Controlled thin film etching is essential for future semiconductor devices, especially with complex high aspect ratio structures. Therefore, self-limiting atomic layer etching processes are of great interest to the semiconductor industry. In this work, a process for atomic layer etching of aluminum oxide (Al2O3) films using sequential and self-limiting thermal reactions with trimethylaluminum and hydrogen fluoride as reactants was demonstrated. The Al2O3 films were grown by atomic layer deposition using trimethylaluminum and water. The cycle-by-cycle etching was monitored throughout the entire atomic layer etching process time using in situ and in real-time spectroscopic ellipsometry. The studies revealed that the sequential surface reactions were self-limiting versus reactant exposure. Spectroscopic ellipsometry analysis also confirmed the linear removal of Al2O3. Various process pressures ranging from 50 to 200 Pa were employed for Al2O3 etching. The Al2O3 etch rates increased with process pressures: Al2O3 etch rates of 0.92, 1.14, 1.22, and 1.31 Å/cycle were obtained at 300 °C for process pressures of 50, 100, 150, and 200 Pa, respectively. The Al2O3 etch rates increased with the temperature from 0.55 Å/cycle at 250 °C to 1.38 Å/cycle at 350 °C. Furthermore, this paper examined the temperature dependence of the rivalry between the removal (Al2O3 etching) and growth (AlF3 deposition) processes using the reactants trimethylaluminum and hydrogen fluoride. The authors determined that 225 °C is the transition temperature between AlF3 atomic layer deposition and Al2O3 atomic layer etching. The high sensitivity of in vacuo x-ray photoelectron spectroscopy allowed the investigation of the interface reactions for a single etching pulse as well as the initial etch mechanism. The x-ray photoelectron spectroscopy measurements indicated that the fluorinated layer is not completely removed after each trimethylaluminum exposure. The Al2O3 atomic layer etching process mechanism may also be applicable to etch other materials such as HfO2.

Atomic layer processes such as atomic layer deposition (ALD) and atomic layer etching (ALE) are needed for future semiconductor manufacturing processes due to the continuous miniaturization of electronic devices. Atomic layer etching, the reverse of ALD, is a technique that can remove precisely material with atomic layer control based on sequential, self-limiting surface reactions.1,2 Therefore, ALE is considered to be one of the most promising techniques to achieve low process variability at the atomic scale.3–5 The basic ALE concept starts with a surface modification step followed by a material removal step. Until recently, most of reported ALE processes are based on halogenation reactions to modify the surface, followed by energetic ion or noble gas atom bombardment to desorb halogen compounds and, therefore, to remove the material.6 This approach has been reported for a variety of materials, such as Si,7,8 Ge,4,9 compound semiconductors,10,11 and metal oxides,12 as well as various carbon films.13,14 However, there are currently two different approaches. An alternative approach to the usage of ion-enhanced or energetic noble gas atom-enhanced surface reactions is the use of thermally activated reactions to remove the modified layer.15 This method, termed “thermal ALE,” enables isotropic etching of materials even at high aspect ratios. The chemistry of thermal ALE is based on the fluorination and ligand-exchange transmetalation reactions. There has been considerable work by George et al. studying thermal ALE processes for etching metal oxides—including Al2O3, HfO2, and ZrO2—using hydrogen fluoride (HF) in combination with different metal precursors as the reactants.15–21 George et al. investigated the mechanism of thermal ALE by principal means of in situ quartz crystal microbalance (QCM) and Fourier transform infrared (FTIR) spectroscopy measurements. In this paper, the thermal ALE of Al2O3 using trimethylaluminum (TMA) as the metal precursor for ligand exchange and HF as the fluorination reactant was examined using in situ spectroscopic ellipsometry (SE) and in vacuo x-ray photoelectron spectroscopy (XPS) measurements. Extensive analysis of etch rates and individual surface reaction mechanism of Al2O3 ALE using these measurement techniques have not been made so far. Ellipsometry has widely served as a powerful noninvasive method for monitoring time-continuous and time-discrete atomic scale processes in situ and even in real time.22–26 Our group have used in situ ellipsometry in the past to investigate several ALD processes.27,28 While ALE modifies the chemical composition of the surface through a material removal, SE can observe these changes in the electronic structure of the surface in situ and in real time. In this work, a detailed study of the impact of process parameters on the etching behavior is presented. Furthermore, we examined the initial stage etch of Al2O3 thin films using in vacuo XPS measurements. The high sensitivity of XPS allowed investigations of interface reactions even for a single ALE pulse. The combination of SE and XPS studies leads to a better understanding of the reaction mechanism during thermal Al2O3 ALE.

The experiments were performed in an FHR-300-ALD reactor (FHR Anlagenbau, Ottendorf-Okrilla, Germany), which allows both ALD and ALE processes. The reactor used has been described previously by Schmidt et al.29 For in situ Al2O3 ALE monitoring, the tool was equipped with a novel rotating compensator ellipsometer M 2000 FI (J. A. Woollam. Co., Inc.), which was mounted under a fixed angle of incidence of 64°. The ellipsometer windows were kept free from depositions by argon (Ar) purging (no shutters are implemented). The accumulated spectra were modeled using CompleteEASE analysis software. The model layers comprised an Si substrate, a fixed layer of 1.6 nm native SiO2, and the Al2O3 film. The optical model of Al2O3 used in SE fitting is the Cauchy model.30,31 The fit parameters of the Cauchy model were determined from the in situ spectrum recorded after 50 ALD cycles. The optical properties were varied respectively for the different process temperatures. Applying the appropriate optical layer stack model, each ellipsometric spectrum was translated into an Al2O3 optical layer thickness. The reactor chamber was also attached to an ultra-high vacuum (UHV) analytic tool (Omicron Nanotechnology GmbH, Taunusstein, Germany) for direct surface analysis with XPS without vacuum break. The combination of an ALE process chamber with a non-destructive analytic tool enables the determination of the chemical film composition during the ALE process under quasi-in situ conditions. Therefore, the surface chemical reactions for a single reactant pulse are observable with submonolayer resolution. The present XPS investigations were operated with a nonmonochromatic Al-Kα x-ray source of 1486.6 eV. The photoelectrons were detected according to their kinetic energy using a hemispherical analyzer at a constant pass energy of 30 eV and an electron emission angle of 45°. The evaluation and synthesis of the acquired x-ray photoelectron spectra were performed with the XPS analysis software unifit (2011).32 Silicon wafers covered with native oxide (denoted nOx in the following) were used as starting substrates. The Al2O3 films were grown on nOx by ALD using TMA and H2O (de-ionized water) as the reactants immediately before the etching process.33 The deposition and etching processes were performed in the same reactor, without air exposure in between. For the measurements in the UHV analytic tool, the substrate wafers were cut into samples with dimensions of 1 by 1 cm. The Al2O3 ALE reactions were accomplished using sequential static exposures of TMA and HF as the reactants. TMA was delivered from a bubbler at 15 °C and introduced into the reactor chamber by using an Ar carrier gas with a flow rate of 75 sccm. HF was derived from HF-pyridine (70% HF, Sigma-Aldrich), which is liquid at room temperature and is a convenient reservoir for anhydrous HF. In situ quadrupole mass spectrometer measurements showed that HF dominates the vapor pressure of HF-pyridine (not shown here). The bubbler containing the HF-pyridine was passivated with fluorine to prevent HF reaction with stainless steel. The HF-pyridine precursor was held at room temperature. An argon purge pulse to remove excess reactant and reaction by-products from the process chamber separated each reactant exposure step.

The ALE of Al2O3 with TMA and HF was assessed with in situ SE analysis, where the layer to be etched is deposited in the reactor via ALD immediately before the etching experiments. Deposition and etching were performed in the same reactor at the same process temperature without air exposure in between. Figure 1 shows the optical Al2O3 layer thickness as determined by in situ SE during 40 cycles of ALD followed by 30 cycles of ALE at 300 °C and 100 Pa.

FIG. 1.

Al2O3 optical layer thickness in the real-time course of 40 cycles of Al2O3 ALD followed immediately by 30 cycles of Al2O3 ALE. HF only experiments were also performed.

FIG. 1.

Al2O3 optical layer thickness in the real-time course of 40 cycles of Al2O3 ALD followed immediately by 30 cycles of Al2O3 ALE. HF only experiments were also performed.

Close modal

One of the basic features of an ALE process is the self-limiting behavior of alternating chemical surface reactions. Consequently, the film thickness decrements after each ALE cycle by a fixed amount, the so-called

The EPC value is an important characteristic for an ALE process. Consequently, the EPC could be studied for varying process parameters, assuming that the film properties are unaffected by changes in the process conditions.

Figure 1 reveals that the etching of the Al2O3 film is linear with the number of ALE cycles. The EPC for Al2O3 ALE at 300 °C is 1.14 Å/cycle. Control experiments were also performed using only HF exposures without TMA pulses. The results in Fig. 1 conclude that HF exposures by themselves cannot etch Al2O3 films at 300 °C. TMA needs to be present together with HF to remove Al2O3 material.

The application of SE measurements enabled the investigation of etch characteristics as a function of process parameter adjustments. The following process parameters were varied to define the characteristic dependences: substrate temperature, process pressure, and pulse time. An individual sample was prepared for each ALE process. After the nOx substrate surface was coated with 100 cycles of Al2O3 ALD cycles using TMA and H2O as the reactants, the Al2O3 sample was exposed to 50 ALE cycles. The etching process was carried out modifying just one process parameter.

Al2O3 ALE was observed to be dependent on the process temperature. Figure 2(a) displays the Al2O3 optical layer thickness change versus the number of ALE cycles for temperatures between 200 and 350 °C. One ALE cycle consisted of a 7 s dose of TMA, followed by a 20 s Ar purge and a 7 s dose of HF immediately followed by a 20 s dose of Ar purge. This timing sequence is designated as (7–20–7–20).

FIG. 2.

(a) Al2O3 optical layer thickness in the real-time course of 50 ALE cycles and (b) Al2O3 etch rate for different process temperatures.

FIG. 2.

(a) Al2O3 optical layer thickness in the real-time course of 50 ALE cycles and (b) Al2O3 etch rate for different process temperatures.

Close modal

In situ SE measurements observed linear thickness loss versus cycle number during the etching process [Fig. 2(a)], even in the initial phase of the etching process. The EPC values calculated from the in situ SE measurements are shown in Fig. 2(b).

The etch rate is 0.55 Å/cycle at 250 °C. A temperature increase of 100 °C leads to a significant EPC increase to 1.38 Å/cycle. These results indicate that the etch rate can be controlled by the process temperature. Lee et al. examined the surface reactions during thermal Al2O3 ALE using QCMs.16 They revealed that higher temperatures produce both a larger mass loss for the removal of AlF3 by TMA and a larger mass gain for the subsequent fluorination of Al2O3 by HF. The percentage of AlF3 removed by the TMA exposures increased from 19% at 250 °C to 74% at 325 °C.

At a process temperature of 225 °C, the Al2O3 etch rate remains near zero (−0.05 Å/cycle). Below 225 °C, the material was deposited; thus, the EPC is presented as a negative value in Fig. 2. Material deposited at 200 °C with TMA and HF as reactants was examined in detail by in vacuo XPS measurements. The film composition shown in Fig. 3 confirms that sequential surface reactions with TMA and HF lead to AlF3 ALD at lower process temperatures. The ratio of F/Al was found to be 3.

FIG. 3.

Composition of an AlF3 ALD film deposited at 200 °C.

FIG. 3.

Composition of an AlF3 ALD film deposited at 200 °C.

Close modal

The increase in film thickness at process temperatures below ∼225 °C is produced by AlF3 growth. Previous studies have shown that the ALD of AlF3 can also be accomplished using TMA and HF as reactants.34 DuMont et al. investigated the relationship between AlF3 ALD and Al2O3 ALE using in situ FTIR vibrational spectroscopy measurements. They reported a dependence between substrate temperature and the coverage of AlCH* and HF* surface species that dictate whether conditions favor Al2O3 ALE or AlF3 ALD.35 During first HF exposure, HF may react with the AlCH3* surface species on the initial hydroxylated Al2O3 substrate to form AlF* surface species in addition to the fluorination of Al2O3 to form an AlF3 or AlFxOy surface layer. The subsequent exposure of TMA causes AlF(CH3)2 products through the ligand-exchange reaction. At temperatures below ∼225 °C, AlF(CH3)2 may remain on the surface and lead to AlF3 growth during the HF exposure. At higher temperatures, AlF(CH3)2 may desorb on the surface and yield AlF3 etching during the TMA exposure. With increasing the temperature, the TMA ligand-exchange reactions with AlF3 are more efficient. The transition between AlF3 ALD at lower temperatures and Al2O3 ALE at higher temperatures is defined by the transition temperature. The transition temperature between growth and removal processes occurred at approximately 225 °C. The thickness change remained close to zero (−0.05 Å/cycle). Above the transition temperature, the film thickness decreased due to Al2O3 etching.

The Al2O3 optical layer thickness change versus number of ALE cycles for process pressures between 50 and 200 Pa at 300 °C is depicted in Fig. 4(a). The Al2O3 etching is linear at all process pressures. The larger Al2O3 etch rates observed at higher process pressures suggest a pressure-dependent reaction mechanism for Al2O3 ALE. For instance, the EPC increases from 0.92 Å/cycle at 50 Pa to 1.31 Å/cycle at 200 Pa at a process temperature of 300 °C.

FIG. 4.

(a) Al2O3 optical layer thickness in the real-time course of 50 ALE cycles for different process pressures at 300 °C, and (b) Al2O3 etch rate vs pressure at 200, 250, 300, and 350 °C.

FIG. 4.

(a) Al2O3 optical layer thickness in the real-time course of 50 ALE cycles for different process pressures at 300 °C, and (b) Al2O3 etch rate vs pressure at 200, 250, 300, and 350 °C.

Close modal

These results indicate that the Al2O3 etch rate can also be controlled by the process pressure. Cano et al. recently explored the effects of HF pressure on the Al2O3 etching rate and Al2O3 fluorination reaction. The etch rate and the fluoride thickness were observed to increase with HF pressure and then level out at higher HF pressures.36 An increase of the fluorination layer thickness results in more etching during the ligand-exchange reaction with TMA. DuMont et al. also reported the dependence of the reactant pressure on the etch rate for the ALE of SiO2 using TMA and HF.35 

Figure 4(b) displays the EPC versus the process pressure at different process temperatures. At 200 °C, the process pressure has no effect on the AlF3 ALD growth per cycle. At higher temperatures, the influence of the process pressure is more significant. At 300 as well as at 250 °C, the EPC increases linearly with the process pressure.

The self-limiting behavior of the TMA and HF reactions was determined by monitoring the film thickness change as a function of exposure time for each pulse of the ALE cycle. The EPC was examined by varying one pulse time at constant process conditions. The results are summarized in Fig. 5. For each combination, an individual Al2O3 film was grown by 50 Al2O3 ALD cycles. Subsequently, 30 cycles of Al2O3 ALE were performed before switching to a different ALE pulse time. The EPC values were calculated versus different reactant exposure and purge pulse times. This approach allowed testing of numerous parameter variations and thus a fast and extensive process development.

FIG. 5.

EPC as a function of (a) HF and TMA pulse time and (b) Ar purge pulse time.

FIG. 5.

EPC as a function of (a) HF and TMA pulse time and (b) Ar purge pulse time.

Close modal

Figure 5(a) examines the self-limiting behavior of the TMA reaction for a 7 s exposure of HF at 300 °C and 100 Pa. A constant Ar purge of 20 s was used after each reactant exposure. This reaction sequence can be denoted as x–20–7–20. Varying the TMA pulsing time, the EPC increases and reaches a satiation value after a TMA exposure of 5 s. Therefore, a TMA pulse length equal to 7 s is needed in order to achieve the surface saturation.

Figure 5(a) also examines the self-limiting nature of the HF surface reaction using different HF exposure times with a 7 s exposure of TMA. This reaction sequence can be denoted as 7–20–x–20. The EPC versus HF pulse time increases and saturates after an HF exposure time of 5 s. HF fluorinates the aluminum oxide and forms an aluminum fluoride layer on the surface. The fluoride layer acts as a diffusion barrier and leads to a self-limiting fluorination reaction. A similar behavior is observed for silicon oxidation, which is described by Deal–Grove kinetics.37 Additionally, even an HF exposure of 1 s ensures an EPC of 0.76 Å/cycle. The self-limiting behavior of the surface reactions is ensured at a reactant exposure time of 7 s for both TMA and HF.

Purge pulses are required to remove the reaction products and unadsorbed molecules in the ALE cycle. Figure 5(b) shows the calculated EPC values for different Ar purge times using a 7 s exposure of TMA and HF, respectively. In either case, the EPC value versus the Ar purge time increases and reaches a saturation after 5 s for purge after the TMA pulse, and after 10 s for the purge after the HF pulse. Therefore, long purge pulses with argon are essential for the Al2O3 ALE process with TMA and HF.

During the thermal ALE reactions, HF fluorinates the aluminum oxide and forms an aluminum fluoride layer on the surface. The TMA precursor molecules accept fluorine and transfer their ligands to the aluminum fluoride in a ligand-exchange reaction.15In vacuo XPS analysis were used to characterize the fluorination of Al2O3 after the very first HF exposure. For these measurements, the initial Al2O3 film was sufficiently thin to allow detection of the underlying Si substrate signal. 20 Al2O3 ALD cycles (approximately 2 nm film thickness) on nOx were exposed to different HF exposure times at 300 °C and 100 Pa. Avoiding ambient contamination and oxidation in the analytic tool ensured constant surface properties during the period between an ALE pulse and the XPS analysis (during transfer between the process chamber and the analytic tool). After the first HF exposure, the sample was immediately transferred to the analytic tool without breaking the vacuum. An individual sample was prepared for each HF exposure time. Figure 6 shows the XPS intensity values of the F 1s peak after the first HF exposure as a function of the HF pulse time.

FIG. 6.

XPS intensity (integrated peak area after Shirley background subtraction and peak synthesis) of the F 1s peak after the first HF pulse as a function of the HF exposure time.

FIG. 6.

XPS intensity (integrated peak area after Shirley background subtraction and peak synthesis) of the F 1s peak after the first HF pulse as a function of the HF exposure time.

Close modal

Figure 6 reveals that the F 1s peak intensity increases with increasing HF exposure time and saturates after 5 s. The saturation behavior corresponds to the results of the SE measurements shown in Fig. 5. The SE and XPS studies reveal that the fluorination reaction is self-limited. The gain of fluorine is in agreement with the fluorination reaction to produce AlF3 or AlFxOy.18 

The Al 2p XPS spectrum of an initial 2 nm Al2O3 ALD film is shown in Fig. 7(a). The initial Al2O3 film grown on nOx is consistent with pristine Al2O3 with an Al 2p peak at 75.4 eV. The Al 2p peak after 7 s of HF exposure at 300 °C is represented in Fig. 7(b). The fluorination of Al2O3 leads to the growth of a subpeak at higher binding energies. The Al 2p XPS peak at 76 eV consistent with oxyaluminum fluoride species and the peak at 77.16 eV may be composed of AlF3.38,39 Oxygen is less electronegative than fluorine - therefore, the Al 2 p binding energy does not shift as much. These results correspond to the FTIR spectroscopy measurements of Al2O3 fluorination from Cano et al.36 They observed the fluorination of Al2O3 to AlF3 or AlOxFy by the reaction Al2O3 + 6HF → 2AlF3 + 3H2O or Al2O3 + zHF → 2AlO(6−z)/4Fz/2 + (z/2)H2O.

FIG. 7.

XPS intensity (after background subtraction and peak synthesis) of the Al 2p peak for (a) initial Al2O3 film and (b) Al2O3 film after HF exposure.

FIG. 7.

XPS intensity (after background subtraction and peak synthesis) of the Al 2p peak for (a) initial Al2O3 film and (b) Al2O3 film after HF exposure.

Close modal

Furthermore, Cano et al. suggested that the fluoride layer on the Al2O3 surface may act as a diffusion barrier to slow down the fluorination of the underlying Al2O3 bulk. In addition, they revealed that higher HF pressure increases the fluoride layer thickness and the fluorination occurs deeper into the Al2O3 bulk.36 In this work, XPS analysis was used to examine the dependence between process pressure and Al2O3 fluorination. Al2O3 ALD films (approximately 2 nm film thickness) on nOx were exposed to 7 s HF at process pressures of 50, 100, 150, and 200 Pa.

The Al 2p spectra after 7 s of HF exposure at different pressures are shown in Fig. 8. With increasing process pressure, the Al 2p peak shifts to higher binding energies and the peak width increases. The change in peak position as well as the change in peak width indicate that at higher process pressure, the formation of Al–F bonds increases and the formation of Al–O bonds decreases. At higher process pressures, the amount of AlF3 is higher than that of oxyaluminum fluoride (Al2OF4, AlOF) compositions. As a result of that, more material can be removed during the subsequent TMA exposure. Therefore, the etch rate is higher at higher process pressures. This XPS analysis corresponds to the results of the SE measurements shown in Fig. 4.

FIG. 8.

XPS intensity (after background subtraction and peak synthesis) of the Al 2p peak for Al2O3 film after HF exposure at different process pressures.

FIG. 8.

XPS intensity (after background subtraction and peak synthesis) of the Al 2p peak for Al2O3 film after HF exposure at different process pressures.

Close modal

In the following, we describe our findings on the Al2O3 ALE etch behavior investigated with XPS measurements. The etching process was done at 300 °C with 7 s HF, 20 s Ar purge, 7 s TMA, and 20 s Ar purge. In the present work, the thermal ALE etch behavior was investigated for initial Al2O3 films that were sufficiently thin to allow the detection of underlying Si substrate signals. 20 Al2O3 ALD cycles (approximately 2 nm film thickness) were deposited on silicon wafers without an oxide surface (HF:Si). In vacuo XPS measurements were done for various cycle numbers, starting from the very first half-reaction (0.5 cycles/precursor exposure). For each cycle number, an individual sample was prepared, which means that each cycle number represents an uninterrupted deposition experiment. Within the first four cycles, we carried out XPS measurements, starting from the very first half cycle equivalent to one precursor pulse and the following purge pulse. The ALE process started with an HF exposure. Figure 9(a) shows the atomic concentrations over the number of ALE cycles up to four cycles. As shown in Fig. 9(a), an alternating change in the atomic concentration can be observed. Higher fluorine and lower oxygen concentrations were measured after each HF pulse. The gain of fluorine and loss of oxygen are consistent with the fluorination reaction to produce AlFx or AlFxOy. The increase in fluorine atom concentration was highest after the first half-reaction. During the very first fluorination reaction, HF reacts with AlCH3* surface species to form fluorine-containing surface species. HF also reacts with the underlying Al2O3 film to form an AlFx layer. Ideally, the whole AlFx layer from Al2O3 fluorination should be removed during the TMA ligand-exchange reaction. However, the experiments indicate that the complete AlFx layer is not removed during the TMA half-reaction. Bounded fluorine atoms can be detected after every completed ALE cycle.

FIG. 9.

(a) Atomic concentrations over the number of ALE cycles up to four cycles and (b) plot of the XPS Al 2p signal (after background subtraction and peak synthesis) as a function of the cycle number.

FIG. 9.

(a) Atomic concentrations over the number of ALE cycles up to four cycles and (b) plot of the XPS Al 2p signal (after background subtraction and peak synthesis) as a function of the cycle number.

Close modal

After each TMA pulse, a higher carbon and aluminum concentration was achieved. The higher carbon concentration after the TMA pulse is attributed to the presence of methyl groups of precursor molecules. The carbon signal can be clearly attributed to the carbon concentration caused by the TMA pulse, since the analytic UHV tool is directly attached to the process tool.

Figure 9(b) depicts the Al 2p signal always measured after the TMA pulse as a function of the ALE cycle number up to four cycles. The initial Al2O3 ALD film is consistent with pristine Al2O3 with a peak at 75.5 eV. The whole Al 2p spectrum shifted toward higher binding energies after the first half-cycle, which can be assigned to Al–F bonds.39 Furthermore, the Al 2p peaks are slightly broader after every HF pulse at higher binding energies since there are more fluorine atoms bound to aluminum. Each TMA exposure results in a significantly higher intensity of aluminum. This behavior confirms that the ligand-exchange reaction removes the fluorinated surface layer only partially during the TMA exposure. During the ligand-exchange reaction, TMA reacts with the AlFx surface layer to form fluorine-containing species [e.g., AlF(CH3)2 and AlF2(CH3)*]. However, the TMA exposure is not able to remove all the HF* species and reform all the AlCH3* species. Fluorine-containing species remain on the surface. During the next fluorination reaction, HF reacts with AlCH3* surface species as well as the remaining fluorine-containing species. The formation of the Al 2p peaks and the decrease of the oxygen intensity after the HF exposures indicate that more AlFx is formed than AlFxOy.

The atomic concentration evolution for selected ALE cycle numbers up to 25 cycles at 300 °C is shown in Fig. 10(a). After five ALE cycles, the aluminum intensity decreases and the silicon intensity increases with increasing cycle number. Already after 20 cycles, the complete Al2O3 layer is etched. However, an aluminum concentration is detectable due to the adsorption of TMA molecules during the second half-reaction.

FIG. 10.

(a) Atomic concentrations over the number of ALE cycles for process temperature of 300 °C up to 25 cycles and (b) plot of the XPS Al 2p signal (after background subtraction and peak synthesis) as a function of the cycle number.

FIG. 10.

(a) Atomic concentrations over the number of ALE cycles for process temperature of 300 °C up to 25 cycles and (b) plot of the XPS Al 2p signal (after background subtraction and peak synthesis) as a function of the cycle number.

Close modal

The fluorine concentration remains at the same level up to ten ALE cycles before continuously decreasing. However, at 25 cycles, the fluorine concentration is still 3.5 at. %, which means that there is still some residual fluorine bound to the surface.

Figure 10(b) depicts the Al 2p signal always measured after the TMA pulse as a function of the cycle number up to 25 cycles. The Al 2p peak shifted toward higher binding energies after the very first ALE cycle due to the formation of Al–F bonds. The peak intensity is lower than of the starting material. The Al 2p peak intensity is slightly higher after the second ALE cycle and remains at this level until four cycles and then the intensity decreases continuously with increasing cycle number. After 20 ALE cycles, the complete Al2O3 layer is etched—the Al 2p peak intensities fall below the XPS detection limit.

In this work, in situ real-time spectroscopic ellipsometry and quasi-in situ x-ray photoelectron measurements were used to explore the sequential, self-limiting thermal reactions of Al2O3 atomic layer etching using trimethylaluminum and hydrogen fluoride as reactants. The sequential, thermal reactions of trimethylaluminum and hydrogen fluoride etched Al2O3 linearly with atomic level precision. In situ spectroscopic ellipsometry analysis confirmed the self-limiting behavior of the surface reactions versus reactant exposure. The Al2O3 etch rate was pressure and temperature dependent. The EPC values varied from 0.92 Å/cycle at 50 Pa to 1.31 Å/cycle at 200 Pa. At higher process pressures, the detected amount of AlF3 is higher than that of AlOF compositions. During the subsequent ligand-exchange reaction, more material can be removed. We also investigated the relationship between AlF3 deposition and Al2O3 etching using in situ ellipsometry measurements. The etch rate revealed a transition temperature between AlF3 deposition and Al2O3 etching at 225 °C. The etch rates varied from −0.4 Å/cycle at 200 °C to +1.38 Å/cycle at 350 °C. The increase in the etch rate results from more efficient ligand-exchange reactions with the fluorinated layer at higher temperatures. In vacuo x-ray photoelectron spectroscopy measurements were carried out to study the film etch behavior. The high sensitivity of these measurements allowed investigations of interface reactions for the corresponding half-reaction. The complete fluorinated surface layer is not removed during the ligand-exchange reaction. The measurements confirmed that the ligand-exchange reaction removes the fluorinated surface layer only partially during the TMA exposure. Bounded fluorine atoms could be detected after every completed cycle. In situ spectroscopic ellipsometry and quasi-in situ x-ray photoelectron spectroscopy measurements are ideally suited for developing atomic layer processes and studying thin film deposition or etching. The combination of these measurements enables a better characterization of etching and growth processes of thin films and gives a better understanding of results achieved with each method.

This work was supported by Sächsische Aufbaubank (SAB) in the joint research project “Atomlagenabscheidung und –ätzen (ALD, atomic layer deposition & ALE, atomic layer etching) (Untersuchungen von Anlagenkomponenten und Technologien für das ALD & ALE Verfahren)” (No. 100319818).

The authors have no conflicts to disclose.

Ethics approval is not required.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
A.
Agarwal
and
M. J.
Kushner
,
J. Vac. Sci. Technol. A
27
,
37
(
2009
).
2.
K. J.
Kanarik
,
T.
Lill
,
E. A.
Hudson
,
S.
Sriraman
,
S.
Tan
,
J.
Marks
and
R. A.
Gottscho
,
J. Vac. Sci. Technol. A
.
33
,
020802
(
2015
).
3.
M.
Leskelä
and
M.
Ritala
,
Angew. Chem. Int. Ed.
42
,
5548
(
2003
).
4.
N.
Marchack
and
J. P.
Chang
,
J. Phys. D: Appl. Phys.
44
,
174011
(
2011
).
5.
C. T.
Carver
,
J. J.
Plombon
,
P. E.
Romero
,
S.
Suri
,
T. A.
Tronic
, and
R. B.
Turkot
,
ECS J. Solid State Sci. Technol.
4
,
N5005
(
2015
).
6.
V. M.
Donnelly
and
A.
Kornblit
,
J. Vac. Sci. Technol. A
31
,
050825
(
2013
).
7.
J.
Yamamoto
,
T.
Kawasaki
,
H.
Sakaue
,
S.
Shingubara
, and
Y.
Horiike
,
Thin Solid Films
225
,
124
(
1993
).
8.
S. D.
Athavale
and
D. J.
Economou
,
J. Vac. Sci. Technol. B
14
,
3702
(
1996
).
9.
T.
Sugiyama
,
T.
Matsuura
, and
J.
Murota
,
Appl. Surf. Sci.
112
,
187
(
1997
).
10.
T.
Meguro
,
M.
Hamagaki
,
S.
Modaressi
,
T.
Hara
,
Y.
Aoyagi
,
M.
Ishii
, and
Y.
Yamamoto
,
Appl. Phys. Lett.
56
,
1552
(
1990
).
11.
W. S.
Lim
,
S. D.
Park
,
B. J.
Park
, and
G. Y.
Yeom
,
Surf. Coat. Technol.
202
,
5701
(
2008
).
12.
D.
Metzler
,
R. L.
Bruce
,
S.
Engelmann
,
E. A.
Joseph
, and
G. S.
Oehrlein
,
J. Vac. Sci. Technol. A
32
,
020603
(
2014
).
13.
Y. Y.
Kim
,
W. S.
Lim
,
J. B.
Park
, and
G. Y.
Yeom
,
J. Electrochem. Soc.
158
,
D710
(
2011
).
14.
15.
S. M.
George
and
Y.
Lee
,
ACS Nano
10
,
4889
(
2016
).
16.
Y.
Lee
and
S. M.
George
,
ACS Nano
9
,
2061
(
2015
).
17.
Y.
Lee
,
J. W.
DuMont
, and
S. M.
George
,
ECS J. Solid State Sci. Technol.
4
,
N5013
(
2015
).
18.
Y.
Lee
,
J. W.
DuMont
, and
S. M.
George
,
Chem. Mater.
27
,
3648
(
2015
).
19.
Y.
Lee
,
J. W.
DuMont
, and
S. M.
George
,
Chem. Mater.
28
,
2994
(
2016
).
20.
Y.
Lee
,
C.
Huffman
, and
S. M.
George
,
Chem. Mater.
28
,
7657
(
2016
).
21.
D. R.
Zywotko
and
S. M.
George
,
Chem. Mater.
29
,
1183
(
2017
).
22.
J. B.
Theeten
,
F.
Hottier
, and
J.
Hallais
,
J. Cryst. Growth
46
,
245
(
1979
).
23.
D. E.
Aspnes
,
W. E.
Quinn
, and
S.
Gregory
,
Appl. Phys. Lett.
56
,
2569
(
1990
).
24.
J. W.
Klaus
,
A. W.
Ott
,
J. M.
Johnson
, and
S. M.
George
,
Appl. Phys. Lett.
70
,
1092
(
1997
).
25.
U.
Kilic
,
A.
Mock
,
D.
Sekora
,
S.
Gilbert
,
S.
Valloppilly
,
G.
Melendez
, and
M.
Schubert
,
Sci. Rep.
10
,
1
(
2020
).
26.
C.
Liu
,
J.
Erdmann
, and
A.
Macrander
,
Thin Solid Films.
355–356
,
41
(
1999
).
27.
M.
Knaut
,
M.
Junige
,
M.
Albert
, and
J. W.
Bartha
,
J. Vac. Sci. Technol. A
30
,
01A151
(
2012
).
28.
M.
Junige
,
M
Geidel
,
M.
Knaut
,
M.
Albert
, and
J. W.
Bartha
, “Monitoring atomic layer deposition processes in situ and in real-time by spectroscopic ellipsometry,” in 2011 Semiconductor Conference Dresden, Dresden, Germany, 27–28 September 2011 (IEEE, Dresden, 2011), pp. 1–4.
29.
D.
Schmidt
,
S.
Strehle
,
M.
Albert
,
W.
Hentsch
, and
J. W.
Bartha
,
Microelectron. Eng.
85
,
527
(
2008
).
30.
J. A.
Woollam
,
B. D.
Johs
,
C. M.
Herzinger
,
J. N.
Hilfiker
,
R. A.
Synowicki
, and
C. L.
Bungay
,
Proc. SPIE
10294
,
1029402
(
1999
).
31.
H.
Fujiwara
,
Spectroscopic Ellipsometry: Principles and Applications
(
Wiley
,
New York
,
2007
).
32.
R.
Hesse
,
T.
Chassé
, and
R.
Szargan
,
Fresenius J. Anal. Chem.
365
,
48
(
1999
).
33.
R. L.
Puurunen
,
J. Appl. Phys.
97
,
121301
(
2005
).
34.
Y.
Lee
,
J. W.
DuMont
,
A. S.
Cavanagh
, and
S. M.
George
,
J. Phys. Chem. C
119
,
14185
(
2015
).
35.
J. W.
DuMont
,
A. E.
Marquardt
,
A. M.
Cano
, and
S. M.
George
,
ACS Appl. Mater. Interfaces
9
,
10296
(
2017
).
36.
A. M.
Cano
,
A. E.
Marquardt
,
J. W.
DuMont
, and
S. M.
George
,
J. Phys. Chem. C
123
,
10346
(
2019
).
37.
B. E.
Deal
and
A. S.
Grove
,
J. Appl. Phys.
36
,
3770
(
1965
).
38.
J.
Chastain
and
R. C.
King
, Jr.
,
Handbook of X-Ray Photoelectron Spectroscopy
(
Perkin-Elmer
, Minnesota,
1992
), p.
261
.
39.
K.
Roodenko
,
M. D.
Halls
,
Y.
Gogte
,
O.
Seitz
,
J.-F.
Veyan
, and
Y. J.
Chabal
,
J. Phys. Chem. C
115
,
21351
(
2011
).