Nucleation and growth describes how new phases of matter form: first, a tiny cluster of atoms or molecules assembles out of a disordered parent phase (nucleation), then that cluster enlarges as more material joins it (growth). This two-step framework explains everything from ice crystals forming in a cloud to the hardening of a metal alloy, and it governs processes in fields as different as pharmaceutical manufacturing, bone formation, and chocolate making. The basic idea sounds simple, but the details of each step, and the surprising ways they can blur together, have kept researchers busy for decades.
How the First Step Works
Imagine a pot of sugar water that has been cooled well past the point where sugar should start crystallizing. The liquid is supersaturated, meaning it holds more dissolved sugar than it can stably support, yet nothing happens for a while. That delay exists because forming a brand-new crystal surface costs energy. Molecules in the liquid are constantly bumping into each other and briefly sticking together in tiny clusters, but most of those clusters fall apart almost immediately because the energy penalty of creating a new surface outweighs the energy benefit of forming the crystal interior. Only when a cluster reaches a certain minimum size, called the critical size, does the balance tip. Past that threshold, every additional molecule that joins makes the cluster more stable rather than less, and the cluster survives. That moment of survival is nucleation.
The energy barrier a cluster must overcome to reach critical size is central to how fast or slow nucleation happens. A highly supersaturated or deeply supercooled system has a lower barrier, so nucleation occurs more readily. A system that is only slightly past the tipping point has a tall barrier, and you may wait a long time before the first nucleus appears. In small volumes of liquid, achieving nucleation tends to require even higher supersaturation or deeper supercooling because there are fewer molecules available to randomly assemble into a viable cluster.
Homogeneous Versus Heterogeneous Nucleation
The description above assumes nucleation happens entirely on its own, out in the middle of the liquid or gas with no help from any surface. That is homogeneous nucleation, and in practice it is relatively rare. Almost every real-world system contains impurities, container walls, dust particles, or other surfaces, and these surfaces dramatically lower the energy barrier. A cluster forming on a flat wall, for instance, does not need to build its entire surface from scratch because the wall provides part of the structure. The reduction in the barrier depends on how well the new phase wets the surface, described by the contact angle between the cluster and the wall. A low contact angle means strong wetting, a much lower barrier, and faster nucleation. A high contact angle means weak wetting and a barrier closer to the homogeneous case.
Monte Carlo simulations have confirmed this picture quantitatively. Flat walls facilitate the condensation of a supersaturated vapor because the free-energy barrier for forming a nucleus on the wall is smaller than the barrier for forming one in open space, by a factor that depends only on the contact angle.
Heterogeneous nucleation is why, in everyday life, crystals or bubbles tend to appear at specific spots: the scratch on the inside of a glass, a speck of dust floating in the air, or the rough edge of a container. Those imperfections serve as launching pads.
What Happens After a Nucleus Survives
Once a stable nucleus exists, growth takes over. The nucleus acts as a sink, drawing dissolved molecules (or atoms, or ions) out of the surrounding medium and incorporating them into its expanding structure. How fast this happens depends on two things: how quickly material can diffuse through the surrounding fluid to reach the growing surface, and how quickly molecules can attach once they arrive.
When diffusion through the liquid is the bottleneck, growth rates are governed by concentration gradients near the crystal face. Detailed computations show that at the surface of a growing crystal, the arrangement of atomic steps matters. Molecules preferentially attach at step edges, where they can bond to the crystal on multiple sides at once, rather than landing on a flat terrace. The spacing between these steps, the degree of supersaturation, and the thickness of the boundary layer between the crystal and the well-mixed bulk liquid all influence how quickly the crystal enlarges.
In systems where attachment at the surface is the bottleneck instead, molecules arrive faster than they can find a good spot to settle in. Growth then depends on the crystal’s surface chemistry and temperature. These two regimes, diffusion-limited and surface-limited, produce crystals with different shapes. Diffusion-limited growth tends to create branching, dendritic structures (think snowflakes), while surface-limited growth tends to produce smoother, more compact crystals.
When Nucleation Does Not Follow the Textbook
The classical picture treats nucleation as a single event: molecules jump from a disordered state straight into an ordered crystalline cluster. Over the past two decades, researchers have found that many systems take a detour. Instead of going directly from dissolved molecules to crystal, the system first forms a dense, disordered droplet or cluster, and crystalline order emerges inside that cluster only afterward. This has been called two-step nucleation, confusingly similar to the broader “two-step” label of nucleation-then-growth but referring to two sub-steps within nucleation itself.
Experiments with colloidal particles, which are large enough to image individually under a microscope, provided some of the earliest direct evidence. In certain colloidal suspensions, amorphous clusters grew to roughly 30 particles and then rapidly crystallized, confirming the two-step pathway in real time.
Protein crystallization has provided another window into this phenomenon. Work on lysozyme and other protein crystals has shown that nucleation proceeds through a mechanism where a density fluctuation (a transient region of higher molecular concentration) appears first, and then a structural fluctuation (the emergence of crystalline order) occurs within that dense region.
These non-classical pathways matter because they change the conditions under which crystals form and the types of crystals that result. If the dense intermediate is long-lived, it may produce a different crystal polymorph than classical nucleation would, or it may produce an amorphous solid instead of a crystal altogether.
Ostwald Ripening and What Happens to Many Small Crystals
Nucleation rarely produces just one crystal. Typically, many nuclei form in a burst, and then they all grow simultaneously, competing for the same pool of dissolved material. As that pool gets depleted, a secondary process kicks in. Smaller crystals, which have higher surface energy per unit volume, begin to dissolve, and the freed material re-deposits onto larger crystals. This is Ostwald ripening, first described in 1896, and it is a thermodynamically driven process that causes the average particle size to increase over time while the total number of particles shrinks.
Ostwald ripening shows up across a startling range of fields, from geology to aerosol science to materials engineering. In nanoparticle synthesis, researchers have identified critical radii below which particles dissolve and above which they continue to grow. For palladium nanocrystals, for example, that critical radius was found to be roughly 5.7 nanometers.
LaMer Burst Nucleation and Making Uniform Particles
If you want particles that are all the same size, controlling nucleation is everything. The LaMer model, developed in the 1950s, describes a strategy: rapidly increase the concentration of dissolved material until a burst of nucleation happens nearly simultaneously, then stop adding material so that the existing nuclei grow without new ones forming. Because all nuclei appear at almost the same moment under nearly identical conditions, they grow at similar rates and end up at similar sizes.
Analysis of LaMer burst data confirms that the final number of nuclei is largely determined by how fast solute is supplied and how fast the nuclei grow once formed. This insight is the foundation of modern nanoparticle synthesis, where uniform size distributions are essential for applications in electronics, catalysis, and medicine.
Nucleation in the Atmosphere
Every cloud droplet starts with nucleation. Water vapor in the atmosphere does not spontaneously condense into droplets at the modest supersaturations found in typical clouds because the homogeneous nucleation barrier is far too high. Instead, droplets form on tiny aerosol particles, dust grains, or other surfaces that serve as cloud condensation nuclei. The availability of these nuclei determines cloud properties like brightness, lifetime, and precipitation behavior, which in turn affect climate.
Where do these aerosol particles come from? Many start as molecular clusters that nucleate from gas-phase precursors high in the atmosphere, then grow as they descend and accumulate additional material. Modeling studies of the southeastern Atlantic found that aerosols nucleated in the free and upper troposphere were the dominant source of boundary-layer cloud condensation nuclei, contributing roughly 41 percent of the total on an annual average as they subsided and mixed into the marine boundary layer.
Whether freshly nucleated particles can grow large enough to serve as cloud condensation nuclei is a critical question for climate models. Observations show that nucleated particles do grow to sufficient sizes, but some global models have found that cloud condensation nuclei concentrations are only modestly sensitive to large changes in nucleation rates, suggesting that growth processes and competition with existing particles play equally important roles.
Bones, Teeth, and Seashells
Living organisms are experts at controlling nucleation and growth. Bone and tooth enamel form through biomineralization, a process where specialized proteins orchestrate where and when mineral crystals appear. In bone and dentin, extracellular matrix proteins modulate the nucleation of calcium phosphate nanoparticles and guide their growth into the micrometer-scale crystals that give these tissues their hardness. One such protein, dentin matrix protein 1, has been shown to facilitate mineral nucleation at specific sites while simultaneously preventing unwanted calcium phosphate precipitation elsewhere.
Nacre, the iridescent lining of certain seashells, is another striking example. It consists of stacked platelets of aragonite (a form of calcium carbonate) interspersed with thin layers of organic material. Specialized proteins and polysaccharides assemble into a three-dimensional organic framework that forms a microenvironment where mineral deposition takes place. The organic matrix does not just provide a surface for heterogeneous nucleation; it actively selects which crystal polymorph forms and controls the orientation of the growing crystals.
Recent work has suggested that biomineralization in these systems proceeds through non-classical pathways. Proteins in bone and dentin interact with dense liquid-like calcium phosphate clusters, modulating when and where those clusters crystallize, rather than templating the classical one-step nucleation of an ordered crystal.
Amyloid Fibrils and Disease
Not all nucleation is desirable. In neurodegenerative diseases like Alzheimer’s, misfolded proteins aggregate into fibrous structures called amyloid fibrils, and nucleation is at the heart of how this happens. The initial formation of a fibril from dissolved protein is called primary nucleation and tends to be slow. But once fibrils exist, a much faster process takes over: secondary nucleation, in which new aggregates form directly on the surfaces of existing fibrils.
Super-resolution microscopy has directly observed this process for the amyloid-beta 42 peptide associated with Alzheimer’s. Monomers were seen growing into relatively large aggregates along the length of fibril surfaces before detaching, providing visual confirmation that fibril surfaces catalyze the creation of new seeds.
The thermodynamics of secondary nucleation are fundamentally different from those of primary nucleation. Researchers have shown that the interaction between dissolved amyloid-beta 42 and existing fibril surfaces reverses the thermodynamic signature of the process: whereas primary nucleation is driven largely by entropy, secondary nucleation benefits from the enthalpic stabilization of peptides that adsorb onto the fibril surface in conformations favorable for aggregation, dramatically lowering the activation energy. This catalytic efficiency is what makes amyloid aggregation so difficult to stop once it starts, and it has become a major focus for therapeutic strategies aimed at slowing disease progression.
Molecular dynamics simulations have explored how peptides interact at fibril edges, observing that peptides attracted to the fibril surface can form small oligomeric structures, including beta-sheets, that then serve as new growth points.
Tempering Chocolate
Cocoa butter can crystallize into at least six different polymorphic forms, and only one of them, known as Form V, gives chocolate the glossy surface and satisfying snap that consumers expect. Tempering, the controlled heating and cooling process used in chocolate making, is essentially a nucleation-and-growth exercise. The goal is to create a large number of Form V nuclei and then let them grow while suppressing the formation of less desirable polymorphs.
Recent work has proposed that tempering should be viewed as a hierarchical crystallization process, where nucleation rate, structural homogeneity, and microstructural organization all play critical roles. Interestingly, certain phospholipids naturally present in cocoa butter have been found to promote Form V crystallization, but they do so through different mechanisms, with nucleation dynamics playing a more influential role than simply achieving the right polymorph.
Poor tempering leads to “bloom,” the whitish coating that appears on old or badly stored chocolate. Bloom is the visible result of Ostwald ripening and polymorphic transformation: less stable crystal forms slowly convert to more stable ones, and fat migrates to the surface. Understanding nucleation and growth at the molecular level has allowed chocolate manufacturers to design tempering protocols that minimize bloom and extend shelf life.
Ice Nucleation by Bacteria
Some organisms have evolved to exploit nucleation rather than just control it. The bacterium Pseudomonas syringae produces ice-nucleating proteins on its outer membrane that can trigger water to freeze at temperatures just below zero, far warmer than the roughly minus 35 degrees Celsius needed for homogeneous ice nucleation in pure water. These proteins provide a surface that mimics the structure of ice, lowering the nucleation barrier so effectively that the bacteria can induce frost damage in plants, which gives them access to nutrients from ruptured plant cells.
The freezing temperature depends on the concentration of bacteria present. Experiments have confirmed that higher cell concentrations raise the temperature at which ice forms, following a logarithmic relationship.
The discovery of bacterial ice nucleation in 1974 had practical consequences beyond plant pathology. Freeze-dried preparations of P. syringae are used commercially in snowmaking for ski resorts, and the same ice-nucleating proteins are a subject of active research in atmospheric science because bacteria lofted into the upper atmosphere may contribute to ice formation in clouds.
Industrial Crystallization and Pharmaceutical Manufacturing
In the pharmaceutical industry, most drug molecules must be crystallized into a specific form to ensure consistent dissolution, bioavailability, and shelf stability. Controlling nucleation is the primary lever. If too many nuclei form at once, you get a powder of tiny crystals that may be hard to filter and dry. If too few form, you get a small number of large crystals that dissolve slowly and unevenly.
One approach uses ultrasound to decouple nucleation from growth. A small crystallizer exposed to ultrasonic energy generates a controlled burst of nuclei (seeds), which are then transferred to a larger vessel where they grow under gentle cooling. By adjusting the power and duration of the ultrasound, manufacturers can control the number of seeds and, consequently, the final crystal size distribution, producing narrow, reproducible size distributions.
The same principles apply across industries. In semiconductor manufacturing, controlling the nucleation of thin-film crystals determines the grain size and electrical properties of the final material. In water treatment, encouraging the nucleation of calcium carbonate can prevent scale buildup in pipes. In each case, the practical challenge is the same: steer nucleation to get the right number of nuclei at the right time, then manage growth conditions to achieve the desired final product.
When Nucleation Fails Entirely
Sometimes the most interesting outcome is when nucleation does not happen at all. If a liquid is cooled quickly enough that crystalline nuclei never have time to form, the liquid can bypass crystallization entirely and solidify into a glass, an amorphous solid with no long-range order. This is nucleation control at its most extreme.
Many metallic glasses require very rapid quenching to achieve vitrification and often do not show a clearly resolved glass transition when reheated. The competition between the cooling rate and the nucleation rate determines whether a given material can be made into a glass. Substances with sluggish nucleation kinetics, such as the silicates used in window glass, vitrify easily. Metals, with their fast atomic mobility and strong tendency to crystallize, generally require cooling rates of thousands or millions of degrees per second, though some specially designed alloys (bulk metallic glasses) can be vitrified at much lower rates.
Understanding this competition has enabled the design of new materials with unusual properties. Metallic glasses, for instance, can be stronger and more corrosion-resistant than their crystalline counterparts precisely because they lack the grain boundaries and defects inherent to crystalline metals.
Watching Nucleation Happen in Real Time
For most of its history, nucleation theory was tested indirectly, by measuring outcomes like crystal size distributions, induction times, or nucleation rates and comparing them to predictions. Directly watching a nucleus form is extraordinarily difficult because the critical cluster is often only a few nanometers across and exists for only a fleeting moment.
The development of in-situ liquid-phase transmission electron microscopy has changed this. By enclosing a thin film of liquid inside an electron-transparent cell, researchers can now watch nucleation and growth events unfold in real time at near-atomic resolution. Studies of gold nanoparticles, calcium carbonate, and iron oxide using this technique have provided direct evidence for crystallization mechanisms and yielded dynamic information about kinetic and thermodynamic parameters that were previously inaccessible.
Computational methods have advanced in parallel. Techniques like umbrella sampling and forward flux sampling allow researchers to simulate rare nucleation events that would take impossibly long to observe in a straightforward molecular dynamics simulation. Comparisons between these methods for the nucleation of hard-sphere crystals have helped validate the simulation toolkit and refine estimates of nucleation barriers and rates. Together, direct imaging and advanced simulation have moved the field from inferring what happens during nucleation to actually seeing and computing it, step by step.