3D nanoprinting via focused electron beams

Additive manufacturing has recently been successfully established in research and industry. Of particular interest are techniques that enable three-dimensional printing without any additional process steps. While such direct-write methods have reached a high level of sophistication for objects with large dimensions, it is very challenging to produce freestanding 3D structures on a microscopic scale. Furthermore, there is considerable interest in the 3D printing of objects on the submicrometer scale as this not only follows the general trend of miniaturization of devices (electronics and microrobotics) but also allows applications where certain functional properties emerge at the nanoscale (i.e., plasmonics).

There are few technologies that can be used to print 3D objects with single structure sizes below 1 μm. In a 2017 review, Hirt et al.¹ compared different additive manufacturing methods for metallic 3D objects. Some of the mentioned techniques directly transfer nanoparticles to the substrate like direct ink writing, electrohydrodynamic printing, or laser-assisted electrophoretic deposition.¹ Others are based on chemical reduction of metal salt solution like meniscus-confined electroplating, electroplating of locally dispensed ions in liquid, or laser-induced photoreduction.¹ While each technique has its individual strengths and weaknesses, the smallest possible feature size is limited by the size of the fabrication tool, e.g., inner diameter of the capillary or the wavelength of the laser light. Even smaller features can be achieved using focused electron/ion beams. The smallest structural dimensions (typically about 20–60 nm) are possible with a technique called focused electron beam induced deposition (FEBID).¹ For example, Fig. 1 shows 3D nano-objects fabricated via FEBID, demonstrating reproducibility [Fig. 1(a)] and the feasibility to direct-write highly complex 3D-geometries with structural dimensions on the nanoscale [Fig. 1(b)]. Further 3D nanoprinted architectures can be found in the literature.^2–4 

In this article, we provide an insight into the state of the art of this emerging 3D nanoprinting technique. Furthermore, we overview the problems that must be solved to expand the application space of 3D FEBID.

Focused electron beam induced deposition is an additive manufacturing technology made possible by the scanning electron microscope (SEM) platform. In the FEBID application, the focused electron beam is used as the writing tool. The nanoscale resolution characteristic of the focused electron beam in an SEM is a critical component required for “nano” printing. Local precursor delivery is another critical aspect of FEBID. Precursor delivery systems are a common accessory deployed on FIB-SEM (focused ion beam/focused electron beam microscope), explaining why a large number of the FEBID experiments reported in the literature are conducted on these platforms. A capillary tube continuously supplies gaseous precursor molecules to the surface, where a fraction of the impinging flux physisorbs on the substrate surface. Once attached, these molecules diffuse on the surface for a finite time which is on the order of the mean residence time⁵ due to a finite probability of surface desorption. Using a focused electron beam, the primary, secondary, and backscattered electrons interact with the adsorbed precursor, which can lead to the dissociation of the molecules.⁶ The fragments of the molecules are either locally deposited or are volatilized and removed by the vacuum pumps. The composition of the deposited material mainly depends on the choice of the precursor type and is discussed in Sec. IV C in more detail. A comprehensive description of the FEBID process, in general, can be found in the literature.^7,8

At this point, we would like to list some advantages of FEBID. First, the adsorption of the precursor gas is not limited to flat surfaces, so FEBID can be performed on almost any substrate morphologies. This allows applications on nonplanar and complex areas² or prestructured layouts. The only requirements on the sample are vacuum and electron-beam compatibility.

Secondly, one can easily control the position/movement of the electron beam with subnanometer precision, which enables complicated exposure patterns. As it will be shown later, this capability to navigate the electron beam in a well-defined manner is a key factor for reliable 3D nanoprinting. These advantages are also attributable to the related technique of focused ion beam induced deposition (FIBID). For Ga⁺ ions, complex three-dimensional nanostructures have been demonstrated as well.⁹ Higher growth rates, increased electrical conductivity, and different functionalities (such as superconductivity¹⁰) are achieved in comparison to FEBID, while larger wire dimensions, ion implantation, and substrate heating/amorphization are found.¹ 3D-nanoprinting with He⁺ (and Ne⁺) ion beams is still at an early stage,¹¹ however, remarkably small diameters of vertical superconducting wires have been demonstrated recently.¹² Since FEBID and FIBID are based on the same principle of charged particle beam induced deposition, both can learn and complement each other. This includes, in particular, growth dynamic for 3D fabrication, differences in material properties, and resolution can be considered as complementary aspects.

Typically, planar (2D)-FEBID is used for TEM lamella preparation (as protection cover and soldering),¹³ circuit editing,¹⁴ etc., and is also used as a lithography mask repair tool.¹⁵ For further FEBID-based applications such as stress–strain sensors,¹⁶ we refer to the literature.^17–20 The extension from two-dimensional deposits into the third dimension opens up many new possibilities, as will be discussed later.

Vertical nanowires (often termed “pillars”) are deposited using stationary electron beam exposure. Real 3D nanoprinting requires the deposition of nanowires having a full range of orientations spanning from vertical to horizontal, with respect to the substrate surface. Such nanowires are often called “segments.” To obtain inclined segments, two approaches are possible:²¹ (1) moving stage or (2) moving beam. In the first case, the substrate surface is tilted relative to the optical axis. With a proper tilting and rotation procedure of the sample stage, further pillars can be deposited on top of previous one, resulting in three-dimensional structures with different inclination angles. Although this procedure has some advantages (cylindrical wires and defined and almost arbitrary angles), it is very time consuming and requires an electron beam refocusing step after each stage movement. Therefore, this “moving stage” approach is limited to geometries with only a low number of individual branches.

In contrast, the “moving beam” approach enables the fabrication of 3D geometries with a much higher degree of complexity as well as curved elements (Fig. 1). Beam deflection coils facilitate the accurate positioning of the electron beam in the XY-plane (plane orthogonal to the optical axis) in discrete steps with subnanometer precision. By slowly moving the electron beam with patterning velocities in the range of tens of nanometers per second, consecutive deposition takes place on top but slightly displaced with respect to the previous deposit. Consequently, the wires lift up from the substrate, where the respective inclination angle depends on the patterning velocity (see Fig. 2).

While FEBID [sometimes also called EBID or electron beam-chemical vapor deposition (EB-CVD)] is a well-established technology to create two-dimensional or simple pillar structures, the unique capabilities of 3D nanoprinting of complex geometries are relatively new. To promote 3D nanoprinting via focused electron beam induced deposition to a broader audience, we suggest the acronym 3D-FEBID (or alternatively, 3BID if other particle beams should be included as well).

Figure 3 shows the number of journal articles published on 3D-FEBID so far to the authors' best knowledge. Here, we exclude the many studies on simple vertical pillar geometries, (review) articles, and bulky FEBID-deposits with a three-dimensional surface morphology, such as obtained by layer-by-layer printing (e.g., FEBID miniature of the mountain, The Matterhorn²²). Furthermore, we omit deposition experiments in Transmission Electron Microscopes as they obey different growth characteristics.

First 3D-FEBID studies were reported in the early 1990s by Koops et al.^23–25 It took several years until the concept was picked up by different groups, exploring individual growth aspects like scan direction²⁶ and patterning velocities^27–29 for various precursor materials. After this short period of increased interest, only little was published on 3D-FEBID in the years 2007–2014. Reasons for that might be the massive problems with unwanted codeposition on the substrate, secondary branch growth below the freestanding segments,^27,30,31 as well as structure bending.³² Furthermore, the related technique of focused ion beam induced deposition (FIBID) made major progress in terms of process control in 3D printing in this period.^33–35 At that time, comparative studies between FEBID and FIBID^36,37 suggested better results in terms of shape integrity by using ions, as deeper understanding of the 3D growth fundamentals via electrons were still at an early stage.

The few articles during these years were focused on material properties like electrical conductivity through suspended nanowires^38–40 and argued with the potential to use 3D-FEBID for plasmonics^37,41,42 or for magnetic applications.⁴³ 

3D-FEBID got a major push in 2016/2017 with the realization of highly complex 3D objects.^2,3 Direct-write architectures consisting of 1296 single segments were demonstrated.² Previously, only geometries with less than 10 connected wires were reported. This evolutionary step was enabled by three developments, namely, (1) exploring proper working conditions,^2,44 (2) applying an alternating or parallel patterning strategy,^2,3 and (3) simulations of 3D-growth.³ These improvements, combined with a better understanding of precursor dynamics in 3D space, opened up new capabilities for complex 3D-geometries and renewed interest in the community, reflected by an increasing number of studies on this topic (Fig. 3). A very important milestone on the road to complex 3D nanoprinting was the development of computer aided design software solutions^4,45 to simplify the applicability of 3D-FEBID. In Sec. III, we will outline the current status of 3D printing via FEBID and discuss requirements and limitations.

The instrumental setup for 3D-FEBID consists of a scanning electron microscope (SEM) and a gas injection system (GIS). Both these required parts are typically combined in a FIB-SEM, which explains why most of the studies in the literature were performed on FIB-SEM systems. However, by installing a GIS, any standard SEM can be converted into a 3D-nanoprinter. Nevertheless, there are some special technical requirements for high-fidelity 3D nanoprinting.

A gas injection system basically consists of a temperature controlled precursor reservoir and a gas injection nozzle to locally deliver gaseous precursor molecules to the substrate. A shortcoming of FEBID is its comparably long process time as the volume growth rates are usually limited by the number of available precursor molecules. The aspect of precursor supply is even more important for 3D-FEBID, since a stable local precursor coverage must be established⁴⁴ for reliable and spatially homogeneous growth. While the delivery of a high number of precursor molecules is ideal, it is limited by constraints like maximal pressure limits in the vacuum chamber and precursor vapor pressure. In order to optimize the precursor coverage within those limitations, a proper GIS-alignment as described in the literature^46,47 is indispensably needed.

For complex 3D nanoprinting, it is necessary to direct the electron beam to any desired spot in the XY-plane. This task is performed by the patterning engine, which directly controls the electron beam and which can handle specific process files (stream files). Typically, those files contain a list of XY-coordinates as well as the exposure times at each spot. A typical (12-bit) patterning engine can address 4096 individual steps in one direction. As the beam displacement (point pitch, PoP) for inclined 3D-growth is in the range of 1 nm and below, the maximum lateral expansion of a single geometry is limited (to 4 μm in the case of PoP = 1 nm). Using a 16-bit pattering engine (65 536 steps) or higher, there is practically no such restriction on the dimensions.

While the basic working principle of 3D nanoprinting via electron beam induced deposition is simple (see Fig. 2), reliable and high-fidelity fabrication is more demanding as multitude of parameters affect 3D-growth. In the following, we only discuss the most important ones, a comprehensive parameter study can be found in Winkler et al.⁴⁴ Instead of specific numbers, we discuss the 3D-growth fundamentals on the basis of the working regimes. The working regime describes the local balance between available precursor molecules and electrons for the dissociation process. There are two extreme situations, which have to be distinguished: (1) electron limited regime (ELR) and (2) molecule (or precursor) limited regime (MLR). In ELR conditions, the deposition rate is limited by the number of electrons, which can be established via low beam currents and/or high precursor coverage realized by a maximized flux and/or appropriate refresh times. While ELR can be realized for small and/or planar structures, 3D FEBID mainly works in MLR conditions as a consequence of the high dwell times. Under such conditions, the local growth rate is determined by direct gas flux adsorption rate and short-range diffusion of precursor molecules. Note that any change in the setup (precursor type, GIS alignment, vacuum pressure, and many more) affects the balance between electrons and precursor molecules and by that the working regime.

For high-fidelity 3D structures, low beam currents (typically <100 pA) should be used,⁴⁴ as reproducible 3D deposition is best achieved when in or close to an electron limited reaction regime. Furthermore, high currents lead to strong proximal codeposition underneath a freestanding segment due to an increased number of transmitted electrons. In contrast to the high local depletion at high beam currents, low beam currents help to maintain a comparably high precursor surface coverage, which allows a better control of the 3D growth. As a result, no significant influence of the scan direction in relation to the directional gas flux was observed,³ which have been reported for strongly precursor limited conditions.²⁶ This low dependency is an essential element toward predictable and reliable 3D nanofabrication.

In principle, 3D-FEBID is possible with any primary electron energy of a typical SEM range (1–30 keV).⁴⁴ Qualitatively, higher energies have the advantages of minimizing codeposition on the substrate⁴⁴ and the highest lateral resolution of the suspended wires.⁴⁸ Note that although a focused electron beam has a diameter on the order of a single nanometer range, the deposited wires exceed these dimensions, as the interaction of the electron beam with the growing deposit leads to the generation of back-scattered, forward-scattered, and secondary electrons.^49–51 In particular, the secondary electrons are largely responsible for the deposition because of a favourable energy range for precursor dissociation. Since the shape and size of the interaction volume depends on the primary energy, the cross-sectional shape of the growing nanowires changes from more circular to elliptical shape with increasing primary beam energy.⁴⁸ Depending on the application, one can use either low primary energies for more cylindrical wires (in the range of 50 nm) or high energies for low codeposition and narrowest segment dimensions (down to 15 nm).⁴⁸ 

The quality of the electron beam (beam focus, astigmatism) has an impact on the vertical growth rate.⁴⁴ It is difficult to define and establish identical beam quality, therefore, the deposition of reference structures (pillars, segment on a pillar) are typically performed^3,44 as a qualitative calibration sequence. Using the so-called immersion mode, which increases the convergence angle of the electron beam, improves the beam focus but has exhibited some disadvantages for 3D-FEBID, especially in terms of reproducibility.⁴⁴ For instance, as the 3D-structure grows in height, the narrow depth of field can more severely defocus the beam and compromise the structure integrity. Nevertheless, in standard mode, the deposition of tall 3D-structures is still possible (up to 10 μm has been demonstrated²), due to the relatively large depth of focus.

While most settings (beam parameter and gas parameter) are kept constant during 3D-nanoprinting, the patterning velocity is used to change the inclination angle of a segment. A slower patterning velocity results in a steeper segment angle and vice versa. Typical patterning velocities are in the range of tens of nm/s, which can go up to 500 nm/s, depending on the precursor type used.⁴⁰ The patterning velocity itself is the ratio of two individual parameters, namely, the (discrete) step size of beam displacement (point pitch, PoP) and the electron exposure time at each pixel (dwell time, DT). Segment inclination can either be changed via the PoP or the DT. These two different approaches are realized in two separately developed software solutions for the generation of pattern files,^4,45 discussed in detail below.

To facilitate complex multibranch geometries, it is essential to use a parallel deposition of the segments. In this mode, multiple segments are deposited simultaneously by cycling the beam position between multiple branches of the 3D-structure on a per pixel basis. In serial mode, each linear element of the object is continuously deposited to completion. The parallel patterning sequence has several advantages as it reduces (1) sample drift issues^2,44 and (2) inhomogeneous structure bending.^2,32 Furthermore, the local precursor coverage at a growing segment is, on average, higher than during serial fabrication because of an additional time to (3) refresh with new precursor molecules³ and (4) dissipate beam induced heat.⁵² 

An important milestone was the development of hybrid finite difference/Monte-Carlo simulations on 3D growth via FEBID by Fowlkes et al.³ Aspects like electron trajectories, precursor coverage/depletion/replenishment and electron dissociation cross sections were included, which made it possible to reproduce experimental results. Ultimately, it became feasible to reverse the experiment/simulation sequence in terms of starting with a 3D-design, followed by 3D simulations that predict the proper patterning file.³ The latter can then be loaded to the microscope where 3D deposition is executed. This approach replaced the cumbersome trial-and-error procedure used before and enabled the upfront design of multibranch structures with a high degree of complexity.^2,3 Furthermore, the simulation revealed valuable insights into 3D growth characteristics, which are not accessible by experiments (e.g., origin of codeposition, directional gas flux influence).³ 

Based on these simulations, a computer-aided design (CAD) program called “3BID” was released as another step to make complex 3D nanoprinting more accessible to researchers.⁴⁵ Besides, the realization of the alternating/parallel patterning sequence (also called 3D-interlacing^2,44 or intermittent mode⁴⁵), the graphical user interface allows the design and editing of vertices that make up complex meshlike structures (Fig. 4).

In parallel, Keller and Huth also developed a program that generates suitable pattern files for 3D-FEBID⁴ [referred to as “F3D” (short for Frankfurt 3D generator) below]. In the following, we compare both programs (3BID and F3D) and highlight their individual advantages. What both generators have in common is that the user has to provide a calibration for the given 3D-FEBID setup (precursor, beam-, and gas settings), which is effectively the relationship of the segment angle to the patterning velocity. To do this, an array of diving board structures⁴⁴ (segments on pedestals, see the orange inset in Fig. 4) with varying patterning velocities is fabricated and the measurements of the inclination angles are passed to the software. One major difference between the two CAD programs/pattern file generators is the realization of the patterning velocity: in 3BID, the point pitch is kept constant and the dwell times are altered, and vice versa for F3D. For example, taking a typical patterning velocity of 100 nm/s, the typical values for PoP/DT are 1 nm (fixed)/10 ms for 3BID^3,45 and 0.1 nm/1 ms (fixed) for F3D,⁴ respectively. No major differences are expected for structures fabricated in a serial patterning sequence, since the PoPs are well below typical growth diameters and the beam progresses in a quasicontinuous way. For structures fabricated with an alternating patterning sequence, the two approaches might lead to different growth rates, considering the difference in refresh times. Figure 5 shows a collage of nine tetrapods fabricated with identical patterning velocity (50 nm/s) but with a wide range of the PoP. As it is evident from the identical structure height, there seems to be no qualitative difference between both approaches. With that, the patterning velocity can be used as universal parameter to set up the inclination angle, at least for the present working conditions. Note that the tetrapod for a PoP of 5 nm is slightly lower, which is in agreement with similar observations, when the beam displacement distance (PoP) is close to the beam diameter.⁴⁴ 

Although both software solutions appear to produce identical results, each of them has unique advantages. For instance, both have integrated compensation strategies to tackle the problem of a decrease in vertical growth rate with increasing structure height. In F3D, the correction is done via a polynomial fit function, whereas in 3BID, individual segments can be corrected on an empirical base. Very recently, beam induced heating was found to be responsible for the reduction of growth rates with increasing structure height or segment length.⁵² 3D-simulations exposed a rise in temperature due to energy dissipation within the segments, which results in an enhanced desorption of precursor molecules. Consequently, the coverage at the beam impact region is reduced and the growth rate suffers.⁵² This beam induced heating results in a downward bending and ultimately to collapses of long single branches. The study also revealed the importance of short-range diffusion of precursor molecules to the growth front during growth.^47,52,53 The implementation of a corresponding compensation module to 3BID, which takes these new findings into account, is currently in progress.

The 3BID code is written in Matlab and available for free download from www.github.com (@FEBiD3D);³ F3D is an executable program written in C++ and available on request.⁴ 

There is a long and still growing list of FEBID materials, enabling the deposition of structures with a large variety of functionalities such as being electrically conducting, semiconducting, insulating, and magnetic or with special functionalities such as being optically active or superconducting. Table I lists those for which 3D-FEBID has already been demonstrated.

As discussed above, a high precursor flux is beneficial, in particular, for 3D nanoprinting (electron limited conditions), therefore, 3D-FEBID works best with precursors that have a high vapor pressure, e.g., MeCpPtMe₃ from which most 3D structures have been grown (see Table I). So far, there is no reason why a precursor suitable for 2D-FEBID should not work for three-dimensional deposits. However, high-fidelity geometries made of insulating materials such as TEOS (tetraethyl orthosilicate) are more challenging, because charging and strong heating due to low thermal conductivity can affect reliable fabrication. Despite these problems, the fabrication of 3D-structures is possible, as shown in Fig. 6.

A disadvantage of many FEBID materials is the incorporation of unwanted elements such as carbon. The origin of these impurities is mainly the precursor molecule itself, since—to ensure the volatility of the targeted element—ligands must be attached to the core atom (see Table I). These contaminations can reduce or even mask the intended material properties.

Several postprocessing approaches are available to increase the purity of two-dimensional FEBID-deposits.^74,75 However, most of these purification procedures are not ideal for 3D structures because of severe structural deformation induced by impurity atom removal. Pablo-Navarro et al.⁷⁶ have shown a high cobalt content (>95 at. %) for vertical pillars from Co₂(CO)₈ precursor after a thermal annealing up to 600 °C in vacuum. The morphological integrity of the nanowire was fully maintained, which was attributed to the comparable low volume loss as a consequence of the already high metal content in the as-deposited state (∼70 at. %). For Au nanopillars, deposited from Me₂-Au(tfac) precursor, Belić et al.⁷⁷ applied an oxygen-plasma postdeposition cleaning procedure, resulting in an Au content over 70 at. %. Up to now, two purification approaches for complex 3D-objects have led to nonporous and compact morphologies, namely, e-beam assisted purification in water vapor² and laser assisted purification.⁵⁹ With the first method, the successful chemical transformation of highly carbon-contaminated structures into pure gold 3D-structures was demonstrated, resulting in plasmonic 3D objects.² Extracting the low density element from the binary composite material leads to extreme deposit shrinkage. Unfortunately, this shape distortion must be accounted for in design phase when using postdeposition purification. In comparison, laser assisted electron beam induced deposition (LAEBID) reduces the carbon content during the 3D-printing by cycling deposition and a photothermal purification step induced from a laser pulse.⁵⁹ Many applications are strongly related to the properties of the deposited material, therefore, we anticipate an increase in research targeting new or improved purification methods or new precursor materials in future.

In the following, we will give a brief review of recent applications and proof of principle studies that show the potential of 3D-FEBID in various fields. Figure 7 illustrates a few of those application concepts (nanoprobes, sensors, plasmonics, and magnetics). The focus is put on the prospects and unique aspects of 3D-FEBID as enabling technology; for a more detailed description, we refer to the respective studies.

The capability to direct-write defined 3D-nanostructures even on exposed or prestructured substrates opens up a wide range of possible applications, e.g., for nanoprobes. Sattelkow et al.⁷⁸ demonstrated a scanning thermal probe concept, where a tetrapod (similar to the ones shown in Fig. 5) was placed at the tip region of an atomic force microscope (AFM) cantilever. Bringing this platinum tetrapod into (thermal) contact with the sample changes the resistivity through the segments as a function of the substrate temperature [Fig. 7(a)]. In this case, the unique selling point is the small volume of the nanoprobes, allowing a fast electrical response as temperature changes. Furthermore, the small tip radius enables AFM imaging with high resolution at the same time.

In the mentioned AFM applications, the 3D structure is brought into contact with the substrate, which requires a certain mechanical stability/stiffness of the fabricated 3D geometry. The mechanical stiffness of 3D-FEBID structures under a load force was evaluated by Sattelkow et al.⁷⁸ and Lewis et al.,⁶¹ proving the suitability for such AFM applications with a suitable AFM-probe design.⁷⁸ 

Using 3D-FEBID structures to design NEMS is also a promising route. Actuation could be accomplished using magnetic FEBID materials and a magnetic field, as described by Vavassori et al.⁶⁷ Arnold et al. demonstrated the actuation of a 3D-structure via electrical AC fields⁶² [Fig. 7(b)]. In this study, the mechanical resonance frequency of the oscillating bridge changes when gas molecules adsorb at its surface. With that, this device can be used as a gas sensor.

The nanoscale dimensions of the 3D-FEBID objects make this technology interesting for applications in the area of nano-optics. 3D-FEBID allows the direct-write of freestanding 3D objects for plasmonic investigations, which are complicated or even impossible with other nanofabrication techniques like electron beam lithography (EBL). However, because of the poor gold content of FEBID gold deposits, a postprocess step is necessary to enable high plasmonic activity. One approach is to use a 3D-FEBID structure as scaffold and cover it with silver,⁴² and the other is to subject the gold-carbon structure to a purification step.^2,79 Using the latter approach, plasmonic hot spots with high intensity have been demonstrated on a single 3D-object² [Fig. 7(c)]. Chiral metamaterials, which facilitate applications with frequencies in the visible regime, have been realized by fabricating Pt helices in large arrays via 3D-FIBID^80,81 and 3D-FEBID.^37,81 As demonstrated in these studies,^37,80,81 the optical properties of the array can be tailored by the geometry of the nanohelices. In a similar study by Kosters et al.,⁷¹ large arrays of SiO_x-FEBID helices were sputter coated with gold, which resulted in an enhanced chiral optical response. Such photonic metamaterials consisting of nanostructures in helical shapes are very promising candidates for future applications in the area of biosensors. Currently, nanomagnetics is a very active application space for 3D-FEBID. The novel capabilities to construct complex geometries allow the direct-write fabrication of artificial lattices, where magnetic frustration can be studied (keyword spin ice). In this context, Keller et al.⁶⁰ and Mamoori et al.⁷² have discussed the switching of magnetic moments of such magnetic 3D-geometries for the precursor HCo₃Fe(CO)₁₂ [Fig. 7(d)]. Sanz-Hernández et al. realized a magnetic conduit by a combination of 3D-FEBID structure as a scaffold for covering with magnetic materials via physical vapor deposition.^63,64

In our opinion, 3D nanoprinting via electron beams has great potential. Although 3D nanoprinting via FEBID has made tremendous progress and has matured from a trial-and-error method to a reproducible direct-write technology in the very recent years, we think that the wider audience is not yet familiar with FEBID's unique and novel 3D nanoprinting capabilities. One step toward more publicity could be the demonstration of eye-catching 3D-nanoarchitectures (e.g., Nano-Louvre² and Bucky-Ball³), presentations at conferences, and showcasing of applications. We also recommend the term “3D-FEBID” as a common and explicit term for this 3D-printing technique to avoid confusions due to duplicated names for the same technology.

A generic FEBID 3D nanoprinting technology for research and development would have a far reaching impact; a capability independent of precursor, design, and the overall size of the 3D architecture. Since the FEBID community is small with only a handful of scientists currently working on 3D-FEBID, several aspects are still unexplored. However, the progress made in recent years has opened up completely new possibilities and attracted increasing interest also outside of the core community. The growing understanding of the fundamentals has helped to overcome many challenges, furthermore, the dedicated 3D-FEBID software solution enables simple access to this technology. Therefore, in the near term, we expect an increasing number of 3D-FEBID users. As long-term vision, the development of a dedicated 3D-FEBID nanoprinter would also pave the way to industrial applications.

In the following, we point out some of the practical challenges and limitations and share our concepts and visions for tackling them. Furthermore, we will mention current developments in the field and share our personal view about the future of 3D nanoprinting and ideas of potential applications.

Advancement in gas injection technology has lagged far behind the continuous improvement seen in scanning electron microscopy. In many cases, self-built solutions are used. Advancements in nozzle design, carrier gas coflow options, and the ability to rapid interchange precursors are needed. Further, to broaden the range of multimaterial structures attainable by FEBID will require multinozzle systems. Currently, the main bottleneck in the 3D-FEBID process is the availability of precursor at the growth front. In this context, proper GIS alignment is critical—isotropic 3D deposition requires a spatially homogeneous gas field.

In addition, new concepts are in development to achieve a high surface coverage of precursor molecules with the help of a cooling stage. The lower temperature of the substrate increases the residence time of the molecules, which leads to higher growth rates for 2D-FEBID as well as for 3D-FEBID as has been demonstrated for nanopillar growth.⁸² In the latter case, the cooled substrate serves as heat sink and can reduce the disruptive effect of beam heating⁵² that was discussed above.

For many industrial applications, a limiting criterion for 3D-FEBID is the demand for high throughput. Unfortunately, 3D-FEBID is a slow process (tens of nanometers per second). It is anticipated that deposition rates can be increased with an optimized GIS system and/or substrate cooling. Multibeam concepts, such as the one used in electron beam lithography,^83,84 could scale production. However, in this case, one has to ensure homogeneous growth rates, which require a new precursor delivery concept or at least a compensation for spatial varying deposition rates. Another route to synthesize larger 3D-objects on a reasonable time scale is the combination with other 3D-printing techniques, e.g., two-photon-lithography or ink jet printing. While other methods allow rapid deposition of microscale objects, 3D-FEBID can finally add 3D features on the nanoscale.⁶¹ This combined approach allows the synthesis of hierarchical structures,⁶⁵ which are not possible with a single technique. For example, the final 3D-FEBID structures may be used for site-specific surface functionalization, e.g., to create locally hydrophobic coatings.

Note that some electron microscopes do not provide the advanced beam control that is required for 3D-FEBID. The inclusion of suitable patterning engines should be a future task for the microscope manufacturer; external patterning engines can also be used as alternatives.

Summarizing the future instrumental needs for 3D-FEBID, there is room for improvements in the gas injection systems and patterning engines, furthermore, a multibeam system and a cooling stage may also be helpful for higher growth rates.

The goal of making a reliable and easy-to-use 3D-nanoprinter available to a wider audience requires further basic research. For this purpose, experimental investigations and simulations must be carried out. One critical aspect is the reduced growth rates observed for long segments^28,37 and tall structures.⁴ As discussed above, beam-induced heating is one aspect responsible for these deviations.⁵² With the knowledge of the cause, work is currently being done on the design of compensation strategies. A respective software module to compensate for the reduced growth rates with structure length/height will be implemented in the 3BID software in the near future. In parallel, there are ideas to correct the patterning velocity on the fly during the deposition process via the sample current signal. A deviation from the expected current signal^52,85 indicates a change of growth condition, thus, real time compensation can be envisioned via a feedback loop, where the patterning velocity can be adjusted in real time. For tall architectures, beam defocus also plays a role. In order to compensate for the resulting degradation in the vertical growth rate, future work could implement an autofocus option that readjust the focal plane during 3D growth.

Research efforts are underway to determine the factors governing the cross-sectional shapes of single segments. A recent study of 3D-FEBID nanowires revealed that the cross-section depends on the wire inclination angle and primary electron energy.⁴⁸ These results serve as starting point for the design of arbitrarily shaped segments via suitable adaptation of the patterning strategy. The wire dimensions/cross sectional areas are of particular interest at the nanoscale as they dictate the properties such as electrical/thermal conductivity, mechanical stiffness, magnetization, or plasmonic modes. With control over the shape, the FEBID nanowires can be customized to fit the respective application. The ability to modulate the dimension arbitrarily within a single wire might also open up new possibilities.

A special type of shape is closed or semiclosed structure as depicted in Fig. 8(a) and previously shown by Sanz-Hernández et al.⁶³ The expansion of 3D-nanoprinting from meshed-styled objects to closed-wall/inclined plane architectures will open up new fields of applications. The integration of such wall-like geometries is on the road map for future versions of the patterning software.

While we have only mentioned shape tuning options that lead to an increase of the wire dimensions, focused ion beam (FIB) milling can be used for precise postfabrication shaping with lateral nanoprecision. In this context, it comes in handy that most 3D-FEBID experiments are conducted in FIB-SEMs, which enable subsequent FIB processing. This subtractive shape manipulation allows customization of wall shapes, as representatively shown in Fig. 8(a), or a reduction of the wire dimensions [Figs. 8(b) and 8(c)]. Other application examples might be postfabrication trimming of plasmonics nanostructures to reduce their size and/or changing sidewall angles for improved optical resonances. Furthermore, FIB processing of FEBID structures can be a useful tool to locally modify the material properties (e.g., via ion implantation).

Another topic is the realization of downward directed segments. Downward growth is not possible, or just to a small extent,³⁵ similar with all 3D printing techniques. Even horizontal wire growth in the focal plane is a challenge as the process parameters respond sensitively to any changes in the low-angle limit. An idea to realize downward pointing segments is the usage of support structures, a common practice in 3D printing technology. The support structures could be fabricated from a material with different functionality (e.g., insulating, while the main geometry is conductive). Alternatively, when identical material is used, the supporting elements can be removed afterward via careful focused ion beam milling from a tilted position.

An open question is whether applying very low primary energies can lead to more cylindrical wire cross sections. This assumption can be rationalized by considering the gradual change in cross sectional wire shape from elliptical toward circular and the smaller interaction volume of low energy electrons in the wire. However, most microscopes have reduced resolution at low energies. While increased proximal deposition occurs at modest beam energies (down to 2 keV), at very low landing energies there could be a turnover peripheral deposition as the SE coefficient turns over.

3D-FEBID is possible with many precursors as shown in Table I. The demonstration of 3D-FEBID for other precursor materials is still pending. A direct comparison of the results from different studies is often complicated because of varying conditions between the microscopes (GIS setup, vacuum system, electron beam setup). An ideal case would be a study that compares many different precursor materials in the same microscope using an identical GIS alignment and the same test-structure (e.g., diving board, as used as calibration geometry in 3BID⁴⁵ and F3D⁴ and by many other studies^2,3,44,59,63 before).

Part of the activities in the FEBID community is the development of new precursor materials, as addressed in the European Union's projects CELINA and ELENA. Precursors designed specifically for FEBID must meet several demands, often conflicting, including high volatility, shelf stability, and purity in the as-deposited state. Recent efforts reported the deposition of new elements and binary alloys derived from heteronuclear precursors.^75,86,87 As a personal view, we are expecting new and better precursors in future, which will expand 3D-FEBID's portfolio and will open up new possibilities for applications.

For some applications, it might become necessary to modify the material properties in an additional process step such as postgrowth irradiation⁸⁸ or exposure to specific gases⁷⁴ with or without electron beam exposure. In this context, the literature reported purification approaches^74,75 for planar FEBID deposits have to be evaluated on their applicability on freestanding 3D-FEBID structures. Furthermore, those and new strategies have to be tested for different materials as well, which is the gateway for a generic material tuning approach in FEBID based 3D nanoprinting. An ideal purification strategy combines 3D nanoprinting and purification in one single process step. A promising candidate for this task is the injection of water vapor during 3D nanoprinting; pure gold structures have already been demonstrated via simultaneous injection of water vapor⁷⁹ (but for planar deposits) on the one hand and for 3D geometries² (but in an additional process step) on the other hand. Although lower growth rates are expected due to competition of precursor and water molecules for adsorption sites, 3D nanoprinting of pure structures such as platinum or gold might become possible from other organometallic precursors.²⁷ 

Also on 3D-FEBID's roadmap is the combination of multiple materials in a single 3D geometry. This adds another dimension to the three-dimensional space via the material properties, what we define here as 4D-nanoprinting. Several approaches seem to be feasible: (1) serial processing with injection of different precursor gases.⁶⁰ This enables the separation of areas with different material properties in 3D (e.g., magnetic/nonmagnetic, conductive/insulating). One concrete example is the coating of ferromagnetic FEBID nanowires with a thin layer of Pt-FEBID to protect the core from oxidation, thus resulting in improved magnetic properties.⁸⁹ One can also achieve material combinations by (2) injecting several precursor gases in parallel,⁹⁰ whereas the stoichiometry can be tuned by the individual precursor flux and the distances of GIS nozzles, respectively.⁹¹ (3) Binary/ternary compounds can also be realized by deposition from heteronuclear precursors.⁷⁵ With all possibilities together (and in combination with postprocessing strategies), a broad range of functionalities could be realized in 3D space, whether strictly separated or gradually changing.

In our view, one of the most important and urgent steps to pave the way toward a generic 3D nanoprinting technology is a user-friendly software for a straight forward generation of patterning files. This development would reduce the initial hurdle for scientists and later on for all other users. With a growing number of 3D-FEBID users, a rapid future advancement of this technology is expected.

At the moment, we intend to join forces to combine both the current software solutions (3BID⁴⁵ and F3D⁴) to a more ubiquitous 3D nanoprinting software environment (with working title “FROG,” acronym for FRankfurt–Oak Ridge–Graz, indicating the involved work groups). The core of the new 3D-FEBID software translates the design to a process file, where advanced functionalities such as correction modules for specific precursors or temperature/shape compensation can be integrated as add-ons due to the modular program design. In the front end, a user-friendly graphical interface should allow a straightforward creation of arbitrary geometries. In this context, traditional 3D-printing software solutions can be taken as inspiration. Furthermore, the import of common 3D file formats (such as .obj or .stl) created via conventional CAD programs should be possible. This will allow the use of huge online libraries of available 3D-models. A respective module will convert those 3D-files into vertices and segments and translate them to 3D-FEBID process files, what is already possible for .obj files with an updated version of the F3D.

In Sec. III E, we have discussed already demonstrated applications of 3D-FEBID. In the following, we will give a more speculative perspective of some potential applications in research and development that could be realized with the capabilities of 3D-FEBID. Many applications are not yet foreseeable, nevertheless, some general concepts, which clearly benefit from the unique advantages of 3D nanoprinting via electron beams are presented.

First, 3D-FEBID structures could serve well as scaffolds. In this case, the nanoarchitectures are covered, for instance, by atomic layer deposition (ALD),⁶¹ chemical vapour deposition (CVD), or physical vapour deposition (PVD).⁶³ The (multi)layers can be used to increase the mechanical stiffness⁶¹ or to add a new functionality (e.g., magnetic⁶³ and optical chirality⁷¹).

Due to small dimensions, nanoeffects can be observed for 3D-FEBID structures. In the field of nanophotonics, for example, three-dimensional plasmonic active objects can be realized. As it has been shown in proof of principle studies, plasmon resonance can be achieved by either covering the FEBID-scaffold with silver⁴² or by purifying the FEBID-material itself.² The novel aspects, namely, freestanding and three-dimensional, make it interesting for fundamental research, in particular, to arrange the location of plasmonic hot spots in 3D-space. As well as for use in science, such plasmonically active 3D-structures can be employed in biosensor concepts or for tip enhanced Raman spectroscopy (TERS), which is currently the focus of our laboratory. In addition, the 3D objects (e.g., helices^37,54,71,81) can be arranged in large arrays for the usage in photonics, i.e., as optical filters or sensing applications.⁸¹ Due to the flexibility in design, heights and shapes can be adapted to fit the targeted frequency.⁷¹ 3D-FEBID might also be suitable to construct metamaterials. Here, one idea is to deposit 3D structures in a suitable arrangement on the cross section of a light fiber to realize a planar lens. Light of a specific frequency is guided through the light fiber and focused to a spot with this type of metalenses. Not only optical metamaterials, but also mechanical metamaterials, which are characterized by an unusual set of physical properties such as a negative Poisson's ratio⁹² might be a future field for 3D-FEBID. Specifically designed 3D-lattices⁹² can be realized, thanks to the ability to fabricate complex three-dimensional geometries, resulting in ultralight, ultratough, and ultrastiff mechanical metamaterials.

Since the 3D objects can also be placed on morphologically exposed areas, the 3D-FEBID technique is ideal for the development of nanoprobes. A recent example is the modification of the tip region of an AFM cantilever to facilitate advanced measurement modes such as conductive AFM or scanning thermal probes,⁷⁸ FEBID-based magnetic tips for magnetic force microscopy⁹³ and magnetic resonance force microscopy.⁹⁴ The typical radius of a 3D-FEBID tip is on the range of a few nanometers, therefore, in parallel to the mentioned advanced modes, conventional height measurements with high lateral resolution can be performed at the same time. Furthermore, the flexibility of the tip shape (e.g., a hammerhead¹¹) might enable access to challenging surface morphologies such as undercuts and sidewalls. In addition to AFM applications, nanotools such as nanotweezers, hooks, or grids can be made to manipulate the sample or grab individual particles, opening up applications in life science.

Three-dimensional nanomagnetic structures are of particular interest, e.g., for spin-ice architectures. This exciting field offers many new ideas for applications, summarized in recent review articles.^75,95

The general suitability of FEBID to construct electron emitters has also been demonstrated.^24,25,96 We think that there is still room for improvement via optimizing the nanoemitter geometry and material properties. Such a miniaturized electron gun (or an array of several emitters) could be again placed on an AFM cantilever and serve as low-energy field emitting electron source for scanning probe lithography.

While most of the mentioned applications are in reach or even in development, a more speculative field for 3D-FEBID is micro/nanorobotics. Here, 3D-FEBID could contribute via deposition of small-scale features such as tools or other functional elements for the micro/nanobots. Additionally, an already finished device can be subsequently modified via the 1-step direct-write approach in combination with the flexibility of 3D-geometry, substrate morphology, and choice of materials. With these strengths of 3D-FEBID also micro/nanoelectromechanical (MEMS/NEMS) systems come into reach as a promising future direction.

One can also speculate about upcoming applications using closed or semiclosed 3D-FEBID structures like shells or hollow spheres. Specifically, the combination of different materials in one structure could unveil novel possibilities for future applications.

Summarizing the presented ideas in this section, 3D-FEBID is expected to be an enabling technology for many applications in science and development in the near future.

3D nanoprinting via focused electron beam induced deposition has made tremendous progress in recent years. The advances presented here have opened up new perspectives in science and technology, which could become a reality thanks to the unique advantages of 3D-FEBID, which are (1) small feature sizes with dimensions down to 10 nm, (2) the complexity of the architectures, (3) the ability to pinpoint the freestanding 3D objects on morphologically challenging substrates, (4) the fabrication in one single direct-write step, and the flexibility in (5) structural design and (6) material choice. After presenting the state of the art of 3D nanoprinting via electron beams and several showcase applications in the areas of nano-optics, nanomagnetics, and nanoprobes, we discussed current challenges and ongoing research. We put special focus on the software solutions for 3D-FEBID as one of the urgent steps. Based on the current status of 3D-FEBID, the anticipated advances in the instrumental setup, basic research, and materials, we have opined a multitude of ideas for future applications. Overall, we anticipate a bright future for 3D-FEBID as a means to a generic 3D nanoprinting technology in research and development.

R.W. and H.P. would like to thank Chris Schwalb, Ernest Fantner (both GETec Microscopy), and Ferdinand Hofer for their continuous support. Gratitude particularly goes to Michael Huth, Ivo Utke, Lukas Keller, Jürgen Sattelkow, Anna Weitzer, and Margit Wallner for fruitful discussions and their input for this perspective article. The same authors also acknowledge the financial support by the Christian Doppler Research Association (CDL-DEFINE), Austrian Cooperative Research (ACR), FFG Beyond Europe project (AIM, No. 11056459). The financial support by the Austrian Federal Ministry for Digital and Economic Affairs and the National Foundation for Research, Technology and Development is gratefully acknowledged. J.D.F. and P.D.R. acknowledge the support from the Center for Nanophase Materials Sciences, which is a DOE Office of Science User Facility. P.D.R. also acknowledges support for 3D plasmonic research from the National Science Foundation under Grant No. NSF DMR 1709275.

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).