The preparation of the stencil mask is one of the crucial steps for SL. Depending on the demanded vacuum condition (the mask may outgas at the growth temperature) and the required resolution of the pattern, various materials could be used for constructing the mask. For example, Kovar alloy has been used for constructing the mask to sputter a patterned film^[22], and refractory metal such as tantalum was used in molecular beam epitaxy-SL growth, both of which are compatible to ultra-high vacuum and can achieve hundreds of μm- to mm-size pattern. Due to the surface roughness and fluctuation, these masks cannot be machined to contact closely with the substrate surface and, hence, they only could be used if the resolution of the pattern is not critical. To address this issue, the low-stress silicon-nitride membrane and silicon wafer with extremely flat surfaces have been widely used as (semi-)rigid stencil masks, which are also compatible to ultra-high vacuum. Thanks to the mature semiconductor techniques, the feature on these masks can reach a resolution down to sub-100 nm. The silicon-nitride layer with thickness of hundreds of nm is usually deposited on a single/both surface(s) of a silicon wafer, which is an ideal material for constructing the mask in SL owing to its great thermal stability, mechanical strength, and chemical inertness. The crystalline silicon mask constructed via anisotropic etching becomes an increasingly popular choice as it allows a more robust mask structure than silicon-nitride membrane of similar thickness, which is a rapid, low-cost, and wafer-scale process^[72-73]. It has been reported that masks made of both the silicon-nitride membrane and silicon wafer can achieve a finest resolution of tens of nm for patterning various metal films^[9,74-76].

Besides the advantages reviewed by Refs[12-13], SL owns the following advantages that quantum material devices are particularly benefitted from:

The usual procedure for fabricating micro-/nano-scale devices, including etching the material to a specific geometry and electrode depositions, relies on techniques such as the photolithography and e-beam lithography. However, in some circumstances, it would be necessary and convenient to fabricate devices in resistless manners. Indeed this manner is equivalent to applying these resist-techniques to fabricate the small structures on stencil masks rather than directly on the substrate or material, which leaves the material and substrate pristine, that is, the SL. In this way, the material deposition processes only include the substrate, evaporated material, and the mask without any further resist, chemical solvents, energy radiation, or mechanical pressure. All the materials involved in SL-deposition could thus be carefully selected to accommodate the ultra-high vacuum. Therefore, the SL maximizes the cleanliness in material deposition and the quality of the device based on in situ fabrication. This compatibility allows the SL to be readily expanded to all of the physical vapor deposition techniques as well.

The stencil masks made by lithography and etching allow for the transmission of atoms, ions, or molecules from the openings to the substrates, which could not only be used in all kinds of physical vapor depositions but also the etching and ion implantation to a certain pattern. Different designated patterns and features can be reliably constructed based on the conventional micro- and nano-fabrication techniques, which could fulfil the modern development of multifunctional device structures. As the stencil mask is placed over the substrate, it can be flexibly manipulated among different fabrication processes to achieve multiplex device structures. For example, one can place the mask directly on a substrate before the deposition to achieve the selective area growth on a substrate; or after multiple continuous film depositions, the mask is then placed for the final-step metal deposition to form a heterostructure devices with electrodes; or only use the mask to deposit some pattern that is sandwiched between two continuous films, forming an embedded nanostructure at the interface between the two materials; etc. Such high flexibility and variety in both the layer structure and patterned structure showcase the power for engineering multifunctional devices.

The stencil mask can be removed after patterning or etching from the substrate-mask assembly and placed onto a new substrate for the next batch of use. Such a process can be reused for many times, which substantially reduces the fabrication costs and provides an efficient patterning replication using different evaporated materials onto all kinds of substrates. Long-term evaporation through the stencil mask will be accompanied by gradual clogging of the openings and apertures due to the adhesion of the evaporated materials onto the mask. However, the clogging can be fully removed by etching or ion-implantation. Moreover, to reduce the clogging, the stencil mask (particular for those with nano-scale apertures) can be coated with alkyl and perfluoroalkyl self-assembled monolayers, which were shown to considerably reduce the adhesion of the evaporated materials onto the mask^[77-78]. It was reported^[78] that such a coating technique can result in an increase in material deposition through the apertures by more than 100%. Therefore, the clogging is not an issue by routinely etching or the self-assembled monolayers coating, improving the performance and lifetime of the stencil mask and hence increasing the throughput of SL.

As stated above, most of the quantum materials are indeed not very stable even in ambient conditions. Multiple device fabrication processes can further degrade the material quality or even eliminate the quantum effect in the material. Conventional fine-scale device fabrication processes, such as coating and baking of resist and the e-beam lithography, can inevitably dope the material more or less, which modifies the position of Fermi level, while the application of high-voltage and scanning of e-beam may irreversibly destroy the crystalline structure of the material. Thus, the more complex fabrication processes, the lower the yield of device fabrication will be, either of which fails can eventually cause the malfunction of the whole device. Therefore, the device yield could be substantially increased if substituting part of the conventional fabrication processes with the SL, or even finishing the entire device fabrication using the SL only. Also, traditional patterning techniques, in particular the e-beam lithography and focused ion beam, are slow when patterning for a large area, and the patterning process needs to be repeated for every different sample, which is expensive and time-consuming. These problems limit the scalability of the devices. Instead of repeating these complex processes, the stencil mask can be made to cover the entire substrate surface and pattern the mesoscopic structures with one-time

deposition. A rigid or semi-rigid stencil mask, such as silicon wafer and silicon-nitride on silicon wafer, has been shown to be sufficiently strong to self-sustain for a large area, which enables the fabrication of mesoscopic devices on a large scale^[76]. Early in 2009, it has already been reported that the SL using a silicon-nitride membrane enables to pattern sub-100 nm superconductive tunneling junction array at a full 100-mm wafer scale with 1 μm alignment resolution^[79].

The applications of SL have been used in patterning various materials and their devices on all kinds of substrates, such as plasmonics, transistor, magnetic and superconductive mesoscopic structures, biosensor, flexible and wearable electronics, solar cells, etc., which have been covered in several other excellent reviews^[12-13]. Below we only emphasize the designated applications of SL in fabricating quantum material devices.

As stated above, so far most of the quantum materials that are potentially applied to electronic devices applications are fabricated as thin film, especially in the 2D limit. It is known that the quantum effects of these materials are significantly affected by the electrical contacts that connect these materials with external circuitry^[80]. However, it remains challenging to create high-quality electrical contacts for these thin quantum materials because several interface issues, leading to high contact resistivity and undesirable contact type^[81]. As stated above, the SL can be used for in-situ growth and device fabrication without breaking the vacuum of the system. Generally, to fabricate a mesoscopic device, the resist will be commonly used in lithography processes. However, it is commonly observed that there will be some residual resist on the contact area after development before the metal deposition^[82], which hinders or degrades the formation of a clean Ohm contact to the quantum materials, or increases the contact resistance. Moreover, the residue on the sample surface will contaminate the sample andcause incidental a doping effect, which locally modifies the electrochemical potential and provides extra scattering sites to the quantum material. Thus, it is believed that the residual resist is one of the causes of the bottleneck for further increasing the sample mobility. This is particularly crucial for 2D materials such as graphene, stanine, and black phosphorus that consist of only a few monolayers of atoms and, hence, very sensitive to surface contaminants. N. Staley et al. developed a lithography-free for fabricating graphene planar tunnel junction using ultrathin quartz filaments as stencil masks^[83]. Possible weak localization behavior and an apparent reduction of density of states near the Fermi energy were observed. Later, W. Bao et al. improved the lithography-free technique for fabrication of clean, high quality graphene devices based on the silicon stencil masks^[84]. As shown in Fig. 2(a), a 100-μm thick silicon layer was used as the stencil mask, which was patterned using e-beam lithography and inductively coupled plasma etching for depositing the electrical contacts. After aligning the resulted mask to the graphene sheet, the entire assembly was transferred into a vacuum chamber for metal deposition as shown in Fig. 2(b). It was found that the metal contacts will typically extend beyond the mask openings by ~ 0.3–0.5 μm due to the extended size of the metal source and the finite mask–device separation. Graphene devices fabricated by this technique show significantly higher mobility values than those obtained by the standard e-beam lithography. This technique has also been extended to fabricate a suspended graphene device, whose mobility can be improved to as high as 120,000 cm²/(V·s). The mask used in this study was reported to be exceedingly robust, which can be used for more than 20 times.

Heterostructures involving quantum materials with conventional materials usually bring about surprising quantum effects at the interfaces. To probe the effect arisen from the quantum material or how the involvement of the quantum material affects the physicochemical properties of the whole heterostructure, convincing evidences could be obtained from a control experiment that can immediately compare a sample with the quantum material and another without. However, such a control experiment may suffer from a problem that the constant variables in the experimental samples and control samples cannot be always kept precisely the same. For example, owing to the fluctuations in material deposition conditions including the flux rate, substrate temperature, surface topography, and vacuum, etc., the quality of samples varies from batch to batch more or less. This will challenge the fabrications of both samples that show the same constant variables and designated control variables. Thus, the conclusion extracted from such a control experiment could be elusive. It seems that this problem may be solved by fabricating a number of experimental and control samples, however, this approach increases the total cost and cannot fully address this problem anyway.

From this point, the SL provides another method that could solve this problem. Take the topological insulator film grown by molecular beam epitaxy-SL as an example. It was theoretically shown that the topological surface states from can act as an effective electron bath that allows oxygen and carbon monoxide molecules more prone to dissociate and get adsorbed on the surface of a noble metal/topological insulator heterostructure via enhancing the molecular adsorption energies^[85]. A desired control experiment would be comparing the surface reactivities of the noble metal/topological insulator heterostructure with the noble metal film only. However, the surface reactivity may also be varied by the delicate details of the surface topographies of the substrate and the noble metal film, which also vary from sample to sample owing to the deposition conditions. Thus, fabricating two or more samples and comparing their differences may not be able to lead to a convincing conclusion. In this case, the SL could be used to address this problem by separating a single piece of substrate into two halves, which could act as the experimental and control samples respectively. In this way, the fluctuations in constant variables should be minimized. Experimentally, half the surface of a single piece of GaAs substrate was covered by a tantalum strip in situ in prior to the deposition of 7-nm topological insulator film Bi₂Te₃. The tantalum strip was then removed in situ to expose the entire sample surface before the deposition of a 9-nm-amorphous-Pd overlayer. In this way, the control experiment could be performed using this single piece of sample, half of which contains the Pd/Bi₂Te₃ heterostructure while the other half is solely the Pd film. After exposing this sample to ambient conditions, one can compare the time-of-flight secondary ion mass spectra, XPS core-level and Auger spectra of the involved elements as shown in Fig. 3. It is clear that Bi₂Te₃ film could enhance the adsorption of organics on the surface of Pd and the oxidation of Pd. Since Te is a congener of O and can act as an oxidizing agent for the metallic Pd, the control sample could be improved

by in situ evaporating Te (nominal thickness of 1.3 nm) onto the entire surface of the above control sample. It was found that the surface concentration of Te and the concentration ratio of Te/Pd in the region with Bi₂Te₃ film are substantially higher than those obtained from the region without. Therefore, by these control samples fabricated using the SL, the surface reactivity of Pd in the region Bi₂Te₃ film can be clearly shown to be enhanced by the underlying Bi₂Te₃ film.

Another example is the direct evidence of the Bi₂Te₃ film-induced interfacial superconductivity in FeTe based on the control sample fabricated using SL. Likewise, the experimental and control samples as shown in Fig. 4(a) are deposited onto a single substrate in order to minimize the fluctuations in constant variables. By measuring the temperature-dependent resistance curves at various regions of this sample shown in Fig. 4(b), it was found that the region fully covered with a Bi₂Te₃ layer shows a single superconducting transition, while the region without Bi₂Te₃, equivalent to a pure FeTe layer, exhibits a semiconducting behavior without any signature of superconductivity. For the middle region that consists of a Bi₂Te₃/FeTe heterostructure and a pure FeTe layer, the resistance curve show similar characteristics in the superconducting transition to that of the Bi₂Te₃/FeTe heterostructure except a different residual resistance. This residual resistance originates from the non-superconducting part in the pure FeTe layer. All these characteristics of the control sample demonstrate thatthe Bi₂Te₃ film is indispensable for inducing superconductivity in the Bi₂Te₃/FeTe heterostructure. Therefore, the SL is helpful for designing an ideal experimental and control sample that could be deposited onto a single piece of substrate. The experiments performed on such a sample could thus provide strong evidences to support the conclusion on the emergent effect.

Devices that could realize designated functions would be particularly benefitted from the SL in terms of, firstly, the clean interface that can dramatically enhance the interfacial coupling of the two constituent materials. Typical examples are the superconductive and magnetic proximity effects, which require atomically flat and clean interface for extended electron wavefunction overlapping and Cooper pairs tunneling. Here, the clean interface is very necessary since the phase coherence lengths in both effects are generally short, ranging from a couple of nm to about hundreds of nm even in the clean limit. If the interface is contaminated by some residual resist and adsorbate, or oxidized by the ambient air, i.e. the heterostructure is formed by some ex-situ technique, such lengths will be substantially shortened, which will diminish or fully eliminate the proximity effect. Secondly, instead of continuous thin-film heterostructures that are normally grown by various physical vapor depositions, discontinuous films and thin-film heterostructures in designated patterns and geometries can be formed using SL. In this way, specific functional devices can be realized without additional etching or deposition processes. Here, these device structures include the Hall bar, Corbino structure, tunneling junction, Josephson junction, arrays of dots and wires, etc., which are widely used and important for engineering the quantum effects.

Using in-situ SL in molecular beam epitaxy, P. Schüffelgen et al. fabricated superconductor-topological insulator-superconductor Josephson junctions and networks of hybrid nanostructures. Such a device aims at the realization of superconductive proximity effect from Nb to the topological insulator, giving rise to the topological superconductor that is promising for constructing the topological qubit in quantum computation^[87]. As shown in Fig. 5(a), a thin Si₃N₄ stencil mask was monolithically integrated to the Si(111) substrate in prior to the epitaxial growth of the topological insulator film (Bi,Sb)₂Te₃ on the Si substrate. The deposition of 50-nm Nb layer was then performed over the mask, which is a stencil bridge for separating the continuous Nb film into two parts. In this way, the (Bi,Sb)₂Te₃ film is covered by the Nb layer except for a narrow strip shaded by the stencil bridge, forming a Nb-(Bi,Sb)₂Te₃-Nb Josephson junction as shown in Fig. 5(b) with a separation of ~ 80 nm. To avoid degradation of the resulted device after exposing to the ambient conditions, the entire device was in situ capped by 5 nm of Al₂O₃ layer. As shown in Fig. 5(c), the transmission electron micrograph of the cross- of this Josephson junction clearly shows the layered structure of the rhombohedral (Bi,Sb)₂Te₃ crystallinefilm, while the two Nb electrodes also exhibit satisfying crystalline quality, which may be benefitted from the atomically clean surface of the (Bi,Sb)₂Te₃ film. Such an in-situ SL technique allows the Josephson junction to be readily measured without further ex-situ fabrication processes, which preserves the most pristine properties of the superconductive proximity effect. The transport measurement over such a device shows a large critical current and ballistic superconducting coherence length to ~110 nm, successfully inducing a superconducting gap as large as 0.6 meV in the (Bi,Sb)₂Te₃ film beneath the Nb electrodes. Further analysis indicates an excellent in-plane transparency of 0.95 between the proximitized (Bi,Sb)₂Te₃ film and the non-proximitized part. Due to such a high interface transparency, the supercurrent could be solely carried by the topological surface states in (Bi,Sb)₂Te₃ across a wide temperature range. Additionally, the Shapiro response of the Josephson junction shows a full suppression of the first Shapiro step at low frequency, which is believed to be a signature of the Majorana bound states in the topological superconductor.

The SL not only provides a technique to fabricate some micro-/nano-scale devices in situ but also can be applied to more complex layout. As shown in Fig. 6, the (Bi,Sb)₂Te₃ film was firstly only selectively grown on the Si(111) surface in the molecular beam epitaxy system. In a second in-situ step, the stencil mask allows to deposit superconductive Nb islands to form parts of the network, which was well aligned to the (Bi,Sb)₂Te₃ network. In this way, a theoretically proposed mock device compromising the topological insulator-superconductor network for topological qubits as shown in Fig. 6(a) was in situ fabricated using SL. Another circuit for using circuit quantum electrodynamics to detect topological superconductivity was also fabricated using a similar SL method^[88]. In this circuit, the topological insulator-based transmons were found to scale with their Josephson junction dimensions, which demonstrates a qubit control as well as temporal quantum coherence. These results made the first step towards the investigations of topological materials in both novel Josephson and topological qubits.

There are some problems and challenges when using SL. First, reflection high energy electron diffraction (RHEED), which is usually used for growthmonitor, may not be compatible with the SL-growth since the stencil mask closely contacts with the substrate. If the selective area is smaller than the RHEED beam size, one still can ensure the film quality by taking all the optimized growth parameters that were obtained using RHEED without capping the substrate with the stencil mask. If the selective area is sufficiently large for RHEED monitor, this method can still be used. Otherwise, a special mask adapter that is compatible with RHEED needs to be designed. A typical example can be found in Ref. [89], which has a tunnel on the reverse side of the mask that allows the entrance and exist of the RHEED beam. Second, the blurring of the edges in the SL-grown film. In practice, the stencil mask can never perfectly contact with the substrate due to the stress, curvature, or topography of the two surfaces no matter the size of the substrate. The narrow gap between the two surfaces can cause the scatterings of the flux and thus the blurring of the selective area, which not only limits the resolution of the pattern but also degrades the quality of the device. So far many methodologies had been used to reduce the blurring, which had been extensively discussed^[13,90-94]. Third, the limitation of the film thickness. After the SL-growth, the undercut profile at the edges of the pattern will, in turn, ensure a discontinuous film during the deposition if the film thickness is less than that of the mask. For a thick film around or thicker that of the mask, it may be difficult to remove the mask after SL-growth. One will need to prepare a thicker mask to address this problem. Also, it is possible that the mask will damage some of the substrate surface more or less. This can be minimized by reducing the relative sliding between the mask and the substrate in particular during the transfer process.

The search of new quantum effects and materials is evolving at a rapid pace and encompasses a board range of disciplines from material science, condensed mater physics, to technology and engineering. The success of material synthesis and device fabrication would be the prerequisite for this exploration. The SL has been shown and tested for many years in a large amount of materials and devices to be extremely powerful, which has been widely used in different areas from fundamental research to industrial application. The application of the SL technique in investigating the quantum materials is so far still in an embryonic stage without many experimental results reported yet. The outcome of applying SL to quantum materials study is evidently versatile and supportive, which is both convincing and exciting. The research areas that strongly rely on the interface control should pay more attention to this technique. Further efforts should be made to adopt the SL to the experimental studies in the emergent materials. Key potentials may refer to the fabrication of a topological superconductor and topological qubits as well as the topological spintronic material that consists of topologically nontrivial matter with various magnets, both of which require controllable interfacial coupling and clean device fabrication process in order to maximize the combined effects.