I was born and reared on a family farm in northwest Illinois and obtained my elementary education in a single-teacher village school. Apparently I had a mild case of dyslexia, for I was virtually unable to read until I had been in school for more than two years. Then, through the valiant tutelage of my mother, who had been a school teacher, I suddenly became able to read and ever afterward did so, well and extensively. Our farming community was not included in any secondary school district, and my parents managed to have me admitted to Kewanee High School which, at the time, was considered to be the most outstanding school, academically, in the area. There, indeed, I received an excellent education from an outstanding group of teachers. I was especially inspired by the Misses Minnie Trask and Wildred Ewan, who posed demanding intellectual challenges and encouraged imaginative but rigorous modes of thought. It seemed that I learned most from having to solve difficult problems and rather little from classroom exposition, however excellent it was. Because of this, the emphasis of my own teaching has been on the posing of meaningful and challenging problems. In this, I have tried to follow the dictum, attributed to Galileo, “you can't teach a person anything, you can only help them to find it within themself”.

Dave, Carol, Mr. Chairman and assembled colleagues and friends, we are here at this symposium to honor Dave Turnbull in his 70th year for his qualities as a human being and his achievements and contributions as a scientist and as a teacher.

A series of important papers by David Turnbull and his collaborators in the late 1940's and early 1950's laid the experimental and theoretical foundation of modern nucleation theory. The elegance, versatility, and generality of the phenomenological approach, coupled with brilliant and insightful experimental confirmation, sparked widespread application which continues today. Much of David Turnbull's subsequent work in other subjects grew directly or indirectly from this work.

My interest in neutron-induced nucleation began with a simple and elegant demonstration in one of David Turnbull's classes in which a drop of water was superheated to about 250°C as it rose in a column of heated oil. As David Glaser, the inventor of the bubble chamber, so ably demonstrated, such superheated liquids are radiation sensitive. Our test system is a simple one. Halocarbon and hydrQcarbon drops are introduced into an aqueous holding gel under pressure at room temperature. As the pressure is released, the drops become superheated. Neutrons of sufficient energy will trigger vaporization of these moderately superheated drop detectors (SSDs), but gammas and x-rays will not unless the homogeneous nucleation limit is approached. We have performed measurements on the neutron energy threshold to produce nucleation in a number of different superheated materials at different temperatures. We have also developed a theory which indicates that of the energy deposited in a critical radius, only about 5% is effective in producing bubble formation. Both theory and experiment are discussed.

For crystals that melt into fluid liquids, superheating of the solid has been detected in only a few cases. In these cases, a crystal was subjected to focused heating in the center. Under a high thermal gradient some interior superheating was possible, but usually only a few degrees or less. A new approach is reported for preparing samples which can exhibit superheating under more uniform thermal conditions. To exclude the vapor from contact with the crystal a continuous coating is developed with a melting temperature in excess of that for the pure core crystal. For single crystal spherical Ag samples with a diameter of about 130 μm the free surface was coated with a continuous higher melting point layer of Au to inhibit surface initiated melting of silver at its melting point. The coated Ag crystals were observed to exhibit superheating of at least 7.5 K above the melting point of pure Ag for time periods of the order of one minute. Under the experimental conditions used, internal stresses could be excluded as the reason for the observed increase in the melting onset temperature of Ag. The superheating duration is related to the interdiffusion between the coating layer and the core crystal. Some microstructural observations and electron microprobe composition scans also relate to the observed superheating behavior. The results seem to disagree with lattice stability models of melting and may open the way to the study of crystals in the metastable superheated state.

Theory for the rate of nucleation of a crystalline solid from its melt is discussed. An expression for the dynamical prefactor of the nucleation rate is presented, which is derived using Langer's field theory of nucleation. The theoretical result generalizes the classical theory of Turnbull and Fisher, and can be tested experimentally. The analysis makes use of the formalism of Ramakrishnan and Yussouff, as extended to solid-melt interfaces by Oxtoby, Haymet, and Harrowell.

Small liquid germanium (ℓ-Ge) droplets (0.3–0.5mm diameter) have been undercooled in a B2O3 flux a maximum of 415 ± 20°C below Tm before recalescence. Although the droplets undercooled to temperatures well below the temperature expected for metastable equilibrium with amorphous Ge, Taℓ, the solidified structures were always polycrystalline. There was no evidence of a continuous liquid-to-amorphous transition at or below Taℓ. The nucleation frequency was calculated to be ˜8 × 106 /cm 3-sec at the maximum undercooling. This should be an upper limit to the homogeneous nucleation frequency of the crystal phase since we did not establish that the nucleation was homogeneous in our experiments. However, this limiting I for Ge indicates that homogeneous nucleation of the crystal or amorphous phase would not become appreciable in nsec laser pulsing experiments until ℓ-Ge is undercooled to well below Taℓ. The structure of solidified Ge and Ge/Sn alloy droplets indicates that dendritic growth does not occur in Ge until the liquid is undercooled at least 300°C, and in Ge/Sn until the liquid is undercooled at least 250°C. Greater undercoolings lead to increased dendritic breakup, which results in grain refinement in the final structure.

The role of nucleation kinetics in affecting glass formation behavior is discussed. Also considered are measurements of homogeneous crystal nucleation in a variety of liquids. For a number of oxide glass-forming liquids, available data indicate pre-exponential factors which are larger than those predicted from classical nucleation theory by factors of 1017 to 1049. Possible sources of this discrepancy are discussed.

Crystallization kinetics and crystal size distributions in Co33Zr67-glasses have been analyzed by quantitative electron microscopy. The polymorphic crystallization of spherical CoZr2 crystals is very suitable reaction for such an analysis. Calculated crystal size distributions at different temperatures were compared to those experimentally revealed. Parameters controlling crystallization were varied within reasonable limits until theoretically calculated and experimentally observed crystal size distributions were in good agreement. It has been found that crystal size distribution can be explained by transient nucleation; the time lag and its temperature dependence can be evaluated. These results are discussed in the light of recent theories on transient nucleation.

Analysis of the Cech-Turnbull small-particle martensitic transformation experiments in terms of heterogeneous nucleation theory defines an exponential nucleation-site potency distribution. Sensitive acoustic-emission detection of martensitic nucleation events shows that the same form of distribution describes the behavior of bulk polycrystals, and the influence of heat treatment on the distribution amplitude can be identified in this way. On the assumption of a random distribution of pre-existing nucleation-site orientations, the effect of applied stress on the effective potency distribution has been calculated, thus accounting for an observed nonlinear stress dependence of the transformation kinetics. The model also predicts a transformation “yield locus” for multiaxial stress.

Small ZrO2 particles were studied by in-situ transmission electron microscopy and the types of events which led to martensitic nucleation were recorded. These metastable particles were single crystals of several hundred nanometers size and were supported on a carbon thin film but were otherwise free from any constraint. At room temperature, they are more than 1200°C below the equilibrium temperature. After dislocation loops were injected in-situ into the originally defect-free particles, a critical loop size was found beyond which the particle transformed spontaneously. It was also observed that Hertzian contacts between two particles led to martensitic nucleation and spontaneous transformation. These results were satisfactorily rationalized by ab-initio theoretical analysis.

David Turnbull has been the most prominent figure in the identification of the metallic glass state. This paper reviews his many contributions: the free volume model for transport in simple liquids, the demonstration of a substantial crystal nucleation barrier in liquid metals, the prediction of the universality of the glass transition, the formulation of criteria for glass formation, the first demonstration of the characteristic glass transition phenomena in a metallic glass, and important contributions to the study of the structure and physical properties.

Inspired by the experiments of Turnbull (1952) on undercooled liquid mercury, many investigators have argued that polytetrahedral short-range order is an important ingredient in the structure of undercooled liquids and metallic glasses. A paradigm for such order is “polytope {3,3,5},” which is a regular lattice of 600 perfect tetrahedra with an icosahedral point symmetry embedded on the surface of a four-dimensional sphere. We present results for the band structure of polytope {3,3,5} in the presence of disclinations, and compare them with band structures obtained from dense random packing models of metallic glasses.

Although the structure of liquids and glasses may appear highly disordered, the atomic motions in these phases are rather strongly correlated. Such a correlation, however, cannot be readily described by a pair density function, and higher order correlation functions are necessary. We suggest that the microscopic stresses defined at each atom are quite useful in describing such high order correlations. We discuss how the concept of the microscopic stress fluctuation helps to describe and understand the glass transition and glass formation phenomena. In a molecular dynamics study the local shear stresses were found to become orientationally correlated below a transition temperature, Ts, which is close to the melting temperature. Below Ts the viscosity shows a rapid increase and transverse phonons become more stable. Thus Ts could be viewed as a dynamical glass transition temperature. The conventional glass transition can be described in terms of the elastic percolation of the orientationally correlated regions. On the other hand, the microscopic stresses in the crystalline solid solutions define the topological stability limit of the crystalline phase, and predict the minimum solute concentration to form stable amorphous alloys.

The differential anomalous scattering technique is outlined and compared to other techniques for studying short-range order in amorphous systems, such as EXAFS. The differential distribution functions obtained for liquid GeBr4 were found to support a model for the liquid state based on the structure of the h.c.p. crystal. Application of the technique to aqueous ZnBr2 also allowed discr imination between structural models.

Metal alkoxide syntheses of the general form or can be used to form soluble alkoxides of most metals. Mutual solution of these alkoxides followed by acid or base hydrolysis, drying and calcining results in the formation of fine-grained (<10Å) amorphous powders at low temperatures. Subsequent hot-pressing or calcining leads to densification and/or crystallization. Compositions in the MgO-Al2O3-SiO2 system with ZrO2 contents to 30 mol% are discussed. Also glass formation of transition oxides derived from metal alkoxides of Hf, Zr, Y and Nb in various combinations are presented.

Turnbull pointed out that good glass formability should correlate with a high reduced glass transition temperature, Trg. The alloy Pd40Ni40P20 has an unusually high Trg of ∼0.66 and indeed it can be formed into a glass at rather low cooling rates (≲ 1 Ks−1 ). In this paper a number of experiments are described which take advantage of the high resistance to crystallization in this alloy to study behavior in the labile temperature range between Tg and Tm. The anomalously low crystal growth velocities arise because the product is a three phase mixture. Nucleation kinetics are derived from crystal size distribution analysis. Heterogeneous nucleation at the surface of melt spun ribbon is characterized; significant surface relief develops on crystallization because of the low number of nuclei. In bulk glass produced by fluxing, good crystal size statistics show that internal nucleation is predominantly also heterogeneous, but with a significant transient time. Measurements of the specific heat of the undercooled liquid reveal that the free energy of crystallization deviates significantly from a linear dependence on undercooling (expected for pure metals) in the way predicted by Dubey and Ramachandrarao.

The process of nucleation and growth in glasses and undercooled liquids is modeled by directly simulating the evolution of the molecular cluster distribution under both isothermal and non-isothermal conditions. Results of that simulation for the nucleation rate during the quench, and for the number of nuclei produced and the volume fraction transformed at the end of the quench are presented. The following three points are discussed: (1) The importance of transient, or non-steady state, nucleation rates on glass formation is assessed by considering three model glass forming systems: lithium disilicate, a relatively good glass former, and two metallic glasses, (Au85Cu15)77Si9Gd14 and Au81Si19. (2) Using experimentally determined values for the steady state nucleation rates and growth velocities for Pd40Ni40P20, it is demonstrated that, in agreement with recent experimental results, this alloy may be cycled at rates on the order of 1 K/sec between the melting and glass transition temperatures without crystallization. Transient effects are shown to be unimportant under these conditions in this system. (3) The effect on glass formation of a non-equilibrium viscosity during the quench due to configurational freezing is evaluated by assuming a phenomenological model for the changing viscosity.

We report the use of microsecond current pulses to produce an amorphous metal alloy. Starting materials consisted of electron beam deposited films of elemental crystalline zirconium and nickel layers with a composition modulation wavelength of 46 nm, a total thickness of about 2.5 μn, and an average composition of Zr38Ni62 deposited onto sapphire substrates. Current pulses had a total ring time of several microseconds and were produced with a low inductance capacitor and a spark gap switch. Integration of current traces showed that 85% of the energy was deposited into the sample within 2 microseconds. The energy in the pulses could be controlled in a range where their effect on the films went from no visible change to total vaporization. Pulses with a peak current in the kiloampere range produced melting in 1 cm wide samples. The cooling rate following this treatment was sufficient to produce amorphous material. We report structural changes resulting from rapid melting and resolidification as well as results of a search for solid state transformations resulting from current pulses below the melting threshold.

Density and x-ray diffraction measurements were performed over a wide compositional range of U-Fe and U-Co metallic glass ribbons. The compositional dependence of both the density and the average nearest neighbor distance exhibits a change in slope at approximately 28 at.% transition metal. This suggests an “ideal” glass composition possessing a minimum excess volume. A qualitative model of atomic packing is proposed, which is based on the compositional dependence of the atomic volume and nearest neighbor distance.