MORRIS COHEN

The idea of interrelationships among processing, structure, property, and performance in materials—a concept that now forms the backbone of modern materials science and engineering ¹ —began to take shape in metallurgy well over a century ago with the advent of the metallographic microscope. ² Processing, properties, and performance of metals and alloys had been known to mankind in one way or another for millennia, but the dawn of metallurgical science as we view it today might well be identified with the emergence of microstructure. That scientific event forged a connecting link to the practice of metallurgical engineering and technology, and all these elements progressively coalesced into what has become the discipline of metallurgy.

Accordingly, to highlight some of the significant accomplishments and anticipations of metallurgical research, we shall regard metallurgy as the science, engineering, and technology of metallic materials. In other terms, it covers the study, production, manipulation, and use of metals and alloys; it is the part of materials science and engineering that not only inquires into the nature of metallic materials but also attempts to harness such knowledge for societal purposes.

The following summary reflects the diverse inputs of some 80 contributors in metallurgical research whose multifaceted cooperation in this task is gratefully acknowledged. To the extent possible, the numerous suggestions and viewpoints have been unilaterally blended into topical themes, specifically selected because of their promise for new ferment in metallurgical research.

It should be emphasized at the outset that virtually all phenomena in metallurgy are incompletely understood, but this is likewise the case for most aspects of any science or technology. Indeed, arguments can be made for needing further research and insight on just about any identifiable subject in materials science and engineering. Nevertheless, some gaps in metallurgical knowledge are long-standing and costly to society. For example, the damage from metal corrosion and other types of failure in service amounts to more than $200 billion per year in 1982 dollars, ³ a loss that is commensurate with the annual federal deficit! One may certainly wonder why the nation puts up with this appalling extravagance. The probable answer is that metals, as reflected in Figure 1, ⁴ exhibit combined ranges of properties not enjoyed by other classes of materials (in stiffness, strength, toughness, and thermal

characteristics, as well as in economic availability) that make them incredibly useful to humanity. In a similar vein, society seems to “accept” the loss of some 50,000 lives per year on U.S. highways because automobiles are so useful. Metals, like automobiles, have become engrained in our way of life; neither should be taken for granted.

Every step in the materials cycle involves processing. In metallurgy, processing comprises all operations that produce, shape, and control the properties of metallic materials to make them perform effectively in service. Metal processing ranges from large scale, as in steel production and refining, to small scale, as in thin-film formation.

Advances in steel refining, particularly through ladle metallurgy involving the injection of calcium, rare earths, and fluxes as well as vacuum degassing, have led to remarkable improvements in compositional control and steel cleanliness. Figure 2 shows the pronounced increase in through-thickness ductility of plate steels that results from the reduction of sulfur content and, hence, of sulfide inclusions. ⁵ This development, growing out of investigations on the multicomponent thermodynamics of sulfur in liquid steel, has spawned the reliable use in pipeline steels and oil-drilling rigs of heavy sections resistant to laminar tearing. The achievement of such metallurgical control in liquid steel is all the more prodigious when one considers the 100- to 200-ton scale of the ladle-refining operations.

Of comparable importance is the potential removal of phosphorus. Rare-earth additions have promise in this respect ⁶ and deserve the kind of research attention that has been given to calcium in its affinity for sulfur. Phosphorus increases the sensitivity of high-strength steels to embrittlement, and its elimination would constitute a major advance in ferrous metallurgy.

An example of solid-state processing that has reached commercial scale in steelmaking is the controlled rolling of steels that are microalloyed with relatively stable carbide- and nitride-forming elements such as niobium, vanadium, and titanium. Here, the precipitation of carbonitride particles is induced by plastic deformation of the austenitic phase during the rolling schedule, and recrystallization of the austenite is then inhibited because the grain boundaries are pinned by the precipitated particles. Further rolling to low finishing temperatures (but still substantially in the austenitic range) leads to a flattening of the austenitic grains, the thin dimension of which contributes to a very fine-grained ferritic structure on further cooling.

The resulting grain refinement, thus attained without the expense of separate heat treatment, contributes both strengthening and toughening, as shown in Figures 3 and 4. ^{7 , 8} This kind of double benefit in structure-property relations is an unusual circumstance in alloy systems, and it forms the basis of the high-strength, low-alloy (HSLA) steels that have been developed during the past quarter century. An additional advantage of these steels is that their beneficial properties are obtainable with low carbon contents and therefore are compatible with excellent welding characteristics for construction purposes. One measure of the technological success of HSLA steels is their increasing utilization in automobiles, despite the general downsizing for fuel efficiency. According to Figure 5, the amount of HSLA steel used per vehicle has been increasing even more rapidly than the lower-density aluminum alloys, plastics, and composites. ⁹

The pioneering publications of Pond ¹⁰ in 1958 and Klement, Willens, and Duwez ¹¹ in 1960 ushered in a dynamic era of research and development on rapid solidification processing (RSP). The major part of this research has been directed to metallic materials, although the phenomena at play are

operative in other classes of materials as well. Cooling rates up to 10⁹ K/s have been reported, but most of the basic effects are observed with cooling rates of 10⁴ to 10⁶ K/s attained by splat quenching, melt spinning, planar-flow casting, atomized droplet solifidication, or self-quenching after surface melting. The corresponding solidification rates, in terms of liquid-solid interfacial velocities, can range up to tens of meters per second, in contrast to about 1 cm/s for a typical mold casting. The microstructures resulting from RSP are characterized by enhanced compositional uniformity, refinement of the microconstituents, high degrees of supersaturation, and retention of metastable phases including metallic glasses. The aluminum-manganese phase with anomalous fivefold symmetry ¹² (see Cahn and Gratias, in this volume) is a startling example of metastability brought to light by RSP. The existence of such a pentagonal structure is distinctly forbidden by long-standing crystallographic theory, thus posing a fundamental dilemma. One rationalization is to regard the strange structure as quasi-periodic instead of truly periodic.

The retention of metastable glassy states in certain alloy systems by RSP offers a classic example of a novel processing method that has paved the way to new regimes of structure, property, and performance relationships. For melt compositions whose glass transition temperatures are about half the respective melting points, a cooling rate of approximately 10⁶ K/s is sufficient to avoid crystal nucleation and thus allow glass formation. Metallic glasses are typically high in strength but are not “glass-brittle.” They tend to undergo localized shear (due to lack of strain hardening) when stressed beyond the yield strength, and thus they resist fracture even though they are not generally deformable like crystalline alloys. The metallic glasses based on metal (Fe, Ni, Co) and metalloid (B, Si, P, C) combinations are of special interest in view of their markedly low magnetic losses, and the processing is potentially inexpensive because of direct casting of the alloy liquid to final strip form. The technological impact of this metal-processing development is discussed below in the section on magnetic alloys.

The extensive refinement of dendritic structures caused by increased cooling rate or growth velocity during the solidification of crystalline alloys is shown by Figure 6. ¹³ The reduction in dendritic-arm spacing not only improves the as-cast strength and ductility ¹⁴ but also promotes compositional uniformity by decreasing the diffusion distances between the regions of microsegregation formed by solute buildup in the last pockets of liquid to solidify between the dendritic arms. Such compositional uniformity is advantageous in raising the incipient melting temperature of alloys intended for high-temperature service, as in the case of superalloys. At the same time, second phases that are likely to precipitate in the microsegregated regions tend to be finer in size and more uniformly distributed because of the rapid solidification. These second phases often appear as intermetallic compounds or nonmetallic inclusions that are embrittling when present in coarse or segregated form but that can be desirable when finely divided and well dispersed.

A beneficial consequence of uniform dispersions is that they pin grain boundaries of the matrix phase and thus inhibit grain growth, as shown in Figure 7. ¹⁵ The effectiveness of this phenomenon is an inverse function of d/f _v , where d is the average diameter of the precipitated particles and f _v is their volume fraction. This means that for the grain boundary pinning to persist at very high temperatures, the distributed phase must be sufficiently

stable (minimal solubility in the matrix phase) to resist Ostwald ripening. Such phases are generally more soluble in the liquid state (e.g., oxides, silicates, and oxysulfides in liquid steel) and are then amenable to uniform dispersal as fine precipitates in the solid state by RSP. The resulting resistance to grain growth permits the consolidation of rapidly solidified particulates and subsequent heat treatment to be carried out at relatively high temperatures without undue grain coarsening. Moreover, with larger-volume fractions of stable second phases, including appropriate intermetallic compounds, dispersion strengthening can be achieved and maintained at elevated temperatures.

An example of second-phase refinement by RSP is shown in Figure 8. ¹⁶ Here the addition of 10 weight percent beryllium to a commercial aluminum-copper alloy leads to extremely coarse second phases upon normal ingot solidification, whereas RSP produces a very fine dispersion of the second phases. In this instance, as much as 10 weight percent beryllium can be dissolved in the liquid alloy and then precipitated by rapid solidification as a finely distributed beryllium-containing compound (approximately Be₃Cu) in the solid alloy, with the maximum equilibrium solubility of the beryllium in solid aluminum being only about 0.03 weight percent. The corresponding improvements in strength and ductility given in the caption of Figure 8 are also worthy of note. The large beryllium addition made possible by RSP is of further potential value because it increases the stiffness and decreases the density of aluminum-base alloys. ¹⁶

Figure 9 compares the high-temperature strength of RSP aluminum alloys with that of conventionally processed high-strength aluminum alloys. ¹⁷ For these studies, the compositions of the RSP alloys were specially developed

to take advantage of RSP; the intermetallic phases involved would be much too coarse and embrittling if they were to form during regular solidification. The strength retention exhibited by the RSP alloys at elevated temperatures is quite striking and is beginning to match or exceed the density-compensated strength of titanium alloys over the temperature range studied. Surely the time is now ripe to apply RSP to titanium alloys and possibly even to niobiumbase alloys.

The RSP aluminum alloys are also unusually resistant to corrosion, ac

cording to the salt-fog data in Figure 10. ¹⁷ The observed weight losses are primarily due to pitting, signifying that a novel phenomenon may be inhibiting the pitting attack. Conceivably, in the RSP aluminum alloys, the pitting sites are sufficiently close together that the localized galvanic cells tend to shield one another. The alloy chemistry per se may also be playing an important role that would not otherwise be revealed except for the compositional uniformity contributed by rapid solidification.

In any event, the operative phenomena are worthy of deep study, consid-

ering their extraordinary chemical effects and prospective benefits for engineering applications. Even more enticing is that RSP magnesium alloys also display enhanced resistance to corrosion ¹⁸ but again the fundamentals are not yet clear.

No less graphic is the effect of RSP on the oxidation resistance of high-temperature alloys, including superalloys; the oxide scales that normally develop on such alloys then become more adherent and protective. With RSP stainless steels, ¹⁹ the fine grain size provides extra paths for chromium atoms to reach the surface quickly and so maintain a chromium oxide layer. If the chromium supply from the base metal is deficient, iron oxide begins to form in the outer scale and renders it less protective. Because of the grain-boundary diffusion mechanism, oxide phases form to some extent at the outer grain-boundary junctures in continuity with the surface oxide layer, perhaps helping to lock the scale to the base alloy. There are also indications that finely dispersed stable phases, such as hafnium and yttrium oxides or sulfides, may participate in the oxidation resistance. Although adequate understanding of the relevant phenomena is still lacking, the effects can be dramatic. Figure 11 ²⁰ shows the results of a severe, cyclic oxidation test on two RSP superalloy specimens, with and without the addition of yttrium. It is obvious here that a fertile field of basic research on oxide-scale formation lies ahead, with farreaching technological implications.

The second-phase precipitates in RSP alloys can be refined even further if the rapid solidification occurs under conditions of solute trapping—i.e., when the interfacial velocity is fast enough (whether by supercooling or by sufficiently rapid heat extraction) to prevent equilibrium partitioning between the liquid and the solid at the growth front. If precipitation is then allowed to take place from the resulting highly supersaturated solid solution rather than during the prior solidification process, the particles can be a factor of 10 smaller than in the absence of solute trapping. As indicated by Figure 12, ²¹ solute trapping (k>k ₀ ) sets in when the interfacial velocity reaches within an order of magnitude of the diffusive velocity (D _L /λ, where D _L is the solute diffusivity in the liquid phase and λ is the diffusion jump distance). The solute trapping becomes substantially complete (k≅1) when the interfacial velocity reaches about an order of magnitude greater than the diffusive velocity. It is evident, then, that one of the future thrusts of RSP research lies in the direction of hitherto unexplored alloy systems and compositional regimes that can be accessed by faster solidification rates, thereby opening up new ranges of structure-property relationships. In this approach, the formation of exotic metastable phases (such as the aforementioned metallic glasses and the “forbidden” fivefold structure in the aluminum-manganese system) should become more common in view of the suppressed equilibrium reactions and the greater opportunities for less stable phases to compete in the kinetics of nucleation and growth. Although the emergence and nature

of such phases cannot be predicted yet, one can confidently anticipate that surprises will continue to arise out of this netherworld of metastability.

The advent of processing methods based on the injection of accelerated ions (10 to 500 keV) into solid metals and alloys is leading to a new world

of metallurgy through the modification of near-surface properties. Figure 13 ²² shows a schematic representation of compositional changes and lattice damages resulting from ion implantation and atomic mixing. The latter process, shown here for atomic interchanges between a thin film and its substrate, can be brought about by the ion beam itself or, alternatively, by a laser beam. Extended solid solutions up to 10 atomic percent of almost any kind of atom in any base metal can be achieved to depths of hundreds of atomic layers. The effective quenching rates are estimated to range up to 10¹⁴ K/s. Ion implantation and laser-beam mixing are compared in Table 1. ²³

Although many alloy systems and compositional variations have been investigated by these surface-modification methods, at least to some extent, ²⁴ much of the research impetus has been directed to semiconductor and electronic materials. On the other hand, the metallurgical community has shown particular interest in potential applications requiring resistance to surface-initiated deterioration in service, such as wear, corrosion, erosion, and fatigue failure of high-performance alloys. ²⁵ Undoubtedly there will be considerable effort to exploit these intriguing possibilities for tools, dies, bearings, nozzles, orthopedic implants, and numerous other industrial components.

However, in parallel with these prospective applications, ion implantation and beam mixing deserve more attention in metallurgical research as pathways for attaining new microstructural states that have not been reached before. The opportunities for extreme supersaturations and metastable-phase

formation in the ion-bombarded surface layers are virtually limitless. Furthermore, high-resolution, high-sensitivity microanalytical techniques for probing such states of matter and their transformation products are now available. The following accounts of experiments illustrate some novel research approaches for taking advantage of ion implantation and beam mixing.

The curves in Figure 14 show the yttrium-concentration/depth profiles in a cobalt-45 atomic percent chromium alloy after low (2×10¹⁴ Y/cm²) and high (2×10¹⁶ Y/cm²) dosages of yttrium ion implantation, followed by an oxidation exposure of both specimens at 1273 K to produce a chromium oxide scale. ²⁶ The fact that the implanted yttrium concentration peaks near the gas/oxide surface in the specimen having the higher yttrium dosage, whereas it peaks near the oxide/alloy interface in the specimen having the lower yttrium dosage, indicates that the scale growth mechanism is fundamentally different in the two cases. This difference signifies that the yttrium-concentration profiles are acting as Kirkendall markers during the scale growth process and demonstrates that the diffusion flux in the higher-dose specimen is carried mainly by the O^2– ions passing through the scale from the outer surface of the oxide to the inner oxide/alloy interface, whereas in the lower-dose specimen the diffusion flux is carried mainly by the Cr³⁺ ions passing through the scale in the opposite direction. Correspondingly, there is a profound contrast in the oxidation rates of the two specimens, as shown in Figure

15; ²⁶ the higher-dose specimen (whose oxidation is controlled by O^2– diffusion) is much more resistant to oxidation than is the lower-dose specimen (whose oxidation is controlled by Cr³⁺ diffusion). Another surprising observation is that microanalytical measurements (by energy-dispersive x-ray analysis) demonstrates that the yttrium in the scale is located almost exclusively in the oxide grain boundaries without detectable presence of second phases. ²⁷ It is tempting to speculate that, in the higher-dose specimens, the implanted yttrium is present in sufficient concentration along the oxide grain boundaries to inhibit the Cr³⁺ ion diffusion that would otherwise select these short-circuiting paths, and so the rate-controlling oxidation mechanism is then shifted to the slower O^2– diffusion.

Clearly, there are new challenges in these findings, not only from the standpoint of ion implantation research per se, but also as a technique for understanding and using the potent oxidation resistance contributed by small percentages of elements like yttrium in high-temperature alloys.

The lattice defects introduced by ion implantation can trap hydrogen, a circumstance that suggests a method for reducing or controlling hydrogen embrittlement in steel. Research along these lines should have great appeal in ferrous metallurgy because the efficacy of high-strength steels is frequently undermined by the detrimental effects of trace amounts of dissolved hydrogen, such as may be picked up even superficially from a moist environment. It has been discovered that such defect traps become very potent when iron is ion-implanted with yttrium. Experiments on the temperature-dependent release of deuterium from defect traps in yttrium-implanted and in iron-implanted iron specimens are compared in Figure 16. ²⁸ It is evident that the deuterium is much more tightly bound to the yttrium-implanted traps than to the iron-implanted traps. The question that immediately arises is whether hydrogen embrittlement in hardened steel might well be ameliorated by ion implantation with yttrium or similar ions; it is a possibility that warrants high-priority study.

Ion- or laser-beam mixing, referred to earlier, can interchange atoms between a surface film and its substrate, often forming a new phase by “skin melting” and rapid solidification. It is conceivable that this mixing process could be made more efficient, and more varied morphologically, if the specimen itself were to consist of a very fine dispersion embedded in the matrix phase, prepared in the first place by RSP. Pulsed-beam mixing under these conditions could lead to further microstructural refinement or the formation of hitherto-unknown metastable phases. This type of surface treatment deserves detailed exploration. Unique mechanical behavior (as well as the corrosion and oxidation resistance described previously) may be expected from still finer structures; this is discussed more specifically in a later section on microstructural refinement.

The use of single crystals in metallurgical research and in microelectronic devices is well known, but it was only in 1982 that monocrystalline alloys first entered service as a critical structural component. It was a high-technology application in every sense of the word, following two decades of research and development on cast superalloy turbine blades. ^{29 , 30} As a result, single-crystal gas-turbine blades are now performing advantageously in both commercial and military aircraft engines. ³⁰

There are high-temperature applications in which the grain boundaries and random grain orientations of polycrystalline superalloys are undesirable, in part because of thermal fatigue failure and creep rupture along such interfaces. This motivated the development of directional solidification during the 1970s to produce columnar grains with controlled orientation and with grain boundaries parallel to the main stress direction. The next major step was to eliminate the grain boundaries entirely; this was made possible and practical by the single-crystal casting process shown schematically in Figure 17. ³⁰ The helical

channel above the water-cooled solidification “starter block” permits only a single growing grain to pass through, normally with a 〈001〉 crystallographic direction advancing parallel to the longitudinal axis of the turbine blade. However, seeding also permits the selection of other orientations for the study of anisotropic properties.

Ordinarily, the optimal composition of superalloys is complicated by a menu of alloying elements to help stabilize the grain boundaries at elevated temperatures. Since this is not necessary with monocrystals, the compositional variations can be focused on property control of the base alloy—e.g., improvement of creep strength as a function of temperature. The progress along these lines has been truly remarkable. Figure 18 ³¹ indicates that the single-crystal processing of superalloy turbine blades has led to a quantum jump in high-temperature operating capability, and the way is now open to substantial further improvements, with corresponding benefits in engine performance and fuel efficiency.

Notwithstanding the sharp advance in gas turbine technology resulting from the single-crystal processing of turbine blades, other metallurgical fronts will see a strong push toward finer and finer microstructures, both in grain size of the matrix phase and in particle size of the embedded precipitates. This is one of the promising aspects of RSP, as described earlier. Extending the range of structure-property relationships through microstructural refinement now constitutes a frontier of metallurgical research reaching down into the nanometer regime.

Dispersed precipitates are effective in obstructing dislocation motion and, hence, in raising the strength of alloys, but such precipitates can also be detrimental to toughness because of void initiation, which promotes shear localization and fracture. It has been estimated, ³² however, that particles smaller than about 20 nm should be subcritical in size relative to void nucleation, whereas particles larger than about 1 nm should be capable of resisting dislocation motion. Thus, a “window of opportunity” exists in this size range for optimum dispersion strengthening and toughening. A further contribution to strength and toughness is obtainable from grain refinement of the matrix phase; in fact, no limit has yet been established in these favorable trends with diminishing grain size. Consequently, they should be exploited, or at least tested, by processing methods now available.

These guidelines have recently been adopted in a comprehensive initiative to increase the fracture toughness of ultrahigh-strength martensitic steels and, at the same time, to test the theoretical and practical limits of the relevant structure-property relationships due to microstructural refinement. ³³ In view of the obvious technological impact of such high-performance steels, several university, governmental, and industrial laboratories are now participating in this research endeavor.

Fine-grained microstructures can be advantageous for plastic-forming operations at elevated temperatures by inducing superplasticity. With rising temperature, grain boundaries in alloy systems tend to lose their strengthening capability and participate in new modes of plastic deformation. If grain growth can be inhibited to maintain a sufficiently fine grain size, typically through the presence of a second phase, as discussed earlier, extensive plasticity may be encountered—e.g., several hundred percent elongation in a tensile test.

Figure 19 summarizes the essential mechanical behavior characteristics of superplasticity. ³⁴ High strain-rate sensitivity (m) is required to achieve stable necking-free elongation, and this condition is favored by an optimum strain rate (usually rather slow), which can be increased with decreasing grain size and with increasing temperature (if grain growth is inhibited). An impressive example of superplastic forming is shown in Figure 20, which illustrates the closed-die forging of a superalloy gas-turbine wheel, with integral blades, in only two steps from a hot-extruded billet of RSP powders. ²⁰ Many other industrial shaping operations based on superplasticity are now in commercial practice. ³⁵

Microstructural models to account for superplastic behavior commonly involve diffusion-aided grain-neighbor switching in the overall process of contributing large macroscopic displacements while generally tending to maintain a steady-state grain size and shape. One such model is idealized in Figure 21. ³⁶ Clearly, the grain boundaries play a crucial role in superplasticity, hence the necessity for a small, stable grain size. Accordingly, the aforementioned thrust toward greater microstructural refinement should lead

to ease of deformation processing at high temperatures as well as to improved mechanical properties at low temperatures—an extremely attractive prospect indeed.

Extending microstructural refinement down to the range of nanostructures has been made feasible by evaporation and fine-particle condensation techniques, as shown schematically in Figure 22. ³⁷ Evaporation from the molten metal, or from a number of separate melts, is conducted in an inert gas at low pressure, and the resulting powders (condensed on a cold finger at the liquid nitrogen temperature) have an average particle size of several nanometers. The powders can then be examined as such or after being cold-pressed into bulk form.

When the compacted specimen grain size is about 5 nm, there are approximately 10¹⁹ cm^–3 boundaries between adjacent grains, and the crystalline and interfacial “phases” have nearly the same volume fraction, assuming a grain boundary thickness of 1 to 2 nm (Figure 23). Observations by x-ray diffraction, Mössbauer spectroscopy, and positron annihilation on compacted nanocrystalline iron indicate that the arrangement of atoms at the grain boundaries is gas-like, rather than crystalline or glass-like; the grain boundary structure appears to have no short-range order when integrated over the

enormous number of adjacent-crystal pairs and their respective orientations. Nanocrystalline, glassy, and normal crystalline properties of various metals are compared in Table 2. ³⁷

By and large, properties that differ by less than 10 percent between the crystalline and glassy states will differ very much more from those of the nanocrystalline state. The phonon-dependent thermal-expansion and specific-heat differences in Table 2 have been interpreted to signify that the grain boundary structure in nanocrystalline metals constitutes a novel state of matter in solids. On the other hand, the remarkable strength of the nanocrystalline iron-carbon alloy is reasonably commensurate with the Hall-Petch strengthening to be expected from grain refinement alone; in fact, this magnitude of strengthening is also found with pure nanocrystalline iron. ³⁸ The 1.8 weight percent carbon in the iron-carbon alloy shown in Table 2 is presumably present as nanocrystals of carbon—i.e., not atomically dissolved in the iron. Such “alloy mixtures” deserve detailed investigation of their structural changes on heating as well as their physical, chemical, and mechanical properties. Experiments along these lines, reaching into the nanoscale grain-size regime and generating a correspondingly high concentration of grain boundary material, represent a new research frontier not only for metals and alloys but for other classes of materials as well. Moreover, the evaporation and condensation method for accessing ultrafine grain sizes in complex alloy systems has the unique feature of not depending on mutual solubility in the liquid state to carry out the alloying.

A different type of microstructural refinement is displayed by multilayered or modulated structures, and the resulting properties are most intriguing. In the solid-state community, the terms “artificially structured materials” and “strained-layer superlattices” have been adopted, particularly when semiconducting, optical, and magnetic phenomena are of primary interest. This field has been reviewed quite comprehensively in a recent report of the Solid State Sciences Committee of the National Research Council. ³⁹ For present purposes, however, we shall focus on the mechanical behavior of metallic systems, in line with the metallurgical emphasis of this chapter.

A number of vapor-deposited compositionally modulated alloy systems (e.g., copper-nickel, gold-nickel, copper-palladium, and silver-palladium) exhibit the strange elastic characteristics shown in Figure 24. ⁴⁰ The biaxial modulus plotted against the wavelength of layered Cu-Ni specimens is determined from a miniature bulge test and is found to peak sharply at a wavelength of about 20 angstroms. ⁴⁰ This “supermodulus effect” represents a phenomenal increase in stiffness, amounting to a factor of 2 to 4 over that of a homogeneous Cu-Ni alloy, and indicates that a significant change compared to stiffness in bulk materials has taken place.

As yet, we have no generally satisfactory theory for this unexpected enhancement in stiffness. One suggestion attributes the effect to coherency strains of approximately a few percent owing to lattice matching between the alternating layers, ⁴¹ but thus far the supermodulus behavior has been observed only with alloy systems in which one of the components is a transition metal. For example, the modulus effect is not seen with compositionally layered copper-gold alloys. ⁴² Another explanation is electronic in nature, namely, that the periodicity of the modulated structure introduces a Brillouin zone which happens to contact a flat portion of the Fermi surface when the layered wavelength is optimal. ⁴⁰ This circumstance is thought to produce a stable energy configuration, which then contributes a large measure of resistance to imposed elastic deformation.

However, more recently, separate elastic moduli (Young’s, flexural, and torsion) have been measured on multilayered Cu-Ni specimens having the same average composition and the same wavelength range as for the biaxial modulus measurements in Figure 24. It is surprising to find that the newer results in Figure 25 show two distinct peaks—at 12 and 28 angstroms—as a function of modulation wavelength. ⁴³ The elasticity interrelationships are such that the double peaks in Figure 25 are consistent with the single-peak biaxial measurements in Figure 24, but now a proper theory of the supermodulus effect must account for sharp enhancements in elastic constants at two optimal wavelengths. Obviously, a bizarre aspect of metal science has emerged here, with fertile opportunities for interactive experimental and theoretical research toward a deeper insight into the basic nature of the metallic state.

Modulated structures arising from the spinodal decomposition of unstable solid solutions are well known, but a relatively new finding is that virgin iron-carbon martensites also decompose spinodally into modulated structures of coherent high- and low-carbon interstitial solid solutions. This happens on aging, even below room temperature, before there is any evidence of the usual carbide precipitation sequences that occur later during tempering. As indicated in Figure 26, ⁴⁴ the carbon content of the high-carbon regions increases progressively (by uphill diffusion) and levels off at approximately 11 atomic percent. (This carbon content corresponds to the composition Fe₈C, which has not yet been established as a carbide phase, but the analogous compound Fe₁₆N₂ is known.) The wavelength of the modulations starts at about 20 angstroms and then undergoes coarsening. During the latter stage,

the carbon content of the low-carbon regions becomes virtually nil, and in view of the coherency between the high- and low-carbon modulations, the elastic strains may reach the incredible level of 17 percent. It has not been realized before that such layered structures and the attendant coherency strains might be playing a substantial role in the classic strengthening of iron-carbon martensites—a matter of vast industrial importance. This long-standing challenge in ferrous metallurgy has defied all theories of steel hardening over the years, but now the existence of fine-scale compositional modulations offers a new outlook for a reinvigorated theoretical attack on the problem.

The use of layered structures for interdiffusion measurements is well known ⁴⁵ and need not be reviewed here. However, there is one related phenomenon that warrants particular attention as a technique for synthesizing bulk metallic glasses. ⁴⁶ In particular, the hypothetical alloy system shown in Figure 27 can form metastable intermediate phases if the more stable intermediate phases are sufficiently slow to nucleate during interdiffusion between layers of the pure metals A and B. ⁴⁷ In such a case, one of the metastable phases may be a liquid, and, if the interdiffusion temperature happens to lie below the glass transition temperature of the supercooled liquid, a metallic glass will then form at the expense of the crystalline A and B components, starting at each A/B interface in the multilayered specimen, as in Figure 28. ⁴⁶ Examples of alloy systems that behave in this way are gold-lanthanum, nickelhafnium, and cobalt-zirconium; in each instance, the first component is a “fast diffuser” whereas the second is a “slow diffuser.” Apparently, this condition tends to inhibit the nucleation of crystalline phases while favoring the rapid formation and thickening of a glassy phase. ⁴⁷

In a real sense under these circumstances, the two crystalline metals A and B dissolve in, or melt into, the growing glassy phase, notwithstanding the relatively low interdiffusion temperature involved. This process can result in the production of homogeneous glassy alloys for the kinds of bulk property measurements and test purposes that have not been feasible before. It is also

evident that, in such metallic systems, bulk glasses can likewise be synthesized from compacted powders of the component metals, thus greatly facilitating appropriate specimen preparation. One can easily visualize new directions of research activity emerging from these recent studies of interdiffusion on a fine scale.

Long-range ordered phases and intermetallic compounds have long been a familiar subject in physical metallurgy, particularly from the standpoints of crystal structure, thermodynamic equilibrium, kinetics of formation, and second-phase strengthening dispersions. Both long- and short-range ordering have been studied in detail. But more recently there has been a new wave of interest in such alloy systems because of the ductility-related mechanical behavior of aluminides and similar phases, especially those having cubic Ni₃Al-type (L1₂) structures. It is a nice coincidence that intermetallic compounds have also come into prominence because of unusual ferromagnetic properties; this is discussed later in the section on magnetic alloys.

Excellent reviews of structure-property relationships in ductile ordered alloys have appeared within the past few years. ^{48 – 53} One reason for this focus has been the surprising degree of ductility and fabricability exhibited by these materials—a kind of mechanical behavior not commonly associated with intermetallic phases. It turns out that many L1₂ compounds are ductile at ambient temperatures both in the monocrystalline and polycrystalline states, but a well-studied exception is polycrystalline Ni₃Al, which suffers from intergranular weakness. It has been found, however, that small amounts of boron will segregate to the grain boundaries, strengthen these interfaces, and thus restore a level of ductility commensurate with that of the crystalline state. The effect of boron on the ductility and strength of polycrystalline Ni₃Al (24 atomic percent aluminum) is shown in Figure 29. ⁵⁴ There is increasing evidence that the beneficial influence of boron is not one of simply displacing a harmful impurity (such as sulfur) from the grain boundaries but rather is a matter of enhancing the electronic bonding across the interface. For this reason the role of boron seems to be quite specific: it works superbly only on the nickel-rich side of Ni₃Al stoichiometry, but not on the aluminum-rich side, and it is not necessary for grain boundary strengthening in numerous other L1₂ alloys. In fact, the interfacial strength and hence intergranular ductility have been shown to increase with the valence difference between the two constituent atoms in a wide variety of L1₂-type A ₃ B phases. ⁵⁵ This idea has been further substantiated by corresponding studies of ternary additions to Ni₃Al alloys. ⁵⁶

Many L1₂ intermetallic phases exhibit anomalous strengthening with increasing test temperature, often up to nearly the disordering temperature; Figure 30 shows several examples. ⁵⁷ This strange mechanical behavior raises lively questions regarding both its origin and its potential use for high-temperature service. It should be emphasized, however, that the anomalous temperature dependence may be displayed to only a minor degree, or even not at all, by some L1₂ intermetallic phases. Nevertheless, we can expect that diffusion-dependent creep will be significantly retarded because of the relatively low atomic mobility typical of ordered lattices, as discussed below.

An important aspect of this increase in strength with temperature is associated with an unusually pronounced effect of temperature on strain hardening. It is evident from Figure 31 that the peak in high-temperature flow stress is primarily due to the operative strain hardening, which sets in at rather small plastic strains. ⁵⁸ A feasible explanation of this extraordinary behavior is the cross-slip mechanism shown in Figure 32 and described in the caption. ⁵⁹ Clearly, these conditions are special and therefore lead to unique phenomena that expand the range and possible usefulness of alloy properties.

Many opportunities unfold for further insights and property improvements through off-stoichiometric variations (Figure 33) ⁶⁰ and multicomponent alloying (Figure 34). ⁵⁶ Generalizations concerning the effects of composition are not yet sufficiently well established, although the field is progressing rapidly. Alloying additions, such as niobium and tantalum, which tend to destabilize the cubic L1₂ phase (relative to other types of close-packed stacking), are effective strengtheners, but usually at the expense of decreased ductility. A related factor is the influence of composition on the difference

in antiphase-boundary energy between the {111} and {100} planes, which governs the driving force for the aforementioned thermally activated cross-slip mechanism. ⁵⁷ Such subtle compositional effects may well account for a continuous range of behaviors from the anomalous temperature dependence of flow stress to the normal dependence observed with the more familiar alloys. ⁴⁹

Research on ordered (Fe,Co,Ni)₃V alloys, paralleling the aluminides, has added considerable breadth and attraction to the subject of L1₂ phases and their properties. ⁵⁰ Here it is found that lowering the electron-to-atom concentration (e/a), say, by substituting iron for cobalt and nickel, stabilizes the cubic close-packed stacking characteristic of L1₂ phases. This, in turn, promotes ductility, often exceeding 30 to 35 percent in tensile elongation. Such high ductility is maintained at elevated temperatures over most of the range

where the anomalous temperature-dependent strengthening is observed. ⁵⁰ On the other hand, when e/a is increased, there is a progressive shift of the stacking sequence in the direction of hexagonal structures, thus decreasing the ductility, as illustrated in Figure 35. ⁵⁰ There is also some indication that the atom-radius ratio (r _A /r _B ) in the A ₃ B compounds may play a supplementary role in determining the stacking sequences, with smaller ratios tending to convert the atomic layering from cubic to hexagonal. ⁶¹

An overall view of the engineering potential of ductile ordered alloys is highlighted by the density-compensated high-temperature strength curves in Figure 36. ⁴⁸ The relatively low density of the aluminides contributes to specific-strength levels that are comparable with some of the strongest high-temperature alloys currently available. At the same time, the creep resistance of the ordered alloys looks attractive at temperatures below the respective disordering temperatures—e.g., 865°C for (Fe_33.3Co_66.7)₃V, 895°C for (Fe₃₀Co₇₀)₃V, and 950°C for (Fe₂₂Co₇₈)₃V. Referred to these temperatures, the creep rates are some two orders of magnitude lower in the ordered state than in the disordered state, reflecting the reduced atomic mobility and stronger binding forces attributable to the long-range order.

Although one cannot do justice here to the expanding literature on ductile ordered alloys, it is certainly evident that a vast field of metallurgical research and development has opened up. The phases involved are capable of tremendous variations in fine structure: planar stacking sequences, fault densities, grain-size and ordered-domain refinements, degrees of disorder, lattice defects quenched in by thermomechanical treatments, and precipitation of second-phase dispersions. The groundwork has also been laid for careful attention to processing methods (including rapid solidification, surface modification, and heat treatment) for compositional, microstructural, and quality control to advance both the underlying fundamentals and the engineering applications of this provocative class of intermetallic alloys. Moreover, the phenomena at play that thus far have been studied largely in the ordered aluminides and vanadides should now be investigated in the analogous silicides and titanides.

The general concept of composite materials has great appeal. It offers the prospect of microstructurally combining various classes of materials, by design, to realize superior sets of properties not obtainable from any of the constituent materials serving alone. Actually, this is an old idea (witness straw in bricks, horsehair in plaster, and reinforcing bars in concrete), but modern engineering requirements have become much more demanding, particularly with regard to specific strength and stiffness for aerospace, military, and similar critical applications. The corresponding R&D effort is enormous; it includes polymer-matrix and ceramic-matrix as well as metal-matrix composites (MMCs), but in the present context we shall deal only with the last of these. It has been reported that, from 1978 through 1986, governmental funding for the MMC program in the United States amounted to almost $300 million. ⁶²

Common structural materials (metals, glasses, wood products) have about the same density-compensated stiffness, whereas nonmetallics such as carbon, boron, silicon carbide, and aluminum oxide enjoy much higher values by factors of 5 to 10. The latter materials are also intrinsically strong, but they tend to be brittle in bulk form. However, if these materials are produced as fibers and aligned, they can contribute both stiffness and strength when suitably embedded in a metallic matrix. In the resulting composite micro

structure, the fibers act as reinforcing elements and are intended to take up most of the axial load being transmitted through the ductile matrix. The matrix is also supposed to accommodate plastically to breaks that may occur in the fibers. Figure 37 shows the fiber strengthening in a number of MMC systems having different fiber-aspect ratios and volume fractions. ⁶³ A comparison is likewise given for the strengthening contributed by similar non-metallic phases in the form of precipitates (up to 0.01 µm in diameter), dispersoids (0.01 to 0.1 µm), and particle reinforcements (0.1 to 10 µm).

Much of the research on MMCs has been directed to the micromechanics and fracture of the relevant two-phase configurations and to experimental ways of incorporating fiber reinforcements in metal matrices. The efforts on both fronts have been extensive and singularly ingenious, yet serious difficulties remain to be resolved before MMCs can reach anything like their envisioned potential. The essential issue lies in proper bonding between the matrix and reinforcing fibers; it is basically an interfacial problem. Surprising as it may seem, this facet of MMCs has not received the broad attention that it deserves.

Polymer-matrix composites (PMCs) have been remarkably successful in fiber reinforcement because, among other things, most materials can be reliably glued with polymers (resins) at modest temperatures, whereas MMCs require elevated temperatures for liquid-metal infiltration or solid-state consolidation. These temperatures may be high enough to start detrimental interfacial reactions between the fibers and the matrix, inasmuch as they are invariably far from thermodynamic equilibrium with each other. Moreover, residual stresses are then introduced because of differential contractions during cooling to room temperature. These difficulties are intensifed by the fact that MMCs are often justified over PMCs because they can withstand use at higher temperatures, but such service tends to add further to the fiber-matrix interactions.

In principle, interfacial reactions may result in a precarious balance between the enhancement of fiber-matrix bonding on the one hand and fiber degradation on the other. These competing effects are represented in Figure 38 for the shear strength between nickel and aluminum oxide as a function of the interaction-zone thickness. ⁶⁴ The right-hand trend is typical of prolonged or cyclic high-temperature exposure; it is also accompanied by a marked deterioration in transverse strength because of the interfacial weakening.

Interestingly enough, when there is near-equilibrium at the fiber-matrix interface, as in the case of directionally solidified eutectics having aligned microstructures, the interfacial bond strength is excellent. The two phases in MMCs of this in situ type are often crystallographically related, which may be a key factor in the observed interfacial stability. Unfortunately, the eutectic route to MMCs imposes severe compositional and phase-ratio restrictions on the alloy systems to be studied and thus detracts from the flexibility that is otherwise characteristic of MMC fiber-matrix combinations.

The importance of interfacial bonding, interdiffusion phenomena, and reaction-product formation in MMCs calls for detailed long-range study by sophisticated microanalytical techniques on carefully prepared and systematically varied fiber-matrix or layered specimens. In this connection there may be useful guidelines in the diffusion studies of fine-scale layered structures discussed earlier in the section on modulated structures. Analogous interfacial problems arise in the thin-film metallurgy of integrated circuits, as described in the following section, where the investigation of diffusion barriers is well advanced. Indeed, there is a thought-provoking case in point that suggests an interplay between the mechanical and electronic properties of MMC interfaces. In the Al/Al₂O₃/graphite junction shown in Figure 39, ^{65 , 66} the Al₂O₃, which is normally in a low-conductive state, can be voltage-switched to a high-conductive state. On peeling the aluminum away from the graphite substrate, it is found that the fracture path runs through the oxide when in the low-conductive state but follows the oxide-graphite interface when in the high-conductive state. Qualitatively speaking, the interface is stronger than the oxide in the low-conductive state and vice versa in the high-conductive state. One possible explanation is that a charge transfer takes place from the aluminum metal to the Al₂O₃ insulator in the former case, and the resulting space charge accumulation at the interface contributes an

electrostatic adhesive force to strengthen the interface. ⁶⁷ This space charge would then be neutralized when the oxide is in its high-conductive state. However, it is also conceivable that the strength of the oxide itself is affected by its electronic state, thus influencing the fracture path without significantly changing the interfacial strength per se. ⁶⁷ This is an open question, but it casts a new light on the bonding between conductors and insulators. Once again, it calls attention to the lure and importance of interfacial research on MMCs.

Even though it is often useful to classify materials into metallic, ceramic, polymeric, and electronic categories, all these groupings have characteristic electronic (or electrical) properties. The electronic materials are typically designated as such because of their unique semiconducting properties and, consequently, their central role in microelectronic devices, but other classes of materials also contribute in essential ways. Indeed, integrated circuit performance may be determined not so much by the switching speed of the contacts as by the time delay for a signal to propagate through the metallic interconnections. ⁶⁸ Moreover, the importance of such metallization in very-large-scale integrated (VLSI) circuits is destined to become increasingly critical in view of the urgent trends toward device densification and related complexities.

The term “metallization” in this context falls within the general realm of thin-film metallurgy and embraces contacts, gates, and diffusion barriers as well as interconnects; examples of these functions are shown schematically in Figure 40 for a metal-oxide-semiconductor transistor. ⁶⁹ The opportunities for metallurgical research in these processing, structure, property, and performance relationships are enormous. It is a field that deserves strong attention by the metallurgical community, not only because of its technological vitality but also because many of the operative phenomena have been investigated in other aspects of metallurgy, such as vapor deposition, microstructures and defect structures, grain boundary diffusion versus lattice diffusion, electromigration, and interfacial reactions.

To take a case in point, the lifetime of integrated circuits is often governed by failure of the aluminum interconnects because of electromigration (current-enhanced atom transport) at high current densities. This is a circuit breakage problem that will become more acute, and perhaps even the ultimate limitation, in the continuing miniaturization of devices to submicron-scale dimensions. ⁶⁸ A common type of interconnect failure stems from the fact that the diffusion flux induced by the electromigration is carried almost entirely

by grain boundaries. ⁷⁰ Accordingly, two directions of metallurgical research suggest themselves: (1) to eliminate grain boundaries by increasing the grain size and (2) to reduce the grain boundary diffusivity.

The former approach invites detailed study of secondary grain growth (discontinuous coarsening) in thin films. ⁷¹ As expected, normal grain growth in thin films is driven by the reduction of grain boundary energy and leads to a maximum grain size on the order of the film thickness. However, if the film surface energy is sensitive to crystallographic orientation, discontinuous coarsening can ensue by the preferential growth of grains whose orientations correspond to the minimum energy, as shown in Figure 41. ⁷¹ A line of investigation that might then be considered lies in the subtle addition of trace elements, possibly by ion implantation, which will augment the anisotropy of surface energy and thus promote secondary growth to relatively large grains having mutually similar orientations. The latter circumstance has added benefits in that a textured microstructure, particularly with the {111} orientation, is found to improve the electromigration lifetime. ⁷² Moreover, the remaining grain boundaries will then be aligned normal to the film surfaces, a geometry that further tends to minimize void formation arising from electromigration.

By the same token, ion implantation can also be used for selective alloying to decrease grain boundary diffusivity. This effect explains why alloying the aluminum interconnects with a small percent of copper significantly retards the electromigration damage. ⁷³ On the other hand, it should be feasible to inject less soluble atoms (e.g., refractory metals), which would segregate

preferentially to grain boundaries without appreciable adverse effect on the electrical conductivity of the interconnect. There is independent evidence that solute enrichment at grain boundaries increases markedly as a function decreasing solid solubility. ⁷⁴ A striking reduction in grain boundary diffusivity by solute segregation was described in an earlier section relative to the marked improvement in oxidation resistance of a cobalt-chromium alloy by yttrium ion implantation.

Another metallurgical phenomenon that is critical to VLSI circuit performance and lifetime involves the interdiffusion between thin-film multilayers; this was pointed up previously in a different context in the section on modulated structures. Of particular note is the undesirable interdiffusion that occurs at aluminum-silicon contacts, such bimetallic systems being rarely at equilibrium. Silicon can dissolve appreciably in solid aluminum at the processing or operating temperatures, causing aluminum to cross the interface into the silicon (often called “spiking”) and thereby degrade or short-circuit the semiconductor junction. ⁷³ Although the kinetics of this process can be slowed by prealloying the aluminum with silicon, much effort has been concentrated on the interposition of diffusion-barrier layers. ⁷⁵ Various barriers have been explored, some with considerable promise, but an orderly rationale has not yet emerged. Barriers under investigation are the silicides of Co, Ni, Pd, and Pt; TiC and TiN; and refractory metal combinations, sometimes with noble or near-noble metals. In cases where undue interaction may take place between the barrier and the aluminum or silicon on either side, multiple barriers are of interest. Here again, an underlying hypothesis is that the detrimental interpenetration is dominated by grain boundary diffusion at the low temperatures involved and that this process can be alleviated by a large grain size or by “blocking” the grain boundaries with relatively insoluble solute or impurity atoms.

The point to be stressed for present purposes is that diffusion barriers in VLSI circuits as well as metallization more generally are rich frontiers for metallurgical research. One can sense that this kind of thin-film metallurgy will soon overlap the previously discussed field of nanoscale multilayered structures in which a high degree of coherency and orientation-alignment (approaching the monocrystalline state) is maintained between the metal layers—setting the stage for unexpected atomic-bonding, mechanical, and electrical properties. It can also be anticipated that other phenomena such as phase transformations and interface-related processes in thin-film and layered alloys will be profoundly different from what is now well known about the bulk systems.

Although ferromagnetism is normally regarded as an integral part of solid-state physics, many of the technologically important magnetic materials are metallic alloys and therefore fall within the general scope of metallurgy. Magnetic materials play a crucial role in modern electronic and electrical devices and enjoy a market of about $2 billion per year (growing at the rate of some 12 percent per year) in components that are 10 to 20 times greater in added value. ⁷⁶ For present purposes we shall call particular attention to

three advances in magnetic alloys that not only constitute important scientific achievements but also are likely to have significant societal impact.

Reference was made to metallic glasses in the earlier section on rapid solidification processing. These amorphous materials, especially those based on iron-boron-silicon and iron-boron-silicon-carbon compositions in the metalloid content range of 15 to 25 atomic percent, display remarkably low magnetic hysteresis and eddy-current losses, mainly as a result of their high electrical resistivity and freedom from magnetocrystalline anisotropy as well as from grain boundaries and other microstructural defects. These materials in ribbon thicknesses have 60-Hz core losses that are less than a third of the best grain-oriented silicon steels, the latter embracing significant improvements in grain-growth and texture control over a period of 50 years. Still more recent research on metallic glasses has shown further reductions in core losses to approximately 1/20 that of the silicon steels. These advances are summarized in Figure 42. ⁷⁷

The implications of these metallurgical developments are great. It is estimated that core losses in distribution transformers alone amount to about $0.74 billion annually in the United States, and that a third of this waste could be saved if amorphous alloy cores were used instead of the present materials. ⁷⁷ A program is now under way to design and build one thousand 25-kVA distribution transformers for testing in service. Twenty-five of these transformers have already been constructed and are operating in various distribution systems. The test results thus far are in accordance with predictions, and the decision has been made to go ahead with production.

Corresponding opportunities for power transformers (500 to 1,500 kVA) and electric motors beckon further into the future. Meanwhile, numerous other magnetic applications for amorphous alloys are at hand: phonograph cartridges, audio and computer tape heads, dynamic microphones, magnetic shielding, and antitheft devices. A useful characteristic of these metallic glasses is that their magnetostriction can be controlled by alloying from zero, which is desirable in many devices, to sufficiently large values for force and displacement gauges. Considering the diverse applications of all these amorphous magnetic alloys, we have another textbook example in which a novel method of materials processing (RSP in this case) has led to the commercial development of previously unavailable materials and, in turn, to sharp improvements, or at least new potentials, in very different materials-dependent technologies.

RSP has also played a direct role in the discovery of the intermetallic compound Nd₂Fe₁₄B as a high-performance permanent magnet ⁷⁸ —eventually superior even to the previously developed samarium-transition metal types. Figure 43 shows how the magnetic energy product (BH) _max of this ternary phase varies with the melt-spinning velocity. ⁷⁸ The peak value of 14 mega-gauss-oersteds (MGOe) corresponds to a solidification rate that yields a rather uniform grain size of about 50 nm, which is presumably appropriate for single-domain particles. Slower rates permit the formation of coarser grains, whereas faster rates tend to inhibit crystallization altogether and an amorphous structure is then retained. Accordingly, the function of rapid solidification in this instance is not to produce a metastable phase but to achieve the proper grain size of a stable phase. In fact, this grain size can also be generated by careful devitrification of the amorphous structure. The magnetic measurements are made on powdered and subsequently compacted specimens.

This discovery has set the stage for more detailed fine-scale microscopy, systematic compositional variations, and powder-metallurgy processing to

obtain magnetically aligned, fully dense bulk materials. ^{79 , 80} With such efforts, the energy product has now been raised to the 40- to 45-MGOe level, shown in Figure 44 for the evolution of permanent-magnet alloys; ⁸¹ there is no reason to believe that a limit has yet been reached. Other rare-earth/transitionmetal/metalloid intermetallic compounds now become attractive candidates for investigations along these lines. Additional research is also needed to attain higher Curie temperatures and better magnetic properties at elevated temperatures. But already the indications are that the high energy densities characteristic of these advanced magnetic intermetallic phases will spur new electronic-device designs toward greater efficiency and miniaturization—e.g., for electric motors, generators, actuators, and electroacoustical pick-ups. ⁸¹ Strong permanent magnets have even been proposed in place of electromagnets for use in accelerators and in electron storage rings for synchrotron radiation. ⁸² All these possibilities constitute another exciting prospect for front-line metallurgical research, development, and application.

Still another research and development sector for magnetic alloys lies in the magneto-optical recording of information. ⁸³ Vacuum-deposited amorphous films of rare-earth/transition-metal alloys, such as GdCo, TbFe, GdTbFe,

and GdTbCo, have the property of switching the direction of magnetization on a very fine scale, under the influence of a weak external field, when heated locally to 100° to 200°C by a pulsed laser-beam signal. After the beam is removed, the reversed domains persist because of hysteresis, as in Figure 45, ⁸⁴ and so the recorded bits are “stored” for subsequent reading by magneto-optical (Kerr polarization rotation) effects. This application is expected to be in wide commercial service soon; it has the advantage of high bit density, noncontact recording and reading, and selective erasability by laser-beam heating in an oppositely biased external field.

The opportunities for creative metallurgical research on these thin-film amorphous alloy systems are immense—e.g., deposition techniques, compositional control, microstructural uniformity, amorphous-state stability and aging effects, corrosion mechanisms and protection, and new alloy chemistries. And once again, we note the emerging impact of thin-film metallurgy.

Although phase transformations pervade virtually all parts of metallurgy, we concentrate in this section on solid-state transformations. Such structural changes are an integral part of physical metallurgy and have been studied with high degrees of sophistication, both theoretically and experimentally. Phase transformations in the solid state are not only intriguing manifestations of nature’s processes, but they are also useful to society for harnessing relationships among structure, property, and performance. Many branches of knowledge and instrumentation are brought to bear on this facet of metallurgy. Four examples are discussed here to illustrate some recent advances and critical issues.

Although solid-state transformations are frequently sensitive to lattice defects in the parent phase and thus tend to nucleate heterogeneously, evidence of homogeneous nucleation has been convincingly demonstrated in the past two years in selected alloy systems. ^{85 , 86} Copper-cobalt alloys are particularly appropriate for this objective in that both the copper-rich parent phase and the cobalt-rich precipitate phase have the same face-centered cubic crystal structure and are mutually coherent, thereby exhibiting low interfacial energy and simplifying its calculation. The corresponding nucleation rates for precipitation in a copper-1 atomic percent cobalt alloy are plotted in Figure 46, as determined experimentally by electron microscopy and as calculated from classical homogeneous nucleation theory involving thermal fluctuations surmounting an interface-related free-energy barrier. ⁸⁵ The indicated agreement, assuming diffusivities based on vacancy concentrations characteristic of the solutionizing temperature, provides good support for the validity of homogeneous nucleation theory. Similar findings have been obtained for more dilute copper-cobalt alloys. ⁸⁵ This signifies that, even though the smallest precipitated particles detected by electron microscopy (approximately 50 angstroms) were considerably larger than the critical nucleus size, the increasing particle-number density versus time was small enough (yet measurable) to permit unimpeded growth of the individual particles to detectable sizes before encountering overlapping diffusion fields between nearest-neighbor precipitates.

On the other hand, with higher supersaturations (e.g., copper-2.7 atomic

percent cobalt), and in other face-centered cubic alloy systems (e.g., copper-1.9 atomic percent titanium and nickel-14 atomic percent aluminum), the nucleation rates may be orders of magnitude larger, and the corresponding number densities of precipitated particles increase into the range where the atom probe/field-ion microscope becomes an exceptionally powerful instrument for quantifying the early stages of precipitation. ⁸⁷ In addition to number densities, this research technique is also capable of measuring small-particle compositions and particle-size distributions below 10 angstroms. However, in this high-particle-density regime, the nucleation stage is overlapped by solute depletion of the parent phase and by Ostwald ripening, both of which act to decrease the observed particle densities by re-solution of some of the previously precipitated particles. Nevertheless, the earliest detectable precipitates are found to have their equilibrium composition.

In a salient advance, these complications have been sorted out analytically, and calculated and experimental results for the critical nucleus size, mean particle size, and number density are compared in Figure 47 for a nickel-47 atomic percent aluminum alloy. ^{87 , 88} Calculated nucleation rates, based on homogeneous nucleation theory, are also shown in Figure 47. The interfacial energy (σ) and diffusivity (D) adopted for these computations can be derived directly from the longer aging times where only Ostwald ripening is operative. There is some uncertainty, however, as to whether the state of coherency characteristic of the nucleation stage is fully maintained throughout the coarsening period, but this does not interfere with the overall interpretation that follows.

It is evident from Figure 47 that the earliest measurements are already at a stage of decreasing particle-number density (N _v ) despite the initially increasing transient nucleation rate (J). This means that Ostwald ripening sets in almost at the beginning of precipitation in this alloy and limits the fineness of the second-phase dispersion that might otherwise be available for maximum precipitation strengthening. This circumstance may also account for the rather constant mean particle size during the initial precipitation process, the growth of some nucleated particles being offset by the re-solution of others. In any event, the subsequent decrease in the calculated nucleation rate reflects the decreasing supersaturation of the parent solid solution as the transformation proceeds, and the critical nucleus size (R*) increases accordingly.

Because of the complexities of the foregoing kinetic analysis, it has been applied so far only to homogeneous nucleation in face-centered cubic systems with coherent interphase interfaces. The next big step must cope to an increasing extent with dissimilar parent/product phases, semicoherent interfaces, and heterogeneous nucleation to establish a more general understanding and a more useful control of precipitation processes in a wider variety of alloy systems. It is a formidable long-range task, to be sure, but recent advances in theory, experimental techniques, and computer calculations are providing both direction and momentum.

It has long been recognized that defects in the parent phase, such as grain boundaries and dislocations, can offer preferred nucleation sites for solid-state diffusional transformations by reducing the interfacial and strain energies associated with the nucleation process. ⁸⁹ However, the relevant mechanisms and kinetics are not well defined as yet. In contrast much progress in heterogeneous nucleation theory has been made in connection with martensitic (displacive and diffusionless) transformations. In view of the prominent shear-like deformation that accompanies this class of structural change, thermally activated homogeneous nucleation can be essentially ruled out; in fact, comparatively potent sites are necessary to start the transformation under the

thermodynamic conditions where these phase changes are known to take place.

A likely type of preferred nucleation site for martensitic transformations consists of a stack or short wall of dislocations such as may exist in a grain boundary or subboundary of the parent phase, where the potency of the site is measured by the number of dislocations in the array. When the transformational driving force is sufficient, these dislocations can dissociate and produce a ribbon-shaped or plate-like particle, as shown schematically in Figure 48. ⁹⁰ One consequence of this diffusionless mechanism is that it defines the dislocation structure of the interface and connects it directly to the crystallographic correspondence and relative displacements between the parent and product phases, the latter being generated simply by the motions of interfacial dislocations. An example of a semicoherent interface for a face-centered cubic to body-centered cubic martensitic transformation is described in Figure 49. ⁹¹

There are far-reaching consequences of this concept of heterogeneous nucleation. First, martensitic transformations can now be couched in the language and phenomenology of dislocation theory, and thus many structural, kinetic, and mechanical aspects can be quantified accordingly. This new approach has been summarized recently in much detail. ⁹² It also points toward a more basic understanding of autocatalytic nucleation, which plays a major

role in the overall kinetics of martensitic transformations, particularly in the heat treatment of steel. In other words, although preexisting nucleation sites are essential for the start-up and early progress of martensitic transformations in ferrous alloys, their density is insufficient without the ongoing generation of new sites to achieve the technological hardening of steel.

The description of martensitic interfaces in the form of dislocation arrays makes it possible to deal with the growth problem in terms of dislocation mobilities under the operating driving-force and obstacle conditions. The nature of such calculations is presented in the next section; and in a later section on displacive diffusional transformations, it is shown how these ideas can be carried over to other phase transformations, including precipitation processes, in which lattice deformation and diffusion participate simultaneously. This establishes a spectrum among phase transformations that have previously been regarded as different in kind.

For convenience of discussion, the propagation of martensitic interfaces can be divided into two categories: thermoelastic transformations, wherein the shape deformation attending the phase change is accommodated elastically by the alloy system; and nonthermoelastic transformations, wherein the shape deformation is accommodated plastically. In the former, a thermoelastic force balance prevails between the driving and retarding forces, and the interface can be made to move progressively and measurably under well-controlled experimental conditions. The interfacial mobility of thermoelastic copper-aluminum-nickel martensites has been investigated comprehensively in this way. ⁹³ It is found that the effects of temperature and applied stress on the interfacial velocity are consistent with the concept of thermally activated interfacial motion, as suggested by the kinetic interrelationships in Figure 50. ⁹³ Just as in the case of mechanically driven slip dislocations, the driving force for interfacial motion has both athermal (τ_µ) and thermal (τ _f —τ_µ) components; similarly, activation energies and activation volumes are derivable from the experimental data, and specified models of interfacial structure can be tested against various modes of lattice friction and obstacles to growth. ⁹³

Nonthermoelastic martensites tend to propagate rapidly, in some instances even approaching the velocity of a shear wave in the parent phase. These transformations are difficult to nucleate and hence are effectively “overdriven” when the martensitic growth process sets in. Dissipation of the latent heat of transformation then becomes a complicating factor and must be taken into account. This has been done for the calculated interfacial-velocity versus particle-size curves plotted in Figure 51. ⁹⁴ During the acceleration stage, the interfacial motion is assumed to be thermally activated at the lower velocities

and phonon-drag controlled at the higher velocities. The marked differences in martensitic growth characteristics between the two iron-nickel alloys in Figure 51 originate from differences in the respective dynamic flow stresses and drag coefficients. ⁹⁴

However, not yet incorporated into these calculations is the quantitative effect of local plastic accommodation, including the associated defect formation and strain hardening, on the resistance to interfacial motion. This is a crucial issue, not only because of the combination of dynamic phenomena at play here but because it may be forcing a general study of the no-man’s-land between continuum and discrete-lattice solid-state science. It is somewhat reminiscent of the crack-tip/plastic-zone problem in fracture mechanics. Furthermore, the plastic accommodation process may hold the key to a deeper understanding of the autocatalytic nucleation that enters so significantly into the overall kinetics of nonthermoelastic martensitic transformations.

The shear-like displacive nature of martensitic transformations allows these phase changes to operate as a deformation mechanism in parallel with ordinary slip processes. ⁹⁵ Applied stress can promote the formation of martensite by contributing to the transformational driving force and by introducing new nucleation sites through accompanying plastic flow. At the same time, the acting stress will favor the formation of those martensitic crystallographic orientations that have optimal displacive components for “yielding” to the stress. The resulting transformation-induced plasticity (TRIP) reflects a novel interplay between the kinetics of a structural change and macroscopic stress-strain behavior. Indeed, there are now cases in which constitutive relations involving transformation plasticity have been derived to predict flow stress as a function of strain, strain rate, temperature, and stress state. ⁹⁵ In austenitic steels having appropriate deformation-induced transformation characteristics, the uniform ductility in a tensile test has been increased about fivefold beyond that of the untransformed parent phase. ⁹⁶

Transformation plasticity offers an attractive mechanism for enhancing fracture toughness at high-strength levels by coming into play in the vicinity of an advancing crack, particularly in alloys that undergo shear instability before rupture. In such instances rather small amounts of mechanically induced transformation occurring in the plastic zone of a crack tip can delay the impending strain localization and thereby increase the fracture toughness substantially. Two examples are shown in Figure 52 for high-strength austenitic steels as a function of normalized test temperature. ⁹⁶ The maximum toughness values correspond to about K _Ic =250 MPa m^1/2 at—70°C, which is very tough for these 1300-MPa yield-strength steels. The beneficial effect on sharp-crack toughness arises not only from the transformation plasticity as such but also from a reduction of the triaxial stress state because of the volume expansion that attends the phase change.

The decrease in fracture toughness at lower test temperatures in Figure 52 is caused by the formation of too much martensite, which in itself is less tough than the parent phase. Hence, the eventual use of transformation toughening will require compositional modifications to decrease the temperature sensitivity of the transformation—e.g., by decreasing the entropy change of the transformation in order to reduce the temperature dependence of the thermodynamic driving force. Manganese is known to affect the transformational thermodynamics in that manner.

It is conceivable that the greatest potential for transformation toughening lies in its applicability to the retained austenite in martensitic steels. This is an unexplored field that warrants intense study, inasmuch as the technological use of ultrahigh-strength steels could be materially advanced even by modest increases in toughness.

One can imagine a potential martensitic transformation under conditions where the driving force is not large enough for nucleation or interfacial motion to ensue. But if some compositional partitioning is permitted at the temperature in question, the driving force may then be sufficient for the phase change to proceed. The velocity of the interface, whether coherent or semicoherent, will depend on diffusion as well as on the resistance to interfacial motion, and this interplay can result in some degree of solute trapping which, in turn, will determine the extent of partitioning and diffusion taking place. It is also possible that a steady-state balance in the functioning of these phenomena may be attained such that there is just enough diffusion to supply the necessary force for overcoming the resistance to interfacial motion. This type of phase transformation is, strictly speaking, neither displacement-controlled nor diffusion-controlled; rather, it represents a coupled displacive-diffusive mechanism of interfacial motion in which both lattice deformation and diffusion participate jointly and interactively.

The above transformational characteristics begin to take on some of the displacive-diffusional features of bainitic transformations, allowing for precipitation processes in the product phase behind the advancing front. ^{97 , 98} This theory of bainite formation has not been definitively tested as yet, but it surely would be worth doing so, especially in view of the favorable me-

chanical properties of bainitic microstructures being reported for high-strength, low-alloy steels. ⁹⁹

However, there appears to be an even broader significance to the concept of displacive-diffusional transformations in that some typical precipitation reactions display both shape and compositional changes. It has now been demonstrated that the well-known precipitation of ∊-carbide from ferrous martensites falls in this category, as shown schematically in Figure 53. ¹⁰⁰ These martensites first develop a compositionally modulated structure of coherent high- and low-carbon bands on aging near room temperature, as discussed earlier in the section on modulated structures, and then the ∊-carbide precipitates with an invariant-plane strain, plate-like morphology composed of both homogeneous lattice and inhomogeneous internal deformations, comparable to a martensitic transformation except that carbon diffusion is also required to form the ∊-carbide. The precipitation is heterogeneous, tends to nucleate along the modulated-band interfaces, and adopts a (012) habit that is close to the (023) habit of the modulated bands. In view of the lattice correspondence between the precipitated phase and its surroundings, the carbide interface is undoubtedly semicoherent. A similar type of displacive-diffusional interpretation has been applied to the precipitation of hydrides in vanadium and zirconium solid solutions. ¹⁰¹

It may be turning out that precipitation processes in general, to the extent that coherent or semicoherent interfaces are involved, combine both lattice-distortive and diffusion-related phenomena (although not necessarily coupled and not necessarily shear-like) with different degrees of kinetic control. In this sense, precipitation reactions are likely to comprise a spectrum of phase transformations in which martensitic transformations constitute only a very special (diffusionless) case. This possibility offers new incentives and guidelines for future research on phase transformations.

There are many innovative areas of metallurgical research not treated in this review that should be recognized as significant opportunities for important advances in metallurgy. Among these are the electromagnetic handling of liquid metals, sensors and techniques for on-line process control, in situ observations of fast events, hydrogen embrittlement, atomic-bonding calculations, and computer modeling of processes and phenomena.

In this chapter we have attempted to integrate various viewpoints and suggestions, received mainly from the metallurgical community, into a valid perspective on some notable accomplishments of modern metallurgical research and on the sustained intellectual excitement that is inspiring further research on this vital class of materials. Metals not only represent a large majority of the elements in the periodic table and therefore constitute a pervasive part of the natural world, but they are generally useful to mankind and therefore invaluable to society. This confluence gives rise to a persistent driving force for probing metals ever more deeply and for using them ever more wisely. It is both a lofty challenge and a proud responsibility for metallurgical science and engineering.

I am extremely grateful for the outpouring of ideas and opinions, both oral and written, that came to me in response to my informal requests for advice on many metallurgical fronts. The resulting communications grapevine operated with amiable efficiency and goodwill. In instances where the submissions could not be included, I am no less appreciative of the cooperation and involvement.

I am also deeply indebted to Marguerite Meyer, Miriam Rich, and John Mara at MIT, who helped so much in the preparation of the manuscript, and to NAE Fellow Dr. Peter Psaras, who waited so patiently to receive it.