Correlative theoretical and experimental investigation of the formation of AlYB14 and competing phases

The phase formation in the boron-rich section of the Al-Y-B system has been explored by a correlative theoretical and experimental research approach. The structure of coatings deposited via high power pulsed magnetron sputtering from a compound target was studied using elastic recoil detection analysis, electron energy loss spectroscopy spectrum imaging, as well as X-ray and electron diffraction data. The formation of AlYB 14 together with the (Y,Al)B 6 impurity phase, containing 1.8 at. % less B than AlYB 14 , was observed at a growth temperature of 800 (cid:2) C and hence 600 (cid:2) C below the bulk synthesis temperature. Based on quantum mechanical calculations, we infer that minute compositional variations within the ﬁlm may be responsible for the formation of both icosahedrally bonded AlYB 14 and cubic (Y,Al)B 6 phases. These ﬁndings are relevant for synthesis attempts of all boron rich icosahedrally bonded compounds with the space group: Imma that form ternary phases at similar compositions. V C 2016 AIP Publishing LLC . [http://dx.doi.org/10.1063/1.4942664]

The formation of these materials in bulk form relies on processing under high temperature and/or high pressure conditions. 1,12,14 Controlling the incorporation of impurities and the formation of competing phases are key challenges to be addressed during synthesis; 15 e.g., the occurrence of YB 6 during synthesis of B 14 -phases has been reported. 10 The influence of such phases, with similar chemical composition to B 14 -phases, on the synthesis of B 14 -phases has not been investigated in a systematic way. Besides synthesis in bulk form 16,17 also thin film synthesis of icosahedral boron-rich solids 15,18 has been reported. The icosahedrally bonded borides should not be confused with other icosahedral phases such as quasicrystals. The following XZB 14 compounds where X and Z are metal atoms (space group: Imma), see Fig. 1, have been synthesized in bulk form: AlYB 14 , 10 23 and AlMgB 14 -TiB 2 . 24,25 The bulk synthesis of crystalline AlMgB 14 17 and AlYB 14 16 has been reported at temperatures above 1400 C, and the crystallographic properties of Al 0.5-0.8 Mg 1.0-1.1 B 14 19 and Al 0.71 Y 0.62 B 14 16 are presented by Matkovich and Economy and Korsukova et al., respectively. Attempts to grow AlMgB 14 thin films by pulsed laser deposition 15 and magnetron sputtering (MS) 26 both at temperatures of 600 C resulted in the formation of X-ray amorphous structures. So far, only post-annealing of films with the composition of AlMgB 14 resulted in the formation of crystalline AlMgB 14 at temperatures of 1000 C. 15 Tian et al. observed three distinct peaks at 13.85 , 27.89 , and 42.37 corresponding to (011), (022), and (033) planes, respectively. 15 For Al-Y-B, direct synthesis of crystalline AlYB 14 thin films at a substrate temperature of T S ¼ 800 C was reported by K€ olpin et al. 18 In that work, a combination of direct current (DC) and radio frequency (RF)-MS was used, but the mechanisms which trigger the growth of crystalline AlYB 14 remain unclear. 18 The phase formation was reported based on the presence of 6 diffraction peaks which are consistent with the AlYB 14 structure: 25.374 , 35.300 , 37.270 , 40.346 , 41.949 , and 42.059 corresponding to (121), (220), (132), (123), (231), and (033) planes, respectively. 18 Yan et al. 27 reported the formation of AlMgB 14 nanocrystals in an amorphous Al-Mg-B matrix in the as deposited state at a substrate temperature of 200 C based on diffraction data reported in 2013. 28 Yan et al. utilized an AlMg compound target which was operated in DC mode and two B targets which were according to the authors sputtered using 350 kHz pulses. 28 This is the first report on the direct synthesis of crystalline AlMgB 14 at a temperature of 200 C. 28 No discussion of the physical mechanism enabling the synthesis at these very low growth temperatures is given. 28 The diffraction peak at 42.3 was associated by the authors with the formation of the AlMgB 14 structure. 28  peak is shifted by 0.3 with respect to (213) position given in JCPDS No. 75-1262. 28,29 Furthermore, Yan et al. 28 present selected area electron diffraction (SAED) data showing a diffuse diffraction ring which is also assigned to (213). For an Al-Mg-B-Ti thin film containing 17.3 at. % Ti only one peak can be found in the diffractogram. 28 This peak position is consistent with the AlMgB 14 structure. No shift with respect to JCPDS No. 75-1262 29 can be seen. 28 No SAED is presented for the thin film containing 17.3 at. % Ti, though. The authors assign the single diffraction peak to textured growth of the AlMgB 14 -phase. 28 However, if nanocrystals are dispersed in an amorphous matrix as suggested by Yan et al. 28 an identical orientation of all nanocrystals appears rather unlikely. Hence, multiple diffraction peaks from the AlMgB 14 -phase should be expected. Hence, both of the above discussed diffraction data sets 28 leave room for alternative phase formation scenarios as presented by the authors. 28 During PVD growth from compound targets the film composition may deviate from the target composition 30,31 which may cause the formation of competing phases. Possible competing phases are YB 12 , 32 YB 6 , 33 AlYB 14 , 16 AlB 12 , 34 and AlB 10 . 35 Considering additional O incorporation in the growing film from residual gas 36 or as a consequence of post growth atmosphere exposure, 37 even more phases have to be considered: Al 2 O 3 , 38 Y 2 O 3 , 39 and YAlO 3 . 40 The diffraction data (Cu K a ) of the previously listed phases are shown in Fig. 2. It is immediately evident that multiple peak overlaps occur and that a positive phase identification only based on X-ray diffraction (XRD) in this material system is challenging.
It was discussed above that for the synthesis of crystalline AlYB 14 thin films, a substrate temperature of 800 C was required. 18 To allow for deposition onto technologically relevant substrate materials for the application as cutting and forming tools, the synthesis temperature of crystalline icosahedrally bonded borides has to be decreased.
It has been reported that by utilizing high power pulsed MS (HPPMS), the deposition temperature can be lowered as compared to DC sputter deposited thin films. HPPMS is a prolific provider of ions formed from the sputter gas as well as from the sputtered and hence film forming species. 41,42 These ions have been reported to enhance the adatom mobility resulting in a reduction of the synthesis temperature for Mo 2 BC from 900 C (Ref. 43) to 380 C (Ref. 41) and for V 2 AlC from 750 C (Ref. 44) to 500 C. 42 During HPPMS, voltages <500 V and peak currents of several A cm À2 are applied to the target over 20-500 ls long pulses at duty cycles <10%, which results in peak power densities of up to several kW cm À2 . 45 Within the pulse on-time (t on ), the plasma density can be increased significantly 46 which results in a higher ionization fraction of sputtered film forming species. 47,48 Ionization fractions of 70% and 40% have been measured for Cu and Ti 0.5 Al 0.5 , respectively. 47,49 The ionization fraction of film forming species is reported to be much larger than the percentage level in DCMS 50 enhancing the surface mobility of adatoms 42,51 at the growing film surface. 41 Enhanced surface diffusion may facilitate growth of high temperature phases at lower temperatures. 41,42,45,51 Hence, we utilize HPPMS to investigate the influence of the substrate temperature on the structural evolution of boron rich Al-Y-B thin films. Special attention is paid to the potential formation of competing binary phases and phases that may form as a result of impurity incorporation. Based on a correlative theoretical and experimental approach, the formation of (Y,Al)B 6 as a possible competing phase differing only by 1.8 at. % in boron concentration to the AlYB 14 phase is investigated.

II. EXPERIMENTAL DETAILS
The Al-Y-B films were synthesized in a high vacuum system with a base pressure of 3 Â 10 À5 Pa by HPPMS utilizing a stoichiometric AlYB 14 (purity 99.9%) compound target with a diameter of 50 mm. The deposition was carried out at an Ar (99.9999% purity) pressure of 0.9 Pa. The timeaveraged power density supplied by a Melec power supply (SIPP2000USB-10-500-S) was 7.1 W cm À2 . The pulse ontime, t on , was 100 ls and the pulse off-time, t off , was 2900 ls which corresponds to frequencies of 333 Hz and a duty cycle, t on /(t on þ t off ), of 3.3%. Thereby peak power densities of 0.3 kW cm À2 over the whole target and 0.9 kW cm À2 in the race track were obtained. The target-substrate distance was 10 cm. A 50 mm sapphire wafer with an (0001) orientation was used as substrate. The substrate temperature, T S , was varied between 675 and 800 C. The substrate was rotated with a constant velocity of 30 rotations per minute. Further information on the deposition system can be found elsewhere. 52 To study the film constitution and grain size distribution, XRD using a Bruker AXS D8 Discover General Area Detection Diffraction System and high resolution transmission electron microscopy (HRTEM) with SAED utilizing an FEI Titan 50-300 PICO with Cs-Cc corrector were performed. XRD measurements were carried out employing Cu K a radiation at a fixed incidence angle of 15 using a collimator with a diameter of 0.5 mm. A 2h-range of 15 -75 was covered during the detector scan using a 2D area detector. The chemical composition was determined by energy dispersive X-ray (EDX) analysis with an EDAX Genesis 2000 analyzer and Time-Of-Flight Elastic Recoil Detection Analysis (TOF-ERDA). The experimental details of the TOF-ERDA measurements can be found in Ref. 53 whereas the detection system and associated systematic uncertainties are discussed in a study by Zhang et al. 54 The sample for TEM investigation was fabricated with a focused ion beam (FIB, FEI Helios 660) followed by a post thinning-process under illumination of 500 eV ion beam (Nanomil, Fischione). Furthermore to investigate the elemental and chemical structure of the thin film EDX spectroscopy mapping as well as electron energy loss spectroscopy spectrum imaging (EELS-SI) of the samples was acquired by analytical scanning TEM (STEM, FEI Titan 3 60-300, equipped with a Super-X EDX detector and a Dual-EELS Quantum ERS GIF).

III. COMPUTATIONAL DETAILS
Density functional theory (DFT) 55 calculations were performed with the Vienna ab initio simulation package (VASP) 56 to determine the energy of formation of the (Al,Y) 2Àx B 14 (x ¼ 0, 0.25, 0.5) and (Y,Al) 1Ày B 6 (y ¼ 0, 0.0625, 0.125) systems. For the calculations generalizedgradient approximation with projector augmented wave (PAW) 57 potential, a reciprocal space integration using the Monkhorst-Pack scheme 58 and tetrahedron method for total energy using Bl€ ochl-corrections 59 were employed. The energy cut off and the electronic relaxation convergence were 500 eV and 0.01 meV, respectively. The used k-point grids were 5 Â 5 Â 5 for the (Al,Y) 2Àx B 14 and 4 Â 6 Â 6 for (Y,Al) 1Ày B 6 , for which a 4 Â 2 Â 2 super cell was used. The Al and Y atoms in (Y,Al) 1Ày B 6 were randomly distributed on an ad hoc basis. Prior to the energy calculations, the atom positions were relaxed with a force convergence of 1 meV/Å . Subsequently, the energy of formation was calculated using where E f is the energy of formation per atom, i the stoichiometric coefficient (negative for reactants, positive for products), E i the energy of the specific compounds (face centered cubic Al, hexagonal Y, rhombohedric B), and n atom the number of atoms in the calculated system.

A. Chemical composition
The sample deposited at 800 C was measured by TOF-ERDA and used as a standard for the EDX measurements. According to the analysis performed with the CONTES program package 60 the film is found to be composed of 3.3 at. % Al, 6.7 at. % Y, 84.8 at. % B, 2.2 at. % O, and 1.6 at. % C. Impurities of F, N, Ar, and H were lower than 1 at. %. The O-content is probably caused by the incorporation of oxygen stemming from residual gas during thin film growth. 36 A local enrichment of oxygen at the film surface may be caused by subsequent exposure of the as-grown film to air. 61,62 The composition of the films deposited at lower T S , 750-675 C, was measured by EDX with an acceleration voltage of 6 kV using a sample quantified by ERDA as standard. The significant difference to the sample deposited at T S ¼ 800 C is an increased Al content of 4.8 at. %. The lower Al content at T S ¼ 800 C can be explained by the Al melting point, which is the lowest among the film constituents. Hence, preferential evaporation of Al takes place at elevated temperatures leading to these low Al-contents. The resulting Al/Y ratio of 0.57 for T S ¼ 800 C is consistent with the composition employed in calculations reported by K€ olpin et al., 18 where the lowest energy of formation for the Al-Y-B system was reported for an Al/Y ratio of 0.5. 18 As the B icosahedra are electron deficient, 1 additional charge is required to stabilize the structure. 18 The phase stability therefore depends crucially on the charge transferred from the metals, e.g., Al and Y, to the boron icosahedra. 18,63,64

B. Structure
The diffractograms of the films for T S between 675 C and 800 C are shown in Fig. 3. With increasing substrate temperature the full width at half maximum decreases by 60% and 53% for peak (I) and (II), respectively, and hence the crystal quality improves. These peaks can clearly be observed at 30.76 (I) and 37.87 (II). All peak positions of the aforementioned peaks fit very well to the AlYB 14 -structure as well as to the binary cubic YB 6 -structure. Other phases, which could have developed due to impurities, such as AlB 12 , 34 AlB 10 , 35  match the diffractograms and are therefore excluded from further discussions. Based on the data of Korsukova et al. 16 AlYB 14 deviates by 0.07 and 0.01 from peak (I) and (II), respectively. Deviations of YB 6 based on Blum and Bertaut 33 are 0.02 and À0.02 from peak (I) and (II), respectively. Hence, a positive phase identification cannot be conducted based on XRD.
In Fig. 4, the STEM-HAADF (high-angle annular darkfield imaging) image as well as the STEM-EELS data of the Al-Y-B thin film grown at a substrate temperature of 800 C is shown. From the STEM-HAADF image depicted in Fig. 4(a), it is evident that a nanocomposite containing nm sized crystals in an amorphous matrix is formed. To identify which crystalline phases do in fact form, energy dispersive X-ray spectroscopy mapping as well as EELS-SI correlative electron microscopy was conducted on the sample synthesized at T S ¼ 800 C. To address the elemental distribution, EDX spectroscopy mapping was performed at high magnification (not shown here). The distribution of B is found to be uniform. Regions that lack the presence of either Y or Al completely could not be identified. Hence, all phases appear to contain B, Al, and Y. However, due to projection effects, one cannot conclude unequivocally that all phases formed do in fact contain all three alloy constituents.
EELS spectrum image (SI) was utilized to further investigate the spatial distribution of the chemical composition of the crystals. In Fig. 4(a) the dashed rectangle indicates the area in which an EELS was recorded. Simultaneous with the SI, a (low resolution) HAADF-STEM signal was recorded for relating the SI details to the microstructural features and is shown in Fig. 4(b). Indicated in Fig. 4(b) are the areas used for averaging spectra from the bright particles and the darker (less dense) matrix, and the resulting spectra are found in Fig. 4(c). Throughout the spectra, all three elements (Y, Al, and B) are present. While the fine structure of the Y and B edges (Y-M at 165 eV and B-K at 201 eV energy loss) remains largely unchanged, the Al-L edge (at $73 eV energy loss) is shown in detail. In Fig. 4(c), the averaged spectra originating from the crystals and the matrix are displayed. As it can be seen, there is a significant difference in the sharpness and height of the initial Al peak at $78 eV energy loss. The peak stemming from the crystalline region is sharper and larger in intensity (independent of crystal orientation) as compared to the amorphous matrix, indicating a different chemical environment for Al and revealing that Al exists in both crystalline particles and amorphous matrix. While two distinctly different signals are observed from this region, the underlying edge profile is rather similar. It should be noted that EELS at this nanometer scale and at these energy losses is not entirely straightforward. Initially, indicated areas used for averaging the spectra from bright crystals and darker matrix (c) STEM-EELS normalized to the area underneath the curves. For better presentation, the EELS spectra are shifted laterally. Both particle and matrix data start with zero intensity. geometrical considerations must be taken into account, such as the converging ($20 mrad half angle) passing through a sample of limited thickness, with potential overlap of multiple particles and embedding matrix in the projected direction. To minimize overlap, the data were recorded at the thin edge of the sample, where it is possible to distinguish and exclude regions with overlapping crystals. Additionally, in these thin regions the Å -sized converging beam is not exciting the matrix and crystal other than at the interface. Additionally, core-loss EELS is a delocalized method such that information is acquired outside of the beam with exponentially reduced probability. Typically, for Al-L losses, the spectrum is acquired from less than a nm from the beam and this causes a mixing of both crystal and matrix components. However, the average spectrum acquired on the crystal exhibits the more crystal like component, while the matrix component dominates in the average matrix spectrum.
This supports strongly the notion that the peaks observed by XRD originate from crystals forming a ternary phase, and not from a potentially competing binary phase, embedded in an amorphous matrix containing Al, Y, and B. Furthermore, the trend that more Y is located in the crystalline areas is consistent with both the formation of AlYB 14 or (Y,Al)B 6 : based on systematic DFT study of the occupancies of Al and Y in Al x Y y B 14 the most stable phase predicted is Al 0.5 YB 14 . 18 To determine which ternary boride is formed HRTEM has been carried out on the crystalline areas. The formation of two different phases is evident. The two different crystal systems are illustrated in the lower part of Figs. 5(a) and 5(b). Fig. 5(a): The orthorhombic AlYB 14 (space group: Imma) containing boron-icosahedra with five-fold symmetry and (b) (Y,Al)B 6 with its body-centered cubic symmetry (bcc) structure are revealed. In Figs. 5(a) and 5(b), both structures are compared to structural data obtained from quantum mechanical calculations. For illustrative purposes, the boron atoms have been left out in both images as they are undetectable in HRTEM due to the high density (87.5 at. %) and therefore small inter-spacing between boron atoms, which result in a very low contrast for the lattice image. The HRTEM data shown in Fig. 5(a) are consistent with the theoretic model in (100) orientation, clearly indicating the phase formation of the orthorhombic AlYB 14. This notion is also corroborated by SAED data shown in Fig. 5(c). The aperture size was approximately 300 nm; therefore based on the STEM-HAADF image in Figs. 4(a) and 4(b), it is expected that diffraction data stem from the whole nanocomposite. The diffuse ring in the SAED data indicates the presence of an amorphous phase which is also consisted with the STEM-HAADF image shown in Fig. 4(a). 9 diffraction rings can be observed and are shown as isolated peaks in the diffractogram in Fig. 5(c) by a rotational average-approach. 73 The diffractogram was calibrated against polycrystalline gold nanoparticles using the same camera length. The corresponding reciprocal values of d-spacings are marked with arrows in Fig. 5(c) bottom. The d-spacings are listed in Table I. It is evident that the most pronounced diffracted intensity stems from the first two rings. To compare the SAED results with the XRD and JCPDS data, these are summarized in Table I. d-Spacings according to SAED and XRD as well as the (hkl) and diffraction data of YB 6 33 and Al 0.71 Y 0.62 B 14 16 are listed in Table I. Additionally, the deviations between SAED and the diffraction data are given. Six SAED rings are assigned to either YB 6 33 or AlYB 14 16 with a maximum deviation of 0.44% between the JCPDS data and the experimentally determined lattice spacings, see bold marked lines. d SAED ¼ 2.316 Å might be assigned to B 2 O 3 69 whereby d SAED ¼ 1.872 and 1.806 Å might be assigned to Y 2 O 3 . 39 Hence these diffraction rings are assumed to originate from oxide formation which is consistent with the chemical composition data discussed above or due to post-oxidation of the TEM sample. For lower d spacing values, more peaks could be assigned but as the data quality decreases, as shown in Fig. 5(c), these diffraction rings are not discussed further as the difference between peak position and JCPDS data are smaller than the measurement error of the SAED. In summary, the SAED data confirm the HRTEM regarding the phase formation as the diffraction data clearly show the formation of both here discussed boride phases: AlYB 14 16 as well as YB 6 . 33

C. Calculations
The above identified ternary hexaboride phase (Y,Al)B 6 (see Figs. 5(b) and 5(c) as well as Table I) has been investigated by DFT. As discussed above, growth of AlYB 14 10,16,18 is challenging. (Y,Al) 1Ày B 6 (y ¼ 0, 0.0625, 0.125) is considered as a potential competing phase for this ternary system at high B contents. YB 6 exhibits a body-centered cubic phase. 33 In this work, calculations have been performed pertaining to the formation of a solid solution where Y is partly substituted by Al, while metal vacancies are introduced in order to study higher boron contents. It should be noted that the Y-B phase diagram suggests significant solubility of vacancies on the metal sublattice of YB 6 . 74 Energy of formation data for (Al,Y) 2Àx B 14 (x ¼ 0, 0.25, 0.5) and (Al,Y) 1Ày B 6 (y ¼ 0, 0.0625, 0.125) has been calculated and is shown in Fig. 6. Al/Y ratio is kept constant at 1. According to the energy formation data, the stability range for (Al,Y) 2Àx B 14 (x ¼ 0, 0.25, 0.5) extends from a boron to metal ratio of 14-6.4, while for a metal to boron ratio of smaller than 6.4 the cubic (Y,Al) 1Ày B 6 (y ¼ 0, 0.0625, 0.125) phase appears to be stable. The required high vacancy concentration on metal sublattices leads to multiple feasible arrangements of metal atoms. Hence, considering possible variations in energy deviations were estimated to be 610% and are shown in Fig.  6. These estimations are consistent with the literature for AlYB 14 . 18 Additionally, the target composition and the composition of the film deposited at 800 C are shown in Fig. 6. From these data, it is evident that fluctuations in film composition may cause the formation of thin films containing both the (Al,Y) 2Àx B 14 phase and the cubic (Y,Al) 1Ày B 6 phase. It may be speculated that the composition modulation is caused by the competitive nucleation of phases with different solubility for the sputtered elements and or by the formation of impurity phases such as oxides modifying the local composition. Both scenarios are expected to result in the formation of the here observed nanocomposite structure.

V. CONCLUSIONS
The phase formation in boron-rich section of the Al-Y-B system has been explored by a correlative theoretical and experimental research strategy. Correlative structure and bonding analysis indicate that the crystals formed in the amorphous matrix are icosahedrally bonded AlYB 14 and cubic (Y,Al)B 6 . HRTEM and SAED data allow the phase identification of coexisting of icosahedrally bonded AlYB 14 and cubic (Y,Al)B 6 . Based on quantum mechanical calculations, we infer that minute composition variations within the film may be responsible for the formation of both icosahedrally bonded AlYB 14 and cubic (Y,Al)B 6 phases. These findings are relevant for synthesis attempts of all boron-rich icosahedrally bonded compounds with the space group: Imma that form ternary phases at similar compositions. Additionally, the boron/metal ratio of the target and the coating as deposited at T S ¼ 800 C is given. The energy of formation data was fitted by cubic functions and provided as a guide for the eye.