Question 5 of 8: Glass-Ceramics — Heterogeneous Nucleation and Phase Separation, Processing Cycle, and Comparison with TTT/CCT Curves for Steel

Question 5: Glass-Ceramics — Heterogeneous Nucleation and Phase Separation, Processing Cycle, and Comparison with TTT/CCT Curves for Steel (20 marks: a–5, b–5, c–10)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

5.1 — (a) Heterogeneous nucleation and phase separation in glass-ceramic formation

A glass-ceramic needs an extremely HIGH number density of crystal nuclei (typically $10^{12}$–$10^{15}\,\text{cm}^{-3}$) to end up with the fine, uniform grain structure that gives it useful mechanical and thermal-shock properties — a density that homogeneous nucleation in an ordinary glass essentially never reaches in a practical time, because the classical nucleation barrier $\Delta G^*=16\pi\gamma^3/3\Delta G_v^2$ is far too large without a catalytic surface to reduce it. Two linked mechanisms are engineered into the glass composition to solve this:

Heterogeneous nucleation. A nucleating agent (commonly $\text{TiO}_2$, $\text{ZrO}_2$, $\text{P}_2\text{O}_5$, or noble-metal colloids such as Ag/Au/Pt) is deliberately added to the melt (typically a few wt%). On the subsequent nucleation heat treatment, this agent precipitates first, as a fine, high-number-density secondary phase that provides a low-wetting-angle substrate for the DESIRED primary crystalline phase to nucleate on — reducing the nucleation barrier by the heterogeneous shape factor $S(\theta)=(2+\cos\theta)(1-\cos\theta)^2/4\ll1$, exactly as for heterogeneous nucleation in a metal (a contact angle close to zero can drop $\Delta G^*$ by orders of magnitude).

Phase separation. Many nucleating-agent glass systems undergo liquid–liquid (amorphous–amorphous) phase separation — by spinodal decomposition or by nucleation-and-growth within a metastable immiscibility dome — BEFORE crystallization even begins. The melt splits into two intimately mixed amorphous phases, one enriched in the nucleating agent. This is doubly useful: phase separation itself typically occurs more readily (at lower undercooling/higher viscosity) than crystal nucleation, so it is easy to trigger on the reheat schedule; and the resulting droplets are extremely fine (tens of nanometres, set by the spinodal wavelength) and enriched in nucleating agent, so once crystallization begins it inherits the phase-separated microstructure's own fine length scale rather than the glass matrix's own (far too coarse and sparse) intrinsic nucleation behaviour.

Together, phase separation supplies the fine-scale compositional heterogeneity and interfacial area, and heterogeneous nucleation on/within that phase-separated structure supplies the low-barrier nucleation sites — the combination decouples the achievable final grain size from the base glass's own (otherwise inadequate) homogeneous nucleation rate, which is the entire purpose of the nucleating-agent addition.

5.2 — (b) Time–temperature processing cycle for a glass-ceramic

A glass-ceramic is produced by first forming the part as an ordinary, fully amorphous glass (melt, then form by casting/pressing/blowing while workable), and only THEN converting it, in the solid state, into a controlled, fine-grained polycrystalline microstructure via a deliberate two-step heat treatment — not by simply cooling more slowly, which would give uncontrolled, coarse, often surface-initiated crystallization instead:

1. Melt and form. The base composition (containing the nucleating agent from part (a)) is melted, formed to shape, and cooled to room temperature as an ordinary, transparent, fully amorphous glass.
2. Reheat to the nucleation temperature $T_n$ and hold. $T_n$, typically just above $T_g$, is where the nucleation rate $\dot N$ is at or near its maximum (nucleation rate and growth rate peak at DIFFERENT temperatures in a typical glass system). Holding here produces an extremely high, uniform number density of very small nuclei — the single most important step for controlling final grain size, since every nucleus formed here becomes one crystallite in the final microstructure.
3. Reheat to the growth (crystallization) temperature $T_c$ and hold. $T_c$, a higher temperature, is where crystal GROWTH rate is high (growth is diffusion-controlled and favoured by higher atomic mobility). Because nucleation is already essentially complete, the existing fine nuclei simply grow and impinge, converting typically 90–98% of the volume into fine crystallites and leaving only a thin residual glassy grain-boundary phase.
4. Cool to room temperature. Cooling is controlled to avoid thermal shock (the crystalline and residual glassy phases generally differ in thermal expansion coefficient), giving the finished glass-ceramic — combining the net-shape formability of a glass with the strength, toughness and thermal-shock resistance of a fine-grained polycrystalline ceramic.

5.3 — (c) Comparing glass crystallization kinetics with steel's TTT/CCT curves

What is the same. Both systems are governed by the same C-curve framework: the overall transformation rate is the PRODUCT of a nucleation rate (rising with undercooling below the equilibrium temperature) and a growth/mobility rate (falling with undercooling as atomic mobility drops), so both plot a "nose" of minimum transformation-start time at some intermediate temperature, with transformation again slower at both higher temperature (low driving force, sluggish nucleation) and lower temperature (low mobility). In both systems the CCT curve for continuous cooling lies to the LONGER-time, LOWER-temperature side of the isothermal TTT curve, because a continuously cooling specimen spends only part of its time at any one temperature, so a real (continuous) cooling schedule must be compared against the shifted CCT curve, never the raw TTT curve, to predict the outcome.

What is different, and why. For steel, avoiding the nose is only one of several possible GOALS (getting all martensite); for glass, avoiding the nose — i.e. NOT crystallizing — is normally the entire point of the process, the opposite intent, with industrial glass-forming and annealing schedules engineered specifically to cool faster, at every temperature, than the crystallization-start time. A good glass former has its nose pushed to very long times (often minutes to hours, at a temperature well below $T_m$ and often close to $T_g$+100–200 K), which is precisely why it is a good glass former; a poor glass former (like a pure metal, or eutectoid steel) has its nose at very short times (steel: ∼1 s at ∼550 °C, from the same C-curve used throughout this paper), because rapid substitutional/interstitial diffusion and simple crystal structures make crystallization fast and hard to suppress.

Mechanistically, steel's nose is governed by solid-state diffusional nucleation and growth of a CRYSTALLINE product from another CRYSTALLINE phase (austenite → pearlite/bainite), with a separate, diffusionless athermal martensite reaction below $M_s$ that has no C-curve behaviour at all (Question 1). Glass's nose is a LIQUID→CRYSTAL transformation whose growth rate is limited by the liquid's own viscosity, which rises by many orders of magnitude as $T$ falls toward $T_g$ — a far steeper mobility collapse than solid-state diffusivity ever shows in steel — so the growth-rate branch of a glass's C-curve falls away much more sharply below its nose than the corresponding branch does for steel. This is also why glass-ceramic processing (Question 5(b)) deliberately splits the nucleation and growth holds into two SEPARATE isothermal steps at different temperatures — the nucleation-rate and growth-rate curves for a glass are typically much more widely separated in temperature than the corresponding curves are for steel, where a single isothermal hold near the nose already gets reasonable rates of both at once.