1 Introduction

A broader understanding of the oxidation processes of semiconductors is necessary, for example, for materials science, sensor, and device development. Thus, the study of the various mechanisms contributing to the oxidation of these compounds as well as the analysis of by-products and other process characteristics are extremely important for the development of these fields. Oxidation of compound semiconductors is one of the current problems in microelectronics because these materials do not naturally undergo oxidation to form an oxide layer. In the literature there is an extensive series of experimental results on the formation of oxides, for example, on the GaAs surface by various methods. The formation of oxides on GaAs surface has been reported using plasma of different composition [1, 2], thermal oxidation in oxygen atmosphere [3, 4], in water vapor [5] and anodization [2, 6]. Each of the proposed methods has its own set of advantages and disadvantages. During thermal oxidation at high temperatures, impurities inside the semiconductor can diffuse. Furthermore, at temperatures above 450 °C, the oxide layer formed exhibits increased roughness [3]. Therefore, the use of oxygen-containing plasma for oxide formation on the surface of such semiconductors is considered more favorable than other methods.

In this work, a simple modification of a basic commercial plasma cleaner was carried out, enabling the oxidation of GaAs crystals, as well as other semiconductors, at room temperature within a short period of time.

2 Experimental

In this work, a simple modification of a basic commercial plasma cleaner model PE-25 from Plasma-Etch Co with an operating frequency of 50 kHz and a maximal power of 400 Watts was carried out. The plasma chamber is a grounded aluminum cylinder with a diameter of 130 mm and a length of 180 mm. The floating electrode is attached to the rear wall through an insulator, see Fig. 1a. We replaced the floating electrode of the plasma cleaner for another one made in laboratory with a reduced area. The additional electrode, made of Al, was placed over the floating electrode as shown in Fig. 1b. It is important to note, as it was established experimentally, with the additional electrode made of another material (we used Mo electrode) oxidation of semiconductors does not occur. We will return to this effect below. All other parameters of the plasma cleaner were left unchanged respect to the offered by manufacturer. We used gaseous oxygen to ignite a plasma; the working pressure of O2 varied from 200 to 300 mTorr. The generator power was varied from 100 to 200 Watts approximately (the plasma cleaner has no power meter, just a basic potentiometer). Plasma treatment was performed for 5 to 60 min; the last value is the maximum possible time in this cleaner. All experiments were performed at room temperature. Figure 1c shows an oxidized GaAs crystal (bottom) compared to the original crystal (top) as an example.

Fig. 1
figure 1

Schematic diagram of electrodes in the original (a) and in the modified (b) plasma cleaner. Original GaAs sample (top) and GaAs sample oxidized for 20 min (bottom) (c)

Full size image

Experimental GaAs (100) and Si (100) samples approximately 5 × 5 mm were cut from standard commercial wafers. Diamond Like Carbon (DLC) films with thickness of around 20–30 nm were deposited on Si substrates by ion beam sputter deposition method with a PECS system from Gatan. Before plasma treatments all samples were cleaned in ethanol in ultrasonic bath for 5 min and dried under a stream of dry nitrogen. Figure 1c shows oxidized during 20 min GaAs crystal (bottom) in a comparison with original one (top).

Several additional experiments were performed with oxidized GaAs: we annealed this sample in a rapid thermal annealing (RTP) system in oxygen atmosphere at 900 °C for 30 min to produce gallium oxide by the arsenic-oxygen substitution reaction.

Obtained experimental samples were analyzed by different analytical methods. chemical composition and depth profiling analysis were performed by Raman spectroscopy with NTEGRA Spectra instrument from NT-MDT Co, and by SIMS methods with TOF-SIMS-V time-of-flight mass-spectrometer from IonTof Co, respectively. Surface relief we analyzed by AFM with solver next atomic force microscope from NT-MDT Co.

The SIMS measurements were performed by sputtering the surface with cesium ions at an energy of 500 eV and an incidence angle of 45° relative to the normal. The analysis of the central part of the 300 × 300 μm etching crater was performed by pulsing beam of 30 keV Bi3+ ions; the analyzed area was 100 × 100 μm. To minimize the matrix effect, secondary CsM+ cluster ions, where M is the element of interest, were recorded. We used relative sensitivity factors (respect to Cs2+ experimental signal), obtained for GaAs and Ga2O3 samples to recalculate experimental ion intensities. The depth of the etch crater was determined with a Dektak XT needle profiler after removal of the sample from the mass-spectrometer.

Raman spectra we obtained in backscattering configuration using blue laser (λ = 473 nm). Laser beam was focused by a 100 × objective on analyzed surface.

AFM measurements were performed in semi-contact (taping) mode. Standard Si cantilever with Si3N4 cover and with radius of the tip around 10 nm was used.

3 Experimental results

Figure 2a shows the main elements distribution at the GaAs sample surface after treatment with oxygen plasma for 30 min in the modified device.

Fig. 2 a
figure 2

SIMS depth profile of main elements in oxide layer formed after 30 min plasma treatment of a GaAs sample in modified Plasma Cleaner. b The thickness of oxide layer formed on GaAs at RT as a function of the time of plasma treatment for 5 min (1), 10 min (2), 30 min (3) and 60 min (4). The inset shows a logarithmic approximation found for experimental points

Full size image

The formed oxide layer with a thickness of about 46 nm has a slightly variable composition: Ga to As ratio varies from 1.5 near the surface to 1.2 in the oxide layer at the interface, while the surface layer (< 5 nm) is much more enriched in gallium, as was noted earlier by other authors [2]. A similar picture is observed for other samples obtained at different treatment times. Thus, for all samples rather sharp boundary of an interface of the oxidized layer with bulk GaAs is observed, estimated on a profile in Fig. 2a at 2.8 nm (measured between 0.84*Imax and 0.16*Imax) that is comparable to the boundaries of the interfaces obtained by epitaxial methods of preparation of thin layers (we are speaking here about the profiles measured by SIMS). The latter suggests that it was a plasma chemical oxidation reaction rather than oxygen diffusion into GaAs that took place. As it is known diffusion from a thin adsorbed layer inside the sample is described by a complementary error function. At the interface between the oxide layer and GaAs, an arsenic-enriched layer is observed in the SIMS profile. Similar results have been observed previously for different GaAs oxidation methods [2, 7]. However, given the matrix effect inherent in the SIMS method, we do not undertake to draw unequivocal conclusions from our data about the formation of such a layer in this case. The surface of the formed oxide is slightly contaminated by aluminum, but we will return to this result later.

Figure. 2b shows the oxygen distribution profiles (or oxide layer thickness) for plasma treatment times from 5 to 60 min. The inset shows the dependence of the thickness of the oxide formed on the plasma treatment time. The experimental data are well approximated by a logarithmic function. For all the samples obtained with different treatment times the contamination of the surface of the oxide layer with aluminum was observed. At the same time, no aluminum is observed inside the oxide layer.

Figure. 3 shows Raman spectra for samples with different plasma treatment times. The typical ratio of LO (~  290 cm−1) to TO (~  267 cm−1) modes for a given spectrometer configuration (backscattering) and GaAs (100) crystal orientation, in which the intensity of the TO oscillation is much smaller or not observed at all, changes abruptly during plasma treatment to almost equal intensities of both modes at 5 min treatment. As the processing time increases, the TO mode intensity increases. This mode ratio is characteristic of nanostructured GaAs [8,9,10,11]. In addition, a peak centered at ~  506 cm−1 is observed for 60 min plasma treatment. This band could be assigned as the ω+ mode resulting from the free-carrier plasmon-LO phonon coupling [8]. At the 260 cm−1, a peak corresponding to vibrations in the GaAsO4 compound can be expected [11]. The lines related to gallium oxide cannot be unambiguously identified, but their presence, for example, in the region of 267 cm−1, 277 cm−1, and 475 cm−1 [10], cannot be excluded either.

Fig. 3
figure 3

Raman spectra for GaAs samples after oxidation with oxygen plasma for different time interval

Full size image

We analyzed the surface of the oxide film with an atomic force microscope: the surface roughness averaged 3 Angstroms, which is almost identical to the roughness of the original GaAs crystal.

Considering the shape of the interface, the kinetics of the oxidation process as well as the presence of aluminum contamination on the surface of the formed oxide layer, we concluded that in this reaction, besides oxygen, the aluminum atoms evaporated from the electrode surface take part. The oxygen discharge with aluminum atoms emits ultraviolet light [12] in addition to visible light. This facilitates decomposition of oxygen molecules adsorbed on the surface into atoms and their chemisorption. Recall that replacing the aluminum electrode located above the sample (at a distance of ~ 10 mm) with a molybdenum electrode leads to the disappearance of the effect of oxidation of the semiconductor. In addition, as has been shown previously [2, 6] it gives an additional contribution to the oxidation reaction at elevated temperatures and/or when a weak electric field is applied. In our case this photocatalytic effect becomes the main oxidation mechanism since the crystal temperature remains at room temperature in all experiments. As an explanation of this photocatalytic effect we consider, among other photonic phenomena, an internal photo effect with formation of free electrons in the conducting band of GaAs. These electrons are then involved in the generation of Ga+ and As+ ions at the GaAs-oxide interface and their subsequent interaction with oxygen. The kinetics of the oxidation process is described by a simple logarithmic equation [2]:

$$\alpha \cdot x\,{ = }\,\ln t + {\text{const,}}$$
(1)

where α is the absorption coefficient of GaAs-oxide. In the cited work [2] the authors note practically disappearance of the photo-oxidation enhancement effect when the film thickness exceeds 25 nm. In our case the effect is observed for a layer almost three times thicker (as minimum). Probably it is determined by longer wavelength of UV, generated by excited Al atoms (396 nm [12]). As a result, UV photons penetrate through the thicker oxide layer and reach the oxide–GaAs boundary. After the discharge is turned off, the aluminum from the plasma is deposited on the surface of the sample and on the electrodes, which is recorded by SIMS.

It should be noted that the large difference in electrode surfaces and the low plasma frequency can theoretically lead to a relatively large potential difference between the floating electrode and the plasma. This potential difference is not enough for ion etching of the sample surface but the potential arising on the surface of the sample to be treated can lead to ion drift in the oxide layer, as is the case in the anodization method [2, 6].

We performed several experiments with silicon crystals under the same oxidation conditions. Treatment in oxygen plasma for 30 min resulted in an oxide film on silicon around 17 nm thick, see Fig. 4. Note that the non-stoichiometry in the 5 nm thick surface layer is due to surface charging during the SIMS measurement, and the associated distortions in the measured secondary ion intensities. A sharp boundary between the oxide and bulk silicon was observed; the roughness of the oxide layer practically coincided with the roughness of the original silicon crystal. The residual atmosphere in the plasma cleaner, and first water vapor, certainly affects the composition of the oxide film (see hydrogen distribution in the profile), but the concentration of impurities is not critical for the properties of the resulting oxide.

Fig. 4
figure 4

Silicon oxide layer formation on Si crystal by oxygen plasma treatment for 30 min at room temperature, measured by SIMS

Full size image

Plasma cleaner is designed to remove organics from the surface of semiconductors, polymers and metals and primarily carbon through the formation of volatile compounds. We used a modified plasma cleaner to etch diamond-like carbon (DLC) films obtained by different methods: vacuum arc deposition and ion beam sputter deposition. Etching with oxygen plasma with power of about 200 W for 30 min completely removes DLC layer of around 20 nm thickness: neither in Raman spectra, nor in SEM image obtained in Z-contrast mode traces of carbon film on the sample surface are visible.

We used GaAs samples with oxide film prepared by plasma to obtain gallium oxide by As–O substitution reaction: samples were annealed in oxygen atmosphere at 900 °C for 30 min. This treatment resulted in formation of polycrystalline monophasic gallium oxide β-Ga2O3 with characteristic white color. Figure 5 shows the Raman spectrum for the annealed sample obtained by blue laser excitation (473 nm). All peaks characteristic for β-Ga2O3 phase are present in the spectrum [13] and there are no vibrations related to Ga–As bonds. On the fracture of the sample no layer with metallic luster associated with GaAs is observed. That is, the oxidation occurred to the full depth of the GaAs crystal (~  400 µ). Surface analysis showed high roughness of both initially polished and unpolished GaAs surfaces after oxidation: the oxide roughness on the previously polished surface was about 170 nm, whereas for the unpolished surface the roughness exceeds 340 nm. We compared the formation of gallium oxide from an original gallium arsenide crystal under the same annealing conditions in oxygen: the absence of an oxide film obtained in plasma results in the resulting gallium oxide having a roughness greater than 300 nm on both sides of the original crystal (polished and unpolished). The original GaAs crystal was doped with silicon with an impurity concentration of 1 × 1019 atoms cm−3. However, surface resistivity of the obtained gallium oxide was more than 109 Ohm. That is, the impurity silicon apparently formed silicon oxide and we ended up with a dielectric.

Fig. 5
figure 5

Raman spectrum for β-Ga2O3 sample obtained from previously oxidized GaAs sample (30 min in plasma) after annealing in oxygen atmosphere for 30 min at 900 °C

Full size image

4 Conclusions

Thus, by a simple modification of a basic plasma cleaner we have obtained an apparatus and a method for producing oxide films on semiconductors, including compound ones, at room temperature. The oxidation takes place through photo-stimulation of the process by UV radiation generated by the appearance of aluminum atoms in the oxygen plasma. The roughness of obtained oxide layers is such that it does not require additional processing of the obtained oxide layer. Contamination of the surface with aluminum can be removed by chemical reagents. The oxide layers that are formed are relatively thin, but the time it takes to form them is also quite short. The oxide layer formed on GaAs surface in oxygen plasma, during further annealing in oxygen atmosphere somewhat reduces the surface roughness of the formed gallium oxide β-Ga2O3.