanol-ethanol and water-water interactions. The result is that measured vapor pressures, boiling points, and solubilities deviate — sometimes dramatically — from ideal predictions. The activity coefficient γᵢ quantifies this deviation by replacing x_i with the effective concentration a_i = γᵢx_i in thermodynamic expressions.
When γᵢ > 1, the component behaves as though it were more concentrated than its mole fraction suggests — it "wants to escape" the solution more than an ideal model predicts (positive deviation from Raoult's law). This typically occurs when unlike molecules interact more weakly than like molecules, as in a benzene-ethanol mixture. When γᵢ < 1, the component is stabilized in solution by favorable interactions with the other species (negative deviation), as seen in chloroform-acetone mixtures where a weak hydrogen bond forms between the components. The key insight is that γ encodes the net energetic and entropic consequences of non-ideal molecular mixing into a single multiplicative correction factor.
Regular Solution theory, the simplest predictive model, assumes that excess entropy of mixing is zero — the non-ideality comes entirely from enthalpy. It introduces a single interaction parameter that captures the energy difference between unlike and like molecular contacts. This gives symmetric activity coefficient curves (γ₁ and γ₂ have the same functional form), which works reasonably well for mixtures of nonpolar molecules of similar size, such as benzene and cyclohexane. However, it fails for systems where molecular size differences or specific interactions (like hydrogen bonding) create asymmetric behavior.
For more complex systems, NRTL (Non-Random Two-Liquid) and UNIQUAC (Universal Quasi-Chemical) models account for the fact that molecules do not mix randomly — local compositions around a given molecule differ from the bulk composition because molecules preferentially surround themselves with energetically favorable neighbors. NRTL uses binary interaction parameters fit to experimental vapor-liquid equilibrium data and handles strongly non-ideal systems including partially miscible liquids. UNIQUAC adds a combinatorial term for molecular size and shape differences, making it effective for mixtures of molecules with very different geometries. These models are the workhorses of chemical engineering: distillation column design, liquid-liquid extraction, and crystallization processes all depend on accurate activity coefficient predictions to determine where phases separate, what compositions coexist at equilibrium, and how many theoretical stages a separation requires.