Electron irradiation induced amorphous SiO2 formation at metal oxide/Si interface at room temperature; electron beam writing on interfaces

Al2O3 (5 nm)/Si (bulk) sample was subjected to irradiation of 5 keV electrons at room temperature, in a vacuum chamber (pressure 1 × 10−9 mbar) and formation of amorphous SiO2 around the interface was observed. The oxygen for the silicon dioxide growth was provided by the electron bombardment induced bond breaking in Al2O3 and the subsequent production of neutral and/or charged oxygen. The amorphous SiO2 rich layer has grown into the Al2O3 layer showing that oxygen as well as silicon transport occurred during irradiation at room temperature. We propose that both transports are mediated by local electric field and charged and/or uncharged defects created by the electron irradiation. The direct modification of metal oxide/silicon interface by electron-beam irradiation is a promising method of accomplishing direct write electron-beam lithography at buried interfaces.


Spectroscopic ellipsometry (SE).
The irradiated spot can clearly be identified on the map of measured ellipsometric angles, Δ, in Fig. 1, revealing that the irradiation caused a significant change in the optical properties of the layers. The difference in Δ between the irradiated and non-irradiated regions is larger than one degree, thus more than an order of magnitude higher than the sensitivity of the measurement.
For determination of the thickness of the amorphous SiO 2 (a:SiO 2 ) the interface was modeled as an effective medium composition of Al 2 O 3 , SiO 2 and Si 18 . Figure 2 shows that both the interface thickness, a, and the volume fraction of a:SiO 2 , b, increase in the irradiated spot. Due to small thickness and the correlation between the Al 2 O 3 and the a:SiO 2 components, there might be a significant error in the determined volume fraction of a:SiO 2 . However, the trend of the increasing interface thickness and volume fraction of a:SiO 2 in the irradiated part is very clear from the ellipsometry results.
Auger Electron Spectroscopy (AES). The samples were produced either by ALD or RF sputtering.
The results did not depend on the production technology and thus in the following the type of sample will not be specified. Figure 3 shows the Auger spectra, N(E), in the 1600 ± 20 eV range (where the Si KLL Auger line appears independently from its chemical state) recorded on the surface of the irradiated (1.32 electrons/s/ A 2 ×20 hours = 9.4×10 20 electrons/cm 2 ) and non-irradiated regions of the sample. It is clear that the intensity of the KLL Auger electrons emitted by Si in oxide environment increased considerably due to the irradiation.  Simultaneously, the intensity of the KLL Auger electrons emitted by Si in metal environment decreased. The rough interpretation of the finding is simple; within the information depth of the Si KLL Auger electrons a fraction of the Si has been converted to SiO x . The x value cannot be accurately determined by AES but it is not far from 2. From the TEM results it is also clear that the compound formed is amorphous. Thus we will sign the compound as a:SiO 2 , which means an amorphous phase with composition close to the stoichiometric. However, since the measured Auger intensity is a weighted sum of the intensities emitted from various depths, the actual depth distribution of the a:SiO 2 produced cannot be derived from this measurement.
For the determination of the depth distribution of the a:SiO 2 produced by the electron irradiation AES depth profiling has been utilized. In AES depth profiling the surface of the sample is etched away in predefined steps by using ion sputtering and the newly exposed surface is analyzed by AES which provides elemental and some chemical information. Figure 4 shows the raw data (measured Auger peak-to-peak intensities in the differentiated, N′(E) curve) obtained from the non-irradiated (a) and irradiated (b) (total electron dose of 9.4×10 20 electrons/cm 2 ) regions of the sample; the two depth profiles show considerable differences. The most important one is that while in the case of the non-irradiated region the intensity of the Si Auger signal in oxide environment (signed by SiO) is just higher than the noise level around the Al 2 O 3 /Si interface, in the case of the irradiated region the same signal is easily measurable, proving that due to the irradiation a:SiO 2 has formed. It should also be noted that longer sputtering time is necessary to remove the layer at the irradiated region than that necessary in the case of the non-irradiated one, demonstrating that the thickness of the top most layer in the irradiated region (pure Al 2 O 3 + altered layer + a:SiO 2 ) of the sample is greater than that of the virgin part. Figure 5 shows the concentration distributions, which best correspond to the depth profiles shown in Fig. 4 measured on the non-irradiated (virgin) and irradiated region of the sample, respectively, calculated by our trial and error evaluation method 19,20 . Note that the irradiation took place at room temperature.  On the other hand, the irradiated region provides a strongly altered structure. Though the surface is covered by a pure Al 2 O 3 layer its thickness is only about 3 nm instead of the 5 nm initial thickness. This cover layer is followed by an Al 2 O 3 and a:SiO 2 mixture in which the concentration of the a:SiO 2 increases toward the interface. The overall thickness of the altered layer and reminder Al 2 O 3 layer is larger than that of the initial Al 2 O 3 layer.

Discussion
Both methods (SE and AES) clearly show that the irradiated part of the sample has considerably changed: a:SiO 2 has been built due to the irradiation. Since the non-destructive ellipsometric measurement cannot produce a:SiO 2 it is clear that the a:SiO 2 is the result of the electron irradiation.
An obvious question is whether the subsequent AES studies combined with ion bombardment might impair the actual distribution of the a:SiO 2 formed during the electron beam irradiation experiment. Considering Fig. 5b we can estimate the overall rate of the process; it is rather low being in the range of 10 −4 -10 −5 a:SiO 2 /5 keV electrons. Thus, the production of SiO 2 during AES measurement can be ignored since the irradiation fluence during the recording of the AES spectra is 300 times lower than that of the irradiation experiment.
The ion bombardment can also produce SiO 2 . The efficiency of the process can also be easily checked experimentally. The measured Auger intensities recorded on the surface before Auger depth profiling cannot be used to determine the distribution of the emitting elements. On the other hand, the depth distributions determined by the AES depth profiling can easily be integrated and they should give the intensities measured at the surface before the AES depth profiling. This integral has been calculated for all measurements and compared with the intensities measured on the surface. The amount of a:SiO 2 calculated from the AES measurement performed on the surface before the AES depth profiling started, and that obtained for the AES depth profile agreed within an error of 20%, meaning that the possible effect of the ion bombardment on the a:SiO 2 content is typically still less than 20%, which will be ignored. Thus, the concentration distributions provided by the AES depth profiling can be accepted as those produced by the 5-kV electron irradiations.
The accurate determination of heating effect of the electron beam irradiation is rather difficult 15 in case of high current densities and small particles. In our case the current density was chosen to be rather low, which allows to make a rough overestimation of the temperature rise, ΔT. It can be supposed that the whole irradiation energy is absorbed by the sample and used for heating. In this case in stationary case the input energy is equal to that carried away by heat conduction which results in the simple equation of ΔT = Pr/K 21 , where P is the irradiation power density, r is the radius of the electron beam, and K is the heat conduction. In the present case ΔT is around 3 K, which is negligible, and consequently all processes are taking places at room temperature.
The basic features of the a:SiO 2 growth (due to the electron irradiation) were similar on samples prepared by different technologies. Thus, we ignore possible effects of the imperfections connected to the sample production technology, and the process will be explained assuming perfect interfaces. The growth of a:SiO 2 can be divided into two steps: creating charged and/or neutral oxygen and the growth process itself.

Electron irradiation induced bond breaking & excitation.
Vast amount of experimental data has been amassed in transmission electron microscopy (TEM) studies concerning various electron beam-induced damages in insulators 15 . Several electron irradiation-induced effects, including phase transformation, decomposition, amorphization, oxidation, reduction, etc., have been identified. For the description of the phenomena two basic models have been developed: a./the mechanical interaction of the bombarding electron with the nuclei, knock-on mechanism, b./the electronic excitation by the electric field, radiolytic processes.
The knock-on process in our case will not be considered since the energy, 5 keV, of our bombarding electrons is far below the threshold of this mechanism 15 . On the other hand, the radiolytic processes have practically no threshold since because of the multiple interactions nearly any type of excitations can happen. E.g., recently the dissolution of boehmite under the high energy electron irradiation (TEM studies) was explained by the electron-hole pairs created during the electron bombardment 22 . This explanation is rather similar to those developed long time ago for the description of low-energy electron-stimulated desorption (ESD) [23][24][25] which can readily be applied for the explanation of our observations. According to these models following the ionization of a core electron, a valence electron from the O decays to fill the resulting hole in frame of the Auger process. Therefore, O 2− transforms to O 0 . It might also happen that by an additional Auger effect a further electron is emitted and O + is formed for a short while [23][24][25] . The lifetime of both O 0 and O + is long, since fast recombination is not possible due to the insulator matrix. During the excitation processes, because of Columbic repulsion, structural changes might also occur. The electron irradiation this way produces neutral and/or charged O in the Al 2 O 3 matrix together with various charged and neutral crystal defects. SiO 2 growth. If Si surface containing O 2 and/or H 2 O is irradiated by electrons, SiO 2 can be produced 3,4 . The description of the process is similar to that of ESD. In this case it is assumed that electrons attach to an adsorbed O 2 molecular precursor to form O 2 − . The O 2 − then decomposes to form O and O − , and one or both of these species cause rapid oxidation of the surface. In our case neutral and/or charged O are produced in the Al 2 O 3 matrix and for the compound formation to take place they should be transported to the interface.
The diffusion of O in α alumina has been studied by Sokol et al. 26 showing, that due to the structural and spin configuration, the defect reaction energy can change by over 2 eV. This behavior affects the equilibrium defect concentrations by many orders of magnitude. Consequently, the diffusion processes in such materials may be more complicated, which has previously been assumed. Århammar et al. reached similar conclusion for amorphous oxides 27 . Still it seems that at room temperature neither the O nor the Si can be transported fast enough by thermal diffusion to explain the growth of SiO 2 layer at the Al 2 O 3 /Si interface.
It is well known that despite the high heat of the SiO 2 formation the native oxide forms on the surface of clean Si at room temperature in air. The thickness of this oxide is about 2 nm. This process is explained by the theory of Cabrera and Mott, which assumes that the electrons can pass freely from the Si to the oxide surface to ionize oxygen atoms. This establishes a uniform field within the oxide, which leads to a shift in the Fermi level of the oxide 28 . The same reasoning can be used if electron is placed to the surface from any other source. Nowak et al. have shown that electron bombardment, providing charges to build up electric field, induces oxide growth on tungsten nanowires at room temperature 29,30 . They use the explanation of Mott and Cabrera; the electric field created reduces the energy barriers for the migration of metal cations or oxygen anions into and through the oxide, allowing significant material transport and thus growth of the oxide layer at low temperature. It is also evident that this is a self-controlled process; after reaching a given thickness the strength of electric field is no more sufficient to drive the diffusion.
Our Al 2 O 3 layer is 5 nm thick that is thicker than the native SiO 2 . In our case, however, the charged defects, charged and neutral oxygen produced by the electron bombardment are distributed evenly in the layer. That is, in our case various distributions of local fields and charged as well as neutral oxygen is produced. It is evident that those oxygen atoms, which are close to the interface can be transported to the substrate Si atoms and form oxide at a high probability; thus the oxide growth starts from the interface. Similarly, if the charged defect is close to the interface, then an electric field of sufficient strength is built up initiating the diffusion of Si into the Al 2 O 3 layer. Since in the defected Al 2 O 3 layer there are charged and/or neutral oxygen atoms, compound formation might take place. The a:SiO 2 grain grows by adding additional Si and O atoms. The primary Si source is the substrate. It should be remembered, however, that the self-limitation process is activated after the a:SiO 2 grain reaches a certain thickness and the local field strength is not sufficient anymore to drive the diffusion. It seems that our particles are somewhat larger than that of the typical thickness of the native oxide. On the other hand, one should also consider that the already produced a:SiO 2 grains are also subjected to electron irradiation producing charged defects, excited oxygen atoms and quasi free Si in the a:SiO 2 grain. The quasi free Si can utilize the local fields to diffuse to the surface of the a:SiO 2 grain resulting in further growth. The probability of this process is lower than that of the primary one, however, and the rate of growth in this phase is much lower resulting in only some additional growth resulting in a 3-4 nm thick a:SiO 2 grains.

Conclusions
We have shown that bombarding 5 nm thick Al 2 O 3 /Si structure by 5 keV electrons at room temperature, amorphous silicon dioxide is produced. The amorphous SiO 2 grains grow from the interface toward the Al 2 O 3 matrix; their amount depends on the irradiated charge. The phenomenon was explained considering electron bombardment-induced bond breaking in Al 2 O 3 , electric field driven diffusion of Si and O in defected regions of the Al 2 O 3 based on the Cabrera-Mott theory.
The direct modification of metal oxide/silicon interface by electron irradiation is a promising method of accomplishing direct write lithography at buried interfaces.

Samples and Methods.
Samples. Samples were made by growing an Al 2 O 3 layer on a Si (100) substrate ALD and RF sputtering.
The ALD layers were made by a Beneq TFS 200 ALD reactor in the plasma-enhanced deposition mode. Trimethylaluminium (TMA -from Sigma Aldrich) together with high purity oxygen gas was used for deposition. Prior to the sample preparation, the deposition chamber had been heated up to 150 °C. During the deposition, the pressure inside the vacuum chamber was 9.5 mbar, while in the reactor chamber 1.1 mbar. The RF power of the plasma was set to be 50 W and the flow rate of the oxygen was 100 sccm. The ALD cycle was the following: 150 ms TMA, which was followed by a 2s purge, then a 2s oxygen plasma at 50 W and at the end of the cycle another 2s purge. The RF sputter deposition was carried out in a Leybold Z400 apparatus evacuated to 5×10 −5 mbar. Sputtering was performed under a mixture of high purity argon and oxygen gases with an applied RF power of 255 W yielding a plasma pressure of 2.5×10 −2 mbar. Oxygen was continuously let into the sputtering chamber at flow rates resulting in a partial oxygen pressure of 6%. The deposited amorphous Al 2 O 3 film has a high refractive index and low absorption coefficient 31 .
Electron irradiation and Auger Electron Spectroscopy. All electron irradiation experiments have been carried out in our standard vacuum system equipped with electron energy analyzer (STAIB OPC 105) and various electron sources. The electron irradiation was facilitated by an electron gun (STAIB EK-10-M) with the following parameters: energy 5 keV, current density up to 6*10 16 electrons/cm 2 /s fluence about up to 6 × 10 21 electrons/cm 2 , angle of incidence 54° (with respect to surface normal). The irradiated area was 100 × 100 μm 2 . For AES analysis the same electron gun was used for the excitation with a current density and energy of 1*10 16 electrons/cm 2 /s and 5 keV, respectively. The fluence during the recording of the Auger spectra is about 300 times less than that used for irradiation.
The Auger spectra, N(E), were recorded in counting mode. The recorded spectrum was numerically differentiated, N′(E), for performing the concentration calculation.
The Auger signals of Al KLL , Al LVV, Si KLL , Si LVV, C and O, were measured; by measuring the high (KLL) and low (LVV) escape depths Auger electrons the quality of evaluation of the depth distributions in AES depth profiling considerably improves. The energies of the Al and Si Auger electrons strongly depend on their chemical environment; the corresponding Auger electron energies (in eV) are shown in Table 1.
Therefore it is easy to determine the metal and oxide fractions of the elements by measuring either the LVV or KLL Auger electrons allowing the determination of the depth distributions for the metal and oxide components separately. AES depth profiling. A low energy ion gun of Technoorg Linda was used for AES depth profiling. The parameters of the ion bombardment used were: energy 1 keV, projectile Ar, angle of incidence (with respect to the surface normal) 80° and specimen rotation during ion bombardment. The ion beam was scanned in an area of 1.5 × 1.5 mm 2 . Using these parameters, the ion bombardment induced roughening and mixing is minimal 32 .
Determination of the concentration distribution from AES spectra. In AES, generally simple expressions are applied for the evaluation of the composition using the measured peak-to-peak amplitudes of the differentiated, N′(E), curve 33 . This expression assumes that the composition within the escape depth of signal electrons is homogeneous; in any other case it cannot be applied. This is however the situation presently since the thickness of the Al oxide layer is only 5 nm, while the inelastic mean free path (IMFP) of the Al KLL and Si KLL Auger electrons in Al 2 O 3 are 3.2 and 3.6 nm, respectively 34 . We used a trial and error approach to determine the composition distribution of our sample 19,20 . The essence of the method is that we assume a composition distribution along the depth and calculate the Auger intensities assuming that the transport of electrons can be described by the exponential attenuation law, not considering the elastic scattering. (Neglecting the elastic scattering creates an error in the range of 10-15%, which will not affect the description of the phenomena.) The composition distributions are varied until the simulated depth profile is close enough to the measured one. If an element emits high (high IMFP) and low energy (low IMFP) Auger electrons, as in the present case, the accuracy of the method is rather good.
In case of depth profiling the above procedure is repeated for all spectra obtained after each sputtering steps assuming that the ion bombardment used for removing the material does not cause serious changes to the material. This is a reasonable assumption since the removed layer thickness is less than 8 nm and all alterations scale with the removed layer thickness 32 .
Spectroscopic ellipsometry (SE). Auger depth profiling uses ions and electrons to reveal the depth distribution of the SiO 2 . Both projectiles may initiate the formation of SiO 2 . Though it will be shown that these are low probability processes, still we have applied SE, a non-destructive method, to verify the presence of the SiO 2 produced by electron irradiation. The SE measurements have been carried out by a Woollam M-2000DI rotating compensator ellipsometer at an angle of incidence of 55°. The microspot option was used to focus the light into a spot with a diameter of approximately 0.2 mm. The surface around the irradiated region was mapped with steps smaller than the spot size, in order to precisely locate the modified spot on the sample surface, and to measure only the irradiated region of the sample.