Fabrication of pristine 2D heterostructures for scanning probe microscopy

2D heterostructures have been a major focus of recent physics, chemistry, mechanics, and electronics research. Novel quantum states of matter¹ including unconventional² and possibly topological³ superconductivity have been evidenced in these heterostructures, and 2D heterostructures are expected to be instrumental in a wide variety of applications from excitonic devices⁴ to chemical sensing.⁵ 2D vertical heterostructures consist of layer-by-layer stacked or synthesized 2D materials, with degrees of freedom in the number of layers, the material composition of those layers, and the relative twist angle θ between each layer. The latter degree of freedom led to a recent surge in the research of Moiré heterostructures⁶ and their associated physics and chemistry.

Early 2D heterostructure research has primarily focused on materials that are chemically insert in air—particularly graphene and hexagonal boron nitride (hBN). At the same time, most 2D materials discovered are chemically reactive and readily degrade under ambient conditions. This includes most of transition metal dichalcogenides (TMDCs);⁷ elemental 2D materials⁸ of boron, silicon, phosphorous, gallium, germanium, arsenic, selenium, tin, antimony, tellurium, thallium, lead, and bismuth; Janus TMDCs such as MoSSe⁹ or MoSH;¹⁰ ferromagnets such as CrI₃¹¹ or CrBr₃,³ as well as a wide variety of layered compounds that are just beginning to be explored in the 2D realm, such as cuprate superconductors.¹² 2D materials have shown a wide variety of properties, including massless Dirac fermions,¹³ non-abelian statistics,¹⁴ unconventional superconductivity,¹² and quantum spin hall effect.¹⁵ Undoubtedly, 2D heterostructures offer the potential to achieve designer materials with tailored properties, discover new physics, and develop new functional devices.

There are two general methods for fabricating such heterostructures: bottom-up synthesis and top-down fabrication. Bottom-up synthesis¹⁶ involves growing heterostructures layer by layer from its constituent parts through processes such as molecular beam epitaxy¹⁷ (MBE), chemical vapor deposition¹⁸ (CVD), and solution phase growth.¹⁹ This synthesis is beneficial for creating large continuous films²⁰ and so far has been the only approach for creating lateral heterostructures^19,21–25 consisting of dissimilar layers in a side-by-side rather than stacked geometry. However, bottom-up synthesis is less flexible as interfacial lattice and/or symmetry matching is often desirable for best growth results. In contrast, top-down approaches, which consist of obtaining (typically through exfoliation) and identifying 2D flakes and then stacking those flakes one at a time, offer much more flexibility,²⁶ especially over the twist angle between each layer and the material species of each layer. As a consequence of the mechanical operations involved in the exfoliation and stacking, a grand challenge in top-down fabrication is preserving the material integrity and preventing surface and interfacial contamination—discoveries of new physics in 2D heterostructures rely on the use of high-quality and pristine materials. Figure 1 shows the chronology of representative development of 2D vertical heterostructures, beginning with the mechanical exfoliation of graphene through to ultra-high vacuum (UHV) assembly. Much of this development has been motivated by the fact that 2D materials are uniquely susceptible to contamination.²⁷ In a single atomic layer, the entire material is exposed to the surface and there is no bulk. If any environmental contamination occurs, the entire material is contaminated. This obviously impacts samples that are chemically reactive in air, but it also can impact samples that are not chemically reactive in air through particle physisorption. For example, after exposure to atmospheric conditions for several days, graphene can be contaminated by environmental hydrocarbons²⁸ [Figs. 2(a) and 2(b)]. The impacts on oxygen and moisture sensitive materials are even more dramatic. Figures 2(c) and 2(d) show black phosphorous samples degrading under ambient conditions over the course of a few days,^29, Figs. 2(e)–2(h) show CrI₃ degrading under humid argon conditions over the course of an hour,³⁰ and Figs. 2(i) and 2(j) show Fe₃GeTe₂ degrading under ambient conditions over the course of just 5 min.³¹ Therefore, many top-down fabrication techniques, which are performed under ambient conditions, succeed only with inert materials such as graphene and hBN. Air sensitive materials are typically only accessible using all-in-vacuum processing techniques, such as bottom-up MBE, or glovebox processing. However, there is increasing interest in searching for new exotic quantum states in 2D heterostructures performed by surface-sensitive characterization techniques such as scanning probe microscopy (SPM).³² At the spatiotemporal resolution limit, surface and interface cleanliness is critical. Because both lateral and vertical heterostructures grown via bottom-up synthesis in vacuum environments are broadly compatible with SPM techniques, we limit the scope of our review to advanced top-down 2D heterostructure fabrication techniques given the enormous challenges in achieving atomically pristine surfaces and interfaces. In addition, SPM techniques impose further requirements for the 2D heterostructures; specifically, the heterostructures often need to be electronically contactable with atomically clean and accessible surfaces and interfaces. To achieve such heterostructures, there are four principles a fabrication process should observe: (1) Dry transfer is preferred, although sometimes the use of solvents is unavoidable; (2) fabrication should be performed under inert or vacuum environments; (3) traditional lithography for making electrical contact should be avoided after the 2D heterostructure fabrication; and (4) ideally, there should be no encapsulating layer on the top of the heterostructure.

The rest of this Review will proceed as follows: We will briefly review the commonly used wet transfer methods and their limited suitability for SPM characterization. We will then review the dry transfer methods currently being employed to fabricate 2D heterostructures for SPM. We will further discuss the technical requirements for performing dry-transfer in an inert-atmosphere glovebox and the techniques being used therein. Then, we will discuss exfoliation and transfer methods in vacuum, particularly UHV environments. Finally, we will review methods for making electrical contacts without the use of traditional wet lithography and provide some future outlook.

For any 2D material transfer method to work, two basic challenges must be addressed. The first is how to pick up the desired flake, and the second is how to deposit that flake. We here define wet-transfer to be methods that use liquids in solving one or both challenges. Several of the first methods developed for 2D heterostructure fabrication used a floating-flake wet-transfer scheme.^33,34 These methods involve coating substrates with polymers that could either be dissolved in solvents³³ or were hydrophobic,³⁴ causing the desired flakes to float to the surface together with a supporting polymer layer where they could then be scooped and deposited. These methods also used solvents in later stages of the fabrication process for removing the supporting polymers. A similar method of injecting water into a salt-promoter layer has also been used to separate and isolate CVD-grown monolayers.³⁵ 

When preparing heterostructures for characterization with SPM, a more common wet-transfer approach^36,37 involves using water or other solvents such as acetone³⁸ to dissolve a sacrificial polymer layer depositing a flake in the desired location. Figures 3(a)–3(h) show one such procedure³⁷ where water is used to dissolve a polyvinyl alcohol (PVA) layer depositing the magic angle twisted bilayer graphene (MATBG)/hBN stack onto a handle [Fig. 3(f)]. This procedure fabricates twisted bilayer graphene devices for characterization using SPM, specifically scanning tunneling microscopy (STM), as shown in Fig. 3(i), but it has significant drawbacks. This procedure cannot be performed in a glovebox or vacuum, as the use of water would contaminate those environments, which means that only materials that are stable in air and water may be used. In addition, a significant amount of post-fabrication treatment is required in this procedure to make the device atomically clean because bubbles can become trapped between the layers,³⁹ individual layers can become wrinkled,^40,41 and environmental contaminants can adsorb to the surface.²⁸ Such a treatment typically involves rinsing the device in heated dichloromethane, acetone, water, and isopropanol alcohol (IPA), in addition to annealing the device in UHV for over 10 h at temperatures above 100 °C.

Dry-transfer techniques must employ different methods to pick up and deposit flakes that do not rely on solvents.⁴² This is typically done by using polymer stamps [made of polydimethylsiloxane (PDMS), polycarbonate (PC), etc.] with relatively low glass transition temperatures. Flakes are picked up and deposited by heating or cooling polymer stamps through their glass transition temperatures, which gives control over the malleability and adhesiveness of the stamp. Many^43–52 2D heterostructures have been prepared using such methods for SPM measurements, particularly STM, and although the details often differ, the overall approach and major challenges are common.

Figure 4 shows a representative dry-transfer tear-and-stack technique⁴³ for creating twisted bilayer graphene to be characterized by STM.⁴⁵ This technique begins by exfoliating monolayer graphene from bulk graphite [Figs. 4(a)–4(c)] using a silicon/PVA/polymethyl methacrylate (PMMA) stamp. Tape, with a window cut to access the graphene [Fig. 4(d)], is then used to peel away the PMMA/graphene layer. Once the monolayer of graphene to be used is identified, half of the flake is brought into contact with a piece of hBN, and the graphene is slowly lifted away tearing at the hBN edge [Figs. 4(e) and 4(f)]. This is known as the tear-and-stack technique, which is a popular^52–57 method for controlling the twist angle between layers in heterostructures where both layers are the same material. A twist angle is then introduced, and the remaining half graphene piece is placed on top of the initial graphene [Figs. 4(g) and 4(h)]. One downside to the tear-and-stack method is that it can introduce unintended strain⁵⁸ when the material is torn. This can be mitigated by using an atomic force microscope (AFM) tip to cut the material in half before lifting off [a step just before Figure 4(e)], a method known as the cut-and-stack technique.^59–62 For 2D heterostructure fabrication involving layers that are not the same material, the same stacking process can be used, but to determine the relative orientation between layers, other methods are required. Second harmonic generation (SHG) provides one method for non-centrosymmetric⁶³ materials such as odd-layer TMDCs.^47,64,65 For materials with anisotropic crystal structures, such as MoO₃, a combination of optical microscopy and AFM has been used to determine the crystallographic axes.⁶⁶ 

When flakes are exposed to polymers (e.g., PMMA) in such dry-transfer processes, polymer residues are left behind on the flakes.⁶⁷ These residues interfere with taking high quality SPM data. To obtain the STM image shown in Fig. 4(i), the sample was annealed at 260 °C in forming gas (10% H₂ and 90% Ar) overnight, which removed the fabrication residues.⁴⁵ Another option to prevent residue contamination entirely involves using clever polymer geometries⁶⁸ to ensure that no polymer has direct contact with a targeted flake surface, as illustrated in Figs. 5(a) and 5(b), in a polymer-contact free method. The goal of the polymer-contact free method is to fabricate a 2D heterostructure without allowing any polymer to contact a surface that is part of the 2D heterostructure, thus ensuring that no polymer residue is left behind. To accomplish this, a base flake is used to build the heterostructure and contact polymer stamps in a very specific geometry. The two stamps consist of PDMS with a hole (stamp 1) and a trench (stamp 2) partially covered by a poly(vinyl chloride) (PVC) thin film [Fig. 5(a)]. The procedure starts by picking up the base flake with stamp 1 such that the PVC film contacts roughly half of the base flake [Fig. 5(b-i)]. A target flake [green in Fig. 5(b-ii)] is then made in contact with and picked up by the part of the base flake that is in contact with PVC. Once this target layer has been picked up, the other half of the base flake is contacted by the suspended PVC film on stamp 2 such that the target flake/base flake heterostructure extends out into the vacant trench of stamp 2 as shown in Fig. 5(b-iii). Stamp 1 is then slowly removed horizontally [Fig. 5(b-iv)], and stamp 2 is flipped over and pressed against a final substrate [yellow in Fig. 5(b-v)]. Finally, by pulling away stamp 2, the target flake/base flake heterostructure is deposited on this substrate and ready to be characterized [Fig. 5(b-vi)]. This approach creates a heterostructure where the interface and the top surface of the target flake have not had any contact with polymers. Optical and STM images, as shown in Figs. 5(b)–5(e), demonstrate the cleanliness of a variety of heterostructures this technique can produce.

In addition to contamination from polymer residues, 2D heterostructures also face substantial contamination risk from the ambient environment, as discussed in the Introduction. The most common method of reducing ambient environment contamination involves moving 2D heterostructure fabrication to an inert atmosphere glovebox, typically filled with either nitrogen or argon (both nitrogen and argon are inert to most materials, although notably nitrogen does react with lithium compounds⁶⁹ commonly used in energy applications). Gloveboxes are primarily used to mitigate water and oxygen induced sample degradation, and, as such, any wet technique that relies on water is inherently incompatible. Volatile solvents such as chloroform^65,70 used in other wet-transfer techniques, while perhaps not being inherently incompatible, are extremely bad for the catalysts in a glovebox, and so techniques relying on these solvents can only be compatible with gloveboxes with solvent traps/scrubbers.⁷¹ To fabricate 2D heterostructures in a glovebox, a transfer station,^71–74 like the one shown in Fig. 6(a), is typically set up inside a glovebox and dry-transfer techniques are employed. By motorizing each of the stage positioners, the fine movements associated with the transfer process can be controlled remotely without being encumbered by the glovebox. Dry transfer techniques are broadly glovebox compatible, and a wide variety of polymers and tapes have been used for glovebox exfoliation and stacking. The technique shown in Fig. 5 is entirely glovebox compatible, and Fig. 6(b) shows four stamps used in gloveboxes utilizing a wide variety of polymers: (i) PMMA,⁷³ (ii) PC,⁷⁵ (iii) Elvacite,⁷⁶ and (iv) PDMS, tape, and polycaprolactone (PCL),⁷⁷ for a variety of 2D materials, including (i) black phosphorous (BP),⁷³ (ii) MoS₂,⁷⁵ (iii) hBN,⁷⁶ and (iv) ZnPS₃.⁷⁷ Fabricating in a glovebox provides two major advantages: first, with fewer contaminants, the surfaces are cleaner, which is necessary for SPM; and second, materials sensitive to oxygen and/or moisture can be incorporated into 2D heterostructures. Indeed, the WTe₂/NbSe₂ heterostructure⁶⁸ shown in Fig. 5(d) was fabricated in a glovebox because both materials are air sensitive.

For air sensitive samples prepared in a glovebox, transferring the samples to external characterization tools presents another set of challenges. There are two strategies that we will discuss briefly here: (1) encapsulating air sensitive samples with other materials, and (2) using a protective suitcase to transfer the sample. Encapsulation is a particularly common method for bottom-up MBE grown samples, but there are also examples of this method in a top-down fabrication regime.^78–81 Inevitably, encapsulation adds significant challenges to SPM measurements, which typically require the sample surface to be exposed. There are a few examples of researchers using STM to scan through a top encapsulating layer of graphene,^79,80 or a layer of MoS₂.⁷⁸ However, this adds additional complexity to data interpretation as signal deconvolution is required. For encapsulation layers in the form of deposited chalcogen (e.g., Te and Se), in situ thermal desorption combined with ion sputtering before SPM measurements can effectively recover a pristine sample surface.⁸² Occasionally, top layers of van der Waals crystals can be locally removed with an STM tip as with graphene⁸¹ or Bi₂Sr₂CaCu₂O_8+x (BSCCO),⁸³ exposing subsurface layers. Another option is to use a protective suitcase to transfer samples. These suitcases can be either evacuated or filled with inert gas, and an example⁷¹ is shown in Fig. 6(c). The biggest challenge with this method is ensuring that the glovebox and the characterization equipment can interface with the protective suitcase. Vacuum versions of these suitcases are also often bulky.

All-in-vacuum systems build on the advantages of inert atmosphere gloveboxes in two important ways. First, all-in-vacuum systems achieve lower levels of contamination and thus cleaner materials and interfaces. As Guo et al.⁸⁴ demonstrated, one way to compare protection methods is to assume that a sample surface has a sticking coefficient of unity and use the partial pressure to compare various sample environments. Using this method, under air, a pristine sample surface becomes fully covered in about 1 ns; under typical inert glovebox conditions [oxygen and water content at 0.1 parts per million (ppm)], the pristine surface still becomes covered in 1–10 ms; and UHV conditions (better than 10⁻⁹ mbar) extend that time to 10³ s and beyond. Even sub-0.1 ppm contaminants present in gloveboxes can be detrimental to 2D material properties and interfacial qualities.⁸⁵ Therefore, the development of methods and instrumentation for exfoliating van der Waals crystals and fabricating 2D heterostructures under vacuum conditions is highly desirable. Second, UHV conditions can enhance exfoliation yields to the extent that Sun et al.⁸⁵ classified using ultra-flat and clean substrates to exfoliate materials in UHV as a universal exfoliation method. Gold substrates⁸⁶ and thin films⁸⁷ have been shown to be particularly effective non-polymer stamps for exfoliating many 2D crystals. Thus, all-in-vacuum systems enable both cleaner samples and access to universal exfoliation methods.

To realize these advantages, all-in-vacuum systems have recently been demonstrated that can exfoliate^85,88–93 and stack^{73,84,87,94–96} flakes into the desired heterostructures. While ideally both exfoliation and deterministic stacking should be completed in vacuum for a complete top-down fabrication process, the process of exfoliating bulk crystals and stacking 2D layers is challenging in vacuum due to limited degrees of freedom for sample manipulation. As such, several systems have been demonstrated that focus only on in-vacuum exfoliation or stacking, and to our knowledge, only two systems have been demonstrated to completely perform both exfoliation and deterministic stacking in an entirely top-down regime in UHV.

In-vacuum exfoliation has been developed by several research groups, beginning with a setup for graphene.⁹¹ One such method involves a transfer substrate consisting of a wheel⁸⁸ covered with an insulating, low outgassing, and residue-free clean-room adhesive tape (UltraTape product) that was used to pick up flakes under ambient conditions. The wheel was then introduced to a vacuum chamber (10⁻⁸ mbar) and rolled over a sample substrate, depositing monolayers of VSe₂. Another method consists of cleaving a bulk crystal in high vacuum⁸⁹ or UHV⁸⁵ and preparing a sample substrate through sputtering, annealing, and plasma cleaning. The substrate material can be flexible, depending on the flakes to be exfoliated, including Ag, Au, Ge, MgO(100), SrTiO₃(100), Al₂O₃(0001), and Si(111). By pressing the substrate against the freshly cleaved material at variable temperatures (between room temperature and 500 K), a variety of materials including MoS₂, graphene, FeSe, and Bi-2212 were exfoliated. Preparation in vacuum can also enhance the size of exfoliated monolayers up to millimeters as demonstrated with graphene.⁹³ 

Monolayers grown using bottom-up synthesis techniques are occasionally stacked using methods compatible with top-down fabrication stacking processes for STM characterizations.³⁵ Imamura et al.⁹⁴ used a high vacuum (10⁻⁶ mbar) system to stack two layers of CVD grown graphene together. The stacking process consists of pressing the two layers together at 200 °C for 1 h. This setup faced the drawback that the twist angle between the two layers is difficult to maintain, and the setup is incapable of deterministic stacking. A more controlled system was developed by Mannix et al.⁹⁵ consisting of an assembly robot in high vacuum (10⁻⁶ Torr), see Figs. 7(a) and 7(b), that robotically stacks layers together, assembling a heterostructure that contains as many as 80 layers [16 layers of MoS₂ are shown in Fig. 7(c)]. The assembly uses a stamp made from several polymers, beginning with PDMS, a support layer of MicroChem lift-off resist, a release layer of poly(cyclohexene propylene carbonate) (PCPC), and an adhesion layer of poly(benzyl methacrylate) (PBzMA), see Fig. 7(d), which it manipulates via heating and cooling through glass transition temperatures to pick up and place each layer. The system has been used for a variety of materials, including metal–organic CVD grown TMDCs, Au thin films, exfoliated flakes of graphite, and hBN, and is expected to be applicable to a variety of materials compatible with van der Waals stacking. One downside to this approach is that since polymers were used, the completed heterostructures have to be cleaned using acetone or chloroform, a process that is incompatible with vacuum.

Two systems have been demonstrated to perform complete exfoliation and deterministic stacking in UHV. Wang et al.⁸⁷ demonstrated a UHV (10⁻¹⁰ mbar) polymer-free transfer setup [Figs. 7(e) and 7(f)] using a SiN_x cantilever [Figs. 7(g) and 7(h)] coated with a thin metal trilayer film for better adhesion to 2D materials. While a PMMA polymer was initially attempted, the PMMA outgassed and released hydrocarbons in UHV as evidenced by bubbles in the fabricated hBN/graphene/hBN heterostructures. To solve this problem, inorganic SiN_x was used, leading to heterostructures with minimal contaminants [Fig. 7(i)]. Within the trilayer metal film, a 1 nm layer of Ta and a 5 nm layer of Pt were used to smooth out the surface and to give a flat place for the Au to bond without clumping [Fig. 7(j)]. A 0.65 nm thick layer of Au was found to be optimal for 2D material transfer at 150 °C.

Guo et al.⁸⁴ demonstrated a UHV (5 × 10⁻¹⁰ mbar) system that is also directly connected to an MBE growth chamber. The assembly chamber [Fig. 8(a)] consists of two stages, a tool stage and a sample stage [Fig. 8(b)]. Samples are brought in on the sample stage, picked up by the tool stage, and then placed in their desired location on the substrate [which is also held on the sample stage as shown in Fig. 8(c)]. The whole assembly is optically accessible via a long working distance objective. Exfoliation is performed using UHV-compatible Kapton tapes onto a silicon oxide substrate (or Au coated silicon), as shown in Fig. 8(d). A sapphire/PDMS/PMMA/polypropylene carbonate (PPC) stamp, which is assembled outside and degassed before being brought into the UHV chamber, is then used to pick up exfoliated monolayers for stacking into heterostructures. With all-in-vacuum operations, this system has produced heterostructures with precise rotational alignment, as shown in the BP/hBN heterostructure in Fig. 8(e), although due to the use of polymers there are still some residues visible in the AFM image as bright specks [Fig. 8(f)].

The characterization of 2D heterostructures, e.g., using STM, often requires them to be electrically contacted. Furthermore, the performance of 2D electronic, optoelectronic, and spintronic devices is greatly influenced by the intricate details of how these 2D materials are electrically interfaced with external circuits. For an in-depth discussion on achieving various types of flexible, ohmic, seamless, or impedance-matched contacts with 2D materials, we refer the readers to the relevant review articles.^97–99 Defining metal contacts via lithography is an indispensable solution; however, the use of polymers inevitably leaves residues and the process of removing resists through lift-off requires the use of various chemicals that are detrimental to air sensitive materials. As a result, interfaces between substrates and 2D heterostructures are often contaminated by the accumulation of resists, polymers, desorbed gasses, and solvents. Most of the time, the contaminants significantly impact the performance and characterization as they tend to modify the electrochemical composition with the introduction of scattering centers, altered potential landscape, reduced charge carriers and mobility.^100,101 Techniques such as annealing and vacuum treatment have shown some success in mitigating contamination issues. However, such a heat treatment introduces further complications to samples that are chemically reactive or heat sensitive.

For SPM characterizations that are sensitive to disorders and contamination at the nanometer to atomic length scale, deposition of contacts on substrates before 2D heterostructure fabrication and in situ patterning and contact formation on as-prepared 2D heterostructures in a vacuum environment are preferred methods as they preserve surface cleanliness and enables uncontaminated contact/sample interfaces. Toward this goal, several techniques have been demonstrated for the latter scheme. These include lithography-free techniques for simple patterning and shadow/stencil mask techniques for more complicated patterning and contacts producing high quality heterostructure devices.^102–104 Most of the techniques involve the use of macroscopic and nanostructured stencil masks in UHV chambers through which the electrical contacts and patterns are formed by subsequent metal deposition using focused ion beams, thermal evaporation, and e-beam evaporation. Through the use of independent mask carriers, evaporation through multiple masks also becomes feasible. These methods are purely mechanical, resistless, and UHV compatible, thus minimizing the introduction of contaminants.

Masks with precise geometry and nanopatterns can be prepared by standard lithography processes.^105–107 A typical silicon mask preparation process is schematically illustrated in Fig. 9(a). Double side polished Si(100) was used as a substrate, and 200 nm of Cr was deposited to serve as an etching mask layer, followed by subsequent deposition of PMMA, an e-beam resist. The other side of the wafer was coated with photoresist to etch away three-quarters of the wafer by using photolithography and KOH etching. The front side of the substrate was exposed to an e-beam to create the patterns for a desirable shadow mask structure. The e-beam resist was developed after exposure, and Cr etchant is used to remove exposed Cr patterns as openings on the mask. Once prepared, those masks can be transferred to UHV to form electrical contact by directly depositing contact material through the mask onto the 2D heterostructures at the desired location [Fig. 9(b)]. The masks so prepared can be used several times before they are destroyed or the apertures are covered with deposited metal. Bao et al.¹⁰² compared the cleanliness of the contacts made by e-beam lithography and shadow masks using AFM. The contacts made by e-beam lithography left a layer of residue, which needed additional sputtering and heat treatments for removal [left panel in Fig. 9(c)], whereas contacts made by shadow masks are much cleaner and need no post-deposition treatments, making the exposure of extended atomically cleaner areas suitable for STM imaging [Fig. 9(d)]. For making few local contacts with no complex patterning, nano-soldering with indium or other alloys of indium with low melting points are used to create ohmic contacts to nanostructures as demonstrated by Girit and Zettl;¹⁰³ main components of their miniaturized soldering setup are shown in Fig. 9(e). A sample with a small bead of indium is placed on a holder, which can be heated up to melt the indium bead. The molten indium is precisely manipulated using a W tip under an optical microscope using XYZ micromanipulators forming a spike to create submicron sized contacts to the nanostructured samples.

Staley et al.¹⁰⁴ used an alternative graphene device fabrication technique by using ultra-thin quartz filaments up to submicron size in diameter as masks. Once the desired 2D material/heterostructure is prepared, a thin quartz filament is placed onto an area of the material as a mask and contact material such as gold can be evaporated on top normal to the surface [Fig. 9(f)], which resulted in exposure of micrometer sized graphene [Fig. 9(g)]. Similarly, graphene based tunnel junctions can also be fabricated by evaporating alumina and metal contacts at angles, taking advantage of the shadows created by the quartz filaments [Figs. 9(h) and 9(i)].

Several similar techniques have been employed in UHV using resistless methods to make electrical contacts for 2D material characterization in STM and other multifold experimental areas. These techniques allow for flexible, inexpensive, fast, and easy to integrate electrical contacts for heterostructure devices in the UHV system; various patterns and geometries can be designed simply by changing the masks in the mask carriers. However, the drawback of this fabrication process is the formation of penumbra during the evaporation process caused by mechanical instability and placement distance between the mask and the substrate, creating smudged patterns due to edge spreading.^108,109 Nonetheless, methods such as SPM-assisted resistless lithography¹¹⁰ and well-designed masks with capacitively controlled alignment¹¹¹ have enabled UHV lithography with sub-micron precision and, therefore, show great potential to be integrated with UHV exfoliation and stacking setups to create functional heterostructure devices all in vacuum.¹¹² 

We have reviewed the recent development of wet- and dry-transfer methods, especially emphasizing those that are UHV compatible for fabricating 2D heterostructures as well as strategies for making clean electrical contacts for SPM characterizations. Fabrication processes in protected environments, both in gloveboxes and in vacuum systems, create significant material advantages and research opportunities. Unconventional air-sensitive 2D materials such as cuprate superconductors, halides, and perovskites can now be integrated into novel and functional 2D heterostructures with preserved material properties and, critically, high reproducibility.^63,113 SPM characterizations that are sensitive to atomic scale surface and interface contamination will significantly benefit from these techniques.

Going forward, we believe there exist significant opportunities for technical development that will address multiple challenges facing ultra-clean 2D material fabrication at present. For example, improving the efficiency and flexibility of in-vacuum fabrication processes will both increase research capabilities and develop scalability necessary for future commercial applications. We can imagine a future closed-cycle autonomous laboratory powered by artificial intelligence and robotic experimental agent that can take and exfoliate bulk materials, identify target flakes, assemble desired 2D heterostructures, perform material characterizations, and refine fabrication strategies using feedback obtained through data analysis. The robotic assembly⁹⁵ system described above provides an example of what the assembly portion of such a system might look like. Substantial effort¹¹⁴ has already gone into developing machine learning algorithms to identify 2D flakes based on optical contrast¹¹⁵ and shape,¹¹⁶ and significant progress has recently been made in self-driven autonomous laboratories for materials discovery in chemistry, biology, and materials science.^117–121 Therefore, we believe that it will be a matter of time before the realization of fully automated and closed-cycle discoveries of novel functional 2D heterostructures.

Another opportunity lies in the development of polymer-less stamps that are compatible with higher vacuum. Using polymers to create stamps is a common fabrication practice in air and glovebox environments, and there is some evidence that these polymers can be used under high vacuum⁸⁸ and UHV⁸⁴ conditions given careful selection and preparation. However, given the significantly lower outgassing rates of inorganic materials compared to polymers,¹²² it stands to reason that higher vacuum levels can be achieved using polymer-free setups like the SiN_x discussed above.⁸⁷ Transfer processes relying on metals or layers of metals, particularly Au^86,87 or Ni,¹²³ have successfully been used to transfer flakes on a variety of materials. Further development of these methods and identification of new stamp materials in UHV environments will further mitigate contamination and thus cleaner surfaces and interfaces.

Finally, there are opportunities to combine bottom-up and top-down fabrication processes in exciting ways. Remote epitaxy¹²⁴ is a process for epitaxial growth that uses a van der Waals material (e.g., graphene) to allow the grown layer to be peeled off. The free layer could then be incorporated into a 2D heterostructure with conventional top-down exfoliated flakes. Such a scheme will become even more powerful when combined with micro-MBE that uses shadow masks to achieve growth resolution on the order of 100 nm,^125,126 allowing for the flexibility of top-down fabrication with the material control of bottom-up synthesis.

Without question, technical advances toward fabricating 2D heterostructures of highest quality will unlock groundbreaking research opportunities that were previously inconceivable and introduce invaluable new research frontiers promised by 2D materials.

J.M., N.S., and X.L. acknowledge the support from the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award No. DE-SC0024291. X.L. acknowledges the support from the Oak Ridge Associated Universities Ralph E. Powe Junior Faculty Enhancement Award.

The authors have no conflicts to disclose.

James McKenzie: Investigation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Nileema Sharma: Investigation (supporting); Visualization (supporting); Writing – original draft (supporting); Writing – review & editing (supporting). Xiaolong Liu: Conceptualization (lead); Funding acquisition (lead); Investigation (equal); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (equal).

Data sharing is not applicable to this article as no new data were created or analyzed in this study.