Electronic structure of epitaxial half-Heusler Co_1-xNi_xTiSb across the semiconductor to metal transition

Half-Heusler compounds are ternary intermetallic (XYZ) compounds, which, depending on the composition and ordering, can exhibit a number of different electronic and magnetic structures. Some of the properties observed include semiconducting,¹ half metallic,² and thermoelectric³ behaviors. Additionally, a number of the half-Heusler compounds have been predicted to be topological insulators,⁴ and recently topologically non-trivial surface states have been observed.⁵ This class of materials can be alloyed in a similar manner to compound semiconductors leading to controllable tuning of their electronic and magnetic properties. Using this technique, promising values of the thermoelectric figure of merit (ZT) have already been shown for half-Heusler based thermoelectrics⁶ as well as Fermi level tuning in full-Heusler magnetic tunnel junctions where room temperature tunneling magnetoresistance of greater than 100% has been reported.^7–9 Additionally, half-Heusler compounds have similar lattice parameters and crystal structure as III–V compound semiconductors, opening the possibility of half-Heusler/III–V semiconductor heterostructures with unique properties. Previous investigations have used molecular beam epitaxy (MBE) to grow high quality thin films of CoTiSb^10,11 and PtLuSb^5,12 on III-V compounds. Using MBE, record high carrier mobilities for the semiconducting half-Heusler, CoTiSb, were demonstrated.¹⁰ Recently, the electronic structure of CoTiSb was investigated using angle-resolved photoemission spectroscopy (ARPES) and scanning tunnelling spectroscopy (STS), where good agreement was found with the density functional theory (DFT) calculated and the experimentally determined band structure.¹³ However, systematic experimental studies of how the electronic structure evolves as a function of alloying within Heusler thin films are lacking.

In this work, we use the semiconducting half-Heusler, CoTiSb, to investigate tunability of electronic properties through alloying. CoTiSb has 18 valence electrons per formula unit, making it a semiconductor with approximately a ∼1 eV bandgap.¹ In contrast, NiTiSb, which has one more valence electron per formula unit, displays metallic behavior with the Fermi level well into the conduction band. By substitutionally alloying nickel on the cobalt site, the electronic band structure should be tunable. Previous efforts have used bulk crystals with density functional calculations to understand this semiconductor to metal transition.^14–16 High defect densities were observed which required the introduction of an impurity band to understand the resistivity data. In addition, no experimental method to directly probe the density of states was performed. Here, the evolution of the electronic structure as a function of nickel alloying is examined using electrical and ARPES measurements on high-quality MBE prepared epitaxial thin films.

Co_1-xNi_xTiSb (001) samples were grown in a VG V80 MBE system on nearly lattice-matched In_0.52Al_0.48As (referred to as InAlAs) buffer layers epitaxially grown on an InP (001) substrate. For electrical transport measurements, an unintentionally doped 400 nm thick InAlAs buffer layer grown on semi-insulating InP:Fe (001) substrates was used. For ARPES and scanning tunneling microscopy (STM) measurements, the InAlAs layer was n-type doped with approximately 10¹⁸ Si atoms/cm³ and was grown on sulfur doped InP (001) yielding an n-type, conductive buffer structure. InAlAs layers were grown in a modified VG V80H III–V MBE system. Buffer layers were then arsenic capped and transferred through air into a dedicated metal MBE system for subsequent growth of Co_1-xNi_xTiSb. After the samples were reintroduced to UHV, the arsenic cap was desorbed to reveal the arsenic-terminated (2 × 4)/c(2 × 8) InAlAs surface. Co_1-xNi_xTiSb was grown by simultaneous evaporation of cobalt, nickel, titanium, and antimony using stoichiometric fluxes with a total flux of 9 × 10¹⁶ atoms/cm² h, giving an approximate growth rate of 2.5 Å/min. All fluxes were measured in situ using a beam flux ionization gauge that is calibrated ex situ by measuring the elemental atomic areal density of calibration sample layers grown on silicon substrates using Rutherford backscattering spectrometry (RBS). Samples were grown at substrate temperatures in the range of 250–380 °C as measured by a thermocouple that is calibrated to the arsenic desorption temperature of arsenic capped GaAs.¹⁷ For samples with x ≥ 0.5 nickel alloying, a thin 1 nm CoTiSb interlayer was used to minimize film reactions with the underlying III–V buffer layer, as nickel has been previously shown to be reactive with GaAs.^18,19

Growth was monitored in situ using reflection high-energy electron diffraction (RHEED). Films were nucleated using 4 nm thick, low temperature (250 °C), seed layer to minimize interfacial reactions. After seeding, the films were annealed and growth was resumed at temperatures up to 380 °C. For growth temperatures above 380 °C, additional spots could be observed forming on the RHEED streaks indicative of roughening, additional phase formation, or interfacial reactions. The crystal structure and electrical properties were analyzed ex situ using X-ray diffraction (XRD), electrical transport, and Seebeck measurements. For XRD, electrical transport, and Seebeck measurement samples, an approximately 5–10 nm thick e-beam deposited AlO_x cap was used as a protective layer to prevent oxidation of the film. Because the AlO_x cap, buffer layer, and substrate are effectively insulating, the electrical conduction is limited to the Heusler layer. The electrical resistivity of the Co_1-xNi_xTiSb series was measured between 20 and 300 K using a standard DC technique in a van der Pauw geometry with indium bonded gold contacts to the samples. Seebeck coefficients of the series were measured at room temperature using indium soldered electrical contacts and thermal paste for the thermocouple temperature probes. All Co_1-xNi_xTiSb films grown for electrical resistivity and Seebeck measurements were approximately 24 nm thick.

ARPES measurements of Co_1-xNi_xTiSb films were performed at beamline I4 of MAX III, part of MAX-LAB in Lund, Sweden, using a Specs PHOIBOS 100 hemispherical electron analyzer. Base pressure of the analysis chamber was approximately 1 × 10⁻¹⁰Torr. All ARPES data were taken at <100 K to reduce the effects of thermal broadening. Binding energies are referenced to the Fermi level as determined by fitting the spectra Fermi edge, in good agreement with the measured position with respect to a tantalum foil in contact with the sample. Samples were transported ex situ using a thin protective antimony capping layer that was thermally desorbed once returned to UHV. Cap removal was confirmed by low-energy electron diffraction (LEED) and measurements of shallow core levels in photoemission.

During growth, a (2 × 1) surface reconstruction is observed in RHEED for x ≤ 0.5 shown in Figs. 1(a)–1(f), similar to that observed in pure CoTiSb.^10,11 For higher levels of nickel alloying, other reconstructions including a (2 × 2) [Figs. 1(g)–1(i)] and disordered (3 × 2) are observed. These variations from the (2 × 1) are likely a result of the difference in the number of dangling bonds at the surface due to the additional electron per nickel atom. A transition from a (2 × 1) to a (3 × 3) surface reconstruction was observed for NiMnSb and was attributed to a transition from manganese/antimony terminated surface to a nickel terminated surface.²⁰ This suggests that as nickel content is increased within the alloyed series the surface termination may be transformed from antimony to nickel termination.

Figure 2(a) shows an XRD 2θ-ω scan for 24 nm thick Co_1-xNi_xTiSb films grown on InAlAs/InP(001) for x = 0.025, x = 0.1, and x = 0.5. Also included is a 16 nm thick x = 0.88 film, with the Sb cap desorbed and replaced with a protective AlO_x cap on top. The sharp peaks at ω = 30.44° and 63.34° correspond to the InP (002) and (004) substrate reflections, respectively, and the Co_1-xNi_xTiSb and InAlAs (002) and (004) peaks are nearly over-laid on the InP peaks indicating the close lattice match. Other than the (00l) peaks and thickness fringes, no additional peaks in the XRD scans are observed. Figure 2(b) shows a scan centered around the (004) reflection. Here, finite thickness fringes can be clearly resolved corresponding to a thickness of 23.3, 24.8, and 23.6 nm for x = 0.025, x = 0.1, and x = 0.5, respectively, in good agreement with the Co_1-xNi_xTiSb film thickness expected from RBS calibrations. These fringes indicate smooth, abrupt interfaces between the Co_1-xNi_xTiSb films and the InAlAs layers. The additional large peak observed in each scan corresponds to the InAlAs buffer layer. Small compositional deviations from the lattice matched In_0.52Al_0.48As led to the small variations in the buffer layer lattice parameter. These XRD patterns combined with the RHEED images indicate an epitaxial cube-on-cube growth with no detectable secondary phases or orientations.

The additional electrons introduced by nickel alloying are expected to strongly influence the electronic properties. Figure 3(a) shows the temperature dependent resistivity for the Co_1-xNi_xTiSb films. For the undoped and lowest nickel doping (x ≤ 0.025), the films exhibit Mott variable range hopping conduction where linear regimes in the log scale of ρ vs 1/T^1/4 data can be observed [Fig. 3(b)].^21,22 A linear region is not observed in the transport data for samples with x ≥ 0.05 suggesting that Mott variable hopping is not the dominant transport mechanism in the more highly doped samples.²³ As the nickel alloying increases, the magnitude and temperature dependence of the resistivity drops until the nickel composition is greater than x = 0.25, when metallic transport is observed. The hopping conduction observed in the x = 0.0 and x = 0.025 samples is likely mediated by localized in-gap states due to disorder, antisite defects, and/or other point defects such as off-stoichiometry defects which are common to Heusler compounds.

The evolution of the thermoelectric properties of the alloyed films is shown in Fig. 3(b). Here, a negative Seebeck coefficient can be observed at all nickel concentrations, consistent with the expectation of nickel being an electron donor. Furthermore, the magnitude of the Seebeck coefficient is largest for small amounts of nickel doping, consistent with previous experimental results for bulk crystals of nickel alloyed CoTiSb measured at 80 K.¹⁴ This large Seebeck coefficient for low nickel doping can be attributed to the large density of states at the Fermi level as it crosses into the conduction band. The small, slightly positive Seebeck observed in the pure CoTiSb film suggests the Fermi level position is close to mid gap with heavy compensation of holes and electrons. While this magnitude is lower than previously reported values for pure CoTiSb (−35 up to −265 μV/K),^24–26 it has been shown to depend strongly on synthesis procedures and post fabrication heat treatments. For CoTiSb films grown using higher growth temperatures, clear n-type conduction in Hall measurements¹⁰ and negative Seebeck values between −100 and −200 μV/K are observed. The lower growth temperature was used for the Co_1-xNi_xTiSb films to reduce the reactions of nickel with the underlying III–V buffer layer and may have resulted in inducing more mid-gap defect states that move the Fermi level towards the center of the gap. In the present study, the lower growth temperature was used for CoTiSb for consistency across the entire nickel doping series.

For a better understanding of the evolution of the electronic structure, ARPES measurements were performed on the substitutional alloyed series on the (001) surface. To be able to transport samples ex situ for ARPES measurements, a thin (∼100 nm) antimony capping layer was used to protect the film surface. Upon reintroduction into UHV, this capping layer was thermally desorbed. Using LEED, a (3 × 1) reconstruction was observed upon initial antimony desorption [Fig. 4(a)] for a sample temperature of 350 °C, as measured by a pyrometer. This reconstruction was not previously observed in CoTiSb,¹³ but was confirmed to be an antimony rich reconstruction by ultraviolet photoemission spectroscopy (UPS) core levels. Upon further annealing, the observed reconstruction transformed to a mixed (2 × 1) and c(2 × 4) [Figs. 4(b) and 4(c)]. Both the (2 × 1) and c(2 × 4) surface reconstructions have been suggested to be antimony terminated but with less antimony coverage than the (1 × 4) surface reconstruction.¹³ To examine the surface following thermal desorption of the antimony cap, STM was performed on a decapped Co_0.25Ni_0.75TiSb sample [Fig. 4(d)]. A well-ordered surface with large terraces could be observed. Additionally, both the c(2 × 4), as seen in LEED, as well as a (2 × 2) surface reconstruction, could be resolved. The recovery of the atomically smooth reconstructed surface upon antimony desorption suggests that the capping layer provides the intended protection of the sample surface but is fully removable upon reintroduction into UHV.

In-plane spectral maps were collected using a photon energy of 104 eV, corresponding to the bulk X point. The location of the bulk X-point was determined by calculating the inner-potential from observed periodicity in the electronic structure in photon energy dependent scans. The measured E-k spectra for x = 0.0 and x = 0.5 films are plotted in Figs. 5(a) and 5(b), respectively. For pure CoTiSb, the band structure is in good agreement with previous experimental reports and those calculated by DFT.¹³ Despite measuring close to the bulk X point, the hole-like valence band maxima at bulk gamma is visible due to surface Umklapp scattering from adjacent surface Brillouin zones.¹³ Notably, no evidence of the bulk conduction band can be seen consistent with the Fermi level being within the gap. As ARPES can only access the occupied states, we are unable to estimate the band gap of CoTiSb thin films from our measurements. However, a lower bound of the bandgap energy of 0.5 eV can be inferred from the observed binding energy of the valence band maxima. For the x = 0.5 film, most features remain unchanged from the pure CoTiSb sample with one notable departure, namely, a clear electron pocket can be seen at the bulk X point (k_ǁ = 0) indicating the Fermi level has crossed into the conduction band, consistent with its metallic behavior seen in electrical measurements. In addition, the bandgap energy has shrunk from >0.5 eV in CoTiSb to ∼0.4 eV for x = 0.5. This reduction in bandgap with nickel alloying is consistent with expectations from DFT, where NiTiSb has a reduced energy gap between the conduction and valence bands compared to CoTiSb.¹⁵ 

To examine at what composition the Fermi level crosses into the conduction band, energy distribution curves (EDCs) averaged over a small momentum region between −0.5 Π/a and 0.5 Π/a for x = 0.0, 0.1, 0.25, 0.5, 0.75, and 0.88 films are shown in Fig. 5(c). For the x = 0.0 and 0.1 samples, only very weak intensity due to inelastic scattering can be observed at the Fermi level. In contrast, for x ≥ 0.25 films, a quasi-particle peak, corresponding to the electron pocket, can be observed near the Fermi level which strengthens with increasing nickel concentration indicating that the Fermi level crosses into the conduction band between x = 0.1 and x = 0.25. This is consistent with the electrical resistivity data where a semiconductor to metal transition is observed for x > 0.1.

The bright streaky RHEED patterns, thickness fringes observed in XRD, and atomically smooth STM images suggest that high quality epitaxial thin films of Co_1-xNi_xTiSb with smooth surfaces and abrupt interfaces can be grown by MBE for all compositions investigated. Temperature dependent resistivity combined with Seebeck measurements show a semiconductor to metal transition for x ≥ 0.1, which is consistent with the appearance of the conduction band at the bulk X point in ARPES for the x = 0.25 sample. The origin of this slow transition may be due to localized in-gap states near the conduction band edge due to disorder, antisite defects, or other off-stoichiometry defects hybridizing with the conduction band.²⁷ This creates a band tail of localized states where conduction initially takes place for pure CoTiSb. As nickel is added to the film, the Fermi level slowly moves towards and then into the conduction band. Once the Fermi level passes through the mobility edge, metallic transport can begin to be observed. This is consistent with the Anderson-like transition proposed for bulk crystals of Co_1-xNi_xTiSb.¹⁴ While no evidence of an impurity band was observed in ARPES, an impurity band would have a finite but low spectral weight and relatively low dispersion making the existence of such a band hard to rule out. Interestingly, the gradual transition from semiconducting behavior to metallic transport indicated by the temperature dependent resistivity has also been reported in CoTi_1-xFe_xSb films, where the thermally activated behavior was observed for x ≤ 0.5²⁸ suggesting that a similar Anderson-like transition may take place in the iron doped films. In addition, similar substitutional alloying would likely be an effective method to adjust the Fermi level position in other half-Heusler systems. A topologically protected surface state was reported in MBE prepared thin films of PtLuSb, though the Dirac crossing point was found to lie well above the Fermi level.⁵ Substitutional alloying, for example, with gold for platinum in PtLuSb, may be an effective strategy to enable observation of the Dirac point using ARPES and produce devices with the Fermi level near the Dirac point. Our present study paves the way to understand the evolution of the electronic structure in alloyed Heusler compounds.

We thank J. Kawasaki and B. Shojaei for help with the growth of InAlAs and D. Read for discussions about the transport data. The growth and transport measurements were supported by the Office of Naval Research through the Vannevar Bush Faculty Fellowship under Award No. N00014-15-1-2845. The ARPES measurements at MAX-Lab and the STM studies were supported by the U.S. Department of Energy under Award No. DE-SC0014388. We also acknowledge the use of facilities within the National Science Foundation Materials Research Science and Engineering Center (DMR 11–21053) at the University of California at Santa Barbara, as well as the LeRoy Eyring Center for Solid State Science at Arizona State University. S.D.H. was supported in part by the NSF Graduate Research Fellowship under Grant No. 1144085.