The length scales accessible to experimental techniques for analysis of the three-dimensional chemistry of solids and feasible for atomistic simulation have converged. The three-dimensional atom probe is capable of reconstructing the chemistry of solids atom-by-atom with subnanometer resolution, while atomistic simulation techniques can calculate similar quantities for ensembles of hundreds of thousands of particles. We present recent progress in the calculation of the interfacial excess of a segregating species by both simulation and experiment.

The distribution of Ni, Al, Ta, and Re in a model Ni-base superalloy is investigated in the as-heat treated alloy by energy dispersive X-ray analysis and three-dimensional atom probe. Re was found to partition mainly to the γ phase in which it forms clusters with sizes of about 1 nm and mutual average distances of 20 nm.

The diffusion coefficients of impurities in ZrTiCuNiBe-bulk glasses with different Be/Ti ratio were measured in the temperature range around the glass transition using the technique of secondary ion mass spectrometry. During long time annealing, the Zr41Ti14Cu12.5Ni10Be22.5 glass (V1) undergoes an amorphous decomposition and the diffusion coefficients decrease by more than one order of magnitude. Two glasses, Zr41Ti8.7Cu12.5Ni10Be27.8 (D1) and Zr41Ti16.5Cu12.5Ni10Be20 (D2), which were produced to represent the decomposition phases of V1, also show a tendency of decomposition and time-dependent diffusion coefficients. Diffusion measurements performed at short annealing times yield only small differences between the temperature dependencies of the diffusion coefficients of the different impurities in all alloys. In general, the size rule is obeyed, i.e. the activation energies of diffusion increase with increasing atom size. The Arrhenius curves are not linear, indicating a significantly higher activation energy above about 580K than below this temperature. No systematic effect of glass composition on the diffusion is observed.

Three-dimensional atom-probe (3DAP) microscopy has been applied to the study of segregation at ceramic/metal (C/M) interfaces. In this article, results on the MgO/Cu(X) (where X = Ag or Sb) systems are summarized. Nanometer-size MgO precipitates with atomically clean and atomically sharp interfaces were prepared in these systems by internal oxidation. Segregation of the ternary component (Ag or Sb) at the MgO/Cu heterophase interface was enhanced by extended low-temperature anneals. Magnesia precipitates in the 3DAP reconstructions were delineated as isoconcentration surfaces, and segregation of each ternary component at the C/M interfaces was analyzed with the proximity histogram method developed at Northwestern University. This method allows the direct extraction of the Gibbsian interfacial excess of solute at the C/M interfaces from the experimental data. A value of (3.2+/-2.0) x 10(17)m(-2) at 500 degrees C is obtained for the segregation of Ag at a MgO/Cu(Ag) interface, while a value of (2.9+/-0.9) x 10(18) m(-2) at 500 degrees C is obtained for the segregation of Sb at a MgO/Cu(Sb) interface. The larger Gibbsian excess for Sb segregation at this ceramic/metal heterophase interface is most likely due to the so-called pdeltaV effect.

The Zr41Ti14Cu12.5Ni10Be22.5-bulk; glass undergoes significant microstructural alterations by an amorphous decomposition during annealing at temperatures well below the onset of crystallization. At the same time the diffusivity in this glass decreases with annealing. The time and temperature dependence of the diffusion coefficients of Al-, B-, Fe- and Co-impurities in the Zr41Ti14Cu12.5Ni10Be22.5-bulk glass were investigated in the temperature range around the glass transition by dynamical secondary ion mass spectrometry. The time constant of the decrease of the diffusion coefficients depends on the atomic size of the impurities, and is determined by a thermally activated process with similar activation energies for all impurities. The degree and time scale of the structural changes of the glass depends on the annealing temperature, and the characteristic times of the decomposition are of comparable magnitude to those found for the decreasing diffusion coefficients. The characteristic times of the decomposition are comparable to those of the decreasing diffusion coefficients. The microhardness increases with increasing diffusion length. We suppose a close relation between the decreasing diffusivity and the time dependence of the decomposition: on the one side the observed decomposition is controlled by the time dependent diffusion, and on the other side the diffusion is strongly reduced by the progressing decomposition.

The three-dimensional (3D) atom-probe technique produces a reconstruction of the elemental chemical identities and three-dimensional positions of atoms field evaporated from a sharply pointed metal specimen, with a local radius of curvature of less than 50 nm. The number of atoms collected can be on the order of one million, representing an analysis volume of approximately 20 nm x 20 nm x 200 nm (80,000 nm(3)). This large amount of data allows for the identification of microstructural features in a sample, such as grain or heterophase boundaries, if the feature density is large enough. Correlation of the measured atomic positions with these identified features results in an atom-by-atom description of the chemical environment of crystallographic defects. This article outlines a data compilation technique for the generation of composition profiles in the vicinity of interfaces in a geometrically independent way. This approach is applied to quantitative determination of interfacial segregation of silver at a MgO/Cu(Ag) heterophase interface.

The results of a three-dimensional atom probe (3DAP) analysis, on a subnanometer scale, of a ceramic/metal heterophase interface, MgO/Cu, are presented. Segregation of Ag, from the Cu (Ag) matrix, at MgO/Cu interfaces is investigated and the Gibbsian interfacial excess of silver is determined; the range is 2.33 x 10(18) to 5.81 x 10(18) m(-2). Also, silver segregation at the same MgO/Cu interfaces is analyzed employing a new approach that utilizes a proximity histogram or proxigram.

Diffusion of B, Fe, Co, Al and Hf in the bulk metallic glasses Zr46.7Ti8.3Cu7.5Ni10Be27.5 (V4) and Zr41Ti14Cu12.5Ni10Be22.5 (V1) was measured by using the technique of secondary ion mass spectrometry for concentration depth profiling. The Al and Hf tracers were selected as substitutes for Zr and Ti because their sizes are similar and the Hf belongs to the same period of periodic table. B, Fe, and Co diffusivities obey different Arrhenius laws above and below about 585 K. The activation energies in the high temperature regime are much larger compared to those in the lower temperature range. For the larger Al- and Hf-atoms no deviation in the investigated temperature region was detected from a simple Arrhenius-type behavior. A trend of an increasing activation energy with increasing size is found.

The Three Dimensional Atom Probe produces a real space map of the elemental identities and positions of atoms field-evaporated from a sharply pointed specimen. The analyzed volume is on the order of 20 nm × 20 nm × 100 nm. This is large enough to enclose microstructural features such as grain- or heterophase boundaries. Correlation of the measured atomic positions with such features results in an atom-by-atom description of the chemical environment of these crystallographic defects. We describe here a method for identifying these interfaces and profiling the composition in the vicinity of the interfaces without any assumptions about the interface geometry. This approach is applied to quantitative determination of interfacial segregation of Ag at a MgO/Cu(Ag) heterophase interface. We discuss the implications of our technique with respect to classical treatments of segregation at interfaces.

The Zr41Ti14Cu12.5Ni10Be22.5-bulk glass undergoes an amorphous decomposition during annealing at temperatures below the onset of crystallization. Accordingly the diffusivity in this glass decreases with annealing. This decrease of diffusivity was investigated by measurements of the diffusion coefficients of Al-, B-, Fe- and Co-, using the technique of secondary ion mass spectrometry. The time constant of the decrease of the diffusion coefficients is controlled by a thermally activated process with similar activation energy for all impurities. The characteristic times for the decomposition are of comparable magnitude to those found for the decreasing diffusion coefficients. We suppose a strong relation between diffusion and decomposition: on the one side decomposition is a diffusion controlled process, on the other side diffusion is strongly reduced by the progressive decomposition.

The decomposition of alloys can lead to concentration fluctuations with small amplitude and spatial extension. A technique of data evaluation, the repeated smoothing procedure (RSP), is described by which atom probe data of such alloys can be analysed. The procedure is applied to a Fe–Ni Invar alloy, annealed at 625°C, and to two Cu–Ni alloys, annealed at 300°C and at 350°C. Decomposition was observed and quantified in both alloy systems.

Grain boundary (gb) self-diffusion in pure and boron-doped Ni3Al was measured using the radiotracer Ni-63 a Serial sectioning technique and sensitive liquid scintillation counting. For comparison grain boundary self-diffusion in pure Ni was reinvestigated in type B-kinetics. It is shown that the temperature dependence of the grain boundary diffusion parameter P = delta D-gb of Ni-63 in pure Ni and in both Ni3Al materials is of the Arrhenius type. The absolute values of P follow the sequence P(Ni) > P(Ni3Al) > P(Ni3Al + B). Ordering of the lattice and the preservation of order up to the gb planes, as predicted in Ni3Al, therefore has a pronounced decelerating influence on gb diffusion. Ni-diffusivity in the doped alloy is about 2-3 times lower than in pure Ni3Al, attributed to the strong segregation of B in Ni3Al gbs, which may lead to an increase in the vacancy formation enthalpy and to a blocking of energetically favourable diffusion paths in the gbs.Applying the semi-empirical relation of Borisov et al. gb energies gamma(gb) were determined for arbitrary high angle gbs in pure and B-doped Ni3Al, resulting in 915 and 870 mJ/m(2), respectively, at 1100 K. The effect of stoichiometry on grain boundary self-diffusion in Ni3AI was investigated for compositions between 73 and 78 at% Ni. A V-shaped concentration dependence of P with a minimum near 75 at% Ni was observed. This behaviour is related to the change in composition and state of order of grain boundaries in non-stoichiometric Ni3Al.

Tracer diffusion of Ti-44 was measured between 1373 and 1126 K in four alloys in the composition range from 25 to 35 at% Al in ordered Ti3Al using the standard precision grinding sectioning technique and polycrystalline samples. The self-diffusion coefficient D*(Ti) was found to be lower than self-diffusion in pure alpha-Ti and to increase very little with Al concentration. D*(Ti) reveals Arrhenius behaviour which is described by the frequency factor D-0 = (2.44 (+1.81)(-1.04)). 10(-5) m(2)/s and the activation enthalpy Q(Ti) = (288.2 +/- 5.7) kJ/mol. Especially, the stoichiometric composition shows no distinguished diffusion behaviour.Interdiffusion was measured in single phase conditions by combining samples of 25 and 35 at% Al. The interdiffusion coefficient (D) over tilde was evaluated by Boltzmann-Matano analysis between 26 and 34 at% Al at different temperatures. Again, only a weak concentration dependence of (D) over tilde and an Arrhenius behaviour were detected. Applying Darken's equation, the self-diffusion coefficient of aluminium D*(Al) was calculated by combining D*(Ti) (X(Al), T) and (D) over tilde(X(Al),T) with the thermodynamic factor Phi(X(Al),T) in Ti3Al. D*(Al) turned out to be smaller than D*(Ti), e.g. by a factor of 6 at 1170 K. The Arrhenius relation is characterized by D-0 = (2.32 (+2.48)(-120)). 10(-1) m(2)/s and Q(Al) = (394.5 +/- 7.5) kJ/mol. Different jump possibilities of the Al atoms are discussed. From the overall diffusion behaviour in the Ti3Al-phase it is concluded that atomic migration proceeds via thermal vacancies. There is no indication regarding the formation of constitutional vacancies. Diffusion behaviour of the components Al and Ti in Ti3Al is compared with our recent results of Al impurity diffusion (SIMS analysis) and self-diffusion in pure alpha-Ti. (C) 1996 Elsevier Science Limited

Grain boundary (gb) self-diffusion in pure Ni-rich Ni3Al was measured between 882 and 1374 K using the radiotracer 63Ni, a serial sectioning technique and sensitive liquid scintillation counting. The results of the gb diffusivity P = δDgb (δ : gb width, Dgb : gb diffusion coefficient) can be represented by the Arrhenius parameters P0 = 3.27 · 1013and Qgb = 168 kJmol. Additionally gb diffusion was investigated in boron-doped (0.24 at%) Ni-rich Ni3Al in the range from 882 to 1352 K yielding P0 = 1.24 · 10−12 m3/s and Qgb = 187 kJmol. The increase in the activation enthalpy Qgb and the decrease of P upon boron-doping is explained by the segregation of B in Ni3Al gbs, which may lead to an increase in the vacancy formation enthalpy and to a blocking of energetically favourable diffusion paths in the gbs. For comparison gb self-diffusion in pure Ni was remeasured yielding Qgb = 112 kJmol. Ordering of the lattice and the preservation of ordering up to the gb planes, as predicted in Ni3Al, therefore has a pronounced decelerating influence on gb diffusion, stronger than on bulk diffusion. Applying the semi-empirical relation of Borisov et al. (Phys. Met. Metallogr., 17 (1964) 80) gb energies γgb were determined for arbitrary high angle gbs in pure and B-doped Ni3Al, resulting in 915 and 870 mJ/m2, respectively, at 1100 K.

Temperature and concentration dependence of the diffusion of Ti and Al has been measured in the ordered {alpha}{sub 2}-phase Ti{sub 3}Al. Ti self-diffusion is almost independent of the Al concentration in the hyperstoichiometric {alpha}{sub 2}-phase. The absolute D*{sub Ti}-values are rather low. Obviously diffusion proceeds by the vacancy mechanism and no constitutional vacancies but Al antisite defects are formed. Chemical diffusion {tilde D} is larger than self-diffusion D*{sub Ti} due to the fact that the thermodynamic factor {Phi} is larger than unity in the {alpha}{sub 2}-phase. Al self-diffusion D*{sub Al} was deduced to be lower than D*{sub Ti} by a factor of two to three in the range of 1,273 K to 1,173 K.