Oxide molecular beam epitaxy has emerged as an effective technique to fabricate complex oxide thin films and novel superlattices with atomic‐level precision. In this chapter, we first briefly introduce the oxide molecular beam epitaxy technique and then show how to use this technique to achieve high‐quality thin films with good stoichiometry. Moreover, we exhibit that the combination of oxide molecular beam epitaxy and in situ angle‐resolved photoemission spectroscopy is indeed a versatile toolkit to tailor and characterize properties of novel quantum materials.
- oxide molecular beam epitaxy (OMBE)
- correlated materials
- angle‐resolved photoemission spectroscopy (ARPES)
In transition metal oxides, the subtle interplay among charge, orbital, lattice and spin degrees of freedom gives rise to a spectrum of fascinating physical phenomena, including high‐temperature superconductivity , metal‐insulator transition , colossal magnetoresistance , and so on. Remarkably, in thin film interfaces and ultrathin films of correlated oxides, emergent physics, which does not exist in bulk crystals, occurs [4, 5]. As a well‐known example, two‐dimensional electron gas with high mobility amazingly emerges at the interface of two‐band insulators LaAlO3 and SrTiO3 . This emergent electron gas was even found to be superconducting . Another example is that strong ferroelectricity and ferromagnetism were found in EuTiO3/DyScO3 superlattices . As Nobel laureate Herbert Kroemer said that ‘the interface is the device’ , these emergent physics may potentially revolutionize our modern technologies.
In order to access these thin film‐based physics, the first and most important step is to grow these oxide thin film structures with high quality. This needs exquisite control of growth, and usually is very challenging. In the past several decades, fortunately, reactive oxide molecular beam epitaxy (OMBE) has been proved to be an effective technique in the growth of some oxides with high quality, though being not easy [10–12]. Recently, the in situ combined system of OMBE and angle‐resolved photoemission spectroscopy (ARPES) [13–17] have shown great potential in exploring intricate many‐body physics based on oxide film structures, which further intensifies the power of OMBE.
In this chapter, we first present basics of OMBE technique. Then, we show how to grow high‐quality films with good stoichiometry, and the power of the integrated OMBE‐ARPES in studying and designing many‐body interactions in complex oxides.
2. Basics of OMBE technique
MBE is a vacuum deposition technique in which well‐defined thermal beams of atoms or molecules react at a crystalline surface to produce an epitaxial film. Originally, it was developed to fabricate GaAs and (Al, Ga)As films , and soon successfully expanded to other semiconductors as well as metals and insulators. In addition to molecular beams coming from individual heated element source, gas molecular may also be introduced into MBE. Including gas oxidants (e.g., oxygen or ozone) can make an OMBE, which is now applied to grow oxides [11, 12].
In 1985, Betts and Pitt began to use this technique to grow LiNbO3 films . Later, motivated by the discovery of high‐temperature superconductivity, OMBE was used to grow complex cuprate thin films. Up to now, it has been broadly employed to fabricate a pool of oxides, including oxide superconductors (e.g., (Ba, K)BiO3, (La, Sr)2CuO4, Bi2Sr2Can‐1CunO2n + 4, etc.), ferroelectrics (e.g., LiTaO3, PbTiO3, etc.), ferromagnets (e.g., (La, Ca)MnO3, EuO, etc.), multiferroics (e.g., BiFeO3, YMnO3, etc.) and superlattices of these phases [11, 12].
While conventional MBE growth occurs in an ultra‐high vacuum, in OMBE growth the induction of active gas oxidants can pose new challenges in the instrumentation as well as the film growth [11, 17]. The presence of oxidant species requires the hardware to be necessarily compatible with an oxidizing environment, thus high‐temperature components (e.g., heater filaments, effusion cells and substrate holders, etc.) need to be made of highly oxidant‐resistive materials. Moreover, adequate pumping is needed to deal with the oxidant gas load. Furthermore, oxygen acts as another variable which needs to be optimized in the growth, and oxygen inside films is tricky to study and manipulate. In addition, the oxidants can oxidize the cell materials such that one cannot get well‐controlled fluxes as planned during growth. These challenges make the use of OMBE in the growth of oxides less mature than the use of MBE in semiconductor growth .
Figure 1 shows the schematic of a typical OMBE system. Single‐element evaporators are used to generate atomic beams for OMBE growth. Knudsen cells and crucibles are chosen for elements with desired fluxes below 2000°C, while electron beam evaporators are adopted for refractory elements (e.g., tungsten, ruthenium and iridium) which require higher temperature to provide the fluxes necessary for the growth. The atomic beams impinge upon the substrate unless they are blocked by shutters which are positioned at the output end of each cell and remotely controlled by a computer. The utilization of shutters enables the elemental fluxes to be supplied to in a continuous or a sequential way. The fluxes can be adjusted by changing the cell temperature, and are in situ measured by a quartz crystal microbalance (QCM). Reflective high‐energy electron diffraction (RHEED) is used in OMBE for the in situ characterization of the growing surface. Due to the grazing angle diffraction, it can provide surface‐sensitive information including thin film crystallinity, roughness, in‐plane lattice constants, growth mechanism and phase purity. If intermediate products or impurity phases are formed, the growth conditions would be adjusted accordingly. Distilled ozone was used as the oxidant. Compared to oxygen, ozone has stronger oxidizing ability and thus needs lower pressure.
Figure 2 displays the photo of DCA R450 OMBE system in Shanghai Institute of Microsystem and Information Technology (SIMIT). It is equipped with 10 changeable effusion cells, and one four‐seat e‐beam evaporator for at most four refractory elements, which can cover all transition metals of interests. It also has in situ QCM to measure the fluxes and real‐time RHEED to directly monitor the growth. Ozone was obtained from self‐made ozone‐generating‐and‐distilling system. Ozone generator would generate a small amount of ozone out of oxygen gas, then silica gel cooled down with liquid nitrogen would absorb ozone. Warming up the silica gel would give out the ozone gas to be used in OMBE growth.
3. Fabricating oxide thin films with good stoichiometry by OMBE
In this section, we show generally how to grow high‐quality oxide thin film with good stoichiometry by OMBE. Starting with choosing the proper substrate, we mainly talk about two growth methods commonly used—absorption‐controlled growth and shutter growth—to achieve high‐quality films.
Epitaxial thin films cannot be obtained without using proper substrates. The substrate not only guides the growth of thin film with the right crystalline structure but also provides the knob of strain which can essentially tune the electronic structure of the material . Figure 3 displays lattice constants of single‐crystalline perovskite substrates which are commercially available and commonly used to grow perovskite or layered perovskite oxide thin films. If growing LaNiO3 films with a pseudo‐cubic lattice constant of 3.84 Å on LaAlO3 (3.75 Å) substrates, an in‐plane compressive strain was applied; if using SrTiO3 (3.905 Å) as the substrate, an in‐plane tensile strain was applied. In addition to the strain, in some cases, the choice of the substrate is vital to obtain high‐quality films. For example, Proffit et al. reported that (1 1 0) orthorhombic CaRuO3 films grown on orthorhombic (1 1 0) NdGaO3 substrates (symmetry matched) exhibit atomically smooth surfaces, whereas films on cubic lanthanum aluminate-strontium aluminium tantalate (LSAT) substrates (symmetry mismatched) show rather rough surfaces . Another example is the film growth of LaNiO3 with polar orientations. As polar discontinuity is suggested to induce surface reconstructions which further lead to bad quality of films [22, 23], metallic Nb‐doped SrTiO3 and iso‐polarity LaAlO3 substrates were shown to be more suitable than the common SrTiO3 in the growth of LaNiO3 films .
As illustrated above, the remotely controlled shutters in OMBE allow elemental fluxes to be supplied to in a continuous or a sequential way. Take the growth of perovskite ABO3 (can be viewed as alternate stacking of AO and BO2 layers along the (0 0 1) direction) as an example. As schematically shown in Figure 4, both shutters of A cell and B cell keeping open in the whole growth make a co‐deposition growth. If the shutters of A cell and B cell alternately turn open (finishing the growth of one AO layer, and then starting the growth of one BO2 layer), we can call this the shutter growth. In either growth, stoichiometry is the most important goal which needs to be achieved.
A well‐known growth method to achieve stoichiometric films is the adsorption‐controlled growth, which was previously used to grow GaAs and recently has been used to fabricate several oxides such as PbTiO3 , BiFeO3  and BiMnO3 . It is used in the growth of some compounds containing volatile species which can re‐evaporate during growth while the others are less volatile. If always making this volatile species excess during the growth as well as optimizing the substrate temperature and oxygen partial pressure, stoichiometric growth can be conveniently achieved. Lee et al. reported the adsorption‐controlled growth of BiMnO3 in which the bismuth oxides are volatile . Figure 5 displays the calculated Ellingham diagram and obtained RHEED patterns . The Bi:Mn flux ratio was fixed to be 3:1. Besides, the substrate temperature and oxygen partial pressure were fully explored to finally expose the growth window (see shadow region II in Figure 5) for phase‐pure stoichiometric BiMnO3 films which were verified by the shiny diffraction spots in the RHEED pattern.
For most oxides, adsorption‐controlled growth unfortunately cannot be applied. To achieve good stoichiometry, generally, one has to adjust the deposition amount based on cycles of combined studies of QCM, RHEED patterns and oscillations, X‐ray diffraction (XRD) pattern fitting, Rutherford backscattering spectroscopy, and so on. In the homo‐epitaxial growth of SrTiO3, Schlom's group reported the empirical method of optimizing shuttered RHEED oscillations to successfully achieve stoichiometric SrTiO3 film within 1% composition deviation [28, 29]. In the shutter growth of SrTiO3, the intensity of diffraction spot would exhibit periodic oscillations at the pace of mechanically closing/opening shutters: in Ti doses, the intensity will decrease monotonically while in Sr doses the intensity will increase. It was shown that if oscillations exhibited smooth sinoidal shape with similar amplitude (Figure 6(a)), the resulted film was investigated to be stoichiometric . If Sr is 10% excess, the combined feature of cusp and shoulder would show up (Figure 6(b)); if Sr is 10% deficient, the amplitude of individual oscillations would oscillate (Figure 6(c)). Thus, in growth, the real‐time performance of RHEED oscillations would infer what to do next to achieve the stoichiometry .
Compared to adsorption‐controlled growth, shutter growth is a more straightforward way to control the film thickness and grow complex oxide structures such as Ruddlesden‐Popper (RP) series An+1BnO3n+1 and various superlattices which display a wide range of physics. By conveniently changing the shuttering sequence of A and B ions to match the layering sequence of the desired RP phase, Haeni et al. and Tian et al. reported the OMBE growth of five RP members of Srn+1TinO3n+1  and Srn+1RunO3n+1 , respectively. Their structures with right‐layering sequences were verified by high‐resolution cross‐sectional TEM measurements (Figure 7).
4. The unique in situ combo of OMBE and ARPES
ARPES can directly visualize electronic band structures of solids, and therefore has emerged as an essential experimental technique to study various novel quantum materials such as superconductors and topological quantum materials [13, 14, 32]. It can be viewed as the ‘k‐space’ microscope, and can provide the essential information about how electrons move inside the material. Based on the well‐known photoelectric effect, an electron inside the solid can absorb an incident photon with a high enough energy and then emit out of the solid. If the kinetic and momentum of the photoelectrons are detected, the band structure of the material (as a function of binding energy and momentum) can be reconstructed in the context of conversation laws and some reasonable assumptions, as shown in Figure 8. Generally being a surface‐sensitive probe, ARPES demands clean and well‐ordered sample surface, which is usually obtained by cleaving single crystals.
Recently, there has been increasing awareness that the in situ combo of OMBE and ARPES (ARPES is in situ connected to OMBE) could be a powerful approach in tailoring many‐body interactions and uncovering the underlying physics of complex oxides. It is a win‐win case: OMBE can provide high‐quality thin films (especially those who cannot be cleaved properly in bulk form, e.g., perovskite oxides) and interesting superlattices, whose clean surfaces ‘naturally’ allow for the in situ ARPES studies; in return, ARPES can characterize the quality of these films which further give feedbacks to the growth, and moreover it can fully explore the band structures of films and study their intriguing physics [15–17, 33, 34].
Figure 9 shows the photo of such an in situ combo system located in Shanghai Institute of Microsystem and Information Technology (SIMIT), Chinese Academy of Sciences. A transfer chamber with high vacuum (∼10 × 10-10 Torr) is used to bridge the OMBE system and ARPES system. Once the growth was finished, the film was immediately transferred to the ARPES chamber within 5 min through the transfer chamber. In so doing, the clean sample surface is expected to be preserved for ARPES studies.
Below, we first present studies on ultrathin perovskite LaNiO3 films (3d system) , SrRuO3 films (4d system)  and SrIrO3 films (5d system)  of which the bulk form lacks natural cleaving planes, as examples to show the power of this in situ combo in studying the many‐body interactions of complex oxides. Generally, when going from 3d to 5d elements, Coulomb interactions among electrons would increase as electrons are more and more delocalized, while the spin‐orbital coupling (SOC) would become stronger due to the heavier elements. Then, we show the studies on (LaMnO3)2n/(SrMnO3)n superlattices .
Bulk LaNiO3, though being strongly correlated with 3d7 configuration of Ni3+, remains a paramagnetic metal at low temperatures. By means of OMBE, King et al. synthesized atomically defined layers of LaNiO3 films down to just one unit cell (UC) thickness, and observed an abrupt metal‐insulator transition in 2‐UC‐thick film by transport studies . The high quality of the films with atomically flat surfaces was revealed by the Kiessig fringes in XRD patterns and atomic force microscope image (Figure 10(d)). Then, in situ ARPES studies were carried out to investigate the competing electronic phases while crossing the transition, as shown in Figure 10(a) and (c). The bulk‐like electronic structure and Fermi liquid characteristics were found to remain almost unaffected by film thickness down to 3 UC (all exhibiting similar electron pocket), which is in contrast to the previous reports that an insulating state emerged in 5‐UC thickness or above and again reflects the high quality of the films. This makes 3‐UC‐thick LNO the thinnest metallic nickelate reported. Reducing the thickness by just one further UC (2 UC), however, causes the spectral weight near EF to be suddenly suppressed. For 1‐UC‐thick LaNiO3 film, no spectral weight was observed at EF, indicative of a full‐charge gap. This evolution was also observed in angle‐integrated photoemission spectra. The authors claimed that the metal‐insulator transition is driven by instability to an incipient order of the underlying quantum many‐body interactions, and demonstrated the power of artificial confinement to harness control over competing phases in complex oxides with atomic‐scale precision .
Perovskite SrRuO3, a prototypical conductive ferromagnetic oxide, exhibits a kink in its band dispersion signalling the unusual electron dynamics therein . The kink could originate from electron‐magnon coupling or electron‐phonon coupling. Uncovering the origin of this kink would hint on the studies of kinks’ origins in many other intriguing systems [37–39] including the cuprate superconductor family . Yang et al. reported the systematic thickness‐dependent electronic structure studies on SrRuO3 films with well‐controlled thicknesses by using the OMBE and ARPES system . Figure 11(a) shows the evolution of band dispersions of SrRuO3 films with reducing the film thickness. Evidently, in all these spectra, the slope of the dispersion near EF is markedly smaller than that of high‐binding‐energy region, namely the kink persists even down to 4‐UC‐thick film. Empirical self energy analyses were carried out to determine the kink energy. As shown in Figure 11(b) and (c), both real part and imaginary part analyses reveal that all films’ kinks are around 62 meV below EF. This is in sharp contrast to the report that reducing the thickness would decrease the Curie temperature , which implies that electron‐magnon should not play a dominant role. On the other hand, the kink energy matches that of in‐phase‐stretching phonon mode proposed by the combined studies of Raman spectroscopy and calculation . Thus, electron‐phonon coupling should mainly contribute to this kink. This work can serve as an example to study the low‐energy excitations of complex oxides.
Perovskite SrIrO3, due to the heavy element of Ir, is expected to have strong SOC which is the key ingredient in building topological quantum materials [43, 44]. Therefore, novel topological phases were proposed in artificial SrIrO3‐based structures . In particular, SrIrO3 was proposed to be an exotic semimetal induced by the delicate interplay between SOC and electron correlations, in which a Dirac nodal ring near the U point would render a non‐trivial topological semi‐metallic state [46, 47]. Based on the OMBE‐ARPES combo, Liu et al. synthesized high‐quality SrIrO3 films, and investigated its low‐lying electronic structure . Figure 12 displays the measured band structure of SrIrO3. In addition to the semimetal state (the Fermi level simultaneously crosses the hole and electron pockets), near the U point, a lifted Dirac node was directly observed, which agrees well with the authors’ calculations. This Dirac node lifting could be due to the selectively breaking of n‐glide symmetry in the hetero‐epitaxial SrIrO3 structure . Iridates would continue acting as the profound platform to explore novel physics that may combine the SOC and electron correlations.
With the powerful capability of the OMBE‐ARPES combo, one can fabricate artificial superlattices which do not exist in nature and study their intriguing emergent physics. Monkman et al. reported the comprehensive investigations on the interfacial electronic structure of (LaMnO3)2n/(SrMnO3)n superlattices as a function of dimensionality . Bulk LaMnO3 and SrMnO3 are anti‐ferromagnetic Mott and band insulators, respectively, and La2/3Sr1/3MnO3 is a ferromagnetic metal that exhibits colossal magnetoresistance around its Curie temperature of 370 K. Monkman et al. synthesized high‐quality (LaMnO3)2n/(SrMnO3)n superlattices with n = 1–3. Transport studies show that n = 1 and n = 2 members show metallic behaviours at low temperatures, while the n = 3 member exhibits a metal‐insulator crossover. Figure 13 illustrates the evolution of the electronic structure and properties of superlattices upon different n. For n = 1 and 2 members, the Fermi surfaces are apparent and consist of two Mn eg‐derived states: a hole pocket around the Brillouin zone corner, and a smaller electron pocket around the zone centre. For the insulating n = 3 member, the spectral weight at EF is highly suppressed, although clear states are still observed below EF. The authors also investigated the near‐EF band dispersions. The n = 1 and 2 members exhibit well‐defined and dispersive bands, whereas the n = 3 sample shows only pseudo‐gapped intensity at EF which is similar to polaronic systems with strong electron‐phonon coupling. For the critical n = 2, there exists a dramatic difference in electronic states compared to that of n = 1. The states near EF are substantially suppressed, and a broad incoherent feature appears below EF, indicating the enhanced correlations. This work provides unique insight into how many‐body interactions could be engineered at correlated oxide interfaces, which is an important prerequisite to exploiting such effects in novel electronics .
These examples reflect the powerful capability of the integrated OMBE‐ARPES system in studying many‐body interactions and resulted novel physics in complex oxides.
5. Conclusion and outlook
In this chapter, we presented the brief inductions to OMBE technique, growth methods and the in situ combo integrating oxide MBE and APRES. We demonstrate that OMBE and the in situ combo formed with ARPES will continue playing a role in unveiling the intriguing many‐body physics of correlated oxide thin films and superlattices as well as exploring novel topological materials in oxide structures.