Two-dimensional electron gases in AlYN/GaN heterostructures grown by metal–organic chemical vapor deposition

GaN-based electronics contribute to a sustainable future by improving the efficiency of optoelectronic, radio-frequency, and high-power devices.¹ For high-electron-mobility transistors (HEMTs) useable for power electronics and high-frequency electronics, the two-dimensional electron gas (2DEG) that forms at the interface of two wurtzite nitride layers with different bandgaps and polarization is key.^2–5 A potential well forms inside the layer with lower bandgap and polarization, the so-called channel layer, and a built-in electric field across the layer with higher bandgap and polarization, the so-called barrier, leading to the accumulation of a high number of electrons in the potential well.⁶ These electrons are spatially separated from their donor atoms and two-dimensionally confined close to the heterointerface, allowing for very high electron mobilities. HEMTs make use of these 2DEGs for high switching speeds and signal amplification. The higher the sheet charge carrier density n_s in the 2DEG, the higher the current and output power of such a transistor. The n_s in classical AlGaN/GaN heterostructures ranges from high 10¹² to low 10¹³ cm⁻², depending on the Al concentration. Higher n_s can be achieved with barrier materials with higher bandgaps and spontaneous polarization, such as Ga₂O₃, which allows for n_s in the low 10¹⁴ cm⁻² range but requires growth in a separate reactor.^7,8 More easily integrable barrier materials that provide charge carrier densities in the 10¹³ cm⁻² range are AlN, or AlN alloyed with group 3 elements.⁹ High-performing AlN/GaN HEMTs are reported in the literature⁴ but suffer from the limitation of the barrier thickness (below 4 nm) due to a high a lattice parameter mismatch to GaN. Highly strained barriers, such as AlN, or strain-relaxation in barriers that exceed the critical thickness suffer from reduced device performance and reliability.^10,11 In contrast, the novel nitride semiconductor AlScN features both high spontaneous polarization, which allows for n_s in the 10¹³ cm⁻² range, and an a lattice parameter matching to GaN, which allows for the growth of unstrained, thick barriers.¹² This is expected to increase the device’s lifetime. AlScN-based HEMTs seem already to outperform standard nitrides,¹³ although device fabrication poses challenges that require regrown ohmic contacts.^14,15 AlScN is stable in the wurtzite phase up to a Sc concentration of only 30% before it transitions to the non-polar rocksalt structure.^16,17 Furthermore, Sc is a rare element that is difficult to extract from ores and purify, making it cost-intensive and not an ideal candidate for sustainable electronics. Alloying AlN with the other group 3 element Y results in AlYN, which could be a viable alternative: It is predicted to be stable in the wurtzite phase up to concentrations higher than 50% and benefit from enhanced piezoelectric characteristics.^18,19 Elemental Y is easier to handle chemically, earth abundant, and cost-effective being up to 100 times less expensive than Sc.²⁰ AlYN features high spontaneous polarization and an a lattice parameter matching to GaN. The standard reduction potential of Y is −2.37 V more negative than that of Sc (−2.08 V) making it more susceptible to oxidation.²¹ Žukauskaitė et al.¹⁸ first synthesized this material by sputter deposition in 2012 and demonstrated its outstanding piezoelectric properties, making it a candidate for RF-filter and micro-electro-mechanical systems (MEMS). In 2023, several research groups focused their attention again on AlYN.²² Solonenko et al.²³ fabricated a bulk acoustic wave resonator made of AlYN. Uehara et al.²⁴ evaluated the enhancement of piezoelectricity by alloying Y with GaN. Tran et al.²⁵ performed thermal conductivity measurements of AlYN and Huang et al.²⁶ reported on AlYN-based photodetectors.

Similar to AlScN, the growth of AlYN by metal–organic chemical vapor deposition (MOCVD) is challenging due to the low vapor pressure of available Y precursors. Only doping concentrations of Y in GaN layers were achieved using Cp₃Y and (n-BuCp)₃Y.²⁷ Thereby, the main technical limitation was the source and gas line heating temperature. At Fraunhofer IAF, we assembled a heated gas mixing system that allows for the MOCVD growth with precursors with extremely low vapor pressure. This paved the way for the growth of the first AlScN,²⁸ which features excellent structural quality and was most recently demonstrated to be ferroelectric,²⁹ and later on also AlYN using the liquid precursor (EtCp)₂(iPr-amd)Y.³⁰ However, the growth of AlYN/GaN heterostructure by MOCVD featuring the presence of a 2DEG could not be proved since the material quality was unsuitable. Wang et al.^31,32 demonstrated that AlYN grown by molecular beam epitaxy (MBE) is a ferroelectric material and that 2DEGs form in AlYN/GaN heterostructures. They achieved an n_s of 4.2 × 10¹³ cm⁻² with an Al_0.93Y_0.07N barrier. The sheet resistance R_sh was 148 Ω/sq and the electron mobility μ was 1000 cm²/(Vs) at room temperature. However, the demonstration of ferroelectricity and the presence of a 2DEG in an AlYN/GaN heterostructures was proven only on a single concentration of Y in AlYN, namely, 7%. It is interesting to understand the effect of Y incorporation on 2DEGs in Al_1−xY_xN/GaN heterostructures and evaluate whether high-quality AlYN layers with suitable characteristics (structural, compositional, morphological, and electrical) can be deposited by MOCVD, which has not yet been demonstrated or reported so far to the best of our knowledge. In this work, we demonstrate that it is possible to deposit AlYN layers with Y concentrations up to 15% under different growth conditions in our customized MOCVD setup, thanks to the implementation of the solid high-purity Y-precursor (MCp)₃Y. We investigate 2DEG properties of Al_1−xY_xN/GaN heterostructures with different Y concentrations and thicknesses by capacitance–voltage (C–V) and Hall measurements and report on good structural and morphological quality by high-resolution x-ray diffraction (HRXRD), x-ray reflectometry (XRR), time-of-flight secondary ion mass spectrometry (ToF-SIMS), and atomic force microscopy (AFM). While AlScN/GaN benefits from higher n_s, AlYN/GaN features higher μ.

For the growth of Al_1−xY_xN by metal–organic chemical vapor deposition, a close-coupled showerhead reactor equipped with the proprietary heating system developed for growth with precursors with an extremely low vapor pressure by Leone et al.²⁸ was used. The employed solid precursor for yttrium was (MCp)₃Y supplied by Dockweiler Chemicals,³³ which was operated at 110 °C bubbler temperature. The vapor pressure of (MCp)₃Y has not been determined yet. However, thanks to its solid nature, it was possible to purify it through repetitive cycles of sublimation, which is an advantage compared to liquid precursors whose purification is more challenging. Oxygen impurities are typical in Y precursors due to the high affinity of Y with O.²¹ Liquid precursors, such as (EtCp)₂(iPr-amd)Y, contain a lot of carbon atoms, leading overall to a higher amount of C that comes with the precursor molecule and easily incorporates into the epitaxial layers. Solid sources, such as (MCp)₃Y, typically have a low vapor pressure. Despite the relatively high bubbler temperature of 110 °C, the maximum achieved growth rate with (MCp)₃Y was 0.017 nm/s. These growth rates are similar to the ones achieved with (MCp)₂ScCl for AlScN.³⁴ Liquid precursors typically allow for growth rates that are approximately five times higher. Trimethylaluminum (TMAl), trimethylgallium (TMGa), and ammonia (NH₃) were used as Al, Ga, and N sources, respectively. The carrier gas was hydrogen or nitrogen. The Al_1−xY_xN/GaN heterostructures were grown on 100 mm Al₂O₃ (0001) or semi-insulating 4H–SiC (0001) substrates with a nucleation layer either consisting of a low-temperature GaN or high-temperature AlN, respectively. An iron-doped GaN lower buffer layer was employed to compensate for residual donor impurities.³⁵ A non-intentionally doped GaN buffer spatially separates the Fe acceptors from the 2DEG to avoid trapping of electrons. The upper part of this buffer serves as a channel and contains the 2DEG. An AlYN barrier, 10 nm thick if not otherwise indicated, was grown on top of the GaN under various growth conditions. This barrier layer was always preceded by a nominal AlN interlayer to improve barrier homogeneity³⁶ and passivated with in situ SiN_x to avoid surface oxidation. We tested growth temperatures from 900 to 1100 °C, varied the molar flow of the Y source (and thus the Y concentration of the barrier layer) by tuning the carrier gas flow to the (MCp)₃Y source from 500 to 1500 sccm (at a bubbler pressure below 200 mbar), and modulated the barrier thickness by changing the growth time.

The surface morphology was examined by atomic force microscope (AFM) in the tapping mode, and the surface roughness was determined as the root-mean-square (rms) average. All AFM measurements reported in this work were obtained on the as-grown epilayers with the SiN_x cap on top. The barrier and cap layer thicknesses were determined by high-resolution x-ray diffractometry (HRXRD) Θ/2Θ-scans of the 0002 and 0004 reflection ranges combined with x-ray reflectometry (XRR). With reciprocal space mapping (RSM), the strain state of the Al_1−xY_xN barrier relative to GaN was determined.

Chemical composition and depth profiles were obtained by time-of-flight secondary ion mass spectrometry (ToF-SIMS) in the dual-beam mode using a sputter beam of 2 keV Cs⁺ ions. A primary beam of 30 keV ¹Bi⁺ ions was employed for the crater analysis. For the determination of the Y concentration, magnetron-sputtered AlYN samples, whose Y concentration was determined with energy elastic recoil detection analysis (ERDA),¹⁸ were used as references by comparing the AlCs⁺/YCs⁺ signal ratios. Due to the lack of standards, oxygen and carbon levels could only be evaluated qualitatively, and the exact concentrations remain unknown.

The presence of a 2DEG was confirmed by capacitance–voltage (C–V) measurements on the as-grown layers with a Hg-probe from which the vertical charge carrier distribution was also calculated.⁶ The dielectric parameter of AlN (8.5³⁷) was used as the effective dielectric parameter ϵ_eff of the SiN_x and the Al_1−xY_xN barrier in these calculations. This allows for fitting the 2DEG peak position in the GaN channel, close to the heterointerface. An in-depth study of the dielectric properties of Al_1−xY_xN will be the subject of future investigation. The same setup was used for current–voltage (I–V) measurements to detect the leakage currents of a mercury Schottky contact to the as-grown heterostructure. Further electrical characterization is performed by eddy-current sheet resistance and contactless Hall measurements with a contactless mobility mapper. The average value of the sheet resistance R_sh, sheet charge carrier density n_s, and electron mobility μ for each 4-in. wafer was determined from 17 evenly distributed measurement points. Alloyed Ti/Al/Au-based contacts were deposited for Hall measurements at 77 K under a liquid nitrogen atmosphere and for temperature-dependent Hall measurements from 7 K to room temperature. The latter were conducted in a Hall measurement setup with a liquid Helium cooling system.

The theoretical n_s of Al_1−xY_xN/GaN heterostructures with different Y concentrations and thicknesses was calculated using the 1D Schrödinger–Poisson equation solver nextnano.³⁸ The material parameters were approximated by fits to literature values as described later on.

The 2DEG properties of SiN_x-passivated Al_1−xY_xN/GaN heterostructures grown on Al₂O₃ substrates are explored by testing: the effect of growth temperature on Y incorporation and electrical characteristics (Sec. I), different Y concentrations (Sec. II), and different thicknesses of the AlYN barrier (Sec. III). Electron transport properties are examined further for the highest performing sample (Sec. IV). Subsequently, Al_1−xY_xN/GaN heterostructures are grown on the better lattice-matched substrates 4H–SiC, allowing for higher crystal quality and enhanced 2DEG properties (Sec. III E).

A carrier gas flow of 500 sccm to the (MCp)₃Y source resulted in approximately constant Y incorporation of 3%–4% at growth temperatures of 900, 1000, and 1100 °C, as listed in Table I. At both 1000 and 1100 °C barrier growth temperatures, the amount of Y incorporated was increased by changing the carrier gas flow to the (MCp)₃Y source from 500 to 1500 sccm at a constant TMAl flow of 4.9 μ mol min⁻¹ and a bubbler pressure below 200 mbar. In our previous study with the precursor (EtCp)₂(iPr-amd)Y, we observed a proportional effect of the growth temperature on the incorporation of Y in AlN layers.³⁰ With (MCp)₃Y, we observe low Y incorporation at lower flows and a strong increase of Y incorporation at the highest possible flow of 1500 sccm at 1000 °C growth temperature. In contrast, a gradual increase and seemingly a saturation of the Y concentration at 8% occurs at 1100 °C. This is illustrated in Fig. 1. While the saturating Y incorporation at higher growth temperatures may be caused by gas-phase parasitic reactions, the reason for the lower than expected Y incorporation at 1000 °C remains unclear.

The ToF-SIMS composition profile of the Al_1−xY_xN/GaN heterostructure grown at 1100 °C with a Y concentration of 8% (1500 sccm) is shown in Fig. 2. The interfaces between the barrier and the SiN_x cap are highly abrupt, while the interdiffusion of atoms between the barrier and the GaN channel is visible. This is a characteristic of MOCVD-grown heterostructures. Here, especially, the back diffusion of Al and Y into the channel is prominent and degrades the nominal AlN interlayer leaving a quaternary interface region.

Interestingly, the electrical characteristics of the Al_1−xY_xN/GaN heterostructures grown at 900, 1000, and 1100 °C with a (MCp)₃Y flow of 500 sccm differed even though the Y incorporation and barrier thicknesses were essentially the same, 3%–4% and ∼10 nm, respectively, as given in Table I. The sheet carrier density n_s determined by Hall measurement increases from 1.64 to 1.77 and finally 2.15 × 10¹³ cm⁻² with increasing growth temperature. A sheet resistance R_sh is 283 Ω/sq lowest and mobility μ with 1246 cm²/(Vs) highest at a growth temperature of 1000 °C. All three heterostructures displayed excellent C–V and I–V characteristics, independent of the applied growth temperature. The threshold voltage shifts to higher negative values for higher n_s as expected. ToF-SIMS depth profiles on all three samples (not shown) show the almost identical Y incorporation and no increased thermal degradation with temperature, probably thanks to the insertion of a nominal AlN interlayer.³⁶ Impurity incorporation is usually facilitated by low growth temperatures and can be reduced by increasing the growth temperatures. Since the precursor molecule introduces a considerable amount of carbon (C/Y = 18) and yttrium reacts quickly with oxygen due to its low standard redox potential, growth at higher temperatures is crucial to increase the purity of the Al_1−xY_xN layers. Reduced incorporation of carbon acceptors that trap electrons that would otherwise be available for the 2DEG increases the n_s of the 2DEG. Cubic inclusions were not observed in the investigated concentration range.

The surface potential was approximated as one-third of the bandgap.⁹ The result is shown in Fig. 4(a). The n_s decreases with increasing Y concentration, as expected. Experimental data obtained by C–V and Hall measurements show the same trend. The n_s decreases from ∼2.8 × 10¹³ to ∼1.6 × 10¹³ cm⁻² from 3.3% to 8.2%. The amount of charge carriers is about half of what was expected. This discrepancy may indicate that the surface potential in the calculation, assumed as 1/3 of the bandgap, was too low. Also, incorporated carbon atoms, expected to be present in elevated quantities, may act as acceptors and trap electrons that are no longer available for the 2DEG. The fact that the n_s determined by Hall measurements remains the same at room temperature (298 K) and under a liquid nitrogen atmosphere (77 K) proves the presence of a 2DEG. The sheet charge carriers are spatially separated from their donor atoms and immune to freeze-out effects. The fact that no charge carriers freeze out indicates that no impurities that act as donors, such as O or Si, are active. The temperature reduction reduces the impact of temperature-dependent intrinsic scattering mechanisms, such as polar optical phonon (POP) scattering and acoustic deformation (AD) scattering, and leads to a mobility increase. This was observed for all examined Y concentrations up to 8.2%. While Al Mustafa⁴⁶ predicted lattice matching for a Y concentration of 16%, the heterostructure with this concentration was highly resistive and did not feature a 2DEG. The experimental data suggest that this barrier suffers from relaxation and that the lattice-matching occurs at lower concentrations. Interestingly, the mobility measured at room temperature (298 K) and then in liquid nitrogen (77 K) increased only by a factor of 1.8 from 707 to 1310 cm²/(Vs) at a Y concentration of 3.3%, and by a factor of 2.8 from 927 to 2660 cm²/(Vs) at a Y concentration of 8.2%, as shown in Fig. 4(b). It seems that alloy scattering, which should increase with increasing Y incorporation, does not degrade the mobility. It is unlikely that the nominal AlN interlayer (∼0.5 nm thick) present in these structures is a binary layer that suppresses alloy scattering; moreover, interface diffusion is considerably high at the given growth temperatures. However, such interlayers were shown to positively impact the interface abruptness in AlScN layers grown with very low growth rates.³⁶ It is possible that the high μ relates to the decreased n_s and reduced electron-electron scattering. The C–V curves in Fig. 5(a) show an abrupt capacitance drop across three orders of magnitude at the threshold voltage V_th, which shifts from −14.05 V at 3.3% to −7.55 V at 8.2% with increasing Y concentration. This can be attributed to the decreasing n_s. The capacitances of all heterostructures are constant above the V_th, and no parasitic capacitances are present. This indicates that the barrier and cap layer form a donor-free dielectric. The charge carrier profiles derived from C–V measurement in Fig. 5(b) show well-confined 2DEGs close to the barrier/channel interface. Leakage currents in the I–V measurement of a mercury Schottky contact to the cap layer were low in the 10⁻⁸ A range for all Y concentrations (not shown).

One feature that makes Al_1−xY_xN/GaN heterostructures interesting for HEMT devices is the a lattice parameter matching with GaN. Wang et al.^31,32 observed similar lattice parameters of Al_1−xY_xN and GaN on 60 nm-thick layers with a Y concentration of 7% by MBE. We expect lattice matching in a Y concentration range of 7%–9%. Since no misfit strain exists in such a structure, the barrier thickness should be tunable without constraints. Y atoms with their atomic radius of 1.80 Å are considerably bigger than Sc atoms (1.50 Å),⁴⁷ and the a lattice parameter increase is more pronounced with increasing Y concentration than with increasing Sc concentration, introducing a considerable amount of strain to the lattice. We expect that already small deviations from the lattice-matched concentration significantly reduce the critical thickness up to which a fully strained Al_1−xY_xN layer can be grown on GaN, generating a very restricted critical thickness window similar to or even more restricted than that of Al_1−xSc_xN.¹² Hence, a series of Al_0.92Y_0.08N/GaN heterostructures with barrier thicknesses of 5.5, 9.9, and 14.7 nm was grown and examined for properties that could confirm this as the strain-free concentration. The barrier thickness was varied by tuning the growth time and determined subsequently by combining HRXRD and XRR. The RSMs in Fig. 6 show that all three barriers are strained to the GaN buffer. The Al_0.92Y_0.08N peak becomes more pronounced with increasing barrier thickness. No relaxation effects are observed for 5.5 and 9.9 nm-thick barriers. The results of the electrical characterization by C–V and Hall measurement are listed in Table II. The presence of a 2DEG was confirmed by the C–V measurement for all three samples. No parasitic capacitances were observed. The n_s,C–_V increases with increasing barrier thickness and leads to a simultaneous shift of the threshold voltage to higher negative voltages, from −7.35 V at a barrier thickness of 5.5 nm and n_s of 1.80 × 10¹³ cm⁻², to −13.1 V at 14.7 nm and an n_s of 2.28 × 10¹³ cm⁻². As shown in Fig. 7, this trend matches simulated values, even though the absolute values of n_s fall short of the calculated ones. The n_s,CV and n_s determined by the Hall measurement of the 5.5 and 9.9 nm-thick barriers are within reasonable agreement, as shown in Fig. 7. The fact that n_s is insensitive to the temperature change highlights that the impurity level in the samples is low. Contrary to the trend observed in C–V measurements, n_s determined by Hall measurements at room temperature of the heterostructure with the 14.7 nm-thick barrier decreases to 1.17 × 10¹³ cm⁻², and consequently, the μ increases to 992 cm²/(Vs) at room temperature. It was not possible to measure this sample at 77 K, probably because of low-quality ohmic contacts. In the I–V measurements of a mercury Schottky contact to this thickness series shown in Fig. 8, we observed an increase in leakage current by two orders of magnitude from 10⁻⁹ to 10⁻⁷ A range for the 14.7 nm-thick barrier. This suggests that the 14.7 nm-thick barrier suffers from some degradation effects. The surface roughness determined by AFM is 1.20 nm, approximately four times higher than that of the heterostructures with thinner barriers listed in Table II. Possibly, 3D structures add current paths that degrade the electrical performance.

For all investigated samples, but especially for the one with high barrier thickness, the formation of an ohmic contact to the 2DEG with the contact stack optimized for AlGaN with low Al fractions was complex. To date, this field is entirely unexplored for Al_1−xY_xN. A conclusive statement on lattice-matching to GaN requires investigation of higher layer thicknesses and multilayer structures. Given that the presence of a 2DEG was observed by C–V, we assume no major relaxation occurs. Surface roughness and ohmic contact formation seem especially critical at high barrier thicknesses.

The high-performing heterostructure with 8.2% Y was examined further with temperature-dependent Hall measurements. Figure 9 illustrates that the μ increases from 1079 cm²/(Vs) at 289 K to 3234 cm²/(Vs) at 7 K, which surpasses the best value of mobility obtained so far in MOCVD-grown AlScN/GaN heterostructures^48,49 and in MBE-grown Al_1−xY_xN/GaN heterostructures.³² The n_s is constant across the whole temperature range, proving that no undesired donors are present. Scattering mechanisms were calculated with the formulas proposed by Jena.^50,51 Since the 2DEG resides in the GaN channel, most parameters required for these calculations could be taken from the mentioned references. Alloy scattering is not considered here, as the reduction in n_s with increasing Y concentration reduces the penetration of the wave function into the barrier layer and reduces the 2DEG’s sensitivity to alloy scattering. The rms of 0.33 nm (in a measurement area of 10 μm²) determined by AFM suggests that interface roughness scattering IFR caused by electric field fluctuations induced by barrier thickness variations is low. However, the low growth rate of 0.017 nm/s likely promotes diffusion of atoms at the interface and leads to increased IFR in that way. According to our calculations, this is currently the limiting factor for the achievable μ. The generally high impurity levels in group 3 nitrides grown by MOCVD with development grade precursors³⁰ cause background residual impurities scattering contribution. Dipole and dislocation scattering should be negligible since the n_s is higher than 10¹³ cm⁻². The contributions of the individual scattering mechanisms add up to the total mobility, according to the Matthiesen rule.

RF devices are usually grown on semi-insulating 4H–SiC substrates, which have higher thermal conductivity and a lower dislocation density, thanks to a smaller lattice mismatch to GaN than Al₂O₃ or Si. We performed several growths of Al_1−xY_xN/GaN heterostructures on 4-in. SiC substrates, resulting in homogeneous heterostructures with good structural quality and morphology. Thickness and ex situ wafer bow measurements pinpointed excellent thickness uniformities with a standard deviation below 4% and wafer bow below 15 μm on the whole 4-in. wafer. Optical microscopy highlighted excellent morphologies free of epitaxial-related defects. The μ map of a Al_1−xY_xN/GaN heterostructure with a Y concentration of ∼8% having μ close to 1400 cm²/(Vs) is shown in Fig. 10. The μ is homogeneous across the wafer with a standard deviation of 3% and sheet resistances of 335 Ω/sq with a standard deviation below 1%. AFM analysis (Fig. 11) confirms that the surface is extremely smooth with an RMS of 0.28 nm on a 10 μm² scan area. HRXRD measurements highlighted good structural characteristics, whereas the full-width-half-maximum (FWHM) of the GaN crystal planes (00.2) and (10.2) were below 200 arcsec, and clear thickness fringes were observable in most of the samples, hinting that the as-grown MOCVD material is suitable for the fabrication of HEMTs, which is going to be started soon and will be reported in the future works.

Based on the presented results, AlYN brings the same key advantage as its counterpart AlScN, namely, the growth of strain-free heterostructures with high n_s, but exhibits different electrical characteristics due to the achievable n_s being lower than that of AlScN. The lower n_s of AlYN/GaN heterostructures with respect to AlScN/GaN leads to a less negative V_th. For a power amplifier utilized in advanced digital communication systems, such as 5 or 6 G, opting for high-energy mobility transistors (HEMT) can yield several advantages, such as operational speed, which is crucial for high-frequency applications, improves energy efficiency, and provides flexibility for achieving broader bandwidths. In addition, due to its lower n_s, AlYN boosts the electron mobility significantly compared to MOCVD-grown AlScN with the same concentration, as plotted in Fig. 12. The MBE-grown Al_1−xY_xN/GaN heterostructure presented by Wang et al.³¹ features an almost equally high electron mobility at approximately three times higher n_s, indicating that increased interface abruptness and reduced impurity incorporation, which are the characteristic advantages of MBE-over MOCVD-grown heterostructures, reflect positively on both n_s and μ. The fact that the low-temperature electron mobility of the 2DEG in the GaN channel is much higher for MOCVD-grown AlYN barriers compared to AlScN barriers suggests that fewer extrinsic scattering mechanisms are at play. A reduction in scattering mechanisms that hamper electron mobility is expected to affect the achievable drift velocity in HEMTs positively. In scenarios where priorities lean toward higher electron mobility over high carrier density, such as in high-frequency communication systems, AlYN emerges as an advantageous choice.

This work demonstrates that Al_1−xY_xN/GaN heterostructures with precisely tunable Y concentration, good structural quality, and electrical characteristics can be grown by MOCVD. Y concentrations up to 16% were achieved with the precursor (MCp)₃Y. Optimum 2DEG properties were achieved in Al_1−xY_xN/GaN heterostructures deposited on a sapphire substrate at a Y concentration of 8.2% with room temperature R_sh of 444 Ω/sq, n_s of 1.30 × 10¹³ cm⁻², and μ of 1079 cm²/(Vs). The low-temperature μ of 3234 cm²/(Vs) at 7 K is the highest reported for Al_1−xY_xN/GaN heterostructure so far. The n_s insensitivity to temperature indicates that no undesired donor electrons are present. Similar Al_1−xY_xN/GaN heterostructures deposited on a S.I. 4H–SiC substrate had μ close to 1400 cm²/(Vs), n_s of 1.4 × 10¹³ cm⁻², and R_sh of 335 Ω/sq or lower depending on barrier thickness and alloy composition. These results showcase that AlYN is a promising candidate for the next generation of HEMTs for enhanced digital communication systems.

Dockweiler Chemicals GmbH supplied the Y precursor for this study. Laytec supported the in situ measurements in the MOCVD process, while the manufacturer of the MOCVD system Aixtron supported the upgrade of the system. The authors thank Nadine Brückner, Hanspeter Menner, Christian Manz, Timileyin Olanipekun, Barbara Weber, and Frank Bernhardt who supported this study with sample preparation and characterization. Furthermore, the authors thank Jannis Jansen for setting up the tool for temperature-dependent Hall measurements. Professor V. Darakchieva at Linköping University is acknowledged for having provided sputtered AlYN samples used for our SIMS calibration. Finally, we are most grateful to Tim Stadelmann and Rüdiger Quay for having supported such innovative research activities.

The authors have no conflicts to disclose.

Isabel Streicher: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Patrik Straňák: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Lutz Kirste: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Mario Prescher: Data curation (equal); Formal analysis (equal); Methodology (equal); Visualization (equal). Stefan Müller: Data curation (equal); Formal analysis (equal); Methodology (equal); Validation (equal). Stefano Leone: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Resources (lead); Supervision (lead); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.