Film deposition by thermal laser evaporation

Several physical vapor deposition techniques such as molecular beam epitaxy (MBE), pulsed laser deposition (PLD) and sputtering have been developed in recent decades, often intended for the study of specific thin-film material classes such as semiconductors or high-T_c superconductors. Whereas classical semiconductivity is linked to a certain part of the periodic table with elements of similar properties, today’s quest for quantum materials with exotic band structures requires the high-purity synthesis of epitaxial layers¹ from very dissimilar constituents with strongly different physical properties. Therefore, a need has arisen for a thin-film growth technique that is ultraclean, versatile, efficient, and able to synthesize layers containing many different elements from virtually the entire periodic system. Such a physical deposition technique would ideally supply individual beams of any element with negligible (i.e., thermal) velocities that react with each other at the surface, thus allowing the straightforward analysis and control of this process, even for complex compounds.

The established film deposition methods excite the source material either indirectly by placing it into a heated crucible or directly by an electron beam. The source material may also be ablated via high-energy laser pulses or sputtered with a plasma. High-quality epitaxial films are often produced by PLD or MBE. For PLD,^2,3 invented in the sixties but developed for and with high-T_c superconductors,⁴ the source material with the nominal stoichiometry of the desired film is ablated by a laser pulse so short that the different volatilities of the constituents can be neglected and, ideally, a film of the same composition is formed. In MBE,^5–8 developed for and with compound semiconductors, the film is assembled from individual elemental atomic or molecular beams on the substrate surface. In the context of complex oxides, it was recognized early on that such a combination of individual beams could allow the precise assembly of arbitrary compounds on a suitable substrate.^9–12 

Thermal laser heating^14,15 offers several advantages for vacuum evaporation. First, laser beams do not contaminate their target. A laser features an almost arbitrary power density and therefore allows very high evaporation temperatures. Second, all energy sources are located outside the vacuum, thus eliminating the need for electrical feedthroughs and allowing a compact, minimalistic mechanical design. As a result, growth configurations with small source–substrate distances become straightforward. In addition, the local heating of the source is efficient because the energy is absorbed directly at the evaporating surface. Finally, the laser beam is not affected by the atmosphere in the chamber—any gas suitable for the growth process can be used. These advantages were recognized and studied^15,16 right after the invention of the CO₂ laser. Subsequently, these studies were quickly abandoned when the evaporated compounds were found to be nonstoichiometric.^17,18

With the availability of affordable high-power lasers, it now becomes feasible to generate individual molecular beams for epitaxy. Whereas long-wavelength lasers are useful for substrate heating,^19–21 the efficient absorption by metals, which constitute most of the elements of the periodic table, requires short wavelengths ideally in the visible or even ultraviolet range.^22,23 For the experiments presented here, a wavelength of 1030 nm was used, provided by a fiber-coupled disc laser with a peak power of 2 kW.²⁴ 

To assess the temperatures required to yield practical growth rates, Fig. 1 shows the vapor pressures of several elements as a function of temperature,¹³ together with their melting points. The green areas denote a superimposed scale of typical deposition rates obtained for the corresponding pressures in an effusion source. The connection between the pressure and the growth rate scales is made by using typical operation temperatures of Ga and Al effusion sources for the 1 ML/s deposition of GaAs or AlAs as marked by the ‘H’ symbols.²⁵ To be on the safe side, the upper limit of 5 ×10⁻³ hPa is chosen as the 1 ML/s reference for the Ga range. Published source temperature vs. growth rate values are rare; one example is marked with an x for Ref. 26. Common crucible materials are limited in temperature, as indicated by the red ranges for pyrolytic boron nitride at ≈1300 °C and aluminum oxide at ≈1700 °C. Singular examples exist both for element–crucible combinations that work at higher temperatures as well as for problematic elements at lower temperatures. Therefore, these boundaries are guidelines rather than strict limits. The blue rectangle denotes the standardized temperature range of a tungsten–rhenium (type C) thermocouple with an upper temperature limit of 2320 °C.

Interestingly, the lowest vapor pressure materials have high melting points, in the range of typical effusion cell evaporation temperatures or even above. This suggests that these elements can be deposited at standard rates by local thermal evaporation without melting the entire source body, even if only modest thermal gradients can be achieved.

To investigate whether these expectations hold in practice, we tested thermal laser evaporation in a simple uncooled trial chamber containing the components shown in Fig. 2. The focused heating laser beam has an incidence angle of 40° to the target surface from above. With its focal point being 62 mm from the source center, the spot size is approximately 1 mm² on the source surface, see inset of Fig. 3. To reduce the coating of the laser entrance window, a beam aperture is mounted at the focal point. A quartz crystal monitor (QCM) is used to measure the deposition rate during the experiment, which is double-checked by measuring the thickness of the grown films. Reflected laser radiation is dissipated at an approximate spot size of 10 mm by a stainless steel chamber flange at the specular beam position.

The source target rests on three tantalum support fingers as shown in Fig. 3. A tungsten–rhenium (type C) thermocouple measures the temperature at the back of the rotationally symmetric source.

An image of the upper side of the target in operation is shown in the inset of Fig. 3. The source target glows uniformly, except for the point where the laser hits. Its support plate is dark, indicating that the plate temperature is much lower. The source is PID-controlled using the thermocouple temperature as input and the laser power as output. The temperature profile recorded during a Mo heating run is shown in Fig. 4. The thin black line denotes the source temperature, and the measured temperature is plotted in blue. Rapid temperature changes with rates of several degrees per second as well as long-term stabilities of better than ±0.1 °C are achieved.

We performed initial deposition tests with Ti and Mo, two elements that are difficult to evaporate precisely with traditional methods, but that are of interest for epitaxy. The results for Ti are plotted in Fig. 1 (blue + markers). The sensor temperatures to evaporate Ti are significantly lower than those used in MBE for similar growth rates, the offset equals approximately 500 °C. This offset offers comfortable headroom to the melting point, even if possibly large temperature offsets between the measured temperature and the actual surface temperature at the contact point of the thermocouple sensor are taken into account.

Raw temperature and film thickness data for Mo laser evaporation is shown in Fig. 5. Again, the blue lines denote the measured temperatures. As the trial chamber used is not actively cooled, deposition was done in 15-min intervals, with intermittent cooling times to allow the chamber—and in particular the QCM—to cool so that precise thickness values could be determined. The small oscillations of the QCM signal are due to temperature fluctuations of the QCM cooling water. The final two depositions were made on the second day. The resulting layer on the Si substrate is shown in Fig. 6. The wafer was cleaved along its diameter, and several thickness measurements such as the one shown in the inset were taken by a scanning electron microscope to determine the film thickness. The results varied from 94 to 98 nm across the wafer, significantly less than the 13% expected for a small, flat surface source emitting with a cosine characteristic.^27,28

The growth rates determined from the depositions in Fig. 5 are shown as red + symbols in Fig. 1. Their temperature dependence is similar to that of the vapor pressure, with an upward deviation of the last data point. We attribute this step to the melting point of Mo because, after the experiment, the target showed a small specularly smooth spot in the center of the illuminated area that was not present on targets heated to lower temperatures. Based on this observation, the temperature offset between the measurement point on the back side and the irradiated spot on the front of the source may be determined by shifting the measured data to the melting point on the vapor pressure curve. The shifted data are indicated by the softer red + markers in Fig. 1. This calibration agrees with the scaling compared to MBE: Whereas the laser source is about four times closer to the substrate than is usual in MBE, its active surface is approximately 100 times smaller than an effusion source. This requires a vapor pressure that is one order of magnitude higher for laser evaporation than for MBE for the same growth rate, in agreement with Fig. 1.

Regarding the slopes of the growth rate, the measured data are steeper than their respective vapor pressure curves. This behavior is readily understood by examining the radiative cooling mechanism of the source, which increases with a dependence of T⁴. The temperature offset between the heated spot and the thermocouple sensor therefore increases with temperature. The temperature offset between the two red + data sets in Fig. 1 is 690 °C, which demonstrates the efficiency of this mechanism at high temperatures. It allows local evaporation at the laser-heated spot at temperatures above the melting point, while the remaining target remains solid. In practice, two temperatures are important: the temperature at the thermocouple contact point should not exceed the eutectic temperature or, in general, the temperature at which the thermocouple and source material bond, react or melt. Similarly, the temperature at the source supports should remain below the eutectic or reaction temperature to avoid sticking or contamination. With the present geometry of a disc-shaped source, both conditions can be met independently either by thickening the source to increase the temperature offset between the evaporating surface and the thermocouple, or by enlarging the source diameter to increase the temperature offset between the evaporating surface and the support points.

One can optimize concurrently for a small temperature offset to the thermocouple for agile and precise temperature control, while keeping the offset large enough that even very low-vapor-pressure materials at high evaporation temperatures can be controlled by a thermocouple not exceeding its temperature limit, as illustrated for the present example of Mo and a type C (W/Re) thermoelement in Fig. 1. A per-target optimization of its shape therefore allows individual optimization with the same given laser beam, making the evaporation from self-supported sources feasible for most, if not all, solid elements in the periodic table.

Such a crucible-free evaporation technique eliminates longstanding problems of film growth by traditional methods, which arise from thermal expansion mismatch or reactions between source material and crucible. Thermal laser evaporation is resource-efficient because the short working distance keeps source-material consumption low. After use, the leftover source material can easily be recycled because it suffers only from possible surface contamination and can be well separated from its holder. The process is also energy-efficient. In the experiments shown here, growth rates of 0.5 Å/s with 5% uniformity on a 50-mm substrate are achieved with 160 W laser power for Ti and 440 W for Mo, values easily dissipated by liquid nitrogen cooling shrouds.

Such thermal laser evaporation can be used for thin-film epitaxy of compounds from elemental or molecular beams as shown schematically in Fig. 7. A source holder containing multiple sources similar to the ones used here is illuminated by several laser beams, either through the same or different shielding apertures at their focal points. The source holder can be transferred in and out of the chamber to grow different compounds sequentially. Preferentially, substrate heating also takes place via laser, thus allowing a wide range of background atmosphere compositions and pressures, as long as the mean free path does not drop below the source–substrate distance so that directed thermal evaporation remains possible. As the evaporating surfaces of the sources and the substrate itself are the hottest surfaces in the chamber, they cannot be contaminated by substances released from glowing filaments or other hot parts inside the chamber such as in MBE. A liquid or liquid nitrogen cooling shroud (not shown in Fig. 7 for clarity) can further reduce background contaminations. With standard enhancements such as shutters and in-situ diagnostics, TLE will be a viable technology for the precision growth of ultraclean films.

Following the above data and arguments, it appears that any combination of solid and liquid elements from the periodic table can be thermally evaporated in the same chamber by thermal laser evaporation. Gaseous elements can be introduced at pressures up to 10⁻³ hPa, where the mean free path of an ideal gas equals the 60 mm working distance of the present experiment.²⁹ Thermal laser epitaxy, which has the ability to provide precisely controlled elemental or molecular beams, is expected to allow the epitaxy of device structures consisting of different and more complex materials than is possible today.

The authors thank Darrell Schlom for valuable discussions and Ulrike Waizmann and Wolfgang Winter for technical support.