Applicability of focused Ion beam (FIB) milling with gallium, neon, and xenon to the fracture toughness characterization of gold thin films

Focused ion beam (FIB) milling is an increasingly popular technique for fabricating micro-sized samples for nanomechanical characterization. Previous investigations have cautioned that exposure to a gallium ion beam can significantly alter the mechanical behavior of materials. In the present study, the effects of gallium, neon, and xenon ions are scrutinized. We demonstrate that fracture toughness measurements on freestanding gold thin films are unaffected by the choice of the ion species and milling parameters. This is likely because the crack initiation is controlled by the local microstructure and grain boundaries at the notch, rather than by the damaged area introduced by FIB milling. Additionally, gold is not susceptible to chemical embrittlement by common FIB ion species. This confirms the validity of microscale fracture measurements based on similar experimental designs.


Introduction
Focused ion beam milling (FIB) has fundamentally changed nanomechanical testing, inasmuch as it has allowed unprecedented experimental designs, which are useful for targeting specific material properties at the nanometer scale [1,2]. For example, FIB-machined samples were pivotal to the discovery of geometric size effects on strength [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20]. FIB-based techniques also have the unique ability to isolate and investigate the mechanical contribution of microstructural features and defects-such as grain and twin boundaries [21][22][23][24][25]-and are increasingly used to study the fracture mechanisms at the nanometer scale, for instance in thin films [26][27][28][29][30][31]. The increasing popularity of FIB techniques has continued, despite concerns voiced from the very beginning over possible detrimental side effects of the ion milling process. Kiener et al. evidenced significant damage to the crystal structure of processed samples, ranging from the nucleation of defects and dislocations to the implantation of ions [32]. This damage has the potential to significantly affect the mechanical properties of the samples. In the worst case, Shim et al. evidenced a drop in strength in a single crystal by one order of magnitude [33,34]. Polycrystals are also affected when the ion species segregates at the grain boundaries and forms a ductile phase. In such a case, Xiao et al. evidenced a drop in strength in aluminum micropillars because of the formation of a eutectic Ga-Al phase at the grain boundaries [35,36]. Some strategies for mitigating FIB damage have already been suggested, such as shadow milling [12,37] or post-treatment annealing [38,39], but they are neither universally effective nor widely used yet. For the time being, most studies merely rely on the assumption that the pristine structure of the samples is preserved throughout the milling process. The aim of the present study is to investigate whether valid measurements of the fracture properties of thin films are possible-for instance by bulge testing [40][41][42]-and to determine the influence, if any, of the ion species and milling process used.

Effect of ion species
The investigations focused on 100-300 nm gold membranes obtained by thermal evaporation. In order to study the influence of ion species on the measured fracture toughness of these films, specimens obtained from a single deposition batch were selectively notched with gallium, neon, or xenon FIBs. Details on the experimental procedure are provided in the "Materials and Methods" section. Representative micrographs showing the notch geometry obtained with each ion species are provided in Fig. 1a-c. Ne-FIB and Xe-FIB yielded the sharpest and the bluntest notches, respectively.
The fracture toughness of the corresponding membranes was measured using bulge tests [40][41][42] and plotted in Fig. 2. Irrespective of the ion species, the fracture toughness data exhibit a correlation with the film thickness. The slightly increasing trend in the Ne-FIB data corresponds to the expectations from Bluhm's and Knott's models [43,44], as well as from previous observations based on Ga-FIB sample preparation [26,45]. For all conditions, the scattering of the data in Fig. 2 is rather low and consistent with the few FIB-free fracture experiments on thin films reported in the literature [46][47][48]. In line with expectations, thinner films exhibit a greater scattering than their thicker counterparts, presumably due to a greater contribution from their surface roughness [28], which introduces some randomness. Because the curves for different ion species partially overlap, it is virtually impossible to ascertain any effect of the fabrication process on the fracture toughness measurements. Note that the Xe-FIB samples-which only encompass three 250 nm samples-fall very well in line with the Ga-FIB ones. While the Ga-FIB and Xe-FIB measurements are virtually indistinguishable, Ne-FIB exhibits some discrepancies. However, there is no consistent trend throughout the investigated thickness range. Possible explanatory approaches and implications for this apparent contradiction are discussed in the next section.

Effect of notch radius
Figure 1a-c illustrates how strongly the geometry of the slit is influenced by the FIB ion species, with the Xe-FIB notches being considerably blunter than the other two ion species. The different notch radii should in theory influence the measurements, as a blunt tip should lead to an overestimation of the fracture toughness [47,49,50]. The effect of the notch radius was investigated in practice by structuring a 150-nm gold sample with an intentionally blunt notch of ca. 500 nm radius. Its fracture behavior was compared to reference samples with sharper notches of ca. 100 nm radius, which had been produced using the same Ga-FIB technique, see Fig. 3a, b. Against expectations, the fracture toughness did not increase with a Article blunter notch, but rather-slightly-decreased from ca. 2.0 to 1.9 MPa m 1/2 (see Fig. 3d). This observation suggests that the notch tip radius does not play any significant role in polycrystalline gold films. In the next section, we will investigate whether the FIB-milling process itself matters.

Effect of ion damage
Scanning transmission electron (STEM) micrographs indeed evidence a FIB-damaged region around the notch. The clustering of crystallographic defects typically results in dark spots on the STEM image, such as those visible ahead of the notch in Fig. 4, which corresponds to dislocation loops and implanted point defects. With the investigated 56-nm-thick gold specimen, a significant concentration of defects can be observed as far as 400 nm away from the notch tip. They rapidly fade away at larger distances. This lateral spread of the damage zone is unexpectedly high when compared to SRIM (stopping range of ions in matter) [51] or molecular dynamics [52] simulations, which are usually regarded as reliable predictions [32,53]. With the experimental parameters of this study, the SRIM simulations shown in Fig. 5 predict a damage radius of 5-25 nm only. In practice, gallium implantation should also be limited to a very thin amorphous layer at the surface [32,54,55], provided the ion beam is perfectly focused and the FIB apertures are in pristine condition. However, gallium is likely to be implanted in a wider zone if extensive beam tails are present, as it is the case here. The combination of a high point defect density and local heating by the ion beam might additionally increase the diffusion rate and lead to a wider lateral spread of the implanted gallium ions [56]. In any case, the 400-nm damage zone remains small compared to the 10 µm length of the slit.
It is usually assumed that FIB damage mechanically weakens [33,34,42,57] the area around the notch and thus reduces the fracture toughness of the membrane. This hypothesis is consistent with our previous AFM observation [26] of brittle fracture processes taking place locally at the notch tip, while a more ductile deformation dominates away from it. In the present study, we attempted to selectively avoid FIB effects by creep loading some FIB-notched samples prior to fracture testing. Creep deformation yielded an intergranular crack extending past the FIB-affected area (see Fig. 3c). Although subsequent bulge testing led to significantly higher K IC values than for plain FIB notches (see Fig. 3d), it is questionable whether the difference can be ascribed to FIB effects. Indeed, the creep-induced cracks are prone to crack branching and crack deflection along grain boundaries oriented perpendicularly to the crack (see Fig. 3c), which could also account for the apparent toughening.
In any case, note that the STEM micrograph shown in Fig. 4 originates from an extremely thin specimen-in order to ensure electron transparency at 30 keV-and therefore corresponds to a lower bound of damage density. Structuring thicker films require higher gallium doses in order to mill notches through the whole gold layer, which has been shown by Kiener et al. to increase the concentration of gallium in the vicinity of the FIB-milled area [32]. In the case of gold, a critical concentration of solute gallium above which it segregates to grain boundaries, as seen in aluminum [58], cannot be ruled out. Due to the low solubility of gallium in gold at room temperature (ca. 3 at. %), brittle intermetallic AuGa phases might form as well [59]. Both the brittle AuGa phases and the embrittled grain boundaries have the potential to limit the fracture toughness of the films, which might explain why its increase slows down at large film thicknesses (see Figs. 2 and 3).

Discussion
It is well known that FIB-induced damage can influence the mechanical behavior of small-scale samples. This can result from a number of different types of FIB damage: knock-on damage-induced dislocations [32], implanted ions [32,60], and gallium embrittlement [35]. Dislocation loops and point defects implanted by FIB can increase the strength of materials through Taylor hardening. Strengthening is usually regarded as detrimental to the ductility of face-centered cubic metals, but this does not hold true for polycrystalline thin films. Indeed, we previously showed that the fracture toughness scales with the yield strength, because the latter controls the necking behavior that is critical for crack propagation [28]. FIB damage from ion implantation would result in a further toughening effect, as the compressive residual stress induced by the strain from the implanted ions would act to resist crack nucleation [60]. These implanted residual stresses scale with the Young's modulus of the material and the implanted dose of ions [61], which further scales with the beam current and exposure time. There are many examples of a gallium-induced embrittled zone in the literature. Best et al. used Ga + , Xe + , and He + beams to prepare notches for fracture toughness measurements on micro-cantilevers of CrN [61]. They observed a smaller fracture toughness when the cantilevers were notched by Xe-FIB instead of Ga-FIB. The physical origin of this effect was not totally accounted for in their study. Furthermore, Xiao et al. demonstrated that the mechanical properties of aluminum micropillars prepared by Xe-FIB and Ga-FIB clearly differed [35]. This was rationalized by the expected embrittlement of aluminum by liquid gallium. Lauener et al. [62] observed the opposite effect from Best et al. [61], where the fracture toughness of Si micropillars fractured by indentation pillar splitting was higher for pillars milled with Xe + compared to Ga + ions. This disparity increased with decreasing pillar size, and this was rationalized by the increase in residual stress from the larger beam tails of the Xe-FIB as the pillar diameter decreased. In terms of dose, although the 250-nm-thick gold films might be expected to be the most affected by gallium-induced embrittlement due to their long processing time, the fracture toughness measurements by Xe-FIB and Ga-FIB were virtually identical (see Fig. 1c). Therefore, the low fracture toughness of the gold films [26-28] cannot be ascribed to gallium embrittlement. Granted, while xenon ions are heavier and have a lower penetration depth than gallium ions of the same energy (see Fig. 5), they also lead to greater amorphization of the surface layer [63]. However, since xenon is a noble gas, it is not expected to bond or react with gold atoms and weaken the film. Instead, xenon rather tends to form nanometer-large bubbles, as was observed in silicon [64]. Similar bubbles are expected to be formed during Ne-FIB processing. The primary difference between Ne ions and Ga-FIB or Xe-FIB is that neon ions are lighter and thus penetrate deeper into the target material (see Fig. 5).
Another significant factor which influences the extent of FIB damage is the milling strategy employed during sample  [65]. The implantation is a function of the duration of the beam in the vicinity of the sample surface, the angle of incidence, beam current, focus of the beam (beam tails) et cetera. This is particularly relevant for the present work, as the films are milled through their total thickness using a simple line cut at normal incidence. This means that no material remains which had experienced normal implantation from the beam, unlike top-down milled notches in other testing geometries, such as micro-cantilevers [57,66,67]. Here, the damage inflicted to the material around the notch mostly originates from the tails of the ion beams. The sharper notch produced by the Ne-FIB-visible in Fig. 1-is the consequence of the tails of a Ne + beam being generally smaller than those of a typical Ga + or Xe + beam. This is because of the narrower current distribution that can be ascribed to its smaller source and spot size [68]. Thus, it is likely that the lateral dimension of the damage-affected zone is smaller with Ne-FIB than with Ga-FIB and Xe-FIB. Apart from this, the inflicted damage is very similar, namely amorphization of the surface layer and an increase of point defect density that might result in clusters or dislocation structures [53]. Nonetheless, the Ga-FIB, Xe-FIB-and to a lesser extent Ne-FIB-based measurements are in very good agreement. This is remarkable, not only because it rules out gallium-induced embrittlement, but also because the geometry of the notches strongly differs, with Xe-FIB yielding the coarsest notches and Ne-FIB the sharpest one, see Fig. 1a-c.
The independence of the fracture toughness measurement from the notch geometry can be rationalized by the intergranular character of crack propagation, previously evidenced in [26,27]. It is likely that cracks initiate from a favorably oriented grain boundary ahead of the notch. As long as at least one such grain boundary is present at the notch, the fracture toughness measurement should not be affected by the actual notch tip radius. Under this assumption, the slightly lower fracture toughness of the specimen with a blunt notch (see Fig. 3d) can be explained  by the presence of more grain boundaries in front of a notch, which increases the likelihood that one of them is favorably oriented for crack initiation. Note that the sensitivity of crack nucleation and propagation on the microstructure present in front of the notch certainly accounts for much of the scatter in the data. Within this scatter, the fracture toughness measurement does not appear to be sensitive to the notch geometry or to damage specific to an ion species. A small effect cannot be ruled out, but it is concealed by larger sources of influence, such as the yield stress and hardening behavior of the film, which control the necking process in front of the notch [28].

Conclusions
The fracture toughness of gold thin films measured by bulge testing was found to be insensitive to the ion species of the focused ion beam (FIB) used for milling the pre-notch. One reason for this is that none of the main ion species (Ga + , Ne + , Xe + ) is likely to cause grain boundary embrittlement by chemically binding to gold, such as in gallium-aluminum. Furthermore, the throughthickness geometry of the slit ensures that there is no remaining contamination from normal exposure to the beam itself. The amount of FIB damage is reduced to that originating from the beam tails. Still, the different FIB technologies introduce different amounts of point defects in front of the notch, which are furthermore spread over different distances. They also yield very different notch radii. The K IC measurements are likely unaffected by this, because a crack typically nucleates at a favorably oriented grain boundary present at the notch. It appears to make little difference whether the density of point defects is high or low around this grain boundary. For these reasons, bulge testing of gold thin films results in consistent fracture toughness measurements irrespective of the ion species that was used.

Materials and methods
The gold thin film specimens were produced by physical vapor deposition onto a 4 by 1 mm sacrificial SiN x carrier membrane (Silson Ltd, UK), using a custom-built thermal evaporation unit. The films were deposited at a rate of 0.6 Å/s on a rotating substrate, and the substrate temperature was maintained at 120 °C during the process. Substrate heating was immediately switched off after completion. The microstructure of the films was columnar (see Fig. 6), with a surface roughness (R a ) of about 2.7 nm. The average in-plane grain size is close to the film thickness. The FIB-milled cross-section (shown in Fig. 6b) indicates a zone II structure with mostly columnar grains but also with some smaller and V-shaped grains [69,70]. The films were finally made freestanding by selectively removing the SiN x sacrificial by reactive ion etching (RIE) for 3 min by applying a 30 sccm flow of CF 4 at 12.67 Pa and a RF power of 100 W at 20 °C.
The next step consisted into introducing a slit in the center of the membranes by focused ion beam milling (FIB), so as to localize the fracture initiation. Unlike publications [26- 28,42], this fabrication step was not limited to using a Ga-FIB. Some of the films were also notched by Xe-FIB and Ne-FIB in order to investigate the effect of the ion species on FIB-induced damage on the resulting fracture toughness.
The reference Ga-FIB slits were fabricated with a 1540 CrossBeam workstation (Carl Zeiss Microscopy GmbH, Jena, Germany) equipped with a standard liquid metal ion source (LMIS). The notches were typically prepared with an ion current of 10 pA at an acceleration voltage of 30 kV. In order to minimize ion damage, beam alignment was performed at remote locations of the sample, where the film is supported by the silicon frame. The freestanding region of the film was only imaged by the electron beam and was never exposed to the ion beam but for milling operations. Milling regular through-thickness notches took 5 to 80 s depending on the film thickness. The resulting notches were typically 100-360 nm wide. After milling, the exact length of each notch was measured from an electron micrograph. On one occasion, a pre-crack was subsequently grown from both ends of the FIB slit by creep loading the membrane at very low pressure (5-10 kPa). Its creep-driven growth was monitored insitu in an AFM to ensure that the final crack extended beyond the FIB-affected zone present around the original notch. An Orion NanoFab system (Carl Zeiss Microscopy GmbH, Jena, Germany) at the Max Planck Institute for the Physics of Light (Erlangen, Germany) was used to prepare notches by Ne-FIB in 100, 200, and 300-nm thick membranes. The ion beam in this system is generated by applying a high electric field to a very sharp tungsten needle with typically only three atoms (a trimer) at its apex. At the same time, the needle is surrounded by neon gas. The electric field at the tungsten trimer is sufficiently high to ionize the neon atoms in its vicinity. This gas field ionization source (GFIS) is not as stable as the LMIS used in Ga-FIB systems. Very high fields are required for the ionization of neon atoms, which raises two challenges. First, it becomes very likely that impurity atoms adhere to the tungsten tip. These may act as alternative apexes, which reduce the intensity of Ne + ions generated at the main tungsten trimer. Second, the tungsten trimer itself is not fully stable at such very high electric field levels. Voltage pulses, which can be applied to remove impurity atoms or excess tungsten atoms can also cause the atoms of the trimer to evaporate. However, when a firm trimer can be formed, stable operation of the source is possible over several hours. The Ne + ions extracted from the source are finally accelerated in an electric field and focused by an electron optics system. In comparison to Ga + ions, they can be focused to smaller spot sizes with smaller beam tails [54] and are typically used to mill finer features. For milling the notches, a neon gas pressure of 6.66 × 10 -6 mbar was maintained at the ion source. Ion currents between 5 and 7 pA were used at an acceleration voltage of 25 kV. A 10-µm long and 1-nm wide line was scanned by the Ne + beam with a step size of 0.25 nm and a dwell time of 1 µs. The milling progress was estimated by monitoring the intensity of the generated secondary electrons. A sudden decrease in this intensity followed by a plateau indicated that milling was completed with the notch extending through the whole thickness of the film. Milling took about 30 to 250 s, respectively, for 100 and 300-nm-thick Au membranes. The resulting notches were 50-110 nm wide. Apart from milling operations, the gold membrane was never exposed to the Ne + ion beam. Imaging for alignment and centering purposes was done with a 3 pA He + beam at 25 kV, since there is no electron beam available in the system. As a matter of precaution, because He + ions can lead to swelling of the exposed material, the beam alignment was performed on the silicon substrate frame outside the membrane area.
The Xe-FIB notches were prepared with a Fera system (Tescan, Brno, Czech Republic) in the microscopy center, ScopeM, at ETH Zurich, Switzerland. In this instrument, the xenon plasma is generated by electron cyclotron resonance (ECR), which couples a magnetic field with microwaves. This technology results not only in faster milling rates, but also a larger spot size than conventional Ga-FIBs. Notching was performed with an ion current of 100 pA at an acceleration voltage of 30 kV. Milling a 10-µm-long notch through a 250-nm-thick polycrystalline gold film took 3 to 4 s.
Mechanical testing of the notched specimens was performed with a custom-built bulge tester previously described in [26][27][28]. The membranes were pressurized at a rate of 250 Pa/s until failure occurred. The stress at failure was calculated from the raw pressure-displacement data using the model by Vlassak et al. [71], taking into account the machine compliance. The fracture toughness was subsequently evaluated using Irwin's model [72], as advocated by [40][41][42]. More details about the experimental parameters can be found in [26][27][28].
Microstructural characterization of the samples was performed with a 1540 CrossBeam workstation (Carl Zeiss Microscopy GmbH, Jena, Germany). Scanning transmission electron microscopy (STEM) imagining was performed with a Helios NanoLab 600i dual beam workstation (FEI, Hillsboro OR, USA).
The Monte Carlo simulations were performed with the SRIM (Stopping and Range of Ions in Matter) software package [51]. A 100-nm-thick gold target was exposed to ion irradiation by 10,000 gallium ions at 30 keV, neon ions at 25 keV or Xenon ions at 30 keV. The trajectory of the ions was plotted as a function of depth, as well as laterally.

Open access
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/ licen ses/by/4.0/.