Technologies are an integral part of our lives and we rely on them for such things as communication, heating and cooling, transportation, and construction. Improvements to technologies have made what they do for us more precise, automated, efficient, and powerful. One area of advancement is in the inclusion of components made from smart materials. Examples of this class of materials include actuators, transducers (a type of actuator), and sensors. Actuators take a signal and turn it into motion, and sensors convert physical changes into a signal. Performance metrics include sensitivity, responsiveness, resolution, and efficiency. Practical considerations include mass, volume, operating conditions, and durability. Smart materials possess some type of functionality that responds to the environment and can be exploited to do useful work. A smart alloy remembers its original cold-forged shape and will return to a pre-deformed shape when heated. Similarly, magnetostrictive alloys change their linear or volumetric shape when a magnetic field is applied which can be used to do mechanical work. When the magnetic field is removed, the material returns to its initial shape and size. Both smart and magnetostriction alloys are lightweight, solid-state alternatives to conventional actuators such as hydraulic, pneumatic and motor-based systems.
Another large area of research is in magnetocaloric materials where the material absorbs or gives off heat in a magnetic field. This type of heat pump has the potential to lower energy consumption by 20-30% over conventional vapor compression technology.1 Particularly attractive are materials that have multiple functionalities that are interconnected because they are often more sensitive to input than their single-function counterparts. In general, the ideal material to fit a particular application needs to exhibit a controlled response to a stimulus that is reversible and repeatable indefinitely. In reality, functionality often diminishes over time due to fatigue or aging mechanisms, and much research is focused on increasing the efficiency and lifetime while keeping costs low. Finding a materials solution that satisfies many of these competing characteristics requires the discovery and design of novel materials.
The crystal growth method and conditions used depend on the chemical and thermodynamic properties of the alloy. Phase diagrams, and the literature they were developed from, are especially critical in assessing and identifying suitable crystal growth methods. When dealing with an alloy that contains volatile, reactive, corrosive, or high meltingpoint components, special attention is needed. In some cases, after an evaluation is made, it is determined that current methods or equipment are not up to the challenge either from a physical limitation or safety standpoint. In these cases, either modified or completely new protocols and capabilities are needed.
Single crystal characterization is ideal because it is in this form that the true nature and origins of the properties can be elucidated without the influences of extrinsic defects such as grain boundaries or second phases. Also, since single crystals can be cut along crystallographic directions, anisotropy of specific properties can be measured. Moreover, careful control of growth parameters can also minimize vacancies, defects and impurities at the microscopic level. These have direct bearing on the mesoscopic scale (i.e., magnetic domain wall movement) and on the macroscopic scale (i.e., bulk magnetic properties).
This paper will review crystal growth challenges of three different smart materials that have been a focus in Department of Energy′s Ames Laboratory material synthesis and processing research initiatives. These materials (GdNi, Gd-Si-Ge, and Ni-Mn-Ga) are currently of interest to materials science research communities, and the capabilities developed to synthesize these materials with well-controlled stoichiometry and high structural quality have accelerated underlying understanding of their functionality. Our broader objective is to develop protocols applicable to other similar classes of materials to produce phase pure, single crystals of reactive and volatile metallic materials. For each of the above materials, we begin with a review of the initial single crystal growth attempts, discuss how specific challenges were addressed and conclude with some examples of characterization that were made possible by successful preparation of single crystal samples.
The class of compounds formed between rare earths (RE) and 3d transition metals (TM) are of interest for their magnetic properties. Materials from this class of compounds benefit from the intrinsic properties of both components, i.e., from the high magnetic moment per atom, strong single-ion magnetocrystalline anisotropy and magnetostriction of the RE partner, and from the high magnetic coupling strength of the moments of the 3d TM partner.2 The total number of these compounds is enormous since it is not uncommon for each RE-TM combination to give rise to multiple intermetallic compounds. Among them have been found outstanding permanent magnets (RECo5), room temperature magnetostrictive alloys (RE-Fe, RE-Ni), and near atmospheric reversible hydrogen adsorption materials (i.e., LaNi5) (Aldrich Prod. No. 685933).2 In these investigations, to avoid crucible reaction with either the RE or TM, the polycrystalline materials were prepared either in an arc furnace or in a levitation furnace.
Here we look at the rare-earth intermetallic material, Gd50Ni50, which exhibits a magnetocaloric effect (MCE), making it a candidate for an alternative to conventional vapor compression refrigeration systems. It also exhibits spontaneous magnetostriction3 at its Curie temperature on the order of 8,000 ppm,4 which is remarkable considering the room temperature magnetostriction of iron (Aldrich Prod. No. 266256) and Terfenol-D are 455 and 2,000 ppm,6 respectively. Single crystals for anisotropic property measurements were needed to fully understand the unusual coupling of its crystal structure to magnetism and the mechanism causing this compound′s remarkably strong linear strain4 along its c-axis. Because this is a congruently melting compound, the Bridgman method of crystal growth is a natural choice; however, it presents the challenge of identifying a crucible material. The particular combination of Gd (Aldrich Prod. No. 263060) and Ni (Aldrich Prod. No. 266965) ruled out common oxide crucibles, such as alumina (Aldrich Prod. No. Z247626), because of gadolinium′s affinity for oxygen and also refractory metal crucibles, such as tungsten and tantalum, because of nickel′s reactivity toward them. In fact, all readily available crucible materials were ruled out because of this alloy′s very reactive nature.
Uhlirova et al.7 prepared single crystal samples by the Czochralski method; however, no details were given concerning the growth. We first used a tri-arc crystal pulling technique,8 which is a sister technique to that of Czochralski, where the charge sits on an actively cooled copper hearth (Figure 1A), which prevents it from fully melting and thus creates a self-crucible. However, sustained growth was not achieved because of arc instability at the low power used to maintain its relatively low melting temperature of 1,280 °C. The arcs were prone to wandering and becoming defocused which created hot and cold spots that favored recrystallization over continuous growth. Also, black body radiation overpowered the viewing lens making it difficult to visually observe the darker button (Figure 1A), which made dipping the seed and monitoring the crystal diameter during growth difficult. Increasing the mass of the button and the size of the tungsten stingers was attempted to increase the operating power needed and thus, improve arc stability. Unfortunately, these alterations, as well as varying growth parameters, only provided minimal improvements; and the resulting ingots were polycrystalline with grains too small to orient and cut for magnetization measurements.
Figure 1.Tri-arc crystal pulling. The button is centered in a well in the copper hearth. A tungsten rod is the seed material and the three arcs are stationary during the run. A) GdNi: visibility is poor due to a lack of black body radiation at 1,280 °C. B) Gd-Si-Ge: in contrast at 1,850 °C, the seed/button interface and size of the overlapping pools are clearly visible.
Solution growth with a self-flux was also tried.8 A Gd-rich starting composition of Gd3Ni2 was chosen, which has a lower melting point than GdNi and also a lower relative amount of nickel, making a pure tantalum crucible feasible. The Gd3Ni2 compound was melted and homogenized at 1,100 °C before slow cooling over a period of just over two weeks to a temperature above the 690 °C peritectic, where the remaining liquid was decanted off the crystals through a tantalum sieve inside of the crucible. Small plates resulted that grew along the a-axis in colonies. Analysis showed the crystals to be contaminated with trapped flux which produced its own magnetization signal and grains were too small for any measurements other than along the a-axis.
Single crystal growth by the Bridgman method is also reported in the literature. Nishimura et al.9 used an alumina crucible but Sato10 did not provide any details. Neither provided characterization beyond magnetic measurements; however, optical and scanning electron microscopy (SEM) with energy-dispersive spectroscopy (EDS) of our crystal growth8 attempt in Al2O3 revealed multiple secondary phases visible at low magnification that incorporated both the aluminum and oxygen from the crucible. Yttria was also tried as it is a thermodynamically more stable oxide than Gd2O3 and should have been inert with respect to Gd.
However, we found the reaction zone to be quite extensive, with yttria dissolving and reacting with the alloy as well as gadolinium substituting for the yttrium in the oxide. Tantalum Bridgman crucibles with a tantalum boride or tantalum carbide coating that were produced inhouse by a procedure developed by Gschneidner, Jr. et al.11 were also tried. The carbide coating was not as successful as the boron coating, and from our analysis8 we found that further processing was necessary to remove the active boron from the surface before use. We took crucibles made by their method and melted a relatively inexpensive pure rare earth in it, such as yttrium. The yttrium reacted with the surface boron and presumably converted TaB to the more thermodynamically stable Ta2B. The yttrium was then etched out of the crucible before use. The crystal growth resulted in single crystal grains suitable for magnetization and resistivity anisotropic measurements. In parallel with this crucible development, we also prepared yttria stabilized zirconia (YSZ) and gadolinia plasma sprayed Bridgman crucibles.8 The crucibles were backed with tungsten for strength. We did not find the YSZ to be an improvement over the yttria crucibles mentioned above; however, the Gd2O3 crucible yielded a remarkably clean ingot with few grains. The large, single phase grains enabled a full suite of magnetization, resistivity, and heat capacity measurements to be done.
Figure 2 shows select isothermal magnetization versus field curves from samples made by three different groups along the b-axis for single crystals. The inverse relationship between magnetization and temperature is maintained in fields above 30 kOe, regardless of growth specifics. Below 30 kOe, there was some minor variation in curve shape that caused some of the scans to cross. The calculated maximum entropy change, -ΔSMmax, for this crystal8 in differential fields of 20 and 50 kOe is slightly higher at 9.82 and 19.2 J/kg K compared to 9 and 17 J/kg K measured on polycrystalline GdNi.12 These comparisons indicate that the magnetic properties of GdNi may not be sensitive to impurities, processing routes, or even long range atomic order. Paudyal et al.4 utilized first principles spin polarized calculations to explain the isotropic behavior seen in single crystals as a function of magnetic field and temperature as arising from the unusual interplay between magnetism and crystal structure.
Figure 2.Isothermal magnetization curves for single crystal GdNi along the b-axis. The data from three different samples follow the trend of increasing magnetization with temperature.
The family of compounds RE5(SixGe1-x)4 has become known for its giant MCE near room temperature. They were also discovered to have unusually large magnetostriction and magnetoresistance, which makes them viable actuators, transducers, and sensors candidates. In this alloy, both Gd and Si raise crucible compatibility concerns. Quartz or silica is normally used for Si, but Gd reacts with oxygen in these materials and they cannot withstand the high melting temperatures of these alloys. Silicon easily wets many crucible materials, requiring a closed crucible system to prevent silicon loss. Tungsten, a weldable refractory crucible material, is resistant to rare earth metal attack and not prone to silicide formation, making it a good candidate.
The first successful growth of large crystals of Gd5Si2Ge2 (Aldrich Prod. No. 693510) suitable for anisotropic materials property investigations was done in a tungsten crucible by the Bridgman method.13 Tungsten crucibles were indeed found to be inert, i.e., no formation of tungsten silicide or germanides. There was a slight solubility of tungsten in the molten alloy which resulted in precipitation of pure tungsten dendrites in the bulk alloy and along the crucible walls. The latter led to strong mechanical bonding between the crucible and ingot. Upon cooling, the ingot developed cracks due to thermal expansion mismatch between Gd5Si2Ge2 and tungsten. Also, the properties varied along the length proportional to the chemical segregation which followed the relation:
for solute redistribution during solidification of an alloy where C is concentration in the solid at position along the growth, where a fraction g of the original liquid has frozen, Co is the nominal composition, and k is distribution coefficient. While anisotropic single crystal measurements were possible with samples harvested from these Bridgman crystal growths, there was still a need to have tungsten and crack-free samples to ensure these features were not adding to or detracting from the true properties of the material. It was also desirable to produce chemically homogeneous ingots with uniform properties. Significant improvement was achieved by utilizing the tri-arc crystal pulling method.14 As described in the previous section, this is a crucible-less method so the problems with crucible contamination and cracks resulting from thermal mismatch were avoided. Also, with no crucible for the Si to wet, it stayed in solution which helped minimize composition variation along the length of the growth.
With the tri-arc crystal pulling method, the ingot size is limited to several millimeters in diameter because of the latent heat of the stingers. To try and keep this heat from melting the surface of the growing crystal, we increased the stinger to crystal distance and reduced the stinger size to better focus the heat. We quickly found, there is limited latitude for both of these variables and even minor changes result in disruption of the delicate heat flow balance at the growth interface and growth cannot be sustained. The length of the growth is also limited to less than 5cm by the button volume/size which is directly proportional to both the crystal-stinger distance and the heat needed for the molten pools to effectively overlap (Figure 1B). Nonetheless, the resulting ingots were of sufficient quality and size for numerous characterization studies. Figure 3 shows the magnetization dependence on crystallographic direction with varying field and temperature for Gd5Si2Ge2. Systematic characterization of the family of R5(SixGe1-x)4 compounds, utilizing both bulk and single crystal samples, led to the discovery of the extreme sensitivity of the crystal and magnetic lattices to the chemical composition, temperature, magnetic field, and pressure15 of many of the members. Also, single crystals have allowed the detailed study of the atypical Widmanstätten features16 found in all alloys of this family, elucidation of its elastic properties17 and response to hydrostatic pressure.18
Figure 3.Magnetic characterization of Gd5Si2Ge2 single crystal along the three crystallographic axes. A) Magnetization as a function of temperature in a magnetic field of H=100 Oe. B) Field dependence of magnetization at T=5 K.18
Field-induced actuators can overcome the disadvantage of traditional thermal, pneumatic, hydraulic, and motor-based shape-memory actuators as they exhibit a fast frequency shape-memory effect in a magnetic field. The lead contenders are intermetallic alloys that adopt the facecentered cubic Heusler structure. These alloys typically undergo a crystallographic disorder-order transition at high temperature and the ordered structure undergoes a reversible martensitic shear transformation at room temperature or below. When the martensite transformation is near the Curie temperature, or coupled to magnetic ordering, applied magnetic fields can drive the structural rearrangements between the low symmetry martensitic phase and the high symmetry austenitic phase. These rearrangements, facilitated by twin boundary motion, result in reversible shape changes. Ferromagnetic Ni-Mn-Ga displays the largest shape changes of all known magnetic Heusler alloys with magnetic field induced strains reaching 10%.19 Also, its magnetic properties are highly composition-dependent, potentially making its response tunable for particular applications. The strain was determined to be crystallographically anisotropic and the martensitic phase to consist of numerous structural variations, many of which differ only in the stacking sequence of their atomic layers. Which variations are present depends on composition and thermal history. Also, the sequence of martensitic, premartensitic and intermartensitic transformations are important to its performance as a shape-memory alloy. When this behavior is present, it is manifested as bumps in differential scanning calorimetry (DSC). One example of this is shown in Figure 4. Even though the bumps appear to be erratic, they are reproduced each time the sample is cycled.
Figure 4.Differential Scanning Calorimetry (DSC) of Ni49Mn30Ga21, 20 °C/min. Features in oval indicate possible premartensitic or intermartensitic transitions and are repeated with cycling.
Alloy preparation of this material is fairly straight forward as the Heusler structure is a solid solution that exists over a large compositional field, but volatility of Mn can be an issue. While a wide phase field is beneficial to alloy preparation, it negatively impacts the ability to grow chemically homogeneous single crystalline ingots due to chemical segregation from non-congruent melting behavior. Samples with uniform and predictable magneto-mechanical responses for scientific research, and ultimately for device applications, require tight control of the growth process especially since the properties and functionality of these alloys are highly composition dependent.
Initial attempts to prepare stoichiometric Ni2MnGa or off-stoichiometric alloys resulted in several grains elongated along the length of the ingot. This initiated a study20 to determine the primary solidification phase field of Ni2MnGa. With this knowledge of the phase equilibria, experiments were conducted to find the proper growth parameters for the Czochralski and Bridgman techniques to produce larger grains, if not single grain ingots. Some success was made with the Czochralski method but consistently large single crystals (1.5 cm diameter × 6 cm long) were first obtained by the Bridgman method from Ni-deficient compositions using a growth rate of 2.0 mm/hr. Although manganese and gallium evaporative losses were easily mitigated by over-pressurization during growth, significant macrosegregation of composition varied along the length of the ingot (Figure 5); and according to Equation 1, resulted in significant differences in properties from one end to the other, even with post-growth annealing.
Figure 5.Electron dispersive spectroscopy (EDS) chemical analysis along the growth direction. The horizontal lines show the nominal composition of the ingot; 5 mm/hr growth rate. Chemical segregation of the elements follow Equation 1.
From here, further improvements to the crystal growth processing were quickly realized to reduce compositional segregation and produce uniform material. Sozinov et al.21 prepared crystals at 30 mm/hr and of sufficient size to make uniform, oriented samples 5×5×10 mm3. Liu et al.22 was successful in obtaining single crystals up to 6-10 mm diameter × 20-80 mm long by the Czochralski method using a cold crucible system with growth rates ranging from 5-30 mm/hr and make oriented samples 2×9×12 mm3. Chu et al.23 grew a crystal by the Bridgman method but their starting materials were in the form of powders, which resulted in multiple grains and minor second phases evidenced by the presence of microstructure. This emphasizes the importance of chemical purity of the starting materials since impurities can be sources of grain nucleation and can impede the magnetomechanical response of the material. No mention of chemical variation along the length was made by either Sozinov or Liu; however, Jiang et al.24 undertook a comprehensive study of parameters such as temperature gradient, zone length, crystal growth velocity and diameter on the growth interface morphology, preferential crystallographic growth direction, and resulting chemical variation along the length. Through the use of optical zone-melting directional solidification, Jiang was able to obtain preferentially <100> oriented ingots 7×100 mm with much less property variation along the length than with the Bridgman method. Later, this group25 used this method to grow single crystals for the purpose of studying the solute partitioning during crystal growth to further eliminate compositional variation along the length. They looked at three compositions (stoichiometric, Ni-rich, and Mn-rich) and were able to produce rods with no obvious macro segregation after the initial 20 mm at a growth rate of 5 mm/hr. A patent was issued in 2009 for the electroslag remelting Bridgman process.26 The advantage of this method is the slag encapsulates the melt so there is no crucible interaction and the feed material is purified by crystalizing out of the slag, which also decreases the number of pores and occlusions. The result is a production-scale crystal essentially free of the defects that are thought to impede the twin boundary or domain wall motion responsible for its large magnetostriction. Sturz et al.27 were also interested in productionscale crystal growth. They designed a crucible that simultaneously grows four to six crystals that are 20-30 mm in diameter × 110 mm long. Starting from polycrystalline as cast material, through the use of a grain selector and post-growth annealing, they were able to produce single crystals with longitudinal axes that were within 10 degrees of the  direction and the composition only varied ±0.6% from end to end. In 2011, Brillo et al.28 reported computer simulation results predicting the thermal profile of their Bridgman-Stockbarger furnace for Ni-Mn-Ga single crystal growth. The benchmark parameters established can be utilized to predict optimum growth conditions in a particular system if the thermal properties and heat transfer coefficients of its components are known.
While the mechanisms that lead to Ni-Mn-Ga large field induced strain are still not well understood and research continues, this particular alloy has completed the cycle of crystal growth development to production development and deployment of commercially available actuators and devices.
As the desire for even smaller, smarter, more efficient, and longer life devices continues, research is starting to revisit previously overlooked candidate materials that were thought to be too corrosive, reactive, or their properties too irreproducible. While the end goal may be commercial production, in the research stage of a material′s development single crystals are fundamental in characterizing its properties and full capabilities unmodified by grain boundaries or impurities. The barriers to the crystal growth of three alloys containing reactive or volatile components as well as displaying varied physical properties as a result of chemical segregation, has been addressed.
This work was supported by the Office of Basic Energy Sciences, Materials Sciences Division of the U.S. Department of Energy under Contract No. DE-AC02-07CH11358 with Iowa State University.