Thermal laser evaporation of elemental metal sources in oxygen Open Access

Different metal oxides feature a wide variety of physical properties, and this enables a broad spectrum of applications of oxides in microelectronics,^1–4 high-power electronics,^5,6 photovoltaics,^7–9 catalysis,^10–13 and display technology.^14,15 Several deposition techniques are available for growing oxide films. In some oxide growth techniques, including oxide molecular beam epitaxy (MBE), chemical vapor deposition (CVD), and thermal laser epitaxy (TLE), metal sources are thermally evaporated in an oxidizing atmosphere, which causes their oxidation.

TLE is a novel technique for growing high-quality oxide thin films. TLE uses continuous-wave lasers for thermally evaporating source materials such as elemental metals,¹⁶ and it uses laser irradiation to heat the substrate.^17,18 The growth of numerous TLE-grown metal films,¹⁹ compound and oxide films,^20–23 and epitaxial metal oxide films²⁴ has been demonstrated.

In oxide TLE, metal sources laser-heated in an oxygen or ozone atmosphere become oxidized.²³ The source oxidation strongly affects the molecular fluxes from the sources because of the differences in the volatilities of the source metals and their oxides. We have observed either an increase or a decrease in deposition rate with increasing oxygen/ozone supply, depending on the source metals.²³ An increase in the deposition rate of metal sources (measured, e.g., for Ti, Co, Fe, Hf, Mo, and Nb) under oxygen atmospheres is typically found if a metal is less volatile than its oxides. The opposite behavior is observed when an oxidized source has a lower volatility than the corresponding metal source (measured, e.g., for Ni and Cu).

Similar observations have been reported in oxide MBE. The formation and evaporation of volatile suboxides in the effusion cell increases the cation flux, causing high cation incorporation in the MBE growth of Ga₂O₃, Sn-doped Ga₂O₃, Si-doped Ga₂O₃, and In₂O₃.^25,26 In contrast, a Sr flux is lowered in oxygen atmospheres due to the formation of SrO, which is less volatile than Sr.²⁷ 

Hoffmann et al. developed a kinetic model that quantitatively describes the formation and evaporation of suboxides from metallic sources in oxide MBE.²⁵ This model anticipates source fluxes in the low-pressure MBE growth of Ga₂O₃ and In₂O₃. By comparing the vapor pressures of metals and their suboxides, those authors also identified which metal sources are susceptible to the formation and evaporation of suboxides.

TLE expands the process parameter space of oxide growth by featuring very high source temperatures T_source and evaporation in many different atmospheres, including ultrahigh vacuum and highly reactive atmospheres.¹⁶ For example, both molecular oxygen and a mixture of molecular oxygen and ozone can be used as process gases for a wide range of pressures $POX$. Therefore, TLE allows the study of thermal evaporation of metals in a significantly wider range of T_source and $POX$ conditions than is possible with alternative film growth techniques. We report here that several evaporation modes can emerge at high $POX$ or high T_source.

In this article, we present and discuss the thermal evaporation of laser-heated metal sources in oxidizing atmospheres with an emphasis on the technologically interesting elements Si and Al. The passive or active oxidation of the Si source varies depending on the oxidation conditions²⁸ and this affects the deposition rate. We find that the formation of volatile SiO greatly increases the deposition rate. However, the formation of less volatile SiO₂ decreases the deposition rate at high $POX$⁠.

In contrast to Si, the deposition rate of Al is found to decrease in the presence of oxygen due to the formation of less volatile AlO_x. With increasing laser power P_L, the deposition rate of Al becomes identical to the rate without oxygen because the majority of the vapor then becomes metallic Al. The absence of a $POX$-dependence of the deposition rate at elevated T_source is attributed to the source oxidation rate being lower than the metal evaporation rate, which is caused by the finitude of the oxygen supply.

Elements, such as Ti, Mo, Hf, Fe, Co, and Nb, show that the $POX$-dependence of the deposition rate follows a power-law relationship. We find that this power-law relation holds when the evaporation of the oxidized source metal dominates the evaporation, the exponent of the power law characterizing this dominance.

High-purity metals were thermally evaporated by continuous-wave infrared laser irradiation with wavelengths of 1030 or 1060 nm. We have found no noticeable difference in the results obtained from those two wavelengths. Cylindrical Si, Ti, Mo, Hf, Fe, Co, and Nb sources were placed without crucibles in Ta-based holders. The laser irradiation of an elliptical area of ∼1 mm² melts the source locally, allowing the remaining solid source to act as a support for the melt. For Al, pellets were placed in an alumina crucible. Substrates were suspended horizontally above the source at a working distance of 60 mm.

High-purity (>6 N) oxygen gas was injected into the chamber via a manual leak valve and a cooled gas injector. An oxygen–ozone mixture was used to achieve a more reactive environment. A glow-discharge continuous-flow ozone generator provided an oxygen–ozone mixture with an ozone content of ∼10 wt. % of the total gas flow. We controlled the background gas pressure from the vacuum base pressure of ∼10⁻⁸–10⁻² hPa by adjusting the gas flow and the rotation frequency of the turbomolecular pumps. Details on the oxide TLE chamber and the growth can be found in Refs. 23 and 24.

Film thicknesses were measured either by cross-sectional scanning electron microscopy or profilometry. The deposition rate was calculated by dividing the film thickness by the deposition time. The deposition time of Si, Ti, Fe, Co, and Nb was 15 and 20 min for Hf and Mo. The deposition times for Al varied from 5 s to 22 h 35 min 33 s at the highest and lowest deposition rates, respectively. The deposition rates listed below are, therefore, time-averaged values. Grazing incidence x-ray diffraction (GI-XRD) was used to structurally characterize the deposited films. GI-XRD patterns of TLE-grown oxide films can also be found in Ref. 23.

We evaporated cylindrical Si sources with a diameter of 12 mm and a height of 8 mm using various values of $POX$ and P_L. Figures 1(a)–1(d) show photographs of SiO_x thin films deposited on 5 × 5 × 1 mm³ single-side polished c-plane sapphire substrates at a $POX$ of 10⁻⁵, 10⁻⁴, 10⁻³, and 10⁻² hPa with 350 W laser irradiation, respectively. With increasing $POX$⁠, the color of the grown films changes from brownish yellow to pale yellow until the film finally becomes transparent. Amorphous SiO has a characteristic brownish-yellow color,²⁹ whereas amorphous SiO₂ is white or colorless. The color change suggests that oxygen-deficient SiO_x (x < 2) forms at low $POX$ conditions. Optically transparent SiO_x (x ≈ 2) films are obtained at $POX$ > 10⁻³ hPa. This is consistent with the GI-XRD patterns shown in Fig. 1(e). The films grown at $POX$ 10⁻³ and 10⁻² hPa have a broad peak at 2θ ∼ 23°, which is assigned to the SiO₂ phase.³⁰ 

Photographs of the Si sources taken after the depositions at $POX$ of 10⁻³ and 10⁻² hPa are shown in Figs. 1(f) and 1(g), respectively. The sources consist of three regions: the melt-pool formed during evaporation, the area surrounding the melt-pool, and the remaining outer area. In the melt-pool region, a smooth Si surface has formed due to melting and recrystallization. A bare Si-like surface is observed in the area around the melt-pool. These two regions are present on all sources. The remaining area shows different surface properties, depending on the evaporation conditions. A brown layer, probably oxygen-deficient SiO_x (x ∼ 1), covers nearly the entire remaining area of the source evaporated at a P of 10⁻³ hPa as shown in Fig. 1(f). When operating the source at P = 10⁻² hPa, optical interference fringes, likely from a transparent SiO_x (x ∼ 2) layer, appear in the remaining area as seen in Fig. 1(g). EDX measurements in these areas confirm the presence of only Si and O.

Figure 2 summarizes the measured deposition rate of Si with varying $POX$ and P_L. The deposition rate of Si increases as $POX$ increases from vacuum base pressure (∼10⁻⁸ hPa) to 10⁻³ hPa for all P_L used, which may be due to the formation and evaporation of volatile SiO from the source.²⁶ However, for P_L ≤ 310 W, the deposition rate decreases at higher $POX$ for 10⁻³ hPa ≤ $POX$ ≤ 10⁻² hPa. Interestingly, it continues to increase for P_L = 350 W. The formation and evaporation of volatile SiO alone cannot account for this observation. Instead, the formation of SiO₂ under high $POX$ conditions is probably responsible for the decrease of the deposition rate as Kalarickal et al. proposed.²⁶ The estimated Si cation flux on the substrate at 350 W is presented in the supplementary material and compared with the maximum SiO flux.

Figure 3(a) shows the vapor pressures of Si, SiO, and SiO₂ as a function of temperature. The higher volatility of SiO compared to that of Si is responsible for the increase of the Si evaporation rate with oxygen supply (Fig. 2). The reported SiO₂ vapor pressures differ significantly in the literature, probably because the vapor pressure is sensitive to the oxygen composition of the material used in the study. Here, we present data from Refs. 31 and 32. The SiO₂ vapor pressure is much smaller than SiO. Therefore, the evaporation rate of Si is expected to decrease once SiO₂ starts to form on the source.

Figure 3(b) shows the phase diagram for the Si oxidation reactions from Ref. 31 displaying two oxidation reactions: (1) formation of SiO (⁠$Si(s)+1/2O2(g)→SiO(g)$⁠), called active oxidation, and (2) formation of SiO₂ (⁠$Si(s)+O2(g)→SiO2(s)$⁠), called passive oxidation, where s and g denote solid and gaseous phases, respectively.^33,34 The hatched area in Fig. 3(b) shows the $POX$ range used in the present study and the corresponding temperature range where passive oxidation occurs. In region R1 [see Figs. 1(f) and 1(g)], the source temperature is above the Si melting temperature (1683 K). The strong temperature gradient along the source¹⁹ enables passive oxidation in R3 where the temperature drops below the transition temperature. At $POX$ ∼ 10⁻² hPa, passive oxidation occurs even at ∼1330 K. This oxidation decreases the deposition rate of Si at $POX$ ∼ 10⁻² hPa in Fig. 2 and yields the optical interference fringes on the source seen in Fig. 1(g). The decrease of Si flux due to passive oxidation was also observed in Ref. 26.

According to the oxidation phase diagram of Si, increasing the laser power, i.e., increasing T_source, avoids the passive oxidation regime. Therefore, the deposition rate continues to increase with higher P_L, even at high $POX$⁠, by sustaining the dominant formation and evaporation of SiO. This understanding agrees with the increase of the Si deposition rate at a P_L of 350 W shown in Fig. 2.

Silicon evaporation can, therefore, serve as a study case of how the formation of a less volatile oxide under oxidizing conditions affects the source fluxes in oxide TLE. This finding is also relevant for oxide MBE and oxide CVD.

Aluminum oxide thin films are versatile dielectric films owing to their high dielectric strength,³⁵ large electron band gap,³⁶ high corrosion resistance,³⁷ and thermal stability.³⁸ Aluminum oxide films can be grown by evaporating Al in an oxidizing atmosphere. Here, we compare the deposition rates of Al in oxygen (⁠$POX$ of 10⁻³ hPa) and at vacuum base pressure (∼10⁻⁸ hPa) as a function of P_L.

The literature values for the vapor pressure of Al, Al₂O, and Al₂O₃ are shown in Fig. 4(a). In contrast to Si and its oxides, even the aluminum suboxide Al₂O has a lower vapor pressure than Al under typical deposition conditions (T_source < 1300 °C for which deposition rates are typically ≤20 Ml/s). Under these conditions, the vapor pressure of Al₂O₃ is approximately four orders of magnitude lower than that of Al. Therefore, a decrease of the deposition rate is to be expected if evaporation is performed in an oxidizing atmosphere.

Figure 4(b) presents the measured deposition rate of Al as a function of inverse incident P_L at $POX$ of 10⁻³ hPa and vacuum base pressure. In the absence of oxygen, the deposition rate has an Arrhenius-type dependence on P_L, which is consistent with previous reports.¹⁹ Photographs of Al sources driven with 150 and 300 W laser illumination under base pressure are shown in Figs. 4(c) and 4(d), respectively. In both cases, the source has been completely molten. Given the high thermal conductivity of Al, one must assume that the entire exposed source surface contributes to the evaporation. Under standard evaporation conditions, the effective evaporation area of the Al source is, thus, constant, independent of P_L. The measured Arrhenius-type dependence of the deposition rate as a function of P_L is, therefore, considered to be caused entirely by the increase of T_source with increasing P_L. The relation of T_source and P_L is given in the supplementary material. Owing to the strong variation of the deposition rate on the P_L, we can experimentally follow this dependence over more than five orders of magnitude.

For Al evaporation in an oxidizing atmosphere, the deposition rate behaves in a more complex way. At lower power (P_L < 220 W), the deposition rate shows a slower increase with power in oxygen compared to that at vacuum base pressure, which is to be expected due to the lower volatility of the suboxide. Figure 4(e) shows a photograph of the Al source being evaporated at P_L of 150 W in a $POX$ of 10⁻³ hPa. The source forms a sphere, which is covered with a crust of AlO_x. Liquid Al source material is visible through the pore where some AlO_x crust has been removed by the evaporation process. The fact that the AlO_x crust is brighter than the liquid Al [Fig. 4(e)] indicates that the emissivity of AlO_x exceeds that of Al. When the AlO_x crust forms in oxygen atmospheres, the energy loss through radiation increases and the temperature is lower than without the crust formation. Therefore, this behavior is also responsible for the decreased slope of the Al deposition rate with oxygen supply at lower power.

AlO_x vapor and Al vapor evaporate from the AlO_x crust and the open pore, respectively. The pore size increases with increasing P_L, which affects the amount of Al vapor from the source and the slope of the deposition rate curve. As the evaporation through the pore is directed mostly along the normal to the pore area, a pronounced angular dependence of the evaporation through the pore is to be expected.

For P_L > 220 W, however, the deposition rates of Al both with and without oxygen supply behave the same as a function of the P_L. This implies that the Al vapor is dominant even at this oxygen pressure and that the deposition rate no longer depends on $POX$⁠. The Al sources illuminated with >220 W in oxygen look and behave like an Al source heated without oxygen supply [see Fig. 4(f)]. The AlO_x crust has disappeared, and the entire source has formed a liquid, in agreement with the elemental Al flux emitted by the source.

The dominance of the Al flux under these conditions can be understood by considering the following two aspects: (1) kinetically, the oxygen supply limits the oxidation rate and thereby the suboxide flux at high temperatures, and (2) thermodynamically, oxidation itself is less favorable at high temperatures. An insufficient oxygen gas supply can limit the formation and evaporation of AlO_x at high P_L, whereas the evaporation of Al continues to increase exponentially with P_L. Consequently, in an oxidizing environment, Al vapor can become the dominant flux from a hot source. Indeed, any metal can evaporate or sublimate as metal vapor, even in an oxidizing atmosphere if its oxidation reaction rate is limited by high temperatures or an insufficient oxygen supply. This observation offers a basis on which to use TLE to control the vapor species over a wide range of $POX$⁠.

We observed that the deposition rate of several other metal sources (e.g., Ti, Mo, Hf, Fe, Co, and Nb) increases with the introduction of oxygen–ozone mixed gas at a given P_L (Fig. 5). Their deposition rates and $POX$ form a power-law relationship (⁠$depositionrate∝POXn$⁠). The colored lines in Fig. 5 are power-law fits to the measured data. The exponent in the equation denotes the reaction order. The empirical rate of metal oxidation or combustion follows the power-law relationship with oxygen pressure. The metal oxidation rate constant in a parabolic regime k_p can be expressed by $kp=kp°⋅PO2n$⁠.⁴⁰ The simplest char combustion model also takes the same nth order form.⁴¹ Therefore, the observed power-law relationship of the deposition rate implies that the evaporation rate is governed by the oxidation rate. This conclusion agrees with suboxide dominated source evaporation in MBE, which also has a power-law dependence on oxygen pressure.^25,26

The presence of a volatile oxide is a necessary condition to obtain such a power-law relationship. Si, Mo, and Nb have volatile oxide phases.²⁵ Our observations indicate that the other elements Hf, Fe, and Co also have volatile oxides. Ti is an especially interesting case. Neither TiO nor TiO₂ have higher vapor pressures than Ti.^25,42 Nevertheless, the T_source required for a vapor pressure of $1.33×10−4hPa$ (⁠$10−4Torr$⁠) is lower for TiO₂ (∼1300 °C) than for Ti (1453 °C).⁴³ Further studies are required to explain this discrepancy between the thermodynamic calculations and the experiments.

Other conditions for the power law to hold are that the oxygen supply not be rate-limiting and that less volatile oxide phases do not form. As shown in Fig. 2(b), the deposition rate tends to deviate from the power-law relation when less volatile oxides form. Similar behavior is observed for Ti, the deposition rate of which decreases when $POX$ increases from 10⁻³ to 10⁻² hPa. In addition, the $POX$-dependence can become irrelevant if metallic evaporation occurs under oxygen-deficient conditions as discussed in the context of Fig. 4(b).

We find the exponents n of the $POX$-dependence to depend on the element as well as on P_L (Table I). The n values listed for Ti, Mo, Hf, Fe, Co, and Nb were determined from the slope of the colored lines in Fig. 5, and those of Si from the slope of the data shown in Fig. 2(b) are restricted to the range from 10⁻⁵ hPa–10⁻² hPa in order to exclude the passive oxidation regime. The value n characterizes a partial reaction order in oxidizing gases. In practice, the oxidation reaction can involve intermediate steps and multiple oxidation states, so that n may vary depending on the material and deposition conditions.

According to Fig. 4(b), metal evaporation can become dominant in oxygen at high T_source, resulting in n ∼ 0. Similar behavior has been observed in char combustion. The transition from high-order behavior (n = 0.6–1) to low-order behavior (n = 0) occurs for temperatures above 1200 K. The low-order behavior is expected when the combustion rate is mainly determined by the desorption of CO.⁴¹ 

Our observations of a decreasing n value of Si from 0.39 to 0.22 with increasing P_L from 230 to 350 W are in agreement with this general trend. Thus, the value of n indicates how dominant the formation and evaporation of oxides are for the thermal evaporation of an elemental source in an oxidizing atmosphere. Consequently, the value of n serves as a guide to understanding the majority of vapor species generated by thermal evaporation.

Oxide formation is a key step for growing oxide films by thermal evaporation of metallic sources. We have characterized the thermal evaporation of metal sources in oxygen and oxygen/ozone atmospheres during oxide TLE. The formation and evaporation of oxides from metal sources significantly affect the source vapor species. If volatile oxides are generated, the source evaporation rate generally increases with oxygen supply, as is, for example, the case for Si. The evaporation rate can also decrease with a greater oxygen supply due to the formation of less volatile oxides. The formation of less volatile oxides may be reduced by raising T_source. For sources with oxides less volatile than the metal, such as Al, the deposition rate often decreases in the presence of oxygen. However, oxygen-deficient conditions at high T_source restore the reduced deposition rate of Al in the oxygen atmosphere, making elemental Al vapor dominate over the oxide vapor. A power-law relationship between deposition rate and oxidizing gas pressure is found for the case where the formation and evaporation of volatile oxides dominate the source evaporation behavior. One can infer the efficiency of source oxidation from the value of the power-law's exponent. These general mechanisms are valid for a multitude of oxide growth techniques using thermal evaporation of oxygen-exposed metal sources such as oxide MBE, oxide CVD, as well as oxide TLE.

See the supplementary material for the estimated Si cation flux from the source at a laser power of 350 W, the theoretical maximum SiO flux through active oxidation (S1), and the relation between laser power and temperature of the Al source (S2).

The authors thank H. Boschker, W. Winter, S. Seiffert, I. Hagel, and K. Lazarus for valuable technical support, as well as editing support by L.-M. Pavka.

The authors have no conflicts to disclose.

Dong Yeong Kim: Conceptualization (lead); Investigation (equal); Visualization (lead); Writing – original draft (lead); Writing – review & editing (supporting). Thomas J. Smart: Investigation (equal); Methodology (equal); Visualization (supporting); Writing – review & editing (supporting). Lena Majer: Investigation (supporting); Writing – review & editing (supporting). Sander Smink: Writing – review & editing (supporting). Jochen Mannhart: Funding acquisition (lead); Writing – review & editing (equal). Wolfgang Braun: Methodology (equal); Project administration (lead); Writing – review & editing (equal).

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