*3.2. XRD Analysis*

Successful production of final devices required the growth of high-quality heterostructure layers with smooth interfaces. To verify the structural quality of our layers, HRXRD measurements were performed on the HEMT heterostructure (including the AlAs sacrificial layer) grown on GaAs substrate. To assess the effect of release and transfer procedures on the quality of the heterostructure, the HRXRD measurement was repeated after sticking the structure to a host sapphire substrate. The X-ray analysis was performed in high-resolution mode with a Bruker D8 DISCOVER diffractometer(Bruker AXS Advanced X-ray Solutions GmbH, Östliche Rheinbrückenstraße 49, 76187 Karlsruhe, Germany) equipped with a rotating Cu anode that was operated at 12 kW (40 kV/300 mA). A parabolic Goebel mirror and a Bartels monochromator were inserted into the primary beam. Standard angular 2*θ*/*ω* scans were recorded to determine the composition and thickness of the layers. Linear scans in reciprocal space were performed to evaluate the degree of relaxation of the layers.

The structures were analyzed by measuring high-resolution 2*θ*/*ω* curves of the 004 diffraction. The thickness and composition of the layers were determined by simulation of the theoretical curves using LEPTOS 3.04 software (provided by Bruker Company). Results of the X-ray measurements are exemplified in Figure 5, which compares the measured and simulated 2*θ*/*ω* curves. The measurements show that the diffraction maxima corresponding to the InGaAs and AlAs layers are clearly distinguished, along with the GaAs substrate maximum. The thickness fringes are also well resolved, indicating a high quality of the analyzed heterostructure. To find out whether the layers of the heterostructure were relaxed, linear scans (not shown here) across the asymmetric 224 diffraction were performed in the perpendicular (*l* scans) and parallel (*h* scans) directions with respect to the sample surface, respectively. It was found that the *l* scan of the 224 diffraction perfectly coincided with the *l* scan of the corresponding symmetric 004 diffraction. The *h* scans across the 224 diffraction maxima were measured at the values of *l* coordinates corresponding to the InGaAs and AlAs layers. The maxima of both curves were found to be precisely at *h* = 2.000. These results clearly indicated that the layers did not undergo any relaxation, and the in-plane lattice parameters of the layers were therefore equal to the bulk value of the GaAs lattice parameter. Both InGaAs and AlAs layers were laterally strained, and the estimated values of the strain were ∼ −1.5% and ∼ −0.15%, respectively. These values were used as the input parameters in the simulation process of the theoretical 2*θ*/*ω* curves.

ℎ = 2.000

2/

2/ **Figure 5.** Comparison of the measured and simulated 2*θ*/*ω* curves.

2/

2/

2/

~ − 1.5% ~ − 0.15%

2/ The 2*θ*/*ω* curve of the 004 diffraction of the heterostructure fixed to the host sapphire substrate is shown in Figure 6. Only the maxima of the GaAs buffer layer and of the ~10 nm thick InGaAs layer were visible, the AlAs separation layer was removed at the releasing step, and the corresponding maximum was missing. The most pronounced feature of the diffraction curve is the overall decrease in intensity by about two orders of magnitude with respect to the initial heterostructure on GaAs substrate (Figure 6). This can be partially ascribed to the size of the sticked heterostructure. The dimension of the measured sample in this experiment was only ~1 mm<sup>2</sup> . The second effect influencing intensity is the possible loss of the planarity of the heterostructure that cannot be avoided during fine manipulation of the sample. The broadening of the GaAs diffraction maximum and disappearance of the thickness fringes seem to support this reasoning. Generally, X-ray diffraction is extremely sensitive to any distortion of the diffracting area. Hence, a slight bowing of the sample, which has no effect on the functionality of an electronic device, can seriously disturb the interference of X-rays and can result in the observed changes in the diffraction curve.

**Figure 6.** X-ray diffraction of the as-grown AlGaAs/InGaAs/GaAs heterostructure (with the AlAs layer) and of the same heterostructure after being released and fixed to sapphire.

The X-ray measurements revealed that the heterostructures prepared on GaAs substrates were of very good quality with smooth interfaces. The release and transfer procedures did not damage the crystalline structure of the heterostructures and the main features of their diffraction curves were preserved.

#### *3.3. PL Analysis*

0

2

4

PL intensity (a. u.)

6

8

The quality of the release and transfer of the heterostructure were also investigated using PL measurement at room temperature. A 488 nm line of argon ion laser served for the sample pumping. PL radiation from the sample was filtered via a quarter-meter monochromator (Monochromator DIGIKROM 240, CVI Laser Corporation, Albuquerque, NM, USA.) and detected by a liquid-nitrogen-cooled InGaAs photodiode. The detector signal was amplified and recorded by a standard lock-in technique. At first, we measured the PL signal of the as-grown heterostructure on the parent substrate. The measurement

1.258 eV

1.05 1.10 1.15 1.20 1.25 1.30 1.35

GaAs/InGaAs transferred

Photoluminescence = 296 K 1.239 eV

GaAs/InGaAs as grown

to sapphire

Energy (eV)

was subsequently repeated when the heterostructure was released and transferred to sapphire. Figure 7 compares both PL signals.

63 64 65 66 67

2 / [°]

 as grown transferred to sapphire

Intensity [cps]

**Figure 7.** PL spectra of the as-grown heterostructure and released heterostructure nanomembrane.

The figure shows that the PL peaks of the as-grown heterostructure shifted after the heterostructure nanomembrane was released and fixed to the sapphire substrate. Each PL spectrum exhibits two resonances that consist of an asymmetric broadened feature at the lower-energy side and a distinct feature at the higher-energy side. According to [24], the broadened feature at lower energies, designated as 11H, originates in optical recombination from the first conduction subband to the first heavy-hole valence subband. The distinctive feature at higher energies, labeled as 21H, results from the recombination of electrons in the second electron subband and optically excited holes. We observed a shift of 19 meV between the measured peaks. The energy shift can be explained with a re-distribution of charge in the nanomembrane. A part of the charge occupies deep levels at the bottom surface of the nanomembrane. This surface was originally the interface between the bottom GaAs layer and the sacrificial AlAs layer; it was exposed as the nanomembrane was released from the growth substrate. This leads to a change in the electric field intensity in the quantum well of the nanomembrane; hence, the PL transition energy is also changed. As the quantum well (QW) lost some of its charge, the conduction and valence band energies of InGaAs were decreased near the bottom InGaAs/GaAs interface. A lower electron concentration in the QW of the nanomembrane corresponds with the electrical characteristics of the transistors described below.

#### *3.4. HEMT Processing*

The HEMTs were processed by standard photolithographic techniques using AZ 5214E positive tone photoresist. At first, MESA etching was performed in an Oxford PlasmaLab apparatus (OxfordInstruments GmbH, Borsigstrasse15a, Wiesbaden, D 65205 Germany) using SiCl4. The etching was stopped on the sapphire substrate. Ohmic contact metallic layers based on Ni/90 nm AuGe /Ni were deposited through a photoresist mask. The NiGe layer was evaporated to achieve a eutectic alloy of 88% Ni and 12% Ge. The layers were lifted off and annealed at 450 ◦C in N2 for 30 s. The ohmic contacts exhibited a contact resistance of 0.3 Ωmm. The gate electrode (prepared using similar lithographic steps) was composed of non-alloyed 15nmTi/30nmPt/50nmAu layers.

HEMTs were simultaneously processed on the growth GaAs substrate and host sapphire substrate for comparison. Basic dc transistor properties were measured. The output characteristics of both types of HEMT suggest that no degradation in the HEMT properties occurred. The input two-terminal measurement showed that the HEMTs processed on the heterostructure nanomembrane transferred to sapphire exhibited slightly decreased leakage current. The results are in Figure 8.

Ω

**Figure 8.** (**a**) Output characteristics of transistors processed on an HEMT heterostructure nanomembrane transferred to sapphire compared with those of transistors prepared on the as-grown heterostructure on GaAs substrate. The inset shows an image of an HEMT on the host substrate. (**b**) Comparison of the gate leakage current in two terminal measurement characteristics of HEMTs on sapphire substrate with those of HEMTs on GaAs substrate.

#### **4. Conclusions**

A 100 nm 2DEG AlGaAs/InGaAs/GaAs HEMT heterostructure was grown by MOCVD on GaAs substrate, released by HF: H2O epitaxial lift-off through a 300 nm AlAs layer, transferred, and attached to sapphire by van der Waals forces.

The heterostructure contained a strained 10 nm 2DEG In0.23Ga0.77As channel with a sheet electron concentration of 3.4 <sup>×</sup> <sup>10</sup><sup>12</sup> cm−<sup>2</sup> and Hall mobility of 4590 cm2<sup>V</sup> −1 s −1 . This channel layer was grown close to the center of the heterostructure to suppress a significant bowing of the heterostructure nanomembrane during and after separation from the growth substrate.

The nanomembrane release rate was studied with respect to the thickness of AlAs (varied between 20 to 500 nm) and to the composition of HF: xH2O (for x varied between 2 to 40) at RT. Thin AlAs layers (<100 nm) did not facilitate the release of the nanomembrane as the etching of AlAs self-terminated.

The as-grown heterostructure and nanomembranes attached to sapphire were characterized by HRXRD, PL, and SEM. The InGaAs and AlAs layers were laterally strained: ∼ −1.5% and ∼ −0.15%. HRXRD showed that the as-grown heterostructure had very good quality and smooth interfaces, and the transferred nanomembrane had its crystalline structure and quality preserved. PL showed that the nanomembrane peak was shifted by 19 meV towards higher energies with respect to that of the as-grown heterostructure.

The 2DEG channel transport properties were measured using HEMTs processed on the as-grown heterostructure and on 1.5 <sup>×</sup> 1.5 cm<sup>2</sup> nanomembranes attached to sapphire. The nanomembrane HEMTs showed no degradation of the output characteristics and their input two-terminal measurement confirmed a slight decrease in leakage current.

This work demonstrated that the properties of the HEMTs were not adversely affected by the transfer of the HEMT heterostructure to sapphire. This is very promising for our follow-up work aimed at the hybrid integration of III-V- and III-N-based devices, in which transistors with excellent dc and high-frequency properties will be attached to host substrates [25].

**Author Contributions:** D.G.: conceptualization, original draft writing, investigation, methodology; E.D.: investigation, data curation; P.E.: investigation, data curation, writing, editing review; R.S.: investigation, data curation; M.B.: investigation, data curation; O.P.: investigation, data curation;

Š.H.: investigation, data curation; M.K.: investigation, data curation; R.K.: investigation, supervision. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Slovak Grant Agency APVV-15-0243, and by the VEGA Grant No, 2/0068/21.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.
