Why thermal laser epitaxy aluminum sources yield reproducible fluxes in oxidizing environments

The importance of oxides in the broad scope of electronic applications has become increasingly clear.^1,2 Oxides comprising aluminum have played a specific and crucial role in this development. Sapphire, the most stable binary oxide of aluminum, is a substrate in rapidly increasing demand due to its low cost, superior properties over silicon, and high-quality wafer availability.³ Sapphire is also a material of interest for power electronics.^4,5 Furthermore, several ternary aluminates have played a central role in some of the most exciting discoveries in the field of oxides. Examples of this are the development of freestanding single-crystal perovskite films enabled by sacrificial Sr $3$Al $2$O $6$ layers⁶ and two-dimensional electron gas at the interface of LaAlO $3$ and SrTiO $3$⁠.^7,8

Molecular-beam epitaxy (MBE) is considered the “gold standard” for the growth of oxides and most electronic materials. The achievable purity, mobility, structural perfection, and atomic control are unmatched by any other technique.^9,10 Despite the central role of aluminum in the field of oxides and the importance of oxide MBE, the operational instability of aluminum effusion-cell sources has remained a longstanding challenge. The instability of aluminum effusion cells is illustrated in Fig. 1(a). Aluminum has a high oxidization potential, and Al $2$O $3$ has an extremely low vapor pressure.¹¹ Therefore, aluminum species react with the oxygen species of the growth atmosphere and deposit Al $2$O $3$ at the cell’s neck. This results in the so-called choking of the cell—a constant decrease in the flux until no aluminum emanation takes place.

There have been numerous attempts to develop a solution to this issue. For example, the recently presented suboxide MBE (s-MBE) has yielded promising results for other group-III oxides with regard to achievable crystal quality, growth speed, and flux stability.^12–14 However, aluminum suboxide effusion cells retain the undesired flux instability known from their metal counterparts. This is illustrated by the decrease in the adatom flux as a function of time measured with a quartz crystal microbalance in Fig. 1(b). This is probably caused by further oxidization of the suboxide like that observed for silicon suboxide sources.^15,16

We will show that thermal laser epitaxy (TLE)¹⁷ meets the challenge of supplying a reproducible aluminum flux in an oxidizing environment. To present the underlying mechanism, we first provide an in-depth analysis of the source’s interaction with the heating laser beam as well as with the oxide environment and then proceed to explore the limits of this new technique.

TLE has recently emerged as a promising approach to the issues stated above. It features a wider parameter space for epitaxial growth than any other deposition technique. As the name suggests, TLE is based on thermally evaporating a source by locally heating it with a continuous-wave laser. This makes in-chamber electronics obsolete, and thus allows—as we will show in this work—much higher background gas pressures than conventional physical deposition techniques.¹⁷ Not long after its development, TLE’s potential for oxide epitaxy was achieved.^18,19 Here, we will investigate in detail the behavior of the aluminum source operated in an oxygen environment.

The cross section of the TLE chamber in which these experiments were performed is shown in Fig. 2. The chamber was operated without substrate heating and was water-cooled to $17 °$C to dissipate reflected laser radiation. Source heating was provided by a fiber-coupled disk laser with $λ=1030$ nm and a peak power of 500 W for experiments with oxygen background pressure and a fiber-coupled disk laser with $λ=1030$ nm and a peak power of 2000 W for experiments in UHV. The laser spot size upon the source was approximately 1 mm $2$ with an approximate Gaussian profile. The distance between the focus of the laser and the source equaled 130 mm. Deposition was performed on $10×10×0.5 mm 3$ Si(111) substrates. The high-purity aluminum metal (4N5) was placed in a sapphire crucible and loaded onto a manipulator arm driven by linear motors attached to the $x$ and $y$ directions, allowing a circular motion of the laser spot on the source as illustrated in Fig. 2. O $2$ was used as an oxidizing agent. The applied oxygen pressure was varied between $10 − 4$ and $10 − 1$ hPa. For each pressure setting, the incident laser power was also varied. The resulting film thickness was measured by a profilometer, and the growth rate was determined accordingly. For each experiment, a new source was used and pre-melted with $≈30$ W incident laser power under UHV conditions to test, ensure reproducibility, and eliminate impurities.

Figure 3 shows photographs of the aluminum source in a $10 − 3$ hPa oxygen environment at 120 W (top) and 300 W (bottom) laser power with (right) and without (left) a circular motion of the source. At low laser powers, the source material coalesces into a sphere, see Figs. 3(a) and 3(b). Furthermore, the formation of a white oxide crust is observable. At lower laser powers without applied motor motion, this crust forms everywhere but at the laser spot, see Fig. 3(a). In the case of a circular motion, the area without oxide formation increases to match the laser spot motion at low laser powers, see Fig. 3(b). At higher laser powers, the effects of the beam motion vanish, and the area without oxide crust formation becomes much larger than the laser spot, see Fig. 3 (bottom). The coalescence of the source is diminished.

The formation of the source’s spherical shape is directly related to the appearance of the oxide crust. At first, the Al coalescence is driven by the strong cohesion of the liquid metal²⁰ and by the small adhesion to the oxygen-terminated Al $2$O $3$ surface of the crucible, which causes a large contact angle. In the absence of oxygen, it is expected that, upon further heating of the aluminum source, the contact angle between source material and crucible decreases. This is due to the decreasing surface tension²¹ and a decreasing wetting angle caused by the formation of a higher-energy metallic surface reconstruction of the sapphire crucible.²² This is indeed found for high fluences of the laser beam, see Fig. 3 (bottom). However, for low fluences, the formation of the oxide crust prevents the equilibrium contact angle that is expected to be reached between the metallic aluminum source material and the crucible. The spherical shape of the molten Al is, therefore, maintained.

From the photographs in the top row of Fig. 3, it is evident that only a small portion of the source surface is comprised of metallic aluminum (dark) and that most of the oxide crust is intact (bright). The metallic surface part of the source coincides with the laser spot, Fig. 3(a), and with the path of the circular motion of the source, Fig. 3(b), for the cases of resting and moving sources, respectively. However, the non-oxidized surface area is much bigger and independent of the source motion in the higher power regime. To investigate whether these two source behavior regimes impact the resulting adatom flux, we compared the growth rate as a function of laser power. We consider the case of a resting source (no motion) in UHV (purple & triangles) under $10 − 3$ hPa oxygen background pressure (blue & squares), and the case of a source motion on a circular track of 1 mm diameter under $10 − 3$ hPa oxygen background pressure (red & circles), Fig. 4. As neither Al, Al $2$O, nor Al $2$O $3$ are volatile at room temperature,¹¹ and as the substrate is not heated intentionally, the growth rate is a direct measure of the flux. In UHV (purple & triangles), the growth rate indeed shows the expected Arrhenius-type dependence on the beam fluence.²³ In the presence of an oxygen background pressure, the growth rate is characterized by two laser-power-dependent regimes that match the regimes (top and bottom) observed in Fig. 3. For a laser output power of $>220$ W, the growth rate is independent of a possible source motion. As a function of laser power, it shows the Arrhenius behavior of a source operated in UHV. For laser output powers $<220$ W, the slope of the growth rate as a function of laser power is more shallow than in the UHV regime, revealing in this case that the formation of the oxide crust plays a central role in the emanating flux. The flux is also much higher for an applied circular motion of the source, implying that an increased non-oxidized surface results in an increased flux. In fact, the increase in flux is proportional to the non-oxidized area as controlled by the applied radius of the circular motion. The top row of Fig. 3 shows an increase in the exposed surface area by a factor of $≈10$ (from 1.4 to 13 mm $2$⁠)and Fig. 4 reveals an increase in the growth rate upon circular source motion by the same factor (from 0.1 to 1 Å/s).

The behavior of the source in the high laser-power regime (⁠ $>220$ W) can be easily understood. The source is sufficiently heated to shift the thermodynamic equilibrium away from oxidizing conditions. This is why the source in this regime looks like previously reported aluminum TLE sources in UHV.^23,24 It can be assumed that the entire source area contributes to the aluminum evaporation, which is why the applied motion no longer plays a role. Thus, it is not surprising that the evaporation behavior follows the established Arrhenius law for the dependence of aluminum sources in UHV on laser power.^23,24 The wide range of available growth rates is worth mentioning in this context. Depending on the applied laser power and the application of a source motion, a growth rate variation over 5 orders of magnitude can be induced with a laser power variation of just over 1 order of magnitude. This deposition regime constitutes the regime with the highest reproducibility.

To understand the more complex behavior of the source in the low laser-power regime (⁠ $<220$ W), we investigated the flux dependence on the laser power for a wider range of applied oxygen pressures. Figure 5 illustrates this dependence for oxygen pressures between $10 − 4$ and $10 − 1$ hPa and sources moving in a circle with a radius of 1 mm. For orientation purposes, the expected Arrhenius dependence of the growth rate on laser power for the evaporation of a source in UHV is marked by a purple dashed line. For the lowest oxygen pressure of $10 − 4$ hPa (green & diamonds), the flux follows the Arrhenius behavior of a non-oxidized source down to relatively low laser powers (⁠ $≈165$ W). This is in agreement with the less oxidizing conditions. Counterintuitively, the flux in the lower laser-power regime increases with increasing oxygen pressure up to $10 − 2$ hPa. For pressures $≥ 10 − 1$ hPa, the flux decreases again.

This behavior changes at $10 − 1$ hPa oxygen background pressure. At low laser powers, the growth rate no longer increases with increasing laser power or increasing oxygen pressure. At higher laser powers (⁠ $>140$ W), we observe that the growth stops (gray line in Fig. 5). Studying the source after growth at high pressures and high laser powers provides a clue to explaining this behavior. Figure 6 shows an optical microscopy image of the source after being exposed to $10 − 1$ hPa at 300 W of incident laser power. Al $2$O $3$ crystals formed on the surface of the source are well visible. We conclude that, at $10 − 1$ hPa oxygen pressure, the oxidization potential becomes so high that oxidization is favorable even at extremely high temperatures (locally $>2000$ K). As a result in the low laser-power regime, less suboxide can contribute to the flux because more Al $2$O $3$ is formed. In the high laser-power regime, the energy supplied by the laser is even sufficient to promote the growth of Al $2$O $3$ single crystals. As a result, the source stops emanating particles.

In summary, we have provided a systematic study of the behavior of aluminum sources used in thermal laser epitaxy under oxidizing conditions. We find that, for laser powers above 220 W and illuminated areas of 1 mm $2$⁠, a reproducible flux is achieved for oxygen pressures up to $10 − 2$ hPa. This exhibits an Arrhenius-law-type dependence as a function of laser output power. We found an upper limit of oxygen pressure useful for film growth on the order of $10 − 2$ hPa. At higher pressures, the source oxidizes and becomes unusable. For laser powers below 220 W, the adatom flux is a function of the applied oxygen pressure and the exposed surface area of the source. In this regime, due to the formation of Al $2$O, the flux increases proportionally to the oxygen flux reaching the exposed source area. By virtue of these mechanisms, TLE can be applied to vary the adatom flux emanating from Al sources over a range unmatched by any other technique. (Note that 5 orders of magnitude of growth rates were demonstrated here.) We expect that this in-depth understanding of the operating regimes of aluminum sources in TLE will pave the way for high-quality sapphire epitaxy. Furthermore, by identifying the oxygen-independent high laser-power regime, we expect that TLE can be applied to grow aluminum-based ternaries.

The authors thank Varun Harbola and Sander Smink for many insightful discussions. The authors also thank Fabian Felden, Ingo Hagel, Konrad Lazarus, Sabine Seiffert, and Wolfgang Winter for technical support.

The authors have no conflicts to disclose.

Thomas J. Smart: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Supervision (supporting); Writing – original draft (equal); Writing – review & editing (equal). Felix V. E. Hensling: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). Dong Yeong Kim: Formal analysis (equal); Investigation (supporting); Methodology (supporting); Writing – review & editing (supporting). Lena N. Majer: Data curation (supporting); Formal analysis (supporting); Investigation (supporting). Y. Eren Suyolcu: Data curation (supporting). Dominik Dereh: Data curation (equal). Darrell Schlom: Investigation (supporting); Methodology (supporting); Supervision (supporting). Debdeep Jena: Conceptualization (supporting); Data curation (supporting). Jochen Mannhart: Conceptualization (supporting); Funding acquisition (lead); Investigation (supporting); Methodology (supporting); Project administration (equal); Supervision (equal); Writing – review & editing (lead). Wolfgang Braun: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal).

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