Ultrathin (1–4 nm) films of wide-bandgap semiconductors are important to many applications in microelectronics, and the film properties can be sensitively affected by defects especially at the substrate/film interface. Motivated by this, an in vacuo atomic layer deposition (ALD) was developed for the synthesis of ultrathin films of Ga2O3/Al2O3 atomic layer stacks (ALSs) on Al electrodes. It is found that the Ga2O3/Al2O3 ALS can form an interface with the Al electrode with negligible interfacial defects under the optimal ALD condition whether the starting atomic layer is Ga2O3 or Al2O3. Such an interface is the key to achieving an optimal and tunable electronic structure and dielectric properties in Ga2O3/Al2O3 ALS ultrathin films. In situ scanning tunneling spectroscopy confirms that the electronic structure of Ga2O3/Al2O3 ALS can have tunable bandgaps (Eg) between ∼2.0 eV for 100% Ga2O3 and ∼3.4 eV for 100% Al2O3. With variable ratios of Ga:Al, the measured Eg exhibits significant non-linearity, agreeing with the density functional theory simulation, and tunable carrier concentration. Furthermore, the dielectric constant of ultrathin Ga2O3/Al2O3 ALS capacitors is tunable through the variation in the ratio of the constituent Ga2O3 and Al2O3 atomic layer numbers from 9.83 for 100% Ga2O3 to 8.28 for 100% Al2O3. The high ɛ leads to excellent effective oxide thickness ∼1.7–2.1 nm for the ultrathin Ga2O3/Al2O3 ALS, which is comparable to that of high-K dielectric materials.
I. INTRODUCTION
Wide-bandgap semiconductors have a vast range of applications in high-power electronics,1 deep ultraviolet optoelectronics,2 and high radio frequency electronics because of several attractive properties including chemical inertness, high critical electric field, and high carrier mobility.3 Among others, gallium oxide (Ga2O3) has emerged as a highly promising semiconductor due to its large bandgap (Eg ∼ 4.8 eV), controllable charge carrier doping,4 excellent thermal stability, and high breakdown field (∼8 MV cm−1).5–7 In particular, the monoclinic β-Ga2O3 phase is the most thermodynamically stable phase8 with applications as high-power devices such as metal–oxide–semiconductors,5,9 Schottky barrier diodes (SBDs),10 optoelectronics,11 and radiation detections (x ray, α and γ particles, and neutrons).12 Moreover, the abundance of oxygen vacancies in Ga2O3 films makes them suitable for use in resistive random-access memory (RRAM) devices,13,14 gas sensors,15 memristors,16,17 and others.18
Further miniaturization in microelectronics following the empirical Moore's law demands ultrathin films that are pinhole-free and defect-free with the desired doping. Therefore, it is imperative to develop a deposition technique for the growth of Ga2O3 ultrathin films with excellent quality. Motivated by this, many methods have been explored for the deposition of thin films of Ga2O3 including electron beam evaporation,19 magnetron sputter deposition,20 chemical vapor deposition (CVD),21 laser ablation,22 mist chemical vapor deposition (mist CVD),23 molecular beam epitaxy,24,25 and atomic layer deposition (ALD).22,23 Compared to other techniques, ALD has the advantages of atomic-precision thickness control attributed to its self-limiting growth mechanism, low defect density through well-defined chemical reactions under optimal ALD conditions, and large-area conformal coating over flat surfaces as well as surfaces of large aspect ratios for industrial applications.26,27 Thin films of Ga2O3 have been made before using ALD. Choi et al.28 deposited 40 nm thick Ga2O3 films with ALD and investigated microstructural, optical, and electrical properties. Comstock and Elam29 demonstrated the ALD growth of Ga2O3 thin films with a thickness of 10 nm on Si substrates with a growth rate of 0.52 Å/cycle. The smaller than expected rate may be attributed to the delayed ALD growth during the initial incubation on the substrates. Shih et al.30 studied the chemical composition and optical properties of 10 nm thick ALD-Ga2O3 films as gate dielectric deposited on SiC substrates. The bandgap of these Ga2O3 thin films was found to be 4.51 eV with an estimated dielectric constant of ∼9.2 at a thickness of 40 nm. This large bandgap and dielectric constant make Ga2O3 good candidates as gate dielectrics and passivation layers in semiconductor-based devices attributed to decreased leakage current. Despite the progress made, the fabrication of ultrathin Ga2O3 films with good dielectric properties remains challenging because of the influence of a defective interfacial layer (IL) at the film/substrate interface.31 For example, Paskaleva et al.32 reported a dielectric constant (ɛ) of ∼12.4 at 1 kHz on Ga2O3 films of thickness of 125 nm, which drops to 4.4 as the thickness is reduced to 15 nm. The strong dependence of ɛ on the film thickness is attributed to the presence of the defective IL on which the oxide films would be defective especially at small film thicknesses. Consequently, the dielectric properties of the Ga2O3 film would be degraded as the layer thickness approaches the ultrathin range of 1–4 nm. Therefore, it is crucial to develop a method that allows the deposition of Ga2O3 ultrathin films with negligible interfacial defects to get excellent electronic and dielectric properties.
Recently, Acharya et al.33 have reported the fabrication of ultrathin Al2O3 films on Al electrodes with and without a defective IL using ALD deposition. The presence of defective IL was found to lead to a defective ALD-Al2O3 ultrathin film and low ɛ ∼ 3–4. When this defective IL is removed, an ɛ of 8.9 that is close to the bulk value of crystalline Al2O3 (∼9.2) was achieved on ALD-Al2O3 films of 3–4 nm in thickness. In a related work, a high ɛ of 8.8–9.4 was achieved on ALD MgO films of thickness of 3–4 nm and attributed to a 0.5 nm thick Al2O3 seed layer on the Al electrode without a defective IL.34 These works illustrate the critical impact of the IL on the dielectric and electronic properties of ultrathin films. In order to eliminate the defective IL, an in vacuo ALD system was adopted in these studies to allow the electrode and oxide films to be deposited without breaking vacuum to avoid sample exposure to the atmosphere.26 Furthermore, dynamic pre-ALD heating was found critical to minimize the exposure of the sample surface to traces of O2 and H2O in high vacuum. In this work, we investigate the growth mechanism and physical properties of ultrathin (1–4 nm) films of Ga2O3/Al2O3 atomic layer stacks (ALSs) on the Al electrode using the in vacuo ALD. We show that the ultrathin films of Ga2O3/Al2O3 ALSs can be obtained under optimal ALD conditions, on which tunable dielectric properties, bandgaps, and carrier concentrations can be achieved by varying the number of the constituent Ga2O3 and Al2O3 atomic layers and their stacking.
II. METHODS
A. Density functional theory simulations
We use the density functional theory (DFT) with projector augmented wave (PAW) pseudopotentials35 as implemented in the Vienna Ab initio Simulation Package (VASP) code.36,37 To correctly describe bandgaps, we use the HSE06 hybrid functional.38 We consider the alloys of Ga2O3 and Al2O3 in the monoclinic (the ground state structure of Ga2O3), corundum (the ground state structure of Al2O3), and orthorhombic structures. For each alloy concentration, we use the lowest enthalpy structure within a certain crystal structure and calculate the direct and indirect bandgaps. The details of the obtained alloys and our approach are reported elsewhere.39,40
B. Ga2O3/Al2O3 ALS sample fabrication
Sample fabrication began with a Si/SiO2 (500 nm) substrate being transferred to the sputtering chamber from a load-lock chamber using a transfer rod where a 50 nm thick Al electrode was deposited. DC magnetron sputtering of the bottom Al electrode was carried out in our in vacuo ALD/sputtering system41 at a base pressure of < 5 × 10−7 Torr, using an Ar plasma 14 mTorr/90 W for Al deposited at 0.5 nm/s and a sample distance of ∼6 cm. To make the dielectric layer, the ALD chamber was preheated to a temperature of ∼225 °C, followed by in vacuo sample transfer into the ALD chamber, and dynamically heated for 25 min with blackbody radiation from the sides of the chamber. We used dynamic heating, meaning that the sample was not in full thermal equilibrium at the beginning of the ALD but that the temperature was in the optimal range (starting at ∼177 °C)27 and would continue to slowly rise toward 225 °C during further ALD growth. This dynamic heating was found critical to minimize metal–insulator IL formation at the Al interface caused by excessive heating times needed to reach proper thermal equilibrium.42,43 The sample was then subjected to alternating 2 s long precursor pulses of H2O and either trimethyl-aluminum (TMA) at room temperature or 0.25 s of tris(dimethylamido) gallium [Ga2(NMe2)6] heated to ∼85 °C. Between each precursor pulse, the ALD chamber was purged for 35 s using N2 and a vacuum pump. The precursor pulses were fully computer controlled to allow ALD-Ga2O3 and ALD-Al2O3 atomic layers to be stacked in any order or amount desired. Once the dielectric layer was grown, the sample was transferred in vacuo back to the high vacuum sputtering chamber to let it cool in a low oxygen environment. Post-cooling, the sample was then removed from vacuum and the second shadow mask was mounted and returned to the sputtering chamber, via the load-lock, for the deposition of the top 100 nm Al electrode.
C. Characterization of Ga2O3/Al2O3 ALS electronic structures
In vacuo scanning tunneling spectroscopy (STS) was taken using an RHK UHV system at ∼10−10 Torr. STS samples were prepared by fabricating a half-cell of the capacitor structure, i.e., the fabrication steps were all the same except the sample was transferred for examination after the dielectric layer was grown without the top electrode. For comparison of the IL defects, a thermal AlOx (Th-AlOx) dielectric was also fabricated for STS. To obtain the Th-AlO layer, The sample was transferred to the load-lock chamber to undergo oxidation via O2 gas at ∼2 Torr for 520 s resulting in ∼1040 Torr s, which is estimated to result in ∼1 nm defective Th-AlOx.42,44 Local density of states (LDOS) was proportionately measured through dI/dV spectra collected using a mechanically cleaved Pt–Ir tip by sweeping a DC voltage from 0 to 1.7–2.3 V with a lock-in amplifier analyzing the 45 mV/5 kHz AC signal on top of the DC signal. Approximately 60–80 spectra were randomly taken as the films were too sensitive for scanning tunneling microscopy scans. Using two bisquare fits, the intersection of the conduction band and bandgap regions are estimated to estimate the barrier height (Eb), as well as the ALD coverage rate.42,45
D. Characterization of Ga2O3/Al2O3 ALS capacitor dielectric properties
A set of ultrathin capacitors composed of ALD-Al2O3 and ALD-Ga2O3 with varying ratios were fabricated using in vacuo ALD. The thickness of the capacitor dielectric layer was in the range of ∼4.4 nm. Since each ALD-Ga2O3 or ALD-Al2O3 atomic layer has a thickness of ∼0.8–0.11 nm,46,47 the resolution of the thickness control is truly atomic. The capacitors were defined using a shadow mask to allow 12 devices to be fabricated on each chip with three different areas, 200 × 200, 200 × 300, and 200 × 400 μm2, for uniformity examination. Electrical measurements of the shadow-mask fabricated samples were carried out using 25 μm tungsten probes in a probe station in conjunction with an Agilent B1500A semiconductor analyzer. The top Al electrode was grounded, while the bottom Al electrode was biased for IV sweeps at low voltage (−500–500 mV). Capacitor vs frequency measurements were taken using the same setup at frequencies of 1, 10, and 100 kHz and 1 MHz in the voltage ranges from −100 to 100 mV. The Agilent B1500A Semiconductor Analyzer was used to perform impedance measurements on our capacitors at different frequencies up to 1 MHz. The conductance and capacitance can be extracted from the measured impedance using the equation Y = G + j2πfC, where Y is the inverse impedance, G is the real part conductance, and C is the imaginary part capacitance. The Drude equation was then used to calculate the carrier concentration (n) of the samples using σ = ηeμ, where σ is the conductivity, e is the electron charge, and μ is the electron mobility. The mobility of Ga2O3 (5.5 cm2/V s) and Al2O3 (2.47 cm2/V s) was obtained from the literature.48,49
III. RESULTS AND DISCUSSION
Figure 1(a) schematically illustrates a metal–insulator–metal device with bottom and top electrodes sandwiching an ultrathin Ga2O3/Al2O3 ALS film made with ALD. Since the oxide nucleation occurs on the surface of the bottom electrode, the IL of ALS/metal (bottom electrode) has a critical impact on the quality of the ALS even under optimal ALD conditions. If a defect IL is formed before ALD as shown in Fig. 1(b), the ALS could be defective because the first few atomic layers in the ALS would be defective. The structures and relevant physical properties of the samples studied in this work are summarized in Table S1 in the supplementary material. Figure S1 in the supplementary material compares the in vacuo STS dI/dV spectra taken on three different ultrathin films of thickness in the range of 0.1–1.1 nm. The Th-AlOx in Fig. S1(a) in the supplementary material was obtained by the in vacuo diffusion of oxygen into the Al bottom electrode, and the thickness of Th-AlOx of ∼1 nm is estimated based on the oxygen pressure of 2 Torr and exposing time of 520 s.42,44 The high defect (vacancies and interstitials) formation is anticipated from the oxygen diffusion process and is confirmed in the noisy and concave shaped dI/dV spectrum at lower bias voltages before the conduction band minimum or Eb is reached. In addition, the low value of Eb ∼ 0.75 eV is indicative of high defect (or carrier) concentration. Interestingly, a comparable and slightly lower Eb ∼ 0.67 eV was observed on 0.1 nm thickness of one atomic layer of the ALD-Al2O3 film using one ALD cycle (1 C) grown on Al with a defective IL formed during a longer than optimal (15–25 min) pre-ALD heating of 75 min [Fig. S1(b) in the supplementary material].42,43 As we increase the ALD-Al2O3 thickness to ∼1.1 nm (10 C), an increased Eb of ∼1.07 eV was observed [Fig. S1(c) in the supplementary material]. This increase is due to the ALS eventually overcoming the initial defective IL to form higher quality Al2O3 layers, which indicates that the defective IL has a more significant impact on the atomic layers nearer the IL.
(a) A schematic illustration of the M–I–M device with top and bottom electrodes sandwiching the Ga2O3/Al2O3 ALS film. All devices were grown on a Si/SiO2 (500 nm) substrate. (b) An ultrathin ALD oxide film grown on a defective IL at the M–I interface, and (c) an ALD oxide film grown on an IL with negligible defects to enable pristine oxide film growth. (d) and (e) Growth mechanism of one ALD cycle of Ga2O3 (left) and Al2O3 (right) with negligible defects in the IL.
(a) A schematic illustration of the M–I–M device with top and bottom electrodes sandwiching the Ga2O3/Al2O3 ALS film. All devices were grown on a Si/SiO2 (500 nm) substrate. (b) An ultrathin ALD oxide film grown on a defective IL at the M–I interface, and (c) an ALD oxide film grown on an IL with negligible defects to enable pristine oxide film growth. (d) and (e) Growth mechanism of one ALD cycle of Ga2O3 (left) and Al2O3 (right) with negligible defects in the IL.
Therefore, an IL with negligible defects would be imperative to the growth of high-quality Ga2O3/Al2O3 ALS as illustrated in Fig. 1(c). In in vacuo ALD, the bottom electrode will be kept in high vacuum to minimize its ambient exposure and, hence, the formation of defective native oxide on its surface. In addition, a dynamic pre-ALD heating for 15–25 min was found critical to prevent the formation of the native oxide on the bottom Al electrode surface in the ALD chamber before ALD-Al2O3 begins as detailed in our previous work using the in situ STS measurement of the Al electrode surface after it was subjected to different heating processes.39,42,43 Figure S2 in the supplementary material shows the sample temperature in the preheated ALD chamber as a function of time using different heating powers. Dynamic heating allows the minimization of the pre-heating time of the metal in the ALD chamber during which the exposure of the electrode surface to trace amounts of O2 and H2O in the ALD chamber can lead to the formation of a defective IL. The presence of a defective IL of sub-nm thickness would lead to lowered Eb values, dielectric constants, and soft dielectric breakdown. In addition, the dynamic heating would allow ALD growth to be performed in the optimal temperature range between 150 and 190 °C as exhibited in the gray region in Fig. S2 in the supplementary material based on our prior42,43 and recent studies. Using the combination of in vacuo ALD and dynamic pre-ALD heating within 15–25 min, an IL with negligible defects can be achieved on which the high-quality Al2O3, Ga2O3, and Al2O3/Ga2O3 ALS can be obtained on an Al bottom electrode. The specific steps of one ALD cycle for the growth of one atomic layer of Ga2O3 or Al2O3 are depicted in Figs. 1(d) and 1(e), respectively. The step-by-step ALD growth of 1 C Ga2O3 (on the left) starts with the introduction of a [Ga2(NMe2)6] pulse into the ALD chamber, which reacts with hydroxylated sample surface in the H2O pulse. The chamber is purged with N2 to pump out the excess reaction by-products. A second pulse of H2O is introduced to react with the new [Ga2(NMe2)6] surface, and another N2 purge is implemented to achieve the Ga2O3 layer with an O–H surface for the next [Ga2(NMe2)6] pulse. This can be repeated for the desired number of ALD cycles with a growth rate of ∼0.1 nm/cycle.47,50 The same process is used for Al2O3 (on the right) using TMA precursor and H2O pulses with a similar growth rate of ∼0.11 nm/cycle.51 Using computer controlled ALD precursor pulses, Ga2O3/Al2O3 ALS can be grown in any desired numbers of Ga2O3 and Al2O3 atomic layers in a selected stacking sequence. It is anticipated that both Al2O3 and Ga2O3 ALS are amorphous. This is based on previous studies utilizing both the same precursor and temperature growth range for the ALD deposition.50,52,53 Furthermore, the amorphous structure has been confirmed on MgO/Al2O3 memristors using high resolution transmission electron microscopy.54
Figure 2 compares the STS dI/dV spectra taken on four Al2O3/Ga2O3 ALS samples grown on Al electrodes with a negligible defective IL at the ALS/Al interface. All four samples have 10 atomic layers (or 10 C) but different ALS designs. Figures 2(a) and 2(b) include the results taken on Al2O3 (10 C) and Ga2O3 (10 C) samples, respectively. On both samples, the bending of the STS dI/dV spectra at lower bias voltages before reaching Eb is much smaller than the cases with a defective IL. Furthermore, higher Eb values of 1.67 and 0.99 eV are obtained for the Al2O3 and Ga2O3 films, respectively. The Eb value of 1.67 eV in the Al2O3 (10 C) in Fig. 2(a) is considerable higher than its counterpart's Eb of ∼1.07 eV in Fig. S1(c) in the supplementary material due to the removal of the defective IL, which confirms the negative impact of a defective IL on ultrathin oxide films in terms of defect contraction. It should be noted that the lower Eb value in the ultrathin Ga2O3 film as compared to that of its Al2O3 counterparts may be ascribed to the lower Eg ∼ 4.8 eV of the Ga2O3 than that of Al2O3 (Eg ∼ 7.0–7.6 eV) since oxide semiconductors with lower bandgaps are expected to have a higher concentration of native defects such as vacancies and interstitials of metal and oxygen.55 This result illustrates the critical importance in eliminating defects at the ALS/metal (bottom electrode) interface in the growth of ultrathin Ga2O3/Al2O3 ALS films. Figures 2(c) and 2(d) compare the STS dI/dV spectra taken on two Ga2O3 (5 C)/Al2O3 (5 C) samples. While both samples have five each of Ga2O3 and Al2O3 atomic layers, the stacking of the two kinds of atomic layers differs by starting either from the Ga2O3 or from the Al2O3 atomic layer in the ALS growth to probe the quality of the ALS/electrode interface. Interestingly, the two samples have comparable Eb values of ∼1.27 (for the former) and 1.23 eV (for the latter), which means that the starting Ga2O3 or Al2O3 atomic layer forms a similar interface with the Al electrode. This observation is important as a confirmation of the comparable ILs with negligible defects at the Ga2O3/Al and Al2O3/Al interfaces.
Representative STS dI/dV spectra measured on samples of (a) 10 C Al2O3, (b) 10 C Ga2O3, 5 C Ga2O3, and 5 C Al2O3 stacking alternatively with (c) Ga2O3 as the starting layer and (d) Al2O3 as the starting layer on the Al electrode.
Representative STS dI/dV spectra measured on samples of (a) 10 C Al2O3, (b) 10 C Ga2O3, 5 C Ga2O3, and 5 C Al2O3 stacking alternatively with (c) Ga2O3 as the starting layer and (d) Al2O3 as the starting layer on the Al electrode.
It should be noted that the value of Eg ∼ 2Eb if the conduction band minimum and valance band maximum are located symmetrically around the zero-bias voltage, which has been confirmed in the samples measured. Therefore, the measured Eb values in Fig. 2(a) can be used to estimate the Eg values of the samples. Specifically, the Eg values for Al2O3 and Ga2O3 are ∼3.4 and ∼2.0 eV, respectively. These Eg values are comparable to the reported ones on ultrathin (∼1.3 nm) Al2O356 and (∼1.8 nm) Ga2O357 films. Furthermore, the observed hard dielectric breakdown in these ultrathin Ga2O3/Al2O3 ALS films is also consistent with those observed in the counterpart epitaxial thin films and serves as another indicator of low defect concentration in the ultrathin Ga2O3/Al2O3 ALS films due to the minimization of the defective IL.43,56,58 The lower Eg for Ga2O3 is consistent with the fact that Ga2O3 has a smaller bandgap than Al2O3.40 The Eg values for the Ga2O3/Al2O3 ALS are expected to be between that for Al2O3 and Ga2O3, which is, indeed, the case for the two Ga2O3 (5 C)/Al2O3 (5 C) samples with Eg values ∼2.46–2.54 eV. This tunability of bandgap of Ga2O3/Al2O3 alloys has been studied using DFT calculations, and the result is shown in Fig. 3 as solid lines for different crystalline structures of corundum (green), orthorhombic (blue), and monoclinic (purple). The experimentally measured Eb values are also included in Fig. 3, and the fitted line (red) shows a similar nonlinear trend predicted by DFT simulation on crystalline Ga2O3/Al2O3 ALS alloys. The measured Eb values with error bars can be found in Fig. S3 in the supplementary material. It should be noted that Ga2O3/Al2O3 ALSs grown in the selected ALD processing window of this work are amorphous, while the DFT simulation of amorphous materials remains challenging so far. However, the bowing of the curves of Eg (or Eb) with alloy composition is independent of the crystal structure, based on which the same nonlinearity is anticipated for amorphous Ga2O3/Al2O3 ALS alloys. The agreement between the experiment and DFT simulation suggests that the ALD-Ga2O3/Al2O3 ALS provides a facile approach for the synthesis of ultrathin films of Ga2O3/Al2O3 with a tunable Eg.
Computed bandgaps (solid symbols) as a function of alloy concentration (left axis). The solid lines are fits showing the nonlinear behavior of the bandgap. The red stars and the fitted nonlinear curve are experimentally measured Eb (right axis).
Computed bandgaps (solid symbols) as a function of alloy concentration (left axis). The solid lines are fits showing the nonlinear behavior of the bandgap. The red stars and the fitted nonlinear curve are experimentally measured Eb (right axis).
Figure S4 in the supplementary material shows specific capacitance vs voltage curves measured on six 40 C (thickness ∼4.4 nm) Ga2O3/Al2O3 ALS capacitors, all have the same bottom and top Al electrodes, at different frequencies (10 kHz, 100 kHz, and 1 MHz) in the voltage range −100–100 mV. The capacitor area for all samples is around 0.06 mm2. The insets in Fig. S4 in the supplementary material exhibit the ALS structures of the six capacitors. Specifically, they include Al2O3 (40 C) [Fig. S4(a) in the supplementary material], Ga2O3 (40 C) [Fig. S4(b) in the supplementary material], Al2O3 (5 C)/Ga2O3 (35 C) [Fig. S4(c) in the supplementary material], and three Al2O3 (20 C)/Ga2O3 (20 C) samples [ Figs. S4(d)–S4(f) in the supplementary material] with different ALS stacking sequences. Within the low voltage range, the capacitance of all six samples remains nearly constant at a fixed applied frequency, while higher capacitance values were observed at lower frequencies. This trend is consistent with that reported on capacitors of high-quality dielectric.59,60 Interestingly, the six samples have different specific capacitance values with the lowest on the 40 C Al2O3 capacitor and the highest on the 40 C Ga2O3 one. The other four alloy Ga2O3/Al2O3 ALS capacitors have their specific capacitance values falling in between. At 10 kHz, the lowest specific capacitance value of ∼1.25 μF/cm2 was seen in the 40 C Al2O3 device, which is in contrast to the highest specific capacitance value of ∼2.1 μF/cm2 measured on the 40 C Ga2O3 device. The three alloy Ga2O3 (20 C)/Al2O3 (20 C) ALS capacitors have ∼1.6–1.75 μF/cm2 in their specific capacitance, while the Al2O3 (5 C)/Ga2O3 (35 C) device has a specific capacitance of 1.9 μF/cm2 closer to that of the 40 C Ga2O3 capacitor. This trend may be attributed to the higher dielectric constant in Ga2O3 than that of Al2O3.61,62 However, in the case of all Ga2O3 dielectric, capacitance has moderate dependence on applied voltage at lower frequencies. In particular, capacitance decreases with increasing voltage, indicating a rise in the leakage current caused by high charge carrier concentration as indicated in its low Eb among the four samples in Fig. 2.33 DC IV curves for the four 40 C samples of Al2O3, Ga2O3, and both 1:1 ratios of Al2O3 and Ga2O3 with the different starting layers are exhibited in Fig. S5 in the supplementary material. The Ga2O3 capacitor has the highest leakage current, which is about four orders of magnitude larger than that of the 40 C Al2O3 capacitor. Interestingly, the leakage currents for the two capacitors with 1:1 ratios of Al2O3 and Ga2O3 fall in between. This is another indication of the increased carrier concentration for the Ga2O3 devices.
Based on the measured capacitance values, the dielectric constant ɛ can be calculated using the equation , where C is the capacitance, d is the thickness of the ALS, A is the area (0.06 mm2), and ɛ0 is the permittivity of free space. The variation in the dielectric constant as a function of frequency in the range from 1 kHz to 1 MHz at room temperature is presented in Fig. 4. The dielectric constants for all six Ga2O3/Al2O3 ALS samples remain approximately constants in the low frequency range up to 100 kHz with a minor variation of 2%–6%. At higher frequencies, the dielectric constants exhibit a monotonic decrease with increasing frequency for all samples studied in this work. This trend can be explained on the basis of dielectric relaxation. The non-interacting interfacial electric dipole inside the dielectric tends to orient or align itself in the direction of the externally applied electric field. At sufficiently low frequencies, these dipoles align themselves promptly in response to the electric field, resulting in high dielectric constants. However, as the frequency rises, the dipoles begin to fall behind the changing electric field, causing a reduction in the dielectric constant.34,63
Variation in the dielectric constant as function of frequency in the range of 1 kHz–1 MHz measured on the six Ga2O3/Al2O3 ALS capacitors. The insets exhibit the ALS structures of the six ALS capacitors including (a) 40 C Al2O3, (b) 40 C Ga2O3, (c) 5 C Al2O3, and 35 C Ga2O3, and (d)–(f) 20 C Al2O3 and 20 C Ga2O3 with different stacking sequences of (d) all Al2O3 layers at the bottom, (e) and (f) Ga2O3 and Al2O3 stacked alternatively with Al2O3 as the starting layer in (e) and Ga2O3 as the starting layer (f), respectively.
Variation in the dielectric constant as function of frequency in the range of 1 kHz–1 MHz measured on the six Ga2O3/Al2O3 ALS capacitors. The insets exhibit the ALS structures of the six ALS capacitors including (a) 40 C Al2O3, (b) 40 C Ga2O3, (c) 5 C Al2O3, and 35 C Ga2O3, and (d)–(f) 20 C Al2O3 and 20 C Ga2O3 with different stacking sequences of (d) all Al2O3 layers at the bottom, (e) and (f) Ga2O3 and Al2O3 stacked alternatively with Al2O3 as the starting layer in (e) and Ga2O3 as the starting layer (f), respectively.
It is evident that the Al2O3 (4.4 nm) capacitor displays the lowest ɛ, while the Ga2O3 (4.4 nm) one exhibits the highest ɛ among the six Ga2O3/Al2O3 ALS samples. For example, at 1 kHz, the average ɛ ∼ 8.3 for Al2O3 and ɛ ∼ 9.8 for Ga2O3 are observed, which are close to the crystalline bulk values of 9.2 for Al2O361 and ɛ ∼ 10 for Ga2O3,62 respectively. This means that the dielectric properties of the ultrathin Al2O3 and Ga2O3 films could be comparable to their crystalline bulk counterparts' by reducing the defects in the films. It should be noted that the dielectric constant obtained for the 4.4 nm thick Ga2O3 films in this work is more than twice the value previously reported (ɛ = 4.4) for a 15 nm thin Ga2O3 film.32 This enhanced value of ɛ for a ultrathin film clearly demonstrates the significance of maintaining an IL with minimal defects to achieve improved dielectric properties. This study, together with the previous studies on ultrathin Al2O3 and MgO capacitors (with a ∼5 C Al2O3 seed layer),33,34 indicates the critical importance of reducing the defect concentration in ultrathin (sub-5 nm) oxide films’ optimal dielectric properties. An effective approach to reduce the defect concentration in such ultrathin films is to control the oxide/metal interface defects to a negligible level.
Taking as reference the measured specific capacitance of ultrathin Al2O3 and Ga2O3 capacitors,34,64 the specific capacitance of the ultrathin Ga2O3/Al2O3 ALS capacitors can be estimated by treating these capacitors as Al2O3 and Ga2O3 capacitors connected in series. The calculated specific capacitance is determined by the equation , where and are the measured specific capacitances of 100% Ga2O3 and 100% Al2O3 in the dielectric layer, respectively. In Fig. 5, the calculated ɛ (black) are compared with the measured ɛ (blue) values from Fig. 4 at 1 kHz. Both the calculated and measured ɛ follow the same trend with the measured ɛ values for the three alloy Ga2O3 (20 C)/Al2O3 (20 C) ALS samples being in close agreement with their calculated values regardless of stacking order. In addition, the measured ɛ values could also be used to calculate the effective oxide thickness (EOT) for these samples using the equation , used for evaluating high-K dielectric materials, where is the thickness of our dielectric, is the dielectric constant of silicon oxide (3.9), and is the dielectric constant of our high-K dielectric material. The obtained EOT values for the same six Ga2O3/Al2O3 ALS are also included in Fig. 5 (red). Specifically, the lowest EOT is ∼1.7 nm for the 40 C Ga2O3 and the highest one is 2.1 nm for the 40 C Al2O3, while the EOT values for the four alloy Ga2O3/Al2O3 ALS fall between the two. These EOT values are in line with that reported on high-K dielectrics like HfO2 with thickness in the range of 3.0–4.5 nm and in the range 10–18.5.65 Hence, our findings suggest that in the ultrathin Ga2O3/Al2O3 ALS, both of their dielectric constant and EOT can be tuned by adjusting the number of Ga2O3 and Al2O3 atomic layers.
Change in the calculated (black) and measured (blue) dielectric constant (at 1 kHz) and EOT (red) with respect to different Ga2O3/Al2O3 ALS structures shown at the bottom of the figure.
Change in the calculated (black) and measured (blue) dielectric constant (at 1 kHz) and EOT (red) with respect to different Ga2O3/Al2O3 ALS structures shown at the bottom of the figure.
The electric conductance with respect to frequency curves is shown in Fig. S6 in the supplementary material on the same six capacitors in a log–log scale. All samples containing Al2O3 exhibit frequency dependent electric conductance. Specifically, the conductance increases with increasing frequency from about 6.0 × 10−8–1.0 × 10−6 S at 1 kHz to 2.0 × 10−3–3.6 × 10−3 S at 1 MHz. In contrast, the conductance of the 40 C Ga2O3 capacitor is 2.1 × 10−3 S that is approximately independent of the frequency up to 105 Hz, followed by a slight increase to 6.8 × 10−3 S at 1 MHz. At lower frequencies, the considerably lower conductance of the Ga2O3/Al2O3 ALS with Al2O3 can be attributed to the presence of highly resistive, large bandgap Al2O3 atomic layers. However, as the frequency increases, the capacitive reactance of the Ga2O3/Al2O3 ALS decreases. This leads to a reduction in impedance, which contributes to an increase in conductance.66,67 The higher conductance in 40 C Ga2O3 ALS may be attributed to the presence of higher concentration of point defects including oxygen vacancy, oxygen interstitial, gallium vacancy, and gallium interstitial, which is expected from its lower bandgap of 4.8 eV in contrast to 7.6 eV for Al2O3.40 These defects provide an additional path for the current to flow between the electrodes of the capacitor and lead to high conductance and leakage current. Figure S7 in the supplementary material shows the conductance of the six Ga2O3/Al2O3 ALS capacitor samples at 1 kHz. The more than 4 orders of magnitude difference in the conductance across the six samples can be clearly seen. Specifically, the 40 C Ga2O3 ALS has the highest conductance of 2.23 × 10−3 S, while the 40 C Al2O3 ALS has the lowest one of 3.57 × 10−8 S. Based on the conductance in these two samples, the carrier concentration (n) values for the 40 C Ga2O3 and Al2O3 samples were calculated using the Drude formula. The n∼1.27 × 1017 cm−3 for the 40 C Ga2O3 ALS is more than four orders of magnitude higher than the n∼4.5 × 1012 cm−3 for the 40 C Al2O3 ALS. The much larger bandgap (∼7.6 eV) of the Al2O3 in the absence of growth defects on an IL with negligible defects can be regarded as an intrinsic semiconductor with very low n values. In contrast, a large n value is anticipated from the much smaller bandgap of 4.8 eV in Ga2O3 ALS due to the presence of interstitials and vacancies of Ga and O. Interestingly, the conductance values for the four alloy Ga2O3/Al2O3 ALS samples are comparable or slightly higher to that of Al2O3 ALS with a narrow range of 4.53 × 10−7 S–8.19 × 10−8 S. This suggests that the much less conductive Al2O3 ALS in these alloy ALS samples dominates the conductance, assuming that the Ga2O3 and Al2O3 atomic layers grow with comparable quality irrespective of their staking sequences in the ALS. Therefore, the Ga2O3/Al2O3 ALS can be viewed as two resistors connected in series, and the one with significantly larger resistance (or smaller conductance), in this case Al2O3, would dominate.
IV. CONCLUSION
In summary, this work reports the first successful fabrication of ultrathin Ga2O3/Al2O3 ALS with thickness in the range of 1.1–4.4 nm using an in vacuo ALD approach for tunable electronic and dielectric properties. A critical advantage of the in vacuo ALD is to enable control of the ALS/Al electrode interface, and negligible interfacial defects have been confirmed at optimal growth conditions. On the Ga2O3/Al2O3 ALS of ∼1 nm in thickness, the electronic structure was studied using in vacuo STS dI/dV spectroscopy to probe the bandgaps Eg. Importantly, the continuous tunability of Eg was obtained by varying the proportions of the Ga2O3 and Al2O3 atomic layer numbers between ∼2.0 eV for 100% Ga2O3 and ∼3.4 eV for 100% Al2O3. With variable ratios of Ga:Al, the measured Eg exhibits significant non-linearity, agreeing well with the DFT simulation, suggesting alloys of Ga2O3 and Al2O3 form well in the Ga2O3/Al2O3 ALS with an atomic control. Furthermore, the comparable Eg values on 10 C Ga2O3/Al2O3 ALS with five each of the Ga2O3 and Al2O3 atomic layers stacking alternatively with either Ga2O3 or Al2O3 atomic layer as the start layer suggests the comparable ALS/Al interface quality in the two cases. This observation is important to confirm that the charge carriers (interstitials and vacancies) in the Ga2O3 are intrinsically associated with its smaller bandgap instead of from the growth defects initiated from the ALS/Al interface. On the Ga2O3/Al2O3 ALS of ∼4 nm in thickness, the impedance was studied as function of frequency up to 1 MHz. Interestingly, high dielectric constants ∼9.8 and 8.3 that are comparable to the values on the corresponding crystalline bulks were obtained on Ga2O3 and Al2O3 ALS, respectively. In addition, the monotonic tunability of the dielectric constant was obtained through varying the alloy proportions of Ga2O3 and Al2O3 in the Ga2O3/Al2O3 ALS and the trend fits well to the calculated values assuming two different capacitors of Ga2O3 and Al2O3 connected in series. The tunable dielectric constants lead to a tunable EOT of 1.7–2.1 nm for the 4 nm thick Ga2O3/Al2O3 ALS, which is comparable to the EOTs of high-K dielectric materials such as HfO2 of similar thicknesses. Therefore, this work illustrates that high-quality ultrathin Ga2O3/Al2O3 ALSs are promising for applications in electronics with tunable electronic and dielectric properties.
SUPPLEMENTARY MATERIAL
See the supplementary material for the data table for the various device structures, additional STS measurements demonstrating the defective growth of Th-AlOx and Al2O3 on defective IL, heating calibration for the ALD chamber, measured Eb values with error, specific capacitance vs voltage curves at different frequencies of the Ga2O3/Al2O3 ALS capacitors, DC IV curves, and electric conductance curves of Ga2O3/Al2O3 ALS samples grown on Al electrodes with a negligible defective interface.
ACKNOWLEDGMENTS
This work was funded by the Department of Energy’s Kansas City National Security Campus, operated by Honeywell Federal Manufacturing & Technologies, LLC, under Contract No. DE-NA0002839. AM and RG also acknowledge the support from the National Science Foundation (NSF) under Grant Nos. NSF-ECCE2314401, NSF-2136676, and NSF-DMR1909292.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Aafiya and Angelo Marshall contributed equally to this work.
Aafiya: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Writing – original draft (equal); Writing – review & editing (equal). Angelo Marshall: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Writing – original draft (equal); Writing – review & editing (equal). Berg Dodson: Writing – review & editing (supporting). Ryan Goul: Writing – review & editing (supporting). Sierra Seacat: Data curation (supporting); Formal analysis (supporting). Hartwin Peelaers: Data curation (equal); Formal analysis (equal). Kevin Bray: Funding acquisition (equal); Writing – review & editing (equal). Dan Ewing: Funding acquisition (equal); Writing – review & editing (equal). Michael Walsh: Funding acquisition (equal); Writing – review & editing (equal). Judy Z. Wu: Formal analysis (equal); Funding acquisition (equal); Writing – review & editing (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding authors upon reasonable request.