Solidification is the phase transformation from liquid to solid and is the first microstructure-forming event in the entire processing chain of virtually every metallic component. The grain size, grain morphology, solute distribution, porosity, and phase constitution of a casting are all established during solidification — and these microstructural features propagate through subsequent hot working, heat treatment, and service to govern final mechanical properties. Understanding the thermodynamics of nucleation, the kinetics of dendritic interface growth, and the solute redistribution that drives microsegregation is therefore foundational to the design and control of casting, welding, and solidification-processing operations.

1. Thermodynamics of Solidification: Driving Force and Undercooling

Solidification is thermodynamically driven by the difference in Gibbs free energy between the liquid and solid phases. At the equilibrium melting temperature T_m, the Gibbs energies of liquid and solid are equal (ΔG = 0) and neither phase is preferred. Below T_m, the solid phase has lower free energy, and the driving force per unit volume for solidification is:

where ΔH_f is the latent heat of fusion (J m^-3), ΔT = T_m − T is the undercooling below the equilibrium melting point, and T_m is the equilibrium melting temperature in Kelvin. This driving force is entirely temperature-dependent: larger undercooling produces a larger driving force and faster solidification kinetics.

1.1 Sources of Undercooling

In solidifying systems, undercooling can arise from several physically distinct sources that may act simultaneously:

• Thermal undercooling: the actual temperature of the melt falls below T_m due to heat extraction. This is the dominant undercooling in most casting operations.
• Constitutional undercooling: solute rejected by the growing solid depresses the local liquidus temperature ahead of the interface below the actual temperature of the melt. This is critical for dendritic morphology selection.
• Curvature undercooling (Gibbs–Thomson effect): the equilibrium melting temperature of a curved surface is reduced relative to a planar surface. For a spherical nucleus of radius r, the effective melting point depression is ΔT_r = 2γT_m / (ΔH_f r), where γ is the solid–liquid interfacial energy.
• Kinetic undercooling: a small additional undercooling required to drive atoms across the solid–liquid interface at the observed rate. Significant only at very high growth velocities (rapid solidification).

2. Nucleation Theory

Before a new solid phase can grow, a stable solid nucleus must form. Classical nucleation theory treats this as a competition between the favourable volumetric free energy change (driving nucleation) and the unfavourable surface energy cost of creating a new solid–liquid interface (opposing nucleation).

2.1 Homogeneous Nucleation

For a spherical embryo of radius r forming in a pure homogeneous melt, the total Gibbs energy change is:

ΔG_total = (4/3)πr³ × ΔG_v  +  4πr² × γ_SL
           |____________________________|   |___________________|
              Volumetric term (negative)     Surface energy term
                  (driving force)               (opposing force)

Differentiating with respect to r and setting d(ΔG)/dr = 0 gives the critical radius:

r* = -2γ_SL / ΔG_v  =  2γ_SL T_m / (ΔH_f ΔT)

ΔG* = (16π γ_SL³) / (3 ΔG_v²)  =  (16π γ_SL³ T_m²) / (3 ΔH_f² ΔT²)

Embryos smaller than r* are unstable: growth increases total free energy, so they dissolve. Only embryos that thermally fluctuate to r ≥ r* become stable nuclei that grow spontaneously. ΔG* is the activation energy barrier for nucleation. For metals, γ_SL ≈ 0.1–0.5 J m^-2, ΔH_f ≈ 10⁸–10⁹ J m^-3, giving r* ≈ 2–10 nm at practical undercoolings.

The homogeneous nucleation rate is:

where A is a pre-exponential frequency factor (~10³⁰–10⁴² m^-3s^-1 depending on atomic vibration frequency and atomic density). Because ΔG* scales as ΔT^-2, the nucleation rate is negligibly small at small undercoolings and rises with extreme rapidity above a threshold undercooling. Measured homogeneous nucleation undercoolings in emulsified metal droplet experiments range from approximately 0.18T_m to 0.20T_m (100–300 K for most engineering metals).

2.2 Heterogeneous Nucleation

In all practical casting operations, solidification initiates on pre-existing heterogeneous nucleation sites: mould walls, oxide films, dissolved inoculant particles, or deliberate grain refining additions. A foreign substrate reduces the activation energy by providing part of the interface and thereby reducing the total interfacial energy cost of forming a nucleus.

For a spherical cap nucleus on a flat substrate with contact angle θ (the angle between the solid–liquid and solid–substrate interfaces):

ΔG*_het = ΔG*_hom × f(θ)

f(θ) = (2 - 3cosθ + cos³θ) / 4

Where: f(θ=0°)  = 0    (perfect wetting; zero barrier)
       f(θ=90°) = 0.5  (hemispherical nucleus)
       f(θ=180°)= 1.0  (no wetting; same as homogeneous)

The contact angle is governed by Young’s equation relating the three interfacial energies at the triple line. Inoculants are effective when they have good lattice matching (low lattice misfit δ < 6%) with the solidifying metal, which minimises γ_{substrate-solid} and reduces θ. For example, TiB₂ particles added to aluminium alloy melts (via Al–5Ti–1B master alloy) nucleate on the (0001) TiB₂ face, which has a lattice misfit of approximately 4.2% with aluminium, giving a large f(θ) reduction and allowing nucleation at ΔT < 2 K.

3. Solidification Front Morphology: Planar, Cellular, and Dendritic

Once a stable nucleus forms, it grows by extracting latent heat into the surrounding melt and, in alloy systems, by redistributing solute across the advancing solid–liquid interface. The resulting morphology of the growth front is one of the most important factors determining final microstructural scale and segregation pattern.

3.1 Constitutional Undercooling and Interface Stability

In a binary alloy with partition coefficient k (= C_s/C_l at the interface, typically k < 1 for most engineering alloys), solidification rejects excess solute into the liquid ahead of the interface. This builds up a solute-enriched boundary layer. The local liquidus temperature in this solute-enriched layer is depressed below the equilibrium liquidus. If the actual temperature gradient G_L in the liquid ahead of the interface is less than the liquidus temperature gradient, a zone of constitutionally undercooled liquid exists ahead of the interface.

The constitutional undercooling criterion for a planar interface to be stable is:

G_L / V  ≥  m_L C_0 (1 - k) / (D_L k)

Where:
  G_L  = temperature gradient in liquid (K m⁻¹)
  V    = interface velocity (m s⁻¹)
  m_L  = liquidus slope (K per wt%)
  C_0  = bulk alloy composition (wt%)
  D_L  = solute diffusivity in liquid (m² s⁻¹)
  k    = equilibrium partition coefficient

When this condition is violated (low G_L/V ratio), the planar interface is unstable. Small perturbations project into the constitutionally undercooled zone, where they experience faster growth than the planar front and are amplified. This is the Mullins–Sekerka instability. As G_L/V decreases, the morphology transitions progressively: planar → cellular → columnar dendritic → equiaxed dendritic.

3.2 Dendritic Growth Mechanism

Dendrites are tree-like branching crystal structures whose primary arms grow along crystallographically preferred directions: <100> in cubic (FCC and BCC) metals, <1010> in hexagonal metals. The driving force for the rapid growth of dendrite tips is the constitutional undercooling in the melt ahead of the tips. Secondary arms nucleate on primary arms driven by the same instability, and tertiary arms may nucleate on secondary arms in coarsely solidified structures.

The dendrite tip velocity is described by the LGK (Lipton–Glicksman–Kurz) model:

V = a₁ (ΔT_total)² - a₂ (ΔT_total)³ + ...  (simplified form)

Full LGK: balances thermal, solutal, and curvature undercooling at tip
where tip radius R* ≈ (Γ / σ*)^0.5 × (1/ΔT)
and Γ = γ_SL T_m / ΔH_f  (Gibbs-Thomson coefficient, ~10⁻⁷ K·m for metals)
σ* ≈ 1/(4π²) (stability constant from marginal stability theory)

Primary and Secondary Dendrite Arm Spacing

Primary dendrite arm spacing (PDAS, λ₁) is the centre-to-centre distance between adjacent primary arms. It is controlled by the solidification conditions at the onset of interface breakdown:

Secondary dendrite arm spacing (SDAS, λ₂) is the dominant microstructural metric in cast components. Unlike λ₁, SDAS is not set at initial growth but evolves continuously by coarsening (Ostwald ripening) throughout the solidification period. Finer secondary arms dissolve to reduce total interfacial energy, feeding coarser arms. The governing relationship is:

λ₂ = a × t_f^(1/3)

Where:
  a   = alloy-dependent coarsening constant (typically 10-60 μm s^(-1/3) for Al alloys)
  t_f = local solidification time (time spent in mushy zone) in seconds

For aluminium alloy A356, typical values are: sand casting t_f ≈ 100–500 s giving λ₂ ≈ 50–80 μm; permanent mould t_f ≈ 20–80 s giving λ₂ ≈ 25–40 μm; high-pressure die casting t_f ≈ 0.1–2 s giving λ₂ ≈ 5–15 μm.

4. Solute Redistribution and Microsegregation

Microsegregation is the non-uniform distribution of solute on the scale of the dendrite arm spacing, arising from the progressive partitioning of solute between solid and liquid during solidification. It is the primary source of compositional heterogeneity in as-cast microstructures and has direct consequences for mechanical properties, corrosion resistance, and hot workability.

4.1 Equilibrium Solidification (Lever Rule)

At the opposite extreme of rapid solidification, if complete equilibrium is maintained throughout solidification (complete diffusion in both solid and liquid at every temperature), the lever rule predicts the fraction solid and composition at any temperature within the two-phase field:

f_s = (C_l - C_0) / (C_l - C_s)  =  (C_l - C_0) / (C_l(1 - k))

At any temperature T:
  C_s = k C_l
  C_l = C_0 / [1 - f_s(1 - k)]   (lever rule liquid composition)

Under equilibrium conditions, the final solid has uniform composition equal to C₀. In practice, this requires infinitely slow cooling and complete back-diffusion in the solid, which never occurs in real casting operations.

4.2 The Scheil–Gulliver Equation

At the other extreme, the Scheil model assumes no diffusion in the solid (realistic for substitutional solutes in metals where D_s << D_l), complete mixing in the liquid (justified by convection), and local equilibrium at the solid–liquid interface. The resulting solute distribution in the solid is:

C_s = k C_0 (1 - f_s)^(k-1)

Where:
  C_s   = instantaneous solid composition at fraction solid f_s
  k     = equilibrium partition coefficient
  C_0   = initial alloy composition
  f_s   = fraction solid (0 to 1)

For the liquid: C_l = C_0 (1 - f_s)^(k-1)
Terminal eutectic fraction: f_eut ≈ 1 - [(C_eut / C_0)^(1/(k-1))]

The Scheil equation predicts that solute-rich liquid persists to very high fractions solid, and that the final interdendritic liquid reaches the eutectic composition even in alloys of sub-eutectic nominal composition. This dramatically extends the apparent freezing range and increases the volume fraction of eutectic phases compared to lever-rule predictions. For example, in an Al–4.5 wt% Cu alloy (equilibrium eutectic at 5.65 wt% Cu), the Scheil equation predicts a eutectic fraction of approximately 10%, versus near-zero under equilibrium conditions.

4.3 Intermediate Models: Back-Diffusion (Brody–Flemings)

The Brody–Flemings model introduces a back-diffusion parameter α = 4 D_s t_f / λ₂² (the Fourier number for diffusion in the solid) that interpolates between the Scheil limit (α = 0) and the lever rule limit (α = 0.5). For interstitial solutes (C, N) with high D_s in steel, back-diffusion is significant and the actual segregation is closer to the lever rule; for substitutional solutes (Cr, Mo, Ni), D_s is very low at solidification temperatures and the Scheil equation is more appropriate.

5. Grain Structure in Castings: Columnar and Equiaxed Zones

A transverse section through a typical industrial casting reveals three distinct macrostructural zones:

5.1 Chill Zone

A narrow outer layer of fine, randomly oriented equiaxed grains forms immediately upon pouring as the melt contacts the cold mould wall. The high heat extraction rate and the nucleating effect of the mould surface create a large number of nuclei simultaneously, producing a fine-grained, isotropic skin layer (typically 1–5 mm thick in steel ingots).

5.2 Columnar Zone

Grains that are favourably oriented (with a <100> direction parallel to the heat flow direction in cubic metals) grow preferentially in the direction of the thermal gradient, at the expense of less favourably oriented neighbours. This competitive growth produces a zone of elongated columnar grains aligned perpendicular to the mould wall. Columnar grain growth is favoured by high temperature gradients and low growth velocities (high G_L/V ratio at the columnar front).

5.3 Equiaxed Zone

Beyond the columnar zone, if sufficient nucleants are present in the melt and if constitutional undercooling is sufficient to sustain them, equiaxed grains nucleate and grow freely in the bulk melt. The transition from columnar to equiaxed growth (the CET) is promoted by: (1) low temperature gradients in the bulk melt; (2) high inoculant particle density; (3) high alloy content increasing constitutional undercooling; and (4) dendrite fragmentation by convection. Most industrial castings target a fine equiaxed structure throughout because it provides isotropic properties and greater resistance to hot tearing compared to columnar structures.

6. Solidification Defects

6.1 Shrinkage Porosity

Most metals contract on solidification (volumetric shrinkage: steel ~3%, aluminium ~6.6%, copper ~4.5%). If this shrinkage is not compensated by liquid metal feeding from risers or adjacent liquid regions, porosity forms. Macro-shrinkage (pipe shrinkage) forms as a concentrated void at the top of ingots where liquid metal last solidifies. Micro-shrinkage (interdendritic porosity) forms in the mushy zone when the dendritic network becomes too coarse to allow liquid flow to the contracting regions, and is the dominant form of porosity in alloy castings with wide freezing ranges.

6.2 Gas Porosity

Dissolved gases (hydrogen in aluminium alloys, nitrogen and hydrogen in steels) have markedly higher solubility in the liquid than in the solid (described by Sievert’s law: [H]_liquid ∝ P_H2^1/2). As solidification proceeds, the solubility decreases sharply at the solidus, forcing gas to nucleate as pores. Hydrogen porosity in aluminium die castings is controlled by degassing the melt with Cl₂, N₂, or Ar gas rotary degassing units prior to casting.

6.3 Hot Tearing

Hot tears (hot cracks) initiate in the mushy zone at high fraction solid (typically f_s > 0.9) when thermal contraction strains cannot be accommodated by the limited remaining liquid in the interdendritic channels. Susceptibility is quantified by the hot cracking criterion proposed by Feurer and later Rappaz:

Hot tearing susceptibility ∝ (dT/df_s) in the critical f_s range 0.9 - 1.0

Alloys with broad dT/df_s (wide mushy zone) and low permeability
of the dendritic network are most susceptible.

Example: Al-1wt%Cu has narrow mushy zone → low HT susceptibility
         Al-3wt%Cu has intermediate → moderate
         Al-0.5wt%Cu near eutectic → re-healing → low
         Maximum susceptibility near "peak alloy" (e.g. Al-2wt%Cu)

Hot tearing in welding (solidification cracking) follows the same mechanism: the solidifying weld pool in the mushy zone experiences thermal contraction strains from the cooling surrounding base metal. Alloys with wide solidification ranges and those producing low-melting-point grain boundary films (Fe–S, Fe–P eutectics in steel; Mg₂Si films in aluminium alloys) are most susceptible. See the hydrogen-induced cracking guide for related weld cracking mechanisms.

7. Industrial Control of Solidification Microstructure

7.1 Cooling Rate and Process Selection

The most powerful lever available to the casting engineer is cooling rate, which directly controls SDAS and all associated properties. The table below summarises achievable SDAS ranges for aluminium alloys across common casting processes:

7.2 Grain Refinement by Inoculation

In aluminium alloys, the industry standard grain refiner is Al–5Ti–1B master alloy, added at 0.5–2 kg per tonne. The active nucleants are TiB₂ and TiAl₃ particles. Effective grain refinement reduces average grain diameter from 2–5 mm (without refiner) to 100–300 μm, dramatically reducing hot tearing susceptibility and improving mechanical isotropy. For more on microstructural control, see the grain boundaries guide and the coverage of bainite and martensite formation in solid-state transformations.

7.3 Continuous Casting of Steel

Modern steelmaking employs continuous casting for over 95% of production. Liquid steel from the ladle flows through a tundish into a water-cooled oscillating copper mould (primary cooling zone), where a solid shell forms. The strand exits the mould with a liquid core and is bent and straightened while passing through secondary cooling water sprays. Typical solidification parameters in continuous slab casting: primary cooling rate 100–1000 K s^-1 in the mould; secondary cooling rate 1–50 K s^-1 in the spray zone; final solidification (metallurgical length) 5–15 m below the mould for slabs 200–250 mm thick. The fully columnar structure with centreline segregation (Mn, C enriched) characteristic of continuous cast slabs is subsequently broken down by hot rolling. For context on the Fe–C phase diagram and how it governs the phases encountered during steel solidification, refer to the dedicated phase diagram article.

7.4 Homogenisation Heat Treatment

Microsegregation established during solidification can be substantially reduced by homogenisation heat treatment: annealing at a temperature well within the single-phase field for sufficient time for solute to diffuse over the SDAS. The required homogenisation time scales as:

t_hom ≈ λ₂² / (π² D_s)

For Al-Cu alloy at 500°C: D_s(Cu) ≈ 5×10⁻¹³ m²/s
For λ₂ = 40 μm:  t_hom ≈ (40×10⁻⁶)² / (π² × 5×10⁻¹³) ≈ 3240 s ≈ 0.9 h
For λ₂ = 100 μm: t_hom ≈ 20,000 s ≈ 5.6 h

This shows clearly why finer SDAS (from faster cooling) allows shorter homogenisation times and reduced energy consumption. For related heat treatment principles, see the annealing and normalising article and the quenching and tempering guide.

8. Characterisation of Solidification Microstructures

Quantitative characterisation of solidification microstructures is essential for quality control and process development. Key techniques include:

• Optical metallography: SDAS measurement by linear intercept method per ASTM E112 after polishing and etching (Keller’s reagent for Al alloys; 2% nital for steels). The most routine quality control measurement in casting facilities.
• SEM/EDX mapping: Elemental maps of Cu, Si, Mg etc. reveal microsegregation patterns and identify intermetallic phases. Wavelength-dispersive spectroscopy (WDS) on EPMA provides higher accuracy for segregation quantification.
• Thermal analysis: Cooling curves recorded during solidification reveal liquidus and solidus temperatures, freezing range, and recalescence events associated with eutectic formation. Used for melt quality assessment and alloy composition verification.
• Hardness testing: Micro-Vickers hardness maps across dendrite arms quantify the mechanical consequences of segregation. See the hardness testing methods article for technique details.
• Charpy impact testing: Columnar-to-equiaxed transition and segregation severity strongly influence low-temperature impact properties. Refer to the Charpy impact test article for testing methodology.

9. Solidification in Welding Metallurgy

The weld fusion zone solidifies from the partially melted base metal at the fusion line, which acts as a substrate for epitaxial nucleation with zero activation energy barrier (complete wetting, θ = 0). Grains in the fusion zone grow epitaxially from the existing base metal grains, with columnar grains growing from the fusion line toward the weld centreline following the steepest temperature gradient (opposite to the heat flow direction). The centreline of the weld pool is the last region to solidify, and it concentrates solute and impurities through the Scheil mechanism, making it susceptible to centreline solidification cracking. For weld microstructure development in the HAZ, see the heat-affected zone microstructure article.

The formation of acicular ferrite in weld metals is a direct consequence of nucleation on oxide inclusions within the solidified weld pool, providing a fine, tough microstructure. Controlling solidification conditions in welding is therefore as important as controlling base metal microstructure. Key parameters include heat input (see the welding heat input calculator), preheat temperature, and interpass temperature.

10. Summary of Key Relationships

Recommended References

Disclosure: MetallurgyZone participates in the Amazon Associates programme. If you purchase through these links, we may earn a small commission at no extra cost to you. This helps support free technical content on this site.

Further Reading