State of the art, trends, and opportunities for oxide epitaxy Open Access

An important foundation of any technological development is the selection of materials whose properties are appropriate to the needs of specific applications. Owing to their wide variety of properties, oxides comprise some of the most promising materials for many key technologies, such as novel computing concepts (e.g., quantum^1,2 and neuromorphic computing)^3,4 and green energy applications (e.g., photovoltaics^5,6 and catalysis).^7–10 For example, these properties cover a broad range of electronic conductivity, from superconductivity^11,12 to ultrawide-bandgap insulating behavior.^13,14 Similarly, ionic conductivity can be very limited or superbly high.^15,16 Oxides also exhibit an array of polar, electric, and magnetic properties.^17,18 Therefore, it is not surprising that some of the most intriguing material properties have been discovered in oxides: high-T_c superconductivity,^11,12 spin-ice frustrated order,^19,20 and multiferroicity, to name but three examples.^21,22

The understanding and application of these properties often necessitate the growth of high-quality epitaxial oxide films. There are numerous modes of epitaxial film growth, which are typically characterized by crystallization processes (e.g., step-flow, layer-by-layer, and Stranski–Krastanov growth) or methodologies used for stoichiometry control.²³ Of the different stoichiometry control methods, adsorption-controlled growth plays an especially important role. Adsorption-controlled growth refers to the epitaxial growth of a film, during which the deposition parameters, i.e., source condition, substrate temperature T_sub, and chamber pressure, result in the volatility of all deposited species but one. This yields self-limiting growth, in which the growth rate is determined solely by the flux of the non-volatile species because all other species desorb if they are present in excess. This growth mode generally results in films with unsurpassed stoichiometric and phase purity.^24,25

Thus, it is no coincidence that modulation-doped heterostructures of GaAs, the highest known electron mobility structures, are grown in an adsorption-controlled manner.^26–30 In contrast to GaAs, epitaxial oxide films usually cannot compete with their corresponding single crystals in terms of perfection and purity. Various approaches have been proposed to overcome this drawback.^31,32 An obvious solution is the adsorption-controlled growth of oxides. However, for most oxide constituents, volatility cannot be achieved in the deposition parameter space available to conventional deposition systems.³³ For some oxides, epitaxial deposition already proves challenging, regardless of the growth mode. For example, pyrochlore iridates require crystallization temperatures far above 1000 °C combined with high oxidation potentials, which are impossible to realize with standard heaters,^34,35 the usual limit of which is around 1000 °C T_sub. This restriction is also what has thus far limited the epitaxial growth of c-plane sapphire.³⁶ Further complications are given by the fact that numerous elements show several oxidation states, which result in a plethora of possible crystal phases. Indeed, the most challenging oxides to grow are those that require high oxidation potentials and high crystallization temperatures. Outside the growth window of adsorption-controlled growth, big vapor pressure differences between the source materials may be problematic for maintaining the stoichiometry of the growing film.

This perspective provides an overview of the main challenges in connection with oxide epitaxy, which are (1) the quest for a universal, versatile, and clean way to transfer any element onto the substrate; and (2) a pure and controlled phase formation, which requires the ability to achieve high oxidation potentials and a high substrate temperature T_sub. We start by evaluating these challenges in more detail and then provide an overview of the state of the art of common oxide epitaxy methods with respect to these challenges. We discuss the most exciting trends to overcome these challenges and the resulting opportunities for oxide epitaxy. Due to space constraints, this analysis is not exhaustive but focuses on those topics that we consider the most important.

Epitaxial growth consists of two fundamental processes. First, in a film-growth system that ensures a high-purity environment, the source material is transferred to a substrate, where it arrives as adatoms. Then, the adatoms on the substrate crystallize in the desired phase and crystal orientation. Figure 1 illustrates these processes and provides desirable features and parameters of their essential components. Numerous mechanisms are used to transport the species from the source toward the substrate, as discussed below. Generally, the material can be introduced into the deposition chamber (Fig. 1, gray) either as a gas (light blue), a liquid, or a solid (blue, green, or yellow). In the latter case, these source materials must undergo a phase transition to be transported to the substrate (burgundy). In most cases, the formation of the epitaxial film with good crystallinity on the substrate requires a higher T_sub, which can be generated by various forms of heaters (red), the most common of which will be discussed.

The transfer of an element to the substrate begins when it is introduced into the growth chamber. For oxide epitaxy, the most relevant gaseous source is usually the oxidizing agent (Fig. 1, light blue). High reactivity, i.e., a high oxidation potential, is desirable. This can be provided, e.g., by activated oxygen or ozone, which exposes the objects in the chamber to the risk of oxidation. Electronic filaments and heaters are particularly easily damaged by oxidation. Even worse effects may result from impurities generated by the oxidation processes of chamber components. Furthermore, oxidizing agents such as ozone pose a potential safety hazard due to their high reactivity.

It is desirable to be able to exchange the source of non-gaseous elements (Fig. 1, green) in the chamber without breaking the vacuum. This increases flexibility and throughput and decreases machine downtime. Once the element is loaded into the deposition chamber, the next challenge is to find a universal, versatile, and clean way for any element to be transferred to the substrate. This can prove difficult due to the elements’ drastically different vapor pressures and reactivities. For most deposition methods, element-specific solutions are, therefore, required for transferring elements to the substrate. In comparison to the film growth of the classic III–V and II–VI compound semiconductors, oxide epitaxy is often accompanied by the desire to access a wide range of elements. The versatility of the element transfer process is, therefore, an important specification.

Once the material reaches the substrate, epitaxial crystallization must be ensured by providing the required T_sub. Thus, in a versatile deposition system, a heater (Fig. 1, magenta) must accommodate a wide T_sub range that provides T_sub stability and homogeneity, ideally for any substrate material and size of interest. Providing very high T_sub proves especially challenging because the efficiency of most heaters declines with temperature, as does the coupling of the heat to the substrate. Simultaneously, more parasitic heat warms the components next to the heater, resulting in an increase in impurities released from the heater and its surroundings.

Therefore, for operation at higher substrate and source temperatures, the effects of parasitic heat must be considered in the design of the recipient system (Fig. 1, gray). The vacuum chamber must be able to withstand local hotspots and dissipate locally high heat loads. Generally, the recipient must provide a clean environment, i.e., background pressures that are as low as possible, together with inertness to the deposition conditions.

The increasing demand for epitaxial oxides is paralleled by the development of a variety of deposition techniques that address the associated requirements and challenges described above in their own unique ways. The film growth techniques most commonly used are pulsed laser deposition (PLD), sputtering, molecular beam epitaxy (MBE), metal–organic chemical vapor deposition (MOCVD), which, for epitaxial growth, is sometimes referred to as organometallic vapor-phase epitaxy (OMVPE), and the closely related atomic layer deposition (ALD). Other successful techniques include metal–organic aerosol deposition^37,38 and chemical solution deposition.^39,40 Several comprehensive reviews of PLD,^41–44 sputtering,^45–47 MBE,^43,48–52 and MOCVD/ALD^53–56 are available that include details of the operation and capabilities of the commonly used oxide epitaxy methods. Here, we focus on how they each address the requirements and challenges identified above.

In PLD, a target material (typically comprising the same stoichiometry as the desired film) is ablated by pulses of a laser beam with a typical wavelength of 248 nm. The target absorbs some of the laser pulses, resulting in a shock wave, ionization, and the formation of a plasma plume. This process is schematically depicted in Fig. 2(a). Details of the complex plasma and phase formation can be found elsewhere.^42,57

Despite the many parameters that influence the stoichiometry of the resulting epitaxial film, such as laser spot size,^58–60 fluence,^61–63 pulse repetition rate,^64–66 and background atmosphere,^67–69 the nominal, yet not always perfect, stoichiometric transfer of the target material to the substrate is a core strength of PLD.⁴² Typically, sintered polycrystalline targets that constitute a major source of impurities are used in PLD.^70,71 These impurities can be strongly reduced by utilizing single-crystal targets, which are available in the required size only for a limited choice of materials. Although the stoichiometry of these targets, given the correct deposition parameters, is translated to a film with high reproducibility and reliability, this comes at the cost of flexibility. Experiments that study varying stoichiometries, dopants, or cation species usually require the synthesis of new targets for every new film composition.

It is generally more complex to control the oxygen stoichiometry in PLD than the cation stoichiometry. Owing to several loss mechanisms, typically only about 50% of the target’s oxygen is incorporated into the thin film.^42,72 This loss of oxygen may be compensated by supplying oxygen as a background pressure because the oxide targets are robust against oxidation. If oxygen-resistant heaters are used as discussed below, background pressures in the range of 10⁻⁷ to 1 mbar may be applied (Fig. 3). The lower limit is set by the target outgassing and the recipient design, whereas the upper limit is set by the mean free path of the ablated species.⁷³ However, the background pressure influences the cation stoichiometry in a complex way, as described in detail in Ref. 42.

Owing to the species-dependent spatial composition of the plasma plume, the stoichiometry of the film may vary across the substrate area. For small samples (≤10 × 10 mm²), this is usually negligible. Upscaling PLD to large wafer sizes is not trivial, but significant progress in large-area PLD has been made over the past two decades.^74–76 In industrial fabrication, PLD is a niche process despite the comparably high achievable growth rates (several 100 nm/h). In research and development, however, PLD is used as a workhorse, in particular, for the growth of complex oxides. Reviews of PLD used for oxide epitaxy can be found in Refs. 41–44.

The sputtering process utilizes plasma generated by electrical discharge. In this plasma, positively charged ions impact the negatively charged target, thereby ejecting the target species. The process is schematically depicted in Fig. 2(b). For enhanced sputter yields, magnetron configurations are used in which magnetic fields trap electrons in helical paths close to the target for efficient ionization of the sputter gas.⁴⁵ 

For oxide targets with at least a ternary composition, one of the constituent species is usually sputtered preferentially, resulting in a depletion of that species in the target. With time, this process equilibrates, and, in the ideal case, the resulting films have the desired stoichiometry. However, precise stoichiometry control in sputtering is challenging compared to PLD. Additional problems are oxygen back-sputtering and the high kinetic energy of the species hitting the growing film, which often causes defects. Again, the targets used for epitaxial oxide sputter processes are common sources of impurities.⁷⁷ The possible high reproducibility and reliability of sputtering also come at the cost of flexibility. Sputtering requires a minimum background pressure to achieve stable plasma. Usually, additional oxygen is necessary to compensate for oxygen loss from preferential sputtering and related loss mechanisms.^78,79 As a result, the typical total background pressure for oxide epitaxy by sputtering lies between 10⁻⁴ and 10 mbar (Fig. 3).^80,81

Generally, sputtering can be upscaled in size, which, together with the stoichiometric transfer capability and comparably low cost, is a major advantage. While complex oxides can be sputter deposited, the most common application of sputtering is for binary oxides. This and the respectable achievable growth rates of several 100 nm/h make sputtering a commonly used method in industry.^47,82 More details on the sputtering of oxides can be found in Refs. 45–47.

In MBE, the individual source materials are thermally sublimated or evaporated, typically from effusion cells, in which the source elements are located in crucibles ohmically heated by filaments. Effusion cells yield directed beams of the evaporated species, the so-called molecular beams. Molecular beams from sources that are hot but not used for growth at a given time are blocked by a shutter. The molecular beams are directed onto the substrates, where they adsorb and form the oxide films. The operating principle is depicted in Fig. 2(c).

Generally, MBE utilizes elemental sources, which have the advantage of having superior available purities compared to compound targets. Hence, of the deposition methods discussed in this section, MBE has the ability to produce the purest films.^43,50 Furthermore, elemental sources provide flexibility. The modularity of the sources enables the combination of different elements available in the chamber. However, the size of the element portfolio is limited by the number of available source ports (typically ⪅9). Furthermore, sources cannot be exchanged easily in MBE. Without expensive special hardware, source changes even require opening the vacuum system to air. Ideally, the crucible material is inert under deposition conditions and does not release impurities. Therefore, the high temperatures needed to evaporate elements with low vapor pressures can drastically restrict the choice of crucible materials. In addition, the heater filaments must withstand deposition conditions while under high electric current densities. Therefore, for elements that yield a significant vapor pressure only above 1800 °C, standard effusion cells are inadequate. In this case, one commonly resorts to electron beam (e-beam) evaporation, which can evaporate most elements. e-beam evaporation, however, has the drawbacks of poor growth rate stability, high maintenance requirements, and proneness to impurity introduction (e.g., via impurities introduced by the filaments or electron-induced dissociation of residual hydrocarbon molecules).

The flexibility of source combinations in MBE comes at the cost of difficult stoichiometry control. The use of elemental sources requires excellent flux calibration to achieve well-controlled stoichiometries. For one single source, this calibration process may already take several hours. For most effusion cells, this requirement does not cause further difficulties because the flux drifts are smaller than 1%/h.^83,84 However, some effusion cells, such as the ones used for Al and Ti evaporation or, even worse, e-beam sources, are not sufficiently stable to enable precise long-term calibration.^83,85,86 To circumvent source instabilities and achieve stoichiometric films, the adsorption-controlled growth mode is preferable.⁸⁷ As described in the introduction, the parameter space most oxides require for adsorption-controlled growth is unfortunately outside the substrate temperature range achievable with a standard MBE system.

Upscaling MBE processes to large substrate areas is rather unproblematic. However, the achievable growth rates in MBE are relatively low, as required in order to maintain flux stability. On the other hand, the low growth rates and the resulting atomic precision make MBE a prime deposition technique for layered oxides. As a result, the strategy in the industry to combat low growth rates (<100 nm/h)⁹⁰ and the resulting low throughput is to perform simultaneous growth on multiple wafers.⁹¹ More detailed information on oxide MBE can be found in Refs. 43 and 48–52.

MOCVD and ALD are called chemical-deposition methods because the growth of the film is based on chemical reactions rather than physical deposition. All species are introduced into the recipient as gases. For those elements that are not available as gases, precursor molecules that contain the desired species are used. If the precursors are liquids, a carrier gas is required in addition. Within the recipient, in this case, also called a reactor, the precursors react under heat intake to form the desired phase on the substrate surface (pyrolysis). The precursor residues and carrier gas should ideally leave the reactor after the reaction of the desired species.⁵⁶ This process is depicted in Fig. 2(d).

For any oxide to be grown epitaxially by MOCVD or ALD, suitable precursors of the involved cations are required that suit the parameter space demanded by the film growth. Indeed, multiple precursors exist for many cations. Using precursors poses the obvious challenge that precursor residues are introduced as impurities into the system and the growing films. This problem is especially relevant at lower T_sub, for which pyrolysis is less efficient.^92,93 In addition, precursors are often a safety hazard.⁹⁴ As precursors enter the reactor as gases, MOCVD and ALD are highly flexible. If a suitable precursor exists, elements can be combined at will for film growth without source swaps. Stoichiometry control of the injected gases is not trivial because it is difficult to predict exactly how the concentrations of the desired cations change as a function of time. One of the precursors can often be chosen to be more volatile, resulting in a growth mode somewhat equivalent to adsorption control,⁹⁵ which is obviously more complicated for complex oxides.⁹⁶ Multi-cation precursors can be a workaround, although this reduces flexibility and their availability is restricted.⁹⁷ Thus, MOCVD is predominately used for binary oxides.

As the growth process in MOCVD and ALD is based on chemical reactions, the mean free path is of no concern. This allows a wide range of oxygen pressures to be applied. Considering that precursor fluxes add to the system pressure, the background pressure is usually much higher than in other deposition methods (10–100 mbar); see Fig. 3.⁹⁸ The high applicable pressures and the usually self-regulating stoichiometry allow very high growth rates for MOCVD (≈2 µm/h).⁹⁹ The unique ability of conformal growth and the effortless upscaling make MOCVD one of the most commonly used tools for the industrial growth of epitaxial oxides.⁵⁶ More detailed information on oxide MOCVD and ALD is available in Refs. 53–56.

All film growth methods discussed here require substrate heating, which should preferably cover a wide range of T_sub and provide excellent control. In other words, there should be only a negligible difference between the real/measured T_sub and the setpoint value, both while maintaining a constant temperature and during temperature ramps. For this process, several types of heaters are used (Fig. 4).

Resistive heaters are the simplest ones. An electric current is applied to a holder block to which the non-conductive substrate is attached [Fig. 4(a)]. In rare cases, the current is applied directly to the conducting substrate. Resistive heaters can be built to be rather oxygen-resilient and compatible with oxygen pressures of ≤1 mbar and to yield a maximum T_sub of up to 1000 °C. Convection increasing with pressure reduces the maximum achievable T_sub in the high-pressure regime. In the case of indirect heating, the maximum T_sub is decreased further by the temperature gradient at the substrate–block interface. This gradient is usually minimized by bonding the substrate to the heater block with metallic paint of Ag, Pt, or In. The limited temperature stability of these paints is an additional limiting factor for T_sub because they may diffuse or evaporate. Even worse, these paints are prone to contaminating the system. This process is exacerbated by the significant parasitic heat of such heaters. The thermal lag of the heating element is the limiting factor for the achievable ramp rates of T_sub. On the other hand, the large thermal mass supports maintaining a constant temperature, both in space and time. When using resistive heaters to which the substrates are glued, the entire heating element must be changed with each substrate change.

If thermal radiation is utilized for substrate heating, heat is usually produced by an electrically heated filament made of a Ni alloy, noble metal, or SiC. To couple the heat into the substrate effectively, the back of the substrate is usually coated with a metal film [Fig. 4(b)]. Depending on the filament in use and the applied pressure (up to 1 mbar), this setup allows for T_sub ≤ 1000 °C. Similar to resistive heaters, the thermal lag of the filament limits achievable T_sub ramp rates. For the filament itself and at high T_sub, the backside coating is also a possible source of impurities.³⁶ As heat radiation is non-directional, parasitic heat management is difficult with such heaters at high T_sub. Filament-based radiative heaters work for all substrate sizes and are usually downward compatible in terms of size. A particular version of thermal radiation-based heaters features quartz lamps [Fig. 4(c)], which has the advantage that no electronic components are required in the chamber. Therefore, pressure is not limited by the heater. Damage to or impurities originating from electronic components are avoided. Quartz lamps enable T_sub ≤ 850 °C. Possible damage to the quartz lamps usually limits the T_sub ramps to a moderate ⪅5 °C/s.

Utilizing coherent light sources such as infrared diode lasers avoids most parasitic heat at the cost of more challenging scalability [Fig. 4(d)]. The issue of insufficient absorption for most oxides also applies to the wavelength of infrared diode lasers (λ = 940 nm), which is well absorbed by metals. Thus, the necessary backside coating limits the achievable T_sub to ≈1000 °C. Unlike non-coherent and resistive heaters, the achievable T_sub change rates are quasi unlimited; the only heated element is the substrate, which has a small thermal mass and can, therefore, quickly change temperature induced either by high-power heating or by radiative cooling. Therefore, T_sub ramps are exceptionally fast (≈100 °C/s). The insufficient absorption of the diode-laser light by oxide substrates is overcome by utilizing CO₂ lasers instead, as these generate infrared light of longer wavelength (λ ≈ 10 µm); see Fig. 4(e). This wavelength is absorbed well by most commonly used oxide substrates (Table I), but not by metals, minimizing the necessity of absorbers for oxide epitaxy. The benefits of CO₂ laser heaters will be discussed in detail in the next section. Substrate heater specs are adapted from Refs. 100–103.

In the previous section, we described the most common oxide epitaxy methods with respect to their ability (1) to provide a universal, versatile, and clean transfer of any element onto the substrate and (2) to achieve sufficiently high oxidation potentials and temperatures to obtain the desired phases at high levels of quality and purity. Numerous efforts are ongoing to overcome the various limitations of existing techniques. In the following section, we will present recent trends that appear to be especially promising.

In hybrid MBE, MOCVD precursors are utilized as one or more molecular sources for MBE systems. Hybrid MBE was first developed in the early 1990s to obtain a stable flux of the low-vapor-pressure element Y, which was not achievable at that time with standard effusion cells.^104,105 In subsequent years, hybrid MBE remained a fringe technique used for other low-vapor-pressure elements such as Zr and Nb.^106,107 Nearly ten years later, the technique gained momentum when it was applied to circumvent the difficulties involved in evaporating Ti. It was then discovered that volatile metalorganic molecules can be utilized to achieve adsorption-controlled growth of SrTiO₃.^108,109 Importantly, Ti is already oxidized in the precursor. Therefore, less potent oxidizing agents suffice during deposition. Despite the impurities potentially introduced by the precursor, the adsorption control achieved by this approach has enabled the growth of oxide films with record electronic mobilities.^110–112 Therefore, it is not surprising that hybrid MBE has since been applied to many material systems with impressive results.^51,113 Most recently, for example, the growth of superconducting Sr₂RuO₄ has been demonstrated with hybrid MBE.¹¹⁴ Nevertheless, hybrid MBE is prone to several of the drawbacks encountered by MOCVD. For example, impurities are easily introduced by the precursor at low growth temperatures.⁹³ As discussed in Sec. II D, precursors are also only applicable in a specific parameter range that ensures their effective pyrolysis.

In suboxide MBE, sources are applied to yield a phase-pure beam of suboxide species. These sources can be comprised of an oxide, a suboxide, or an oxide–metal mixture.^33,115,116 Oxide sources were first used in MBE in the 1980s with the aim of providing a stable flux of easily oxidized elements.¹¹⁷ Shortly thereafter, the potential to use a highly volatile oxide of the low-vapor-pressure element boron was realized.^118,119 Subsequently, however, oxide sources were used mainly for obtaining stable fluxes of easily oxidized elements.^120–126 Suitable sources were identified by trial and error until thermodynamic modeling more recently allowed ideal oxide, suboxide, or oxide–metal mixed sources to be identified—first only for specific elements, then for all elements.^33,127 This was the basis for a more systematic understanding and investigation of suboxide MBE. The development of suboxide MBE is motivated by the idea to utilize the volatility of suboxides to enable the evaporation of low-vapor-pressure elements as high-vapor-pressure oxides at much lower temperatures and to gain access to the adsorption-controlled growth regime while maintaining stable fluxes.¹¹⁶ Indeed, suboxide MBE has been utilized to grow adsorption-controlled Ga₂O₃ with unmatched crystal quality at MOCVD-like growth rates.¹¹⁵ In addition to the high growth rates obtainable with such sources, their flux is still reproducible and sufficiently controllable to grow complex ternary systems that have a high sensitivity to exact flux ratios.¹²⁸ This newly gained knowledge about suboxide MBE has allowed ternary oxides with low-vapor-pressure constituents such as KTaO₃ and SrMoO₃ to be achieved.^129,130 Currently, the remaining drawbacks of suboxide MBE are the limited choice of source elements and the lack of high-purity oxide and suboxide source materials.^33,131

Whereas the two trends discussed in Secs. III A and III B represent developments of commonly used film-growth technologies, thermal laser epitaxy (TLE) is a separate technique. Indeed, the evaporation of source materials by lasers has a long history. In 1962, the growth of tellurides from compound sources evaporated with a ruby laser was first presented at the Optical Society of America meeting but not published until 1965.¹³² The same principle was quickly applied to growing metallic films from elemental sources.¹³³ Upon the development of Nd:YAG lasers, they were commonly used for laser evaporation. Subsequently, this technology gained momentum and was used, for example, for growing tellurides.^134–136 The technical boundary conditions at that time caused these lasers to be pulsed.¹³⁷ Note that, because the wavelengths of the beams were not sufficiently short and the power densities of the pulses were small, these laser beams sublimated or evaporated the source materials rather than ablating them as done by PLD.¹³⁸ Owing to the low growth rates and restrictions on film quality caused by the unfavorable deposition conditions used in the 1970s and 1980s for pulsed laser evaporation, this technique was mostly abandoned for PLD, by which epitaxial films of high-T_c superconductors could be grown with great success.^139–141 Here, we note that the advantages of avoiding pulsed evaporation of source materials were already recognized in the 1960s. CO₂ lasers were found to enable rather high growth rates and more uniform films.¹⁴² However, apparently due to high equipment costs and technical issues, the popularity of this technique eventually faded after 1990. A review of the work on laser evaporation in that era can be found in Ref. 138.

The concept of thermally evaporating sources by using lasers was revived in the late 2010s at the Max Planck Institute for Solid State Research in Stuttgart.^143–145 The term “thermal laser epitaxy” (TLE) was then defined by the following characteristics: Evaporation (sublimation) of one or more sources by lasers, the entrance window being protected by an optical component, and the sources and substrate being exchangeable in situ.¹⁴³ Because in PLD, the target is ablated rather than evaporated, PLD and TLE differ in their basic principles. In TLE, the substrate is also preferably optically heated to ensure the maximum available pressure range.¹⁴³ This approach is schematically shown in Fig. 5(a) without the optical element protecting the entrance window of the source lasers. Figure 5(b) shows a 3D rendered image of the inside of a state-of-the-art TLE chamber, which, together with recent technological improvements, is described in the following. Beams of continuous-wave (CW) fiber lasers with a wavelength of ≈1 µm and a typical power of ⪅2 kW are aimed at elemental or compound sources, with one laser usually used for one source. Optical components mounted outside the deposition chamber allow the spot sizes on the sources to be adjusted. This beam control is important for evaporation from solid surfaces as well as from melt pools. Ideally, freestanding solid sources are used for evaporation, which is frequently possible for source materials that sublimate. For source materials that evaporate from the melt, tightly localized heating by the laser allows the formation of a melt pool contained in the source material itself, with some exceptions to be discussed below. The short thermal time scales in TLE make it possible to start, stop, and change the flux of the elemental sources quickly by changing the laser power, thus making shutters optional. As discussed below, the flux of the evaporated source material for a given geometry is a function of the incident power of the laser beam, which allows an estimation of the flux with first order accuracy.¹⁴⁴ At present, TLE systems allow source materials to be changed in the same way as substrates. This avoids breaking the vacuum for source exchange and offers significant advantages in terms of throughput and flexibility. For (oxide) TLE systems, CO₂ lasers are currently the ideal choice for most oxide substrate heating purposes (Table I).

As discussed in Sec. II E, the ≈10 μm wavelength of the CO₂ laser light is well absorbed by most commonly used substrates (Table I), but not by metals. Given the high laser powers and, particularly, the laser power densities available at present, the accessible T_sub parameter space of any deposition method is vastly extended if it is equipped with a CO₂ laser heater, as illustrated in Fig. 3. The large T_sub space available by CO₂ laser heating opens up new possibilities for oxide film growth via the adsorption control mechanism, for example, by using MBE, PLD, or TLE (see Sec. IV C). The long wavelength of CO₂ lasers eliminates the need for beam absorbers thermally coupled to the substrates [Fig. 4(e)]. In this way, a significant source of impurities is avoided,³⁶ which is a considerable advantage over diode laser heaters. The excellent T_sub control and high T_sub change rates of diode laser heaters are also achieved by CO₂ laser heaters. Historically, CO₂ laser heating was introduced for the PLD growth of YBa₂Cu₃O₇.^146–148 This technology was soon applied to substrate heating in CVD,^149–151 where CO₂ lasers had already been used for the enhancement of precursor dissociation.^150–153 

A drawback of using CO₂ lasers is the substantial investment costs, which are partially compensated by the higher available throughputs enabled by the high T_sub ramp rates. Improved equipment reliability and the commercial availability of deposition systems (first by SURFACE for PLD, to our knowledge)¹⁰⁰ have recently revived interest in CO₂ laser heaters. Therefore, it is not surprising that they are currently being used with all major epitaxy techniques.^{146–151,154,155}

The new trends discussed in the previous section offer unique approaches to tackle the challenges of oxide epitaxy. Synthesis routes are being established that were previously thought impossible. In our view, CO₂-laser substrate heating and TLE open a vast and fertile parameter space, and we will now focus on these two technologies. We do so by analyzing the characteristic features of TLE with respect to the transfer of source materials and the role of the oxidation potential of the background atmosphere during TLE growth. Then we present the opportunities provided by CO₂ laser heaters for any film-deposition technology.

Generally, there are two fundamental strategies to ensure the desired stoichiometry of a mixture of species arriving on a substrate. First, one may try to transfer the stoichiometry of a target directly to the film (sputtering and PLD) or, second, one may adjust the flux of independent elemental sources to yield the desired stoichiometry (MBE, MOCVD, ALD, and TLE). As discussed above, stoichiometric targets usually introduce impurities into the system and inherently make it difficult to explore a multitude of different element combinations. Ensuring a stoichiometric transfer over time further proves problematic because the target species with a higher vapor pressure depletes preferentially. The greater the vapor pressure differences among the involved species, the greater the depletion effect. The challenges reported for the growth of KTaO₃ by PLD illustrate this issue. In this specific example, the K deficiency caused by the high vapor-pressure discrepancy is compensated by using a K effusion cell, an approach referred to as hybrid-PLD.¹⁵⁶ 

Figure 6 compiles the vapor pressures of the chemical elements using a color scheme for the temperatures needed to achieve an elemental vapor pressure of 10⁻³ mbar. It is apparent in Fig. 6 that these vapor pressures cover a wide range. As discussed above, MBE elements that require T (p = 10⁻³ mbar) > 1800 °C necessitate the undesirable use of electron-beam evaporation. In comparison, the vapor pressure of the elements plays a minor role in MOCVD/ALD, where the challenge consists of finding a precursor that fulfills the deposition conditions. Table 1 of the Supplementary Material compares the (commercial) availability of source materials for MBE, MOCVD/ALD, and TLE sorted alphabetically, not considering elements that are gaseous under ambient conditions. For TLE, it has been found that all elements used so far are compatible with film growth by sublimation or evaporation.¹⁴⁴ As the elements yet untested by TLE lie within the parameter range of the already tested ones, there is good reason to assume that nominally the entire Periodic Table is suitable for TLE (disregarding potential safety concerns).

Technically, any CW laser is suitable for TLE if its light is absorbed by the source material. Indeed, we have successfully tested several different lasers for TLE. We have found CW fiber lasers operating at λ = 0.5–1 µm and power outputs of ⪅2000 W to be especially suitable. They enable the deposition of any element of interest. Considering Table 1 of the supplementary material, most elements in TLE chambers do not require a crucible and can, therefore, be used as freestanding sources. In this ideal case, contamination by crucible material is obviously excluded. Recent results for the epitaxial growth of sapphire by TLE reveal that, even for elements that require a crucible (Al in this case), the localized heating of the source by the laser leads to film purities even exceeding the purity of the underlying substrate single crystal.³⁶ 

In TLE, the flux of an elemental source is controlled by the incident laser power of the laser beam. This relationship is depicted in Fig. 7 for elements with widely different vapor pressures, using the same color scheme as shown in Fig. 6. It is evident that the laser power required to achieve a specific growth rate does not necessarily scale with the element’s vapor pressure. It is also clear that the source diameter plays a crucial role, as illustrated by comparing the iron fluxes. Additional factors that determine the required laser power are the element’s light absorption (or absorptance for larger irradiated areas) and its thermal conductivity. Despite the complexity invoked by this assortment of factors determining the relationship between laser power and flux, there seems to exist a rather linear relationship between the power of the incident laser beam on the source and the resulting local source temperature for the elements plotted in Fig. 7.¹⁴⁴ To accurately assess the dependence of the flux on the laser power on a general basis, for example, in cases where the source material reacts in an oxidizing environment, an in-depth understanding of the evaporation or sublimation dynamics and oxide formation is required. Obviously, this is often complicated. Workarounds exist, such as the use of high laser powers, for which the source is less easily oxidized.^85,157 This results in growth rates so high that for some applications, these may even be excessive. In our investigations, we have not yet observed any negative effects when using very high growth rates of up to 3 µm/h.

Usually, the growth of complex oxides requires not only a controllable but also a stable flux. For some elements, this is currently a shortcoming of TLE. Flux variations are caused by changes in thermal mass over time, the evolving shape of the source during deposition, and changes in the source surface morphology. Controlling this evolution requires further research. Nevertheless, an elegant and powerful solution exists for the deposition of oxide films with complex stoichiometry because the extended parameter space provided by TLE with CO₂-laser substrate heating enables the adsorption-controlled growth of most oxides. In that case, only the flux of the least volatile species must be controlled to stabilize the composition of the growing film. The applicability of this principle was recently demonstrated by the TLE growth of superconducting Sr₂RuO₄ films with excellent phase purity and very high residual resistivity ratios up to R_{300 K}/R_{2 K} = 88.¹⁵⁸ 

Finally, we point out an intriguing potential of TLE, which is provided by the capability to evaporate or sublimate nominally any element and the compatibility of TLE with robotic operation. These properties of TLE render it able to grow films and heterostructures of any element or compound with a very high level of automation. For example, one may envision artificial intelligence systems that design new films, heterostructures, compounds, or even devices and then fabricate them by robotic TLE film growth and a possible patterning process, with a subsequent AI-controlled analysis closing the loop. This principle is schematically shown in Fig. 8.

In summary, TLE offers several unique opportunities to work with source materials. The availability of practically any element as a source paves the way to the epitaxial growth of oxides containing elements such as refractory metals, the elemental fluxes of which were previously difficult to produce. The practical in situ source exchange, paired with the availability of virtually all elements, further allows the growth of compounds that could not previously be grown by MBE. In addition, it provides unmatched throughput enabled by easy source exchange and high growth rates, which allow numerous compositions to be tested. Matching the ever-increasing speed of modeling materials’ properties by ab initio techniques,¹⁵⁹ TLE offers rapid experimental feedback. Furthermore, the exceptional purity available through localized heating of sources is valuable for improving any (oxide) material for research and electronic applications.

Generally, two challenges are associated with oxidation in film growth: the stabilization of metastable oxidation states and oxidation states that require a high oxidation potential. The wide T_sub range of CO₂ laser heaters allows access to a broad process window in which metastable oxides can be grown epitaxially. In addition, the extremely high cooling rate allows meta-stable phases to be preserved by suppressing further reactions or phase transitions during cooling.^160,161

Figure 9 displays the oxygen pressure (P_O2) required to achieve specific oxidation states of all elements considered relevant for film growth, based on a thermodynamic analysis of binary oxide formation reactions. The majority of these data were extracted from Shang et al.’s work,¹⁶² whereas data for Rh⁴⁺ were separately calculated based on input from Ref. 163. Underlying Ellingham diagrams for each element and oxidation state are provided in Fig. 1 of the supplementary material. Each panel in Fig. 9 is divided by groupings of integer oxidation states and sorted by the oxygen pressure necessary to achieve the respective oxidation states at 1000 °C. The required oxygen pressure is shown for three examples of T_sub at 500 °C (blue), 1000 °C (turquoise), and 2000 °C (magenta). The data show the general trend that high temperatures necessitate high oxygen pressure. The green line marks the maximum equivalent oxygen pressure compatible with MBE (≈10⁻⁵ mbar) for the case of using distilled ozone; see Eq. (1). Whereas many oxidation states are achievable, several of these states lie firmly outside the MBE parameter space. The limitation of the applicable oxygen pressure in MBE results from the mean free path and the necessity to protect electronic parts within the chamber.

The compact design of TLE chambers elevates the pressure corresponding to the mean free path equaling the source substrate distance by an order of magnitude as compared to MBE for standard cases. Avoiding electronic components in the chamber also does not limit the available growth atmosphere. Growth at pressures as high as 10⁻¹ mbar has been tested successfully. Thus, TLE allows previously unattainable oxidation states to be achieved over a wide T_sub range. Elements, such as Ag, that cannot be oxidized at all in an MBE system are potentially within reach for growth in oxide form by TLE.

The importance of the ability to cool quickly has been discussed above. Any laser heating system possesses this ability, and CO₂ laser heaters are especially practical for this purpose. The key capability of CO₂ laser heaters is the avoidance of absorbers, which eliminates thermal stress between absorber and substrate [Fig. 4(e)]. Table I lists substrates and their ability to absorb the 10 µm beam of a CO₂ laser. If the absorption spectrum of a desired substrate material is known, its suitability can obviously also be predicted. If one wishes to use a substrate material that absorbs only weakly at this wavelength (e.g., a metal), one may resort to backside coating of the substrate, e.g., by amorphous Al₂O₃, which is clean and rather stable regarding heat and oxidation.

Generally, two sources of problems during epitaxial growth are eliminated if absorbers are avoided: the absorber itself and its parasitic heat. The excellent T_sub control and fast cooling are illustrated in Fig. 10 for the case of a sapphire substrate heated to 1700 °C in a vacuum in a TLE chamber equipped with cryo-shrouds. The top panel shows the temperature setpoint T_set, the center panel shows the real change of T_sub as a function of time measured by a backside pyrometer, and the bottom panel shows the resulting change of background pressure. Note that transferring the substrate has a greater impact on chamber pressure than heating the substrate. As a result of these excellent vacuum conditions, CO₂ laser heating allows highly pure (oxide) films to be grown.

The most significant benefit of CO₂ laser heating is the range of available T_sub because this provides several opportunities for oxide epitaxy. First, it allows the in situ preparation and termination of the most commonly used and commercially available out-of-the-box substrates, which makes ex situ preparation (chemically or thermally) superfluous.^164,176 The capability of in situ substrate preparation is accompanied by the ability to produce clean interfaces.³⁶ In addition, many oxides require a high T_sub in order to form in the first place, e.g., pyrochlore iridates or sapphire.^34–36,165 As described above, any oxide phase forming at >1000 °C is beyond the capability of most conventional heaters, whereas CO₂ lasers allow these T_sub as well as the typical temperatures required for adsorption-controlled growth to be reached with ease. In many cases, the only factor limiting the achievable T_sub by CO₂ laser heating is the melting point of the substrate.³⁶ However, in some cases, undesired interdiffusion of substrate and film species can impose an additional obstacle (see the supplementary material, Fig. 2). Considering the volatility of all oxide binary species and the T_sub accessible by CO₂ laser heating, the single remaining obstacle to growing oxides by adsorption-control is the oxide’s phase stability at the necessary T_sub.³³ 

An example that represents emerging opportunities is the homoepitaxial growth of c-plane sapphire. High temperatures beyond 1000 °C are necessary for the phase formation and adsorption-controlled growth.^36,165 The high oxygen pressures available in TLE chambers allow for high growth rates even in the adsorption-controlled regime. Furthermore, TLE sources are able to overcome the issues of MBE Al sources in oxidizing environments.⁸⁵ 

Many of the trends and opportunities discussed here are neither specific nor restrictive to the epitaxial growth of oxides. For example, chalcogenides, pnictides, nitrides, carbides, and other material systems pose comparable challenges, which implies that we expect many of the potential solutions presented here to also apply to those systems. The currently available temperature regime of many film-growth technologies for non-oxide materials may be broader (<2000 °C) if a low reactivity of the involved species allows conventional heaters to reach higher temperatures. To provide a glimpse into materials beyond the oxides, we now briefly discuss implications for two other material systems: nitrides and carbides.

Compared to oxides, the growth of transition-metal nitrides and nitride ternaries has not been extensively explored. This is due in part to the fact that they often require deposition conditions outside of the standard parameter space of most deposition techniques. However, many of these nitrides have valuable properties, and epitaxial growth is crucial to unlocking their full potential.¹⁶⁶ Recently, for example, the promising properties of refractory metal nitrides for superconducting qubits have resulted in increased efforts to grow high-quality samples of these nitrides.^167,168 However, their epitaxial growth requires several parameters that remain challenging for standard deposition systems, namely, that refractory metals are difficult to evaporate and high-quality growth requires high T_sub.¹⁶⁸ Both challenges are easily overcome by combining TLE and CO₂ laser heating. A longer ongoing effort is the epitaxial growth of c-BN and h-BN, which faces the same complications, i.e., B is difficult to evaporate, and forming the desired phase requires high T_sub.^169,170 It is expected that TLE can also significantly improve achievable film qualities in such cases.

Similarly, whereas the epitaxial growth of SiC is well established, the high-T_sub required in film growth implies that higher purities of the film are likely to be achieved with CO₂ laser heaters.¹⁷¹ The growth of transition metal carbides and ternary carbides is arguably even less well understood than that of nitrides. The main challenge in this case is the lack of reactivity of the carbon species, which is commonly supplied as C₆₀.¹⁷² In TLE chambers, carbon can be evaporated from a graphite source, which paves the way to resolving this issue.¹⁴⁵ Another option is to utilize a gaseous carbon source of high reactivity, which the inertness of the TLE chamber would probably allow. Much like SiC, many other carbides require high temperatures in order to form, e.g., MoC >1800 °C.¹⁷³ As described above, these T_sub are readily achieved by CO₂ laser heating.

For one carbon-based material, there have been especially intense efforts to identify a method that can deliver high-quality epitaxial films. That material is diamond, whose unique properties not only make it the possibly ultimate ultrawide-bandgap material but also provide the basis to realize intriguing physical phenomena such as single-photon emission.^174,175 The extreme environments needed for diamonds to form result in equally extreme parameters required for epitaxial deposition. The epitaxial films achieved so far lack the desired quality and growth speed.¹⁷⁴ Beyond a doubt, the TLE growth of diamond films will be challenging, but this new deposition technique provides a promising new approach to meet that challenge.

This perspective provides an overview of two central challenges currently faced by oxide epitaxy. First is the difficulty of finding a universal, versatile, and clean way to transfer any element as individual atoms or molecules onto a substrate. Second is the phase formation in the growing film, especially regarding the high oxidization potentials needed to achieve many desired compounds and the high temperatures required for numerous oxide phases to form. Common oxide epitaxy methods have been evaluated, and some recent trends to overcome these challenges have been reviewed here. We have determined that CO₂ laser heaters combined with TLE are one promising approach to overcome these challenges. To our understanding, this combination possesses the potential to achieve significant strides in oxide epitaxy, especially because of the high available purities.

Most notably, CO₂ laser heaters are compatible with any epitaxy method. In particular, CO₂ laser heating enables a huge parameter space to be explored (Fig. 3) and higher film purities to be achieved, making it the most promising substrate heating system for future research. However, if CO₂ lasers are to become the predominant form of substrate heating even outside of research, it will be necessary to upscale their applicability to any wafer size.

TLE, in its current form, is still a relatively young technique. This implies that, along with the opportunities discussed here, the emergence of obstacles has to be expected. Currently, for some elements, the flux stability of TLE is not satisfactory. Corresponding improvements may utilize moving sources⁸⁵ or a variable beam size to suit the respective source.¹⁵⁸ Each of these approaches promises higher flux stability, so it remains to be seen which approach will be successful. Despite this obstacle, TLE presents a currently unmatched research opportunity for epitaxial film growth. As the first results are published, including results based on materials relevant to the industry, it will be exciting to see whether TLE will find a place in the industrial growth of epitaxial (oxide) films.

A general trend to overcome the weaknesses of individual growth methods is the effort to combine their respective strengths. In this paper, we have discussed the combination of MOCVD and MBE (hybrid MBE) and briefly mentioned the use of MBE effusion cells on PLD chambers (hybrid PLD).³² We expect hybrid systems of TLE and other epitaxial film growth systems to emerge. Possible scenarios also include replacing e-beam evaporators in MBE chambers with TLE sources and using low-temperature effusion cells in TLE systems.

See the Supplementary Material for the following: Table 1 lists sources available for MBE, TLE, and MOCVD/ALD; Fig. 1 shows the Ellingham diagram on which Fig. 9 is based; Fig. 2 shows the SIMS profile of a Sr2RuO4 film grown by TLE on LSAT at elevated temperatures, resulting in the diffusion of the substrate species into the first few nanometers of the films; and Table 2 compares the capabilities of the different methods with respect to handling sources.

The authors thank Varun Harbola, Hans Boschker, Y. Eren Suyolcu, and Darrell G. Schlom for many insightful discussions and Wolfgang Winter, Ingo Hagel, and Sabine Seiffert for the technical support. The authors are grateful to Tobias Schwaigert, Shun-Li Shang, and Zi-Kui Liu for assistance in establishing Fig. 9. The authors also thank Matthew R. Barone, Veronica Show, and Joe Falson and his team for additions to Table I, and Tolga Acartürk and Ulrich Starke for supplying data for Fig. 2 of the supplementary material.

W.B. and J.M. are partners in epiray, a company that sells CO₂ laser heaters and TLE systems, W.B. also serves as CEO of that company, and B.D.F. as a consultant of epiray.

F. V. E. Hensling: Conceptualization (lead); Data curation (equal); Project administration (equal); Visualization (lead); Writing – original draft (lead). W. Braun: Conceptualization (equal); Data curation (equal); Project administration (equal); Writing – review & editing (equal). D. Y. Kim: Data curation (equal); Writing – review & editing (supporting). L. N. Majer: Data curation (equal); Visualization (supporting). S. Smink: Conceptualization (supporting); Data curation (supporting); Writing – review & editing (equal). B. D. Faeth: Conceptualization (supporting); Data curation (equal); Visualization (equal); Writing – review & editing (equal). J. Mannhart: Conceptualization (equal); Project administration (lead); Visualization (equal); Writing – review & editing (lead).

The data that support the findings of this study are available within the article and its supplementary material.