Classical Nucleation Models for Drug Precipitation

Why a nucleation barrier matters

A solution can be supersaturated without precipitating immediately. Supersaturation provides a thermodynamic driving force for forming a solid phase, but a new crystal also creates an interface with the surrounding solution. Creating that interface costs free energy. Classical nucleation theory (CNT) describes the competition between this surface cost and the bulk free-energy gain, and uses the resulting barrier to predict a nucleation rate.

This tutorial focuses on the formation of a critical nucleus, not on the subsequent enlargement of crystals. The distinction matters: the nucleation rate predicts how often nuclei form under specified conditions, whereas an experimental induction time can also include the time needed for a nucleus to become detectable. The introductory material on precipitation induction time and metastable zones establishes that induction can depend on both nucleation and observation; here, the aim is to derive the classical rate dependence and understand its sensitivity.

Building the classical free-energy barrier

CNT makes a simplified, spherical-nucleus model. It assumes a uniform solution with fixed \(S\) and \(T\), a spherical nucleus with the bulk solid’s molecular volume, and a size-independent, isotropic interfacial energy \(\gamma\). It also treats formation of a nucleus as a thermodynamic fluctuation and neglects molecular-scale structural detail. For a homogeneous nucleus, no foreign surface assists its formation.

Under the ideal activity relation, the chemical-potential advantage of transferring one molecule from the supersaturated solution to the solid is \(k_BT\ln S\). The bulk free-energy decrease per unit solid volume is therefore \(k_BT\ln S/v_m\). A spherical nucleus of radius \(r\) has surface area \(4\pi r^2\) and volume \(4\pi r^3/3\). Its net formation free energy is the positive interfacial cost minus the bulk driving-force benefit:

For \(S>1\), the surface term dominates at small \(r\), so small clusters tend to disappear. At larger \(r\), the negative volume term can dominate and growth becomes thermodynamically favorable. The maximum of \(\Delta G(r)\) separates these regimes. Differentiate with respect to \(r\) and set the derivative to zero:

The nonzero stationary point is the critical radius \(r^*\). The second derivative there is negative, so it is a maximum of the formation free energy. Substitution gives the barrier \(\Delta G^*=\Delta G(r^*)\):

A cluster smaller than \(r^*\) is more likely to dissolve than to persist in this model; a cluster larger than \(r^*\) can lower its free energy by growing. The critical radius is a dividing size, not a prediction of the final particle size.

Using the barrier expression, the central rate relationship is

The exponent is dimensionless: \(\gamma^3v_m^2\) has units of energy cubed, as does \((k_BT)^3\). The dependence on supersaturation is through \((\ln S)^{-2}\) in the barrier, while the dependence on interfacial energy is cubic. Thus even though \(S\), \(\gamma\), and \(v_m\) enter a comparatively simple expression, small changes in them can lead to large changes in \(J\).

Why predictions respond sharply to parameters

It is often more useful to compare rates by taking their ratio than to calculate two very small exponential terms separately. For fixed \(T\) and \(J_0\), a change in the barrier by \(\Delta(\Delta G^*)\) changes the rate by the exponential of the negative barrier change in thermal units. This makes the scale of a parameter perturbation interpretable even before an absolute rate is known.

The same sensitivity affects predicted induction times. If nucleation events occur independently at a constant rate \(J\) throughout a solution volume \(V\), the expected event count over time \(t\) is \(JVt\). Under this constant-rate Poisson model, the probability of no event by time \(t\) is \(\exp(-JVt)\), so the mean time to the first event is \(1/(JV)\). This is a first-nucleation result, not a complete model of an experimentally observed induction time.

Assumptions, diagnostics, and alternatives

CNT is useful because it links measurable or hypothesized thermodynamic properties to a rate law, but its assumptions should be checked against the intended use. In particular, the ideal relation \(k_BT\ln S\) assumes that the activity ratio adequately represents the chemical-potential difference. For nonideal solutions, \(S\) should be based on activities rather than concentrations alone. The spherical shape and constant \(\gamma\) are approximations, especially for nanometre-scale clusters whose structure may differ from the bulk crystal.

Homogeneous nucleation is also a strong assumption. A vessel wall, impurity, excipient, or pre-existing particle can reduce the barrier. In a common idealized heterogeneous model, a spherical-cap nucleus on a flat surface has a barrier equal to the homogeneous barrier multiplied by a wetting factor \(f(\theta)=(2+\cos\theta)(1-\cos\theta)^2/4\), where \(\theta\) is a contact angle. Since \(f(\theta)\) is between zero and one for the usual cap geometry, heterogeneous nucleation can be much faster. But the contact angle and relevant surface population may be uncertain, so this expression is not a universal correction.

Useful diagnostics include checking whether predicted rates change plausibly when \(S\), \(\gamma\), \(v_m\), and \(J_0\) are varied over credible ranges; confirming dimensional consistency; and comparing predictions across sample volumes. If constant homogeneous \(J\) is appropriate, first-event times should scale inversely with volume. Strong departures can indicate heterogeneous sites, changing supersaturation, time-dependent conditions, or a detection lag. A fit that requires an implausible prefactor or an interfacial energy that changes greatly across a narrow range of conditions is evidence that the model may be absorbing unrepresented processes.

The principal alternative is an empirical induction-time or nucleation-rate relationship fitted directly to data. It can be more practical when classical parameters are unavailable, but its fitted coefficients may not transfer reliably to other temperatures, media, or formulations. CNT provides a mechanistic structure and explicit parameter sensitivities, but its apparent mechanistic detail does not remove uncertainty in \(\gamma\), the kinetic prefactor, or nucleation pathway. Neither approach should be treated as a substitute for checking whether the experimental measurement detects the first nucleus or a later visible event.

The practical lesson is to report uncertainty and sensitivity, not just a single CNT rate or induction-time estimate. In particular, interfacial energy enters cubically in the barrier and then exponentially in the rate. A narrow uncertainty range for \(\gamma\) can therefore be more consequential than a much larger relative uncertainty in a parameter that enters only through the prefactor.