Electronic Properties of Atomic Layer Deposited HfO₂ Thin Films on InGaAs Compared to HfO₂/GaAs Semiconductors

Abstract

This paper demonstrates how the treatment of III-V semiconductor surface affects the number of defects and ensures the conformal growth of the high-k dielectric thin film. We present the electrical properties of an HfO₂/InGaAs-based MOS capacitor, in which growth temperatures and surface treatments of the substrate are two key factors that contribute to the uniformity and composition of the HfO₂ thin films. A remarkable asymmetry observed in capacitance versus voltage measurements was linked to the interface defects and charge redistribution, as confirmed from X-ray photoelectron spectroscopy. The GaAs substrates that were etched with only NH₄OH showed a large frequency dispersion and a higher surface roughness; however, the HfO₂ thin films grown on GaAs pre-treated with both NH₄OH etching and (NH₄)₂S passivation steps produced a desirable surface and superior electronic properties.

1. Introduction

New technologies are needed to compensate for the scaling issues associated with the electrical limitations of silicon complementary metal–oxide–semiconductor (CMOS) devices as their channel lengths reach the nanometer regime. This is due to the electrical barriers breaking down or losing their insulating properties due to such effects as thermal injection and quantum-mechanical tunneling [1]. Therefore, new materials and processes are being heavily researched to solve this scalability issue. Processes such as the use of various substrate preparation techniques [2,3], dielectric material deposition techniques [4,5], and metal gate deposition techniques are being investigated to provide optimal gate/oxide interfaces as well as metal gate/dielectric interfaces. As with all CMOS/MOSCAP devices, low current leakage, low threshold voltage, and high capacitance, along with clear accumulation, depletion, and inversion regions, are desired [6,7]. Hafnium oxide (HfO₂) gate dielectric material paired with a high electron mobility substrate, InGaAs, grown under optimal conditions, will serve as a feasible replacement for silicon oxide gate dielectrics for scaling down CMOS devices. The search for technology beyond the CMOS 22 nm and 14 nm nodes has now called for research activities on high-k gate dielectrics on channel materials with high carrier mobility, which are especially found in III-V compound semiconductors, such as InGaAs [8,9,10,11].

Due to the presence of dangling bonds in semiconductor materials that easily interact with dust particles and other debris to create electronic defect states in atmospheric environments, these processes are typically performed in high-quality clean room environments [12]. Even in these clean room environments, there are a number of steps involved in preparing semiconductor materials for various types of implementations including chemical surface treatments and material depositions/evaporations in order to remove native defects. These treatments and material depositions, often used in conjunction with one another, can effectively serve as a solution for the aforementioned dangling bonds. Every chemical has a different purpose and provides a different effect including, but not limited to, freeing a surface of contamination, etching away at semiconductor native oxides, or capping dangling bonds so they do not create defect states. There are various cleaning chemical reagents such as acetone and methanol that are commonly used to degrease a surface. Various corrosive etching chemicals are used to etch away at semiconductor native oxides that create further electronic degradation, which react with moisture in the air causing oxidation [13]. A common etching material is ammonium hydroxide (NH₄OH), which effectively etches away at many different native oxides of III-V semiconductors such as gallium arsenide (GaAs) [14].

There are many ways to chemically passivate a semiconductor material, making it less susceptible to oxidation and electronic defect states caused by environmental factors. Ammonium sulfide [(NH₄)₂S] is amongst the most commonly used chemicals for passivation purposes of materials such as GaAs, as its sulfide solution caps off the dangling bonds that can attract defects [15,16]. These chemicals (and combinations of them) assist in the fabrication process of MOSFETs and such devices by improving the surface morphology to allow better conformal growth of thin films on semiconductor materials.

In this paper, we report on the fabrication and characterization of HfO₂ high-k gate dielectrics on InGaAs. We present the remarkable asymmetry observed in capacitance versus voltage (C–V) characteristics of the atomic layer deposited (ALD) HfO₂ on InGaAs. This asymmetry is explained in terms of the interface charge redistribution as confirmed by X-ray photoelectron spectroscopy (XPS). On the other hand, we investigated the effects of etching GaAs with NH₄OH followed by passivating with (NH₄)₂S, on the growth kinetics, surface properties, and electronic properties of HfO₂ thin films on GaAs.

2. Materials and Methods

HfO₂ thin films were grown on n-type InGaAs substrates using an Ultratech Cambridge/Nanotech Savannah ALD system. Three temperatures (150 °C, 175 °C, and 200 °C) for the HfO₂ growth were studied based on previous knowledge. Although the precursor used to grow the HfO₂ thin films, Tetrakis (dimethylamido)hafnium (IV), has a breakdown in chemical composition above 150 °C, previously performed work has suggested that high-quality films are still achieved between 150 °C and 250 °C [6,8]. As-received InGaAs substrates were ultrasonically degreased in subsequent baths of acetone, methanol, and deionized water for one minute each. This was followed by a three-minute etch in a 25% concentration ammonium hydroxide (NH₄OH) solution to reduce the native oxides on the surface of the InGaAs substrate that would produce unfavorable electrical characteristics. The ammonium hydroxide etch was followed by a one-minute bath in deionized water to terminate the etching process that while reducing native oxides, also increases surface roughness. These degreased and etched InGaAs substrates were then split into two groups. One-half of the samples received a self-cleaning layer before the HfO₂ thin film growth, while the other half did not and were labeled as-grown. That is, the AG samples were used directly from degreasing and etching for HfO₂ thin film growth. The self-cleaning step was performed using a self-terminating reaction using trimethylaluminum (TMA) at 280 °C [16,17,18].

GaAs samples were cut from n-type GaAs (100) wafers and degreased with acetone, methanol, and isopropanol. After being cut and cleaned, the samples were split up into four different sets, where one set went through a simple etching of native oxides in 29% NH₄OH for 3 min. The second set was processed through chemical passivation and was placed in 10% (NH₄)₂S for 20 min at room temperature. The third set of samples was treated with both etching and passivation immediately after one another. The samples were then transferred for deposition to the ALD system. A self-cleaning process was performed on all samples to further reduce any native oxides on the GaAs substrate by pulsing 20 cycles of Tetrakis (dimethylamido) hafnium (IV) at 300 °C. Following the self-cleaning process, 100 cycles (~10 nm) of HfO₂ using Tetrakis (dimethylamido) hafnium (IV) were grown at 150 °C. A film thickness of 10 nm was chosen due to the current and future wave of semiconductor technology and the desire to create quality sub-10 nm structures and films. After the HfO₂ deposition, all samples underwent a rapid thermal processing (RTP) treatment at 700 °C for 30 s in a N₂ atmosphere to improve the electronic properties [10].

Titanium (10 nm)/gold (50 nm) metal contacts for electrical characterizations of the devices were deposited using an e-beam/thermal evaporation system. The titanium metal serves as an adhesion layer for the gold metal contacts. Ti/Au top metal contacts with 200 nm diameters were deposited for electrical characterization of HfO₂/GaAs. The ohmic contact electrodes on both semiconductor crystals were performed using Au deposition. The electrical characterizations of the devices were carried out using the Keithley 4200-SCS semiconductor characterization system using Signatone probe holders and tungsten probes.

The chemical nature of the surface of the substrates and HfO₂ films was characterized eX situ by a Kratos Axis Ultra DLD X-ray photoelectron spectroscopy (XPS) system using a monochromatic Al Kα (1486.7 eV) X-ray source at a power of 150 W and a hemispherical energy analyzer. All spectra were acquired using a 20 eV pass energy and charge neutralizer to prevent charging of the HfO₂ film. All spectra were also corrected to the C 1 s peak at 284.6 eV. A Shirley background correction was typically used, and components used for peak fitting were a Lorentzian–Gaussian (30:70) mix. The software used for all data analysis was the Kratos Vision 2.2.7. Surface and interfacial roughness, film density, and film thickness were measured using X-ray reflectivity (XRR) with a standard four-circle diffractometer with Cu-Kα radiation. XRR data were fit using Parratt’s recursive formalism with roughness [19,20,21,22], where density values are within ±20% of bulk. The HfO₂ films were amorphous, and the thickness of the layer was controlled by number of cycles of Tetrakis (dimethylamido) hafnium (IV).

3. Results and Discussion

Atomic force microscopy (AFM) was performed on all samples to compare the surface morphology of the self-cleaned to as-grown samples and also to clarify which growth temperature produces the least amount of surface roughness. The self-cleaning process does not significantly change the surface roughness of the HfO₂ thin film at low temperatures (i.e., 150 °C and 175 °C) as compared to the as-grown samples (see Figure 1a–d) but does improve the surface roughness at high temperatures (i.e., 200 °C), as shown in the AFM images (see Figure 1e,f).

X-ray reflectivity (XRR) scans can elucidate the density, thickness, and conformity of the HfO₂ thin films, as well as the surface/interface roughness of the film and/or substrate. XRR data and fits are shown for the as-grown samples (see Figure 2 top plots) and the self-cleaned samples (see Figure 2 bottom plots), where the best fit parameters are presented in

Table 1. Growth temperature, surface treatment, HfO₂ film thickness, surface and interface roughness from the ALD grown InGaAs/HfO₂.

Growth Temperature	Surface Treatment	HfO2 Film Thickness (nm)	HfO2 Film Density (g/cc)	HfO2 Surface Roughness (nm)	InGaAs/HfO2 Interface Roughness (nm)
150 °C	As-grown	31.35	9.11	0.74	0.70
	Self-cleaned	31.6	9.08	0.67	0.64
175 °C	As-grown	30.17	9.16	0.73	0.68
	Self-cleaned	30.1	9.01	0.71	0.72
200 °C	As-grown	27.9	9.03	0.66	0.63
	Self-cleaned	28.16	9.05	0.61	0.58

. High-quality HfO₂ thin films with their anticipated thickness near 30 nm from the ALD process were grown as shown in Table 1.

Figure 3 shows the XPS data of the sample surface in order to infer the binding energy of the chemical species. The energy peaks due to As_3d, In_3d, and Ga_2p of the InGaAs substrate are indexed. We explored the As_3d and Ga_3d levels. It is believed that the hump in the binding energy between 14 and 16 eV is due to the combination of Hf 4f _7/2 and Hf 4f _5/2 for Hf metal. However, the peaks observed around 18 and 20 eV are due to HfO₂ (ref: for their respective Hf4f _7/2 and Hf4f _5/2) which is clear evidence for a single chemical state for Hf as HfO₂ [9].

Capacitance versus voltage (C–V) plots yield information about the capacitance of each device over the different modes of operation, as well as information about the interfacial trap densities. Low-frequency dispersion in the accumulation region followed by hysteresis and a small mid-gap bump [10] in the as-grown samples (see Figure 4a) and the self-cleaned samples (see Figure 4b) were observed. This can be attributed to the interfacial defect of the heterostructure. The flat band voltage (VFB) dispersion (at the measured frequencies) becomes smaller as the growth temperature increases. This may be attributed to the difference in the interfacial layers with increasing growth temperatures. However, the remarkable asymmetry observed in the C–V measurements, especially in the reverse bias condition, is linked to the presence of an ionic state at the interface, as evidenced by the presence of Hf at the interface.

Figure 5 shows the capacitance vs. voltage (C–V) characteristics for the four different surface pre-treatments at both 10 kHz and 1 MHz for HfO₂ thin films grown on GaAs. Qualitatively, it can be seen that the samples that underwent etching in NH₄OH and passivation in (NH₄)₂S exhibit the least amount of frequency dispersion, while the samples that were simply etched in NH₄OH display the largest amount of frequency dispersion. This can be attributed to NH₄OH being such a corrosive chemical. Also, Figure 5 shows that the sample that was both etched and passivated displays the largest capacitance (~225 pF at 3 V) at 1 MHz, which is attractive to technological companies whose goal is to produce devices that operate efficiently at high frequencies. When a negative voltage is applied to the upper electrode, the depth of the depletion layer increases, which leads to a decrease in capacitance, since the total capacity is the series-connected capacity of the dielectric layer and the depletion layer.

Figure 6 shows the dual sweep characteristics of each chemical surface pre-treatment at 10 kHz, where it can be seen that similar gaps, and thus trap defects, exist between the sweeps for each surface pre-treatment. Therefore, it can be estimated that each chemical surface pre-treatment is equally effective at reducing trap defects, although this may be attributed to the RTP treatment that each set of samples underwent.

Figure 7 shows the current vs. voltage (I–V) characteristics for HfO₂ thin films grown on GaAs that underwent the four various chemical surface pre-treatments. As seen in the figure, the sample that underwent both the etching and passivation processes exhibits the lowest leakage current throughout the entire range of −3 V to +3 V. The samples that were only etched in NH₄OH or only passivated in (NH₄)₂S both displayed leakage currents that were closely similar at −3 and +3 V but experienced some trade-off near −1.5 and 0.5 V. NH₄OH is typically known as a very corrosive chemical that is effective at etching away at native oxides and dangling bonds that create defect states. The passivation solution simply caps the dangling bonds off, with a small amount of them still possibly being present under the solution. The sample that only underwent a degreasing step exhibits the highest leakage current, which can be due to the fact that native oxides were still present [19]. Such MIS structures are relevant not only for field-effect transistors but also for photodetectors, both charge-coupled devices and photodiodes with a tunnel-transparent dielectric, as discussed recently [20].

X-ray reflectivity was used to obtain the HfO₂ thin film properties such as surface and interfacial roughness, film density, and film thickness, as described earlier.

Table 2. XRR best fit values for HfO₂ thin films grown on GaAs that were treated with four various surface pre-treatments. Values are presented for film thickness, film density, and surface roughness. The sample that was etched and passivated produced the closest theoretical thickness and a smooth surface.

Surface Pre-Treatment	HfO2 Film Thickness (nm)	HfO2 Film Density (g/cc)	HfO2 Surface Roughness (nm)
Degrease Only	10.4	8.3	0.5
NH4OH	10.6	8.9	0.6
(NH4)2S	10.9	8.8	0.5
NH4OH and (NH4)2S	10.4	8.1	0.5

shows the best fit values for each thin film property. The XRR data and best fits use Parratt’s recursive formalism with roughness [21,22], where the density values are within ±20% of bulk. The data in Table 2 show that the sample that was both etched and passivated produced an HfO₂ film that was closest to the theoretical thickness of 10 nm and had the lowest film density and surface roughness. The sample that was solely degreased displayed a similarly low surface roughness as well. Sole use of the etching chemical pre-treatment produced the roughest surface as expected, which is mitigated by the addition of the passivation chemical pre-treatment. The sample that was solely passivated produced the thickest sample.

4. Conclusions

In summary, we grew HfO₂/InGaAs high-k dielectrics using the ALD technique, which demonstrated asymmetrical C–V characteristics that can be explained due to the interface charge redistribution, as confirmed from XPS data, as well as numerical modeling of the OH terminated InGaAs (001) interface as it is exposed to HfO₂ using the first principles quantum mechanical approach. Our results show the key role of the bounded oxygen in the interface area for the observed electrical properties of the HfO₂/InGaAs interface. The HfO₂/InGaAs interface studies demonstrate the right choice of gate metal for InGaAs and are consistent with other reports [23,24,25,26]. Chemical surface pre-treatments are effective in producing desirable leakage current and capacitance properties for HfO₂ thin films grown on GaAs. Although both passivation and etching processes separately produce good surface and electrical properties of HfO₂ thin films grown on GaAs, combining both NH₄OH and (NH₄)₂S at the correct concentrations into a fabrication process will produce even better results. Our findings are of great importance for several applications.