ORDER/DISORDER IN ELECTRODEPOSITED ALUMINUM-TITANIUM ALLOYS

The composition, morphology, and crystallographic microstructure of Al- Ti alloys electrodeposited from two different chloroaluminate molten salt electrolytes were examined. Alloys containing up to 28 % atomic fraction Ti were electrodeposited at 150 °C from 2:1 AlCb-NaCl with controlled additions of Ti2+. The apparent limit on alloy composition is proposed to be due to a mechanism by which AhTi forms through the reductive decomposition of [Ti(AlCLt) 3 ] . The composition of Al-Ti alloys electrodeposited from the AlCl 3 -EtMeImCl melt at 80 °C is limited by the diffusion of Ti2+ to the electrode surface. Alloys containing up to 18.4 % atomic fraction Ti are only obtainable at high Ti2+ concentrations in the melt and low current densities. Alloys electrodeposited from the higher temperature melt have an ordered LU crystal structure while alloys of similar composition but deposited at lower temperature are disordered fee. The appearance of antiphase boundaries in the ordered alloys suggests that the deposit may be disordered initially and then orders in the solid state, subsequent to the charge transfer step and adatom incorporation into the lattice. This is very similar to the disorder-trapping observed in rapidly solidified alloys. The measured domain size is consistent with a mechanism of diffusion-controlled domain growth at the examined deposition temperatures and times. The surface morphology of the electrodeposits studied by electron microscopy (SEM) using a and field-emission scanning microscopy using a Electrodeposits also examined by x-ray diffraction (XRD) using a Siemens D-500 diffractometer and Cu-Ka radiation. The lattice parameters of the deposits were refined using the copper substrate reflections as an internal standard. Cross-sections of alloys deposited onto Cu wires were formed by first overplating with copper using a cyanide copper strike followed by bright from copper sulfate electrolyte The alloy composition was determined by performing quantitative energy dispersive x-ray spectroscopy (EDS) on polished cross-sections using pure aluminum and titanium as standards. Deposits were typically 5 pm to 20 pm in thickness. Transmission electron microscopy (TEM) performed on selected deposits prepared for both plan view and cross-sectional analysis. The plan view were detached from a planar Cu affixed to which were deposited at similar current densities. Deposit (e) shows relatively well-defined superlattice reflections, while those in (c) show significant broadening. The primary difference between these two samples is the thickness of the deposit, and as a result, the time that the deposit is exposed to the elevated temperature of the molten salt. Although there does not appear to be a systematic correlation between the deposition time and superlattice reflection shape, it is clear that the domains are better defined in the deposit that was exposed for the longest time at 150 °C.


INTRODUCTION
The excellent high temperature properties, low density and corrosion resistance of aluminum-based intermetallic compounds has led to the consideration o f these ordered alloys for structural applications, with applications envisioned for monolithic and, more likely, composite materials. Several processing methods have been investigated including solidification from the liquid and gas phase, mechanical alloying, reaction sintering of elemental powders and electrodeposition. Electrodeposition is an attractive method for fabricating intermetallic compounds since undesirable compositional inhomogeneities are very limited in scale and grain sizes are typically very small. Although electrodeposition is very different from traditional approaches used to fabricate these alloys, a number of analogies exist between it and rapid solidification. For example, electrodeposition from direct addition o f Ti(AlCl4)2, prepared by reacting Ti metal with AICI3 using the procedure of Brynestad and co-workers [15].
Al-Ti alloys were electrodeposited from 2:1 AlCh-EtMelmCl using a standard three-electrode cell. Coils o f 0.10 cm diameter aluminum wire (Alfa Aesar, 99.999 %) were used for the counter and reference electrodes. These electrodes were immersed in the melt with the same composition as the bulk melt, but were separated from the bulk melt by a porosity E glass frit (Ace Glass). The aluminum electrodes were cleaned with a mixture of concentrated H2SO4, HNO3, and H3PO4, rinsed with distilled H2O, and dried under vacuum before use. Anhydrous titanium (II) chloride (Aldrich, 99.98 %) was used as received. All experiments were carried out in a nitrogen gas-filled glove box (VAC Atmospheres NEXUS system) with O2 and H2O < 5 ppm. The electrodeposition of Ti-Al alloys was performed with an EG&G PARC Model 173 potentiostat / galvanostat equipped with a Model 179 digital coulometer plug-in module.
The surface morphology o f the electrodeposits was studied by scanning electron microscopy (SEM) using a JEOL JXA-840 and field-emission scanning electron microscopy (FE-SEM) using a Hitachi S-4700. Electrodeposits were also examined by xray diffraction (XRD) using a Siemens D-500 diffractometer and Cu-Ka radiation. The lattice parameters of the deposits were refined using the copper substrate reflections as an internal standard. Cross-sections of alloys deposited onto Cu wires were formed by first overplating with copper using a cyanide copper strike followed by bright copper from a copper sulfate electrolyte [16]. The alloy composition was determined by performing quantitative energy dispersive x-ray spectroscopy (EDS) on polished cross-sections using pure aluminum and titanium as standards. Deposits were typically 5 pm to 20 pm in thickness. Transmission electron microscopy (TEM) was performed on selected deposits prepared for both plan view and cross-sectional analysis. The plan view films were detached from a planar Cu substrate and affixed to a Mo support ring. Cross-sectional cuts, 0.75 mm thick, were cut from the overplated samples and were then dimple polished down to 20 pm thickness. Plan view and cross-section samples were subsequently milled with 4.5 kV Ar ions at 5° incident angle. Portions o f the samples with electron transparency were examined at 300 kV by TEM using a JEOL 3010. Figure 1 is a graph showing the composition o f alloys deposited onto copper substrates as a function of Ti2+ concentration and current density in 2:1 AlCL-NaCl [14] and from 2:1 AlCL-EtMelmCl at a single Ti2+ concentration [11]. For low Ti2+ concentrations in AlCL-NaCl, the alloy composition is dependent upon the applied current density. An alloy having a titanium concentration of 25 % atomic fraction is deposited only at low current densities. As the current density is increased, the Ti partial current density becomes limited by the diffusion of Ti2+, and the Ti content o f the alloy drops. At a Ti2+ concentration o f 150 mmol L"1, the current density dependence is virtually gone, and an alloy containing 25 % atomic fraction titanium is deposited at nearly all current densities. The highest titanium concentration observed is 28 % atomic fraction. Similar behavior has been reported by Uchida and co-workers [17] in the AICI3-NaCl-KCl eutectic at 200 °G.  [11]. This lower value can be attributed to both the lower diffusion coefficients inherent to ambient temperature melts, as well as to the formation of polymeric Ti2+ species at the lower temperature.

Deposit Composition
There are two unusual aspects of the alloy composition data shown in Figure 1.
The first is the apparent 25 % atomic fraction limit on the titanium concentration of the Ti-Al electrodeposits, an observation which has been reported by two independent groups [14,17]. It is somewhat surprising that even at the higher Ti2+ concentrations in the melt, alloys having a higher titanium content are not accessible. Generally, increasing the concentration o f an electroactive species results in an increased composition o f that species in the alloy, particularly when that species is the minor component. The independence of alloy composition on current density is also unusual and is generally the result o f compound formation or an indication that the two deposition reactions have There are obvious problems with this mechanism. In acidic chloroaluminate melts, the electroactive aluminum species is AI2CI7 while the more stable AICI4 is reduced at potentials about 0.5 V more negative [21]. Since these alloys fraction. The nodules show crystallographic faceting, although no symmetry characteristic of a particular growth direction was observed. It is clear that the nodule size decreases with decreasing current density and increasing titanium content. Typically one would expect an increase in grain size as the deposition current density or overpotential is decreased. However, in this case the grain refinement is driven by the incorporation of Ti into the alloy rather than the deposition overpotential. This nicely demonstrates the impact that impurities and alloying additions have on deposit morphology.
Interesting structural features can be seen in alloys electrodeposited at very small current densities. Since small current densities generally lead to poorly nucleated deposits, a higher current density pulse is typically required to properly nucleate the surface. Figure 2 Figure 2(f). The step faceting is clearly seen in each nodule, confirming that the nodules are single crystals. It also suggests that although the macro-deposit grows by three-dimensional nucleation and growth, each nodule grows primarily by two-dimensional, layered growth.

Deposit Crystal Structure
The crystal structure of several Al-Ti electrodeposits was determined by both xray and electron diffraction. X-ray patterns from deposits produced from both melts are shown in Figure 3. The current density and deposition time for each deposit are listed with each diffraction pattern. There was no evidence of the equilibrium DO22 A^Ti phase (body centered tetragonal, a0 = 0.384 nm and c0 = 0.858 nm) in any of the diffraction patterns or the high temperature variant A^Tig identified by van Loo and Rieck [22] (tetragonal, a0 = 0.388 nm and c0 = 3.384 nm). All of the patterns show the fundamental face-centered cubic (fee) reflections in positions very close to that of pure aluminum. The lattice parameters calculated from these reflections vary from 0.405 nm for pure aluminum to 0.400 nm for the deposits containing 25 % atomic fraction Ti. The shrinkage of the lattice with increasing Ti addition is expected as the smaller Ti atoms substitute for A1 in the fee lattice. The relative position o f the 111 fundamental reflections in Figure 3 reflects the Ti content of the electrodeposit. The alloys electrodeposited from the higher temperature AlCL-NaCl melt. Figure 3 The superlattice reflections shown in Figure 3(b-e) are typically very broad, the extent o f the broadening appears to be influenced by the deposition conditions. Extensive broadening o f the superlattice reflections might make them difficult to detect by x-ray diffraction. A report in the literature describing a glancing angle x-ray examination of Ti-Al alloys electrodeposited from eutectic AlCL-NaCl-KCl containing Ti3+ also shows the fundamental reflections for fee A1 for compositions up to 27.5 % atomic fraction Ti. However, the authors were unable to confirm the LI2 structure due to the lack of superlattice reflections in the x-ray diffraction pattern [17]. Whereas broadening o f fundamental reflections generally indicates limited crystallite size, the broadening of superlattice reflections suggests that the Lh-ordered domains are very small. One might expect that a low deposition current density might lead to more extensive ordering since surface rearrangement could occur under the reduced flux o f metal atoms. This appears to be contradicted by alloys (c) and (e) which were deposited at similar current densities. Deposit (e) shows relatively well-defined superlattice reflections, while those in (c) show significant broadening. The primary difference between these two samples is the thickness of the deposit, and as a result, the time that the deposit is exposed to the elevated temperature o f the molten salt. Although there does not appear to be a systematic correlation between the deposition time and superlattice reflection shape, it is clear that the domains are better defined in the deposit that was exposed for the longest time at 150 °C. I a F2 = 0 for superlattice reflections (fee) [3] where I is the intensity of a given reflection, F is the structure factor, fa is the scattering factor for the A1 lattice sites, and fa is the scattering factor for the Ti lattice sites. The scattering factor for a lattice site is the weighted average of the atomic scattering factors o f the atoms that occupy it. Since some o f the titanium lattice sites must be occupied by aluminum atoms in the substoichiometric alloys (assuming all lattice sites are occupied), the difference in the scattering factors for the two sites and, therefore, the superlattice reflection intensities must decrease. This is in qualitative agreement with the experimental observations. It is not possible to ascertain if the apparent absence of superlattice reflections in the 3.6 % atomic fraction Ti specimen is due to low intensity or to a disordered crystal structure because o f the loss o f superlattice reflection intensity at low Ti concentrations.
To confirm the chemical disorder in the Al-Ti alloys deposited from AICI3 -EtMelmCl melt, electron diffraction analysis o f selected electrodeposits in both plan view and cross-section was performed [11]. A typical diffraction pattern taken on the [001] zone axis from a deposit containing 18.4 % atomic fraction Ti is shown in Figure   5 [24].

DISCUSSION
The Al-Ti phase diagram [25] indicates that the solubility of Ti in fee A1 is about 0.7 % atomic fraction at 660 °C and decreases rapidly at lower temperatures. The line compound Al3Ti (DO22) is the next phase in order of increasing Ti content. Thus, the equilibrium structure of electrodeposits containing less than 25 % atomic fraction Ti is two-phase fee A1 (saturated with Ti) and DO22 Al3Ti. Likewise the 25 % atomic fraction Ti deposit should be single-phase DO22 Al3Ti. Consequently, the structures observed in 930 Electrochemical Society Proceedings Volume 2004-24 this study are non-equilibrium in two senses: they are single phase and the crystalline phase is either disordered fee or ordered L I 2 [29] showed that the energy difference between binary ALTi in the two crystal structures L I2 and D O 22 is quite small, about 2.4 kJ mol'1. Asta has reported similar energy differences in his ab initio study of phase stability in the Al-Ti system [30].

rather than D O 2 2 . The deposition o f an ordered LI2 phase is not surprising because small amounts of transition elements have been found to change the equilibrium structure o f ALTi-based alloys from D O 2 2 to L I 2 , which suggests that there is a very small difference in free energy between the two phases [26-28]. Fu
The magnitude of the energy difference between AbTi in the L I 2 and D O 22 structures suggests that it should be possible to form LI2 ALTi in a binary alloy using non equilibrium processing. The chemically disordered fee structure obtained from the ambient temperature melt is a significant departure from the ordered LI2 crystal structure deposited from AlCh-NaCl at 150 °C and suggests that larger departures from equilibrium can be obtained at lower deposition temperatures. Asta has calculated the energy difference between disordered ALTi and the equilibrium D O 2 2 structure to be 16.9 kJ mol"1 (i.e., 14.6 kJ mol' 1 between disordered AI3T1 and the LI 2' structure).
Metastable or unstable structures are formed by raising the free energy o f the starting materials and then removing that energy rapidly enough to ensure that the atomic mobility is sufficiently low to prevent or limit subsequent transformations [31]. The many fabrication processes which allow one to maintain these structures can generally be divided into three categories: quenching, molecular deposition, and external action (deformation, irradiation, or chemical attack) [32]. In rapid solidification the cooling rate is primarily a function o f the smallest dimension of the sample and the medium used to remove the heat. Unless an amorphous phase is retained* nucleation seems to limit the degree o f super-cooling to about 30 % o f the melting temperature; i.e., AT/Tm ~ 0.3 [33] .
Based on this limitation, it is generally impossible to produce a metastable crystalline phase from the melt having an excess free energy relative to the equilibrium form larger than ~ 0.3 AHm (-3.0 kJ mol'1), where AHm is the enthalpy o f melting [34,35]. The fact that the LI2 phase has not been observed in rapidly solidified binary alloys [36,37] is most probably due to the fact that the energy difference between the LI2 and DO22 crystal structures (2.4 kJ mol*1) approaches the excess free energy limit o f this technique (3.0 kJ mol'1).
Based on the formation energies discussed above, one would not expect A^Ti to form a disordered fee structure by rapid solidification. Boettinger has developed the theory that describes the trapping o f disorder in intermetallic phases by rapid solidification [38]. Equation [4] describes the critical solidification velocity, V, which results in the crystallization of a disordered alloy, where Tc is the critical order-disorder transformation = JiL-1 Vd Tm [4] temperature, Tm is the melting point of the solid, and Vd is a kinetic rate parameter for crystallization. Vd is typically taken as the speed o f liquid interdiffusion (Dl/X) where Dl is the liquid diffusion coefficient and X is the jump distance. The Tc/Tm for LI2 A^Ti is estimated to be about 1.8 [25,30]. Assuming a Vd value of 200 cm s'1 ( D l = 6 x 10*6 cm2 s'1, X = 3 x 10'8 cm), one calculates a critical solidification velocity o f 160 cm s'1. Such a velocity is typically difficult to achieve in rapid solidification from the melt.
In condensation, vapor deposition, or sputtering, the extent of supercooling possible is much greater that that from the melt. In addition, the enthalpy of vaporization is generally an order of magnitude larger than that o f melting. Consequently, one can produce metastable crystalline phases having an excess free energy on the order of -50.0 kJ m o l1, making vapor condensation a potentially much more powerful method for crystalline phase retention than melt solidification [34,35]. While exact correlations between electrodeposition and vapor deposition have not been developed, the excess free energy possible in alloys produced by these two techniques is similar [34,39]. Electrodeposited alloys are rarely in equilibrium and one can assume that at least a portion of the activation overpotential is used to increase the free energy o f the deposit relative to that o f the equilibrium phase. The extra energy reauired to electrodeposit AETi in the L h (2.4 kJ mol'1) and disordered fee (16.9 kJ mol'1) structures rather than the equilibrium DO22 structure would require a minimum overpotential of 9 mV and 64 mV respectively. These are clearly accessible deposition overpotentials and add some justification for the metastable structures observed in these electrodeposits.
Although it is clear that the excess free energy observed in these Al-Ti alloys can be achieved through electrodeposition, the actual mechanism by which these structures are electrochemically formed remains unclear, particularly since these deposits display crystallographic features of distinctly different length scales. The grain size of all electrodeposits examined in this study is on the order o f 0.5 pm to 5 pm. Deposits formed at 80 °C are chemically disordered while those formed at 150 °C have LI2 domains measuring 5 nm to 10 nm in size. The image in Figure 6(b) suggests that these domains appear to have grown through a first order nucleation and growth process, independent of the electrocrystallization process. This type of domain structure is quite common in rapidly solidified alloys wherein the disordered phase produced by the solidification process transforms to the equilibrium ordered phase quite rapidly during solid-state cooling to room temperature [40]. Although this often occurs at temperatures that do not allow for significant bulk diffusion, the activation barrier to the nucleation of ordered domains is rather small because both the nucleus and matrix have essentially the same crystal structure and composition [41]. Consequently, nucleation tends to be homogeneous and independent of lattice defects. CU3AU is the classic system for studying first-order compositional order-disorder transformations. Similar to electrodeposited AhTi, CU3AU forms an LI2 structure in the ordered state. When quenched from above the critical (order-disorder transformation) temperature to room temperature, C113AU retains the disordered structure. Annealing the disordered alloy below the critical temperature, causes ordering to occur through the nucleation and growth o f ordered regions at different points throughout the crystal. These small ordered regions grow by consuming the disordered material until they impinge to form a domain structure, similar to that shown in Figure 6(b). There is considerable experimental evidence in the literature suggesting that antiphase domain growth is analogous to classical metallurgical grain growth [42][43][44][45] in that it follows the relation, D2-Dg = 2 K t [5] where D is the domain size (diameter), and D0 is the domain size at t = 0. The rate constant K has an Arrhenius-type temperature dependence and is proportional to the free energy of the domain boundary and its mobility. In a given domain, the boundary advances when the atoms in anti-phase positions in the domain being consumed jump the required distance and become in-phase with the other atoms in the growing domain. In a binary ordered alloy, both atomic species will be required to jump for the boundary to advance. Consequently, the process of antiphase domain growth is typically viewed to be diffusion-controlled. The activation energy for domain growth in CU3 AU has been measured at 184 kJ mol"1 which is consistent with bulk diffusion in the alloy [42,43]. Cahn has made a direct comparison of the interdiffusion coefficient and the time derivative of the square of the domain size in CU3AU and determined that they are essentially identical in the temperature range o f 300 °C to 375 °C [46].
Assuming that the Ti-Al alloys deposit in the disordered state and that ordering occurs in the solid state, the degree of chemical order-disorder would be directly linked to the deposition temperature and the time at which the deposit remains at the elevated temperature during deposition. At first glance it would seem that there is little difference between 80 °C and 150 °C with regards to diffusional processes, both temperatures are rather low. If we assume that K in equation [5] is thermally activated (i.e., K = K0 exp (-Q/RT)), then the activation energy (Q) required to produce a 5 nm domain size at 150 °C and a disordered deposit (< 0.1 nm domain size) at 80 °C is 138 kJ mol'1. This is consistent with a thermally activated diffusion process. This analysis can be taken a step further with knowledge of the A1 and Ti diffusivity in fee A^Ti. Unfortunately, this data does not exist and diffusion coefficients from chemically related systems are only available for temperatures exceeding 350 °C. Extrapolation of high-temperature data to temperatures less than 150 °C may not be valid since alternative mechanisms such as grain boundary, dislocation, and surface diffusion may become dominant at lower temperatures. However, since the domains shown in Figure 6(b) are much smaller than the grain size, it is likely that a vacancy-exchange mechanism is operative during the ordering process and that high-temperature data can be used to quantify domain growth. The kinetics for Al self-diffusion as well as the diffusion of Al in pure Ti is sufficiently fast to account for the ordering phenomena observed in the electrodeposited alloys whereas the kinetics for Ti diffusion in pure Al as well as the interdiffusion of Ti and Al in AbTi would not appear to support ordering at these relatively low temperatures. Although one might expect the AI3U diffusion data to best describe our experimental system, the former data is based on an alloy that has the ordered DO22 structure. The chemical ordering as well as the smaller lattice volume inherent to this structure may alter the diffusion kinetics; consequently diffusion in disordered AI3H may be expected to behave differently. l/T (K*1)

SUMMARY
The composition, morphology and crystallographic structure of Al-Ti alloys electrodeposited from two different chloroaluminate molten salt electrolytes were examined. Alloys containing up to 28 % atomic fraction Ti were electrodeposited at 150 °C from 2:1 AlCl3-NaCl with controlled additions o f Ti2+. The apparent limit on alloy composition is proposed to be due to a mechanism by which AI3T1 forms through the reductive decomposition o f [Ti(AlCl4 )3] . The composition o f Al-Ti alloys electrodeposited from the AlCb-EtMelmCl melt at 80 °C is limited by the diffusion of Ti2+ to the electrode surface. Alloys containing up to 18.4 % atomic fraction Ti are only obtainable at high Ti2+ concentrations in the melt and low current densities. Al-Ti electrodeposited from the higher temperature melt has an ordered LI2 crystal structure while alloys of similar composition but deposited at lower temperature are disordered fee. The appearance of antiphase boundaries in the ordered alloys suggests that the deposit may be disordered initially and then orders in the solid state, subsequent to the charge transfer step and adatom incorporation into the lattice. This is very similar to the disorder-trapping observed in rapidly solidified alloys. The measured domain size is