FIB-SEMs combine a scanning electron microscope (SEM) and a focused ion beam (FIB) in a single instrument and are increasingly used to prepare cross sections and TEM lamellae of biological samples as well as of other thermally low conductive materials such as polymers . The easily induced heat damage is increasingly being reported . Despite the importance of this topic, there is not a large number of recent papers published looking at the underlying physics in the field of focused ion beams. A broader look at current literature on heat transfer induced by particle beams highlights that heat damage is not only problematic for FIB processing, but also presents challenges for light interactions with biological tissue as well as focused electron beam-induced deposition (FEBID) . The general approach to assess the beam-induced heat damage and undesired artifacts, regardless if working with ions , photons , or electrons , compares experiments to models based on heat transfer and to Monte Carlo or finite element simulations . Open source programs that assess heat deposition and diffusion are readily available to assess damage in light–tissue interactions . For electron beams, multidimensional models predicting electron beam-induced heating effects that lead to noticeable changes in nanostructure deposition geometry during FEBID processes have recently been published . These approaches, a thorough understanding of the parameters that govern the beam-induced heat damage as well as open source software, would also be beneficial for the FIB community, especially with the increase in FIB milling of thermally low conductive materials.

Recent work assessed the ion beam-induced heat damage and how this can be limited by FIB parameter choice . The reported results suggested that ion beam-induced heat can be minimized by using lower ion beam currents and by reducing the beam overlap as well as blurring the beam . The drawback of this approach, however, is increased processing times due to the small current (picoampere range). Other approaches that were successfully used to avoid or reduce heat damage include working at cryogenic sample temperatures, , short beam dwell times , as well as employing different scan strategies . This work builds on previously reported experiments and looks at the effect of the ion energy on the ion beam-induced sample heating to maintain nanoampere beam currents and fast processing times. Suitable ion beam parameters for lower-energy ions (5 keV) are assessed, and the results are compared to the conventionally used 30 keV ions.

The interactions of 5 keV gallium ions and the effect on heat accumulation due to ion impacts with the sample (skin) were investigated using the Monte Carlo simulation program SRIM , the program COMSOL (finite element analysis platform), and a numerical analysis using the forward time–centered space method to solve the 3D heat equation. This approach is discussed in detail elsewhere . The results are experimentally tested by milling a TEM lamella and assessing the ion beam-induced heat damage in collagen.

Purified collagen was selected as the experimental test material for two main reasons. First, it is the principal component of skin, for which well-documented literature values are readily available . Second, collagen’s fibrillar structure, visible by microscopy, is denatured by heat to give gelatin that lacks any fixed structure , making heat damage easily recognizable.

Despite the focus on Ga ions impacting in skin (simulations) and collagen (experimental), the broader results presented here are true for any type of instrument and ion species, which includes gallium FIBs, plasma FIBs, helium ion microscopy FIBs, as well as low-temperature ion source and magneto-optical trap ion source FIBs.

Figure 1: (A) Ion trajectories for 5 keV gallium ions (Ga+) in human skin. SRIM/TRIM assumes a beam with a single impact point. The projected range Rproj is marked as the dashed line in the ion trajectories plot. The peak phonon and ionization positions are marked in the collision cascade plots (B) as a dotted line.

Table 1: SRIM calculation parameters showing the total energy loss to phonons and ionization per incident gallium ion in skin.

Figure 2: COMSOL simulations of 5 keV gallium ions (Ga+) interacting with collagen. Top row: Single ion impact simulations. (A) Top view of a single 5 keV gallium ion trajectory in collagen and (B) the single ion vertical mid-plane temperature isosurface. (C) Trajectories and temperature isosurfaces for six consecutive ion impacts with a 1 nA ion beam current for 5 keV Ga+ in collagen.

To qualitatively assess the effect of multiple ion impacts within a short timeframe, six ion impacts for 5 keV ions were simulated. For a 1 nA beam, ions hit the sample on average every 160 ps. A scanned beam will generally not have moved away from any impact site before 100 ns have elapsed. These 100 ns are currently the typical minimal dwell time for most FIB machines. Consequently, for a 1 nA beam, approximately 625 ions will impact the same scan point before the focused ion beam spot has been moved one DAC step further. For a system with 1 nm placement accuracy, the sample, if not cooled to room temperature before those 160 ps have elapsed, will be heated continuously at this location. As long as minimal dwell times cannot be shortened below 1 ns, there will be heat accumulation within a single scan point when using nanoampere beam currents and nanometer spot sizes.

Figure 3: Illustration of the numerically solved three-dimensional heat equation for 5 keV Ga+ ions in collagen for different ion beam currents, (A) 1.4 nA and (B) 29 pA, assuming a 50 nm Gaussian beam profile. The surface temperature (top view) surface and the temperature of a cross section (side view, below) through the middle of the ion beam spot have been plotted after an irradiation time of 990 ns. The color scales are fixed to 340 K max for better comparison.

The previously reported model

is used to estimate the ion beam-induced changes in sample temperatures. A detailed description of the model is given in the previous publication .

Figure 4: SEM micrographs showing cross sections that were cut into collagen using (A) 30 keV gallium ions with 1 nA beam current and 1.4 × 1013 ions·cm−2 and (B) 5 keV gallium ions with 1.4 nA beam current and 1.8 × 1013 ions·cm−2. Unnaturally smooth areas within the cross section cut using 30 keV cannot be observed in the cross section that was cut with 5 keV.

Figure 5: SEM micrographs showing cross sections that were cut into collagen using 5 keV and different ion beam settings: (A) 29 pA, 20% overlap, 50 nm blur; (B) 70 pA, 20% overlap, 50 nm blur; (C) 0.12 nA, 50% overlap; (D) 29 pA, 20% overlap, 200 nm blur; (E) 70 pA, 20% overlap, 200 nm blur; and (F) 0.12 nA, 20% overlap, 200 nm blur. The scale bars correspond to 10 µm.

Reducing the ion beam energy reduces the amount of energy that is transformed into heat for each ion impact per unit of time. In addition to a lower energy loss to heat within the sample when reducing the ion energy, the final beam diameter is increased for lower energies in comparison to higher energies. The area underneath the ion beam is increased; therefore, the smaller resulting local ion dose and dose rate both play a significant role in additionally reducing heat damage. This reduces the probability by a factor of r2, with r being the ion beam radius, of an ion impact in close proximity to the previous impact and reduces any potential accumulative increase in local temperature.

Being able to use a higher ion current in the nanoampere range in comparison with the previously suggested heat-reduced approach (where the beam current was reduced) addresses the issue of increased patterning times and cross sections with unfeasibly large area.

Figure 6: TEM micrographs showing cross sections of collagen prepared with 5 keV gallium ions. The collagen fibers are visible in the TEM lamella prepared with currents in the nanoampere range.

Cryo FIB preparation has become a key enabling technology in recent years and reports on successful cryo FIB preparation of biological specimens at higher ion energies can be found throughout the literature . Heating has not been observed when working with frozen hydrated samples. The thermal conductivity of ice is 2–3 W·m−1·K−1. FIB-induced heat damage would not be expected for cryo FIB processes, since the thermal conductivity of cryo samples is orders of magnitude higher than that of the thermally low conductive materials used in this report.

The effects of acceleration voltage and ion energy on the heat damage during FIB processing was assessed using Monte Carlo simulations, finite element simulations, time-dependent numerical modelling approaches, a model based on Fourier’s law of conductive heat transfer, as well as cross-sectioning and TEM lamella preparation. Heat damage that extends into surrounding tissues could be avoided while using high currents when the sample was processed using lower ion energies (such as 5 keV).

By reducing the ion energy from 30 to 5 keV, the loss to heat can be reduced by a factor of six, while the sputtering speed is only reduced by a factor of two. The reduction in energy allows the operator to use nanoampere beam currents, which addresses the drawback of the previously reported approach of unfeasibly long patterning times. The results suggest that an additional reduction in ion beam current and applying blur and reduced overlap, when using low ion energies, can further reduce the induced heat. Using that combination might be required for processing materials exhibiting an even lower thermal conductivity than skin.

The theoretical work presented here assumes an isotropic and homogenous and flat specimen. Samples are generally more complex, exhibiting different sample topographies as well as material anisotropies. These factors are likely to play a crucial role, and future work should be carried out to better understand the effect of complex heat transfer mechanisms. In addition, other FIB processing parameters, such as scan strategies, should be incorporated and the effect evaluated in future work to further reduce the local heating, as demonstrated by Schmied et al. in an earlier study . Schmied et al. showed that the dwell time plays a significant role in reducing local heating . The effect of the ion dose rate/energy converted to heat per time is, thus, an important parameter and should be carefully investigated in future experiments.

In addition, the effect of the ion incidence angle and associated changes in interaction volume shape as well as heat generation need to be investigated further. The ion beam usually hits the sample at glancing angles during TEM lamella preparation and cross-sectioning, which leads to anisotropic changes in the interaction volume. Whilst the heat generation here was minimized, with irreversible heat damage limited to 5 nm, elevated temperatures (<20 K difference) can be observed to about 10 nm beyond the ion impact point. This is a significant distance when compared to a prepared TEM lamella, which has a typical thickness of 50 to 100 nm. Whilst this temperature does not cause any damage to this sample, elevated temperatures beyond the impact point can potentially cause heat damage, especially when working with materials with a lower thermal conductivity or a lower melting point than collagen. Future experiments should carefully evaluate if changes to the ion incidence angle can lead to a further reduction in ion beam-induced heat damage. The effect of the different interaction volume sizes and shapes for various ion species should be studied to better understand where the proposed 1D model works and where a 3D model needs to be applied. Mutunga et al. showed that the 1D model overestimates the induced temperature in comparison to the 3D model . A careful evaluation needs to be carried out in future to assess if this is the case for FIBs as well. This is especially important considering the rise in different ion species, which are becoming more popular now, especially the lighter ion species where the assumption of a flat disc may not hold. Allen et al. showed that He ions can be beneficial when working with thin films as they mostly transmit through the sample. The ion beam-induced heat for various ion species in bulk as well as thin samples (where ions might transmit through the sample) should be systematically investigated and evaluated with 1D and 3D models. The proposed further studies will help develop faster and more robust approaches, especially with the increase in biological and thermally low conductive materials that are processed by FIBs today.

Preparation of porcine collagen (UAEC no.1700000190) followed standard methods, as previously described . Briefly, skin was degreased and digested in acetic acid by pepsin. Solubilized collagen was purified by NaCl precipitation at acidic and then neutral pH. Collagen sponges were prepared using purified type-I collagen, glutaraldehyde fixation, and freeze drying, also as previously described .

The program SRIM (Monte Carlo simulation) was used to determine the heat flux that 5 keV gallium ions (Ga+) induce in skin. 50000 ions were simulated for each incident ion energy. The methods “surface sputtering/monolayer collision steps” was selected as the calculation type. The plotting window size was selected in such a way that all ions trajectories in the simulation were completely followed. The calculation parameter output was evaluated to determine the energy losses of each ion for the different interactions. SRIM combines the values for heat and energy required for secondary electron creation in the ionization output value. Therefore, the exact contribution to heat in the inelastic losses (ionization) cannot be determined from the program directly. The presented value will overestimate the energy loss to heat when including the inelastic losses. The following values were used to describe collagen: specific heat cp = 3540 J·kg−1·K−1, density ρ = 1079 kg·m−3, thermal conductivity k = 0.29 W·m−1·K−1, and thermal diffusivity α = 1.36 × 10−7 m2·s−1. The SRIM determined heat flux is later used as input parameter for the COMSOL simulations and the numerical modelling of the heat equation.

The finite element simulator COMSOL was used to simulate the interaction of 5 keV Ga+ with collagen using the time-dependent differential equation of heat conduction for a stationary, homogeneous, and isotropic solid,

The time evolution of a single gallium ion’s energy deposition into collagen due to the ion’s electronic and nuclear stopping was studied using COMSOL “Particle Tracing Physics”. The dissipated ion energy is converted to thermal energy at each mesh element encountered along the ion’s path in the collagen. For each of these mesh elements the “Heat Transfer Physics’ is used to distribute the resultant thermal energy throughout the rest of the sample. The spatial and temporal evolution of the ions and their heat trail are stored and reassembled after the “Solve” is completed to generate various “Results” plots and images of the sample history.

A 20 nm collagen cube for the 5 keV incident gallium ion case study was chosen, which is based on the ions’ projected range in the collagen material and on a volume representative of the lamella that might ultimately be fabricated in a FIB. The geometry was constrained to room temperature on all sides except at the sample top, which was thermally insulated, as in a vacuum. As the ion passed through the sample, the energy gained by the sample due to the traversing ion’s electronic and nuclear stopping was accumulated. This deposited energy was then conducted away as heat. This heat was distributed to all mesh elements in the vicinity of the ion’s path based on the density, thermal conductivity, and specific heat of collagen.

Multiple cross-sections (15 µm × 10 µm × 200 nm) were cut into the non-resin embedded collagen sample using the FEI Quanta 200 3D at the Queensland University of Technology, Brisbane, Australia. All cross-sections were processed using 5 keV ions. One cross-section was cut with 1.4 nA, 50% overlap to assess if reducing the ion energy is sufficient to avoid heat damage when processing using nA beam currents. 1.4 nA was the highest workable ion beam current at 5 keV ion energy for the device. To further assess the effect of the beam current on sample heating, cross-sections were cut using 0.12 nA, 70 pA as well as 29 pA. For this set of experiments a 20% overlap and 50 nm blur were required to allow patterning with 29 pA and 70 pA to avoid an excessive amount of points which could not be computed by the Quanta 200 3D. The blur was chosen to be smaller than the width of the interaction volume, so that the latter remained the larger parameter and to allow comparison to the previous study. The effect of a blur/overlap combination was evaluated in a second set of experiments by cutting cross-sections using 0.12 nA, 70 pA as well as 29 pA with 20% overlap and 200 nm blur. The beam diameter for 5 keV and 1.4 nA, which is given as 246 nm, exceeds the 200 nm blur and an additional measurement is therefore not meaningful here. The blur was achieved by overfocusing the beam. All cross-sections were prepared using 1 µs dwell time.

SEM images of the prepared cross-sections were recorded with 5 kV acceleration voltage, 90 pA beam current, 1 µs dwell time, 128 frames integration filter and 1024 × 882 pixel resolution with an Everhart–Thornley detector.

To verify if the collagen was heat-damaged when processing the sample using nanoampere range currents and different incident ion energies, TEM lamellae for 5 keV ion energy was prepared using the Quanta 200 3D and compared to the previously reported results .

First, a protective platinum layer was deposited onto the desired area using 5 kV and 0.23 nA. The cross sections were cut using 5 kV acceleration voltage with 1.4 nA ion beam current, 50% beam overlap. The lifted-out lamella was pre-thinned at 50.5° (front side) and 53.5° (backside) using 5 kV, 0.6 nA, and 50% beam overlap. The final thinning was performed using 5 kV, 0.23 nA, and 50% overlap at 50.5° (front side) and 53.5° (backside). Polishing was not performed as the thinning was already performed with 5 kV acceleration voltage.

The FIB-prepared TEM lamella were analyzed using a Jeol 2100 TEM operating at an acceleration voltage of 200 kV.