Flash electropolishing for TEM: Reducing FIB‐induced defects in tungsten with protocols for new materials

Abstract

Focused ion beam (FIB) milling has become the dominant approach for site‐specific transmission electron microscopy (TEM) specimen preparation; however, FIB damage remains a critical limitation for reliable microstructural characterisation, particularly in radiation effects studies. Tungsten is especially susceptible to FIB damage due to its high nuclear stopping power, which promotes the formation and strong diffraction contrast of FIB‐induced ‘black spot’ defects that are indistinguishable from very fine irradiation‐induced loops/defects resulting from low to intermediate temperature neutron irradiation. In this work, flash electropolishing is systematically evaluated as a post‐FIB treatment for minimising preparation‐induced artefacts for TEM analysis of tungsten‐based alloys. Using a range of non‐, ion‐, and neutron‐irradiated tungsten materials, the effectiveness of flash electropolishing has been assessed through direct comparison with conventional FIB and plasma‐FIB preparation including low‐energy Ga, Ar, Xe ion cleaning. The results demonstrate that flash electropolishing effectively removes FIB‐damaged layers and ‘black spot’ defects, thereby enabling reliable observation of irradiation‐induced dislocation structures. Key processing parameters governing flash electropolishing quality – including lamella thickness, applied voltage, polishing duration, electrolyte chemistry, and cathode geometry – have been systematically evaluated, and clear criteria were established for determining when flash electropolishing is required to ensure reliable microstructural analysis. This work also provides practical guidance for implementing flash electropolishing as an artefact‐controlled specimen‐preparation approach for TEM characterisation of FIB‐produced specimens. The systematic protocol can be extended to other, non‐tungsten materials.

LAY DESCRIPTION: Flash electropolishing is demonstrated as an effective post‐FIB treatment for removing preparation‐induced damage in tungsten TEM specimens, including FIB‐induced ‘black spot’ defects that can obscure irradiation microstructures. Through systematic comparison with conventional and plasma‐FIB methods across irradiated and non‐irradiated tungsten alloys, key processing parameters and reliability criteria are established. The work provides a practical, artefact‐controlled preparation protocol for accurate radiation‐damage characterisation and offers a broadly adaptable strategy for other material systems.

1. INTRODUCTION

Transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM) are among the most powerful characterisation techniques for examining microstructures in materials, enabling direct observation and analysis of defects, interfaces, second phases, and chemical distributions with nanometre to atomic‐scale resolution. ¹ , ² , ³ As TEM continues to advance toward higher spatial resolution and increasingly sensitive diffraction‐ and spectroscopy‐based analyses, the quality of the acquired data has become ever more dependent on specimen preparation. In many cases, specimen preparation – not microscope performance – now represents the dominant limitation in achieving accurate and reproducible microstructural analyses. ⁴ , ⁵  In this work, we present a systematic investigation and provide detailed, practical guidance to enable independent implementation and reproducible application of flash electropolishing as an artefact‐controlled TEM specimen preparation method, focusing on tungsten (W) materials.

Focused ion beam (FIB) milling has become a ubiquitous TEM specimen preparation method due to its site specificity, minimal material consumption, and compatibility with a wide range of materials, including metals, ceramics, semiconductors, and complex heterostructures. ⁶ , ⁷ , ⁸ , ⁹ For neutron‐irradiated materials, FIB‐based site‐specific lamella preparation minimises specimen volume and radioactivity, allowing ‘hot’ samples to be examined in microscopy facilities with controlled low‐activation environments. ¹⁰ , ¹¹ These advantages have made FIB preparation the standard approach for TEM sample preparation in materials research for a wide range of applications. However, FIB milling is intrinsically an energetic ion–solid interaction process and therefore introduces preparation‐induced artefacts, known as FIB damage. ¹² , ¹³ These artefacts commonly include amorphisation, curtaining, implantation of source ions, residual strain, point defect clusters (‘black spot’ defects), dislocation loops, surface roughness, redeposition layers, and chemically modified near‐surface regions, and may also induce phase transformations. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ Notably, previous work by Prosa and Larson on FIB‐prepared atom probe tomography (APT) multilayer specimens reported Ga implantation and ion‐beam‐induced intermixing due to ballistic damage during specimen preparation. ¹⁸ The affected layer can range from a few nanometres to several tens of nanometres in thickness, depending on ion species, energy, incidence angle, material response, and the experience of the operator. Critically, FIB‐induced features can closely resemble microstructural characteristics of interest (e.g., irradiation‐induced defects, dislocation structures, precipitates, and other microstructures) making it difficult to distinguish specimen‐inherent microstructures from preparation‐induced artefacts. ¹⁹ , ²⁰ , ²¹

Post‐FIB cleaning approaches, including reduced‐ion‐energy final polishing, low‐energy ion milling (e.g., nanomilling or precision ion polishing systems, PIPS), or the use of plasma focused ion beam (PFIB) sources in place of Ga, can reduce FIB‐induced damage but often leave residual defects. ²² , ²³ , ²⁴ , ²⁵ , ²⁶ Recently, flash electropolishing (or flash polishing) has emerged as a powerful and complementary post‐FIB treatment that can potentially address FIB artefacts. ²⁶ , ²⁷ In flash polishing, electrochemical removal occurs for extremely short durations (typically milliseconds) under carefully controlled voltage, temperature, and electrolyte conditions. This approach enables the selective removal of the ion‐damaged surface layer from FIB‐prepared lamellae while preserving the underlying microstructure. Unlike conventional twin‐jet electropolishing of bulk TEM discs, ²⁸ , ²⁹ , ³⁰ flash‐polishing is specifically adapted to micrometre‐scale FIB lift‐out lamellae and, when properly optimised, effectively suppresses thickness variations and etching or pitting. Importantly, flash polishing is generally applicable to a wide range of conductive metals and alloys. By removing FIB‐induced damaged layers, this approach enables specimens to achieve TEM microstructural characterisation quality comparable to (or in some cases exceeding) that of classical electropolished foils, while retaining the site specificity and practicality of FIB lift‐out. It is noted, however, that FIB lift‐out specimens are relatively small, which significantly limits assessment of microstructural variability. A well‐prepared electropolished 3 mm TEM foil provides an electron‐transparent area equivalent to nearly 50–100 FIB specimens, meaning that many FIB lift‐outs would be required to achieve comparable statistical representativeness.

Early demonstrations of flash polishing on neutron‐irradiated tungsten lamellae established its feasibility and highlighted the dependence of polishing behaviour on crystallographic orientation, electrolyte chemistry, and support grid materials. ³¹  More recently, flash polishing has been systematically applied to a range of irradiated metallic alloys, conclusively demonstrating that FIB‐induced ‘black spot’ defects, implanted ions, redeposition of protective‐layer materials, and surface amorphisation can fully mask true microstructures and that these artefacts are removed only after flash polishing. ³² , ³³ , ³⁴ , ³⁵ Despite these advances, transferable guidelines for artefact‐controlled flash polishing, as well as clear criteria for determining when flash polishing is required across different material systems and irradiation conditions, remain limited. Here, we evaluate the effectiveness (removed thickness vs. polishing time) of flash polishing, assess its impact on the visibility of irradiation‐induced defects in tungsten (well‐known for its susceptibility to FIB‐induced ‘black spot’ defects) and discuss the critical processing parameters governing the success and stability of flash polishing. By establishing clear links between preparation conditions and the resulting microstructures, this study seeks to provide practical guidance for reliable TEM characterisation of microstructures in materials.

2. MATERIALS AND METHODS

2.1. Flash polishing equipment and procedures

Flash polishing was conducted using a temperature‐controlled electrolyte and a pulsed electrical bias. The flash polishing equipment at Oak Ridge National Laboratory (ORNL) was modified from a setup originally developed at the University of Tennessee. ²⁷ The equipment, setup, and general procedures are summarised in this section, with additional details of the W polishing recipe described in a later section. The flash polishing apparatus is shown in Figure 1a and consists of a 0–60 V DC power supply, a relay timer with a time range of 1 ms to 9.999 s, self‐closing gold‐coated tweezers (Ted Pella® PELCO® 5375‐G) used to hold TEM gold grids (Ted Pella® PELCO® 1GG75) with FIB lift‐out specimens, and a cathode basket (4.5 cm length × 3.5 cm width × 6 cm height) fabricated from 304 stainless‐steel mesh (aperture 4.5 mm; wire diameter 0.7 mm). Electrolyte temperature was routinely measured prior to polishing using a K‐type thermocouple (measurement range: −200°C to 1000°C) connected to a digital thermometer to ensure consistent flash polishing conditions.

FIGURE 1.

Flash polishing setup and procedure. (a) Schematic of the flash polishing apparatus. (b) Schematic of the flash polishing circuit. (c) Schematic illustrating immersion of a TEM gold grid containing a FIB lift‐out specimen into the electrolyte within the cathode basket during flash polishing. (d) Post‐polishing cleaning procedure for W and W alloys, in which the sample is sequentially transferred through cleaning solutions to remove residual electrolyte. (e) Air‐drying configuration, with the TEM grid held by tweezers and propped on the edge of a membrane box prior to TEM characterisation.

A schematic of the flash polishing circuit is shown in Figure 1b. The circuit was connected to the DC power supply with the voltage set between 18 and 20 V, and the relay timer was programmed to control polishing durations of 1–10 ms. The electrolyte solution was prepared in advance and, when required for other metallic alloys, cooled using an ethanol bath contained in a foam‐rubber‐insulated 1 L glass measuring cup with liquid nitrogen or dry ice to maintain a low and stable temperature during polishing. The W flash polishing experiments in this study were performed at room temperature, and an electrolyte consisting of 1 g NaOH dissolved in 100 mL deionised (DI) water was used. The TEM gold grid was held with gold‐coated tweezers and immersed in the electrolyte within the cathode basket (Figure 1c). For consistent polishing, the tweezers were oriented parallel to the straight edges of the grid (Figure 1b), and the sample was positioned near the centre of the cathode basket, equidistant from the surrounding mesh.

Upon triggering the timer, the electrical bias was applied for the prescribed duration, after which the sample was rapidly withdrawn from the electrolyte and immediately transferred to a series of cleaning solutions (Figure 1d). For the W polishing recipe, the cleaning sequence consisted of DI water, ethanol, acetone, and ethanol, with the sample immersed in each solution for 1 min to remove residual electrolyte. During cleaning, the tweezers were moved vertically, perpendicular to the solution surface, to enhance rinsing efficiency and minimise disturbances that could dislodge the grid. Finally, the grid was air‐dried for approximately 5 min by propping the tweezers on the edge of a membrane box (Figure 1e) prior to TEM characterisation.

2.2. Tungsten materials and flash polishing conditions

W is a critical structural and plasma‐facing material for irradiation environments such as nuclear fusion reactors; ³⁶ however, it is well known to be particularly susceptible to FIB damage, which can significantly affect the reliability of microstructural analysis in irradiated W, particularly for small dislocation loops and ‘black spot’ defects. In this work, a range of irradiated and non‐irradiated W materials (e.g., single crystal W with ⟨100⟩ and ⟨110⟩ surface orientations, polycrystalline pure W (>99.99 wt.%), polycrystalline K‐doped W (28 wppm K), ITER‐grade W, and W–3 wt.% Re alloys) were selected for flash polishing. To develop an optimised W flash‐polishing recipe, more than 40 FIB lift‐out specimens were processed under systematically varied conditions, including initial specimen thickness, polishing voltage, polishing duration, and stainless‐steel cathode basket mesh size. Flash polishing conditions that yielded acceptable or high‐quality specimens are summarised in Table 1. Unsuccessful polishing conditions are provided in Table 2 to support reproducibility and future development of flash polishing recipes for W and other material systems. This paper focuses exclusively on the flash polishing methodology; detailed microstructural characterisation of these materials is ongoing and will be reported in future publications. Further details on material processing routes and irradiation conditions are available elsewhere and can be provided upon request. ³⁷ , ³⁸ , ³⁹

TABLE 1.

Summary of flash polishing parameters applied to FIB lift‐outs from W materials mounted on Au grids using a room‐temperature electrolyte (1 wt.% NaOH in deionised water).

Material	Irradiation condition	Voltage (V)	Time (ms)	SEM‐measured initial target thickness (nm)	STEM–EELS‐estimated post‐polishing thickness (nm)
Single crystal W <100>	10.8 MeV W ion irradiated (17°C, 0.008 dpa)	18	5	400 ± 30	92 ± 18
Single crystal W <110>	10.8 MeV W ion irradiated (17°C, 0.008 dpa)	18	10	400 ± 30	69 ± 14
Polycrystal W	Non‐irradiated	18	1	200 ± 30	55 ± 11
Polycrystal W	Non‐irradiated	18	3	400 ± 30	>128
Polycrystal W	Non‐irradiated	18	5	400 ± 30	74 ± 15
Polycrystal W	Neutron irradiated (567°C, 0.33 dpa, 0.3 wt.% Re/Os)	18	5	400 ± 30	90 ± 18
Polycrystal W	Neutron irradiated (787°C, 0.73 dpa, 0.6 wt.% Re/Os)	18	5	400 ± 30	>109
Polycrystal W	Neutron irradiated (1131°C, 0.70 dpa, 0.6 wt.% Re/Os)	18	5	400 ± 30	84 ± 17
Polycrystal W	Neutron irradiated (850°C, 0.42 dpa, 5 wt.% Re/Os)	18	5	400 ± 30	73 ± 15
Polycrystal K‐doped W	Non‐irradiated	18	5	400 ± 30	84 ± 17
Polycrystal K‐doped W	Neutron irradiated (850°C, 0.42 dpa)	19	5	400 ± 30	>120
W–3 wt%Re	Non‐irradiated	20	5	400 ± 30	86 ± 17

Note: The target initial thickness (measured by SEM) and post‐polishing thickness (measure by STEM‐EELS) refer to the thickness of the thinned window within the FIB‐prepared lamella. Post‐polishing thickness >100 nm is qualitative and primarily indicate relatively thick regions rather than precise measurements.

TABLE 2.

Summary of unsuccessful flash polishing conditions that resulted in poor TEM sample quality or complete removal of thin regions of the TEM foils.

Material	Irradiation condition	Voltage (V)	Time (ms)	Cathode basket mesh size
Polycrystal W	Non‐irradiated	13, 16, 19, 20, 24	1–6, 50	D = 0.7 mm; A = 4.5 mm
Polycrystal W	Non‐irradiated	18	5	D = 0.4 mm; A = 1.0 mm
Polycrystal W	Irradiated	19	1–3, 20	D = 0.7 mm; A = 4.5 mm
Polycrystal W	Irradiated	18, 19	2, 5	D = 0.4 mm; A = 1.0 mm
Polycrystal K‐doped W	Non‐irradiated	18	5	D = 0.4 mm; A = 1.0 mm
W–3 wt%Re	Non‐irradiated	18, 22	5	D = 0.7 mm; A = 4.5 mm

D: wire diameter; A: aperture size.

2.3. FIB sample preparation and TEM characterisation

FIB specimens were prepared at the Low Activation Materials Development and Analysis (LAMDA) laboratory at Oak Ridge National Laboratory (ORNL) using a Thermo Fisher Scientific (TFS) Versa FIB/SEM operated at 30 keV with Ga ions for trenching and thinning. The target initial thicknesses (200 or 400 nm) prior to flash polishing were measured from SEM cross‐sectional images, with a variation of ±30 nm and an estimated SEM measurement uncertainty of 10%. No low‐voltage (2–5 keV) Ga‐ion cleaning was applied prior flash polishing unless otherwise noted. For samples subjected to low‐voltage Ga ion cleaning, the polishing was performed at 5 keV for 2 min per side, followed by 2 keV for 5 min per side. One flash‐polished specimen was further milled using 600 eV Ar ions at ±15° from glancing incidence for 15 min on each side in a Fischione NanoMill system to evaluate whether low‐energy Ar⁺ milling assists in cleaning the W specimen or introduces additional ion‐beam‐related damage. STEM characterisation of irradiation‐induced defects was performed using a JEOL JEM‐2100F TEM operated at 200 kV. TEM specimens were tilted to the nearest low‐index zone axis to acquire on‐zone STEM bright‐field (BF) images for examining diffraction contrast arising from FIB preparation and ion‐ or neutron‐irradiation‐induced defects. The post‐polishing thickness of the TEM foils was determined using the electron energy‐loss spectroscopy (EELS) log‐ratio method, ⁴⁰ which typically has an uncertainty of approximately ±20%. Using the Malis model, ⁴⁰ the inelastic mean free path (λ) for pure W was estimated to be 66.72 nm at a measured collection semi‐angle of 20.8 mrad. In this study, the relative thickness (t/λ) of the post‐electropolishing ranged from 0.82 to 1.94, corresponding to an absolute thickness of approximately 55–130 nm. It should be noted that for t/λ > 1–1.5, the log‐ratio method may underestimate the specimen thickness, with errors increasing due to the more significant contribution of plural scattering. ⁴¹ , ⁴² Under these conditions, a linear relationship between the inelastic mean free path and the ratio of the total intensity of the EELS spectrum to that of the zero‐loss peak may no longer hold. ⁴³ , ⁴⁴ More accurate thickness determination in the range t/λ > 1–1.5 would require independent calibration of the mean free path; ⁴² , ⁴³ , ⁴⁴ however, such calibration was not performed in this work. Therefore, for t/λ > 1.5 (i.e., >100 nm in this study), the estimated thickness should be regarded as qualitative, serving primarily to indicate that the samples are relatively thick rather than to provide precise values.

3. RESULTS

Figure 2a and b compares ⟨100⟩ on‐zone STEM bright‐field (STEM‐BF) images of non‐irradiated polycrystalline W specimens prepared by conventional FIB processing and by FIB extraction followed by flash polishing, respectively. The conventional FIB sample was extracted and thinned at 30 keV and subsequently cleaned at 5 and 2 keV using a Ga ion source. The flash‐polished specimen was extracted at 30 keV and then subjected to flash polishing. In the FIB‐only sample (Figure 2a), dislocation lines are visible; however, they are obscured by a diffuse, blurry surface layer, accompanied by numerous nanoscale ‘black spot’ features distributed throughout the foil near‐surface. In comparison, the flash‐polished sample (Figure 2b) exhibits a much cleaner microstructure, with no observable ‘black spot’ features and significantly sharper dislocation contrast.

FIGURE 2.

Comparison of on‐zone STEM‐BF images (Z = <100>) of non‐irradiated polycrystalline W samples with similar foil thicknesses (70–100 nm) prepared using different methods: (a) conventional FIB preparation with final Ga‐ion cleaning at 5 and 2 keV; (b) flash‐polished sample; (c) flash‐polished sample followed by final Ga‐ion cleaning at 5 and 2 keV; and (d) flash‐polished sample followed by 600 eV Ar‐ion cleaning.

To investigate the origin of the ‘black spot’ features, additional post‐polishing ion cleaning experiments were performed, including 5 and 2 keV Ga ion cleaning (Figure 2c) and 600 eV Ar ion cleaning (Figure 2d). The results show that low‐energy Ga ion cleaning reproduces ‘black spot’ features, although without the continuous blurry surface layer observed in the conventional FIB‐only condition. In addition, 600 eV Ar ion cleaning also induces ‘black spot’ features, but with a substantially lower density and much weaker dark contrast.

As shown in Figure 3, in addition to non‐irradiated polycrystalline W, a wide range of W materials, including ion‐ and neutron‐irradiated single‐ and polycrystalline W, W–3Re, and K‐doped W were successfully processed using the flash polishing approach. High‐quality on‐zone STEM‐BF images were consistently obtained by fine‐tuning the applied voltage (18–20 V) and polishing duration (1–10 ms). All flash polish experiments were conducted at room temperature using the same electrolyte composition (1 wt.% NaOH in deionised water). Overall, both irradiation condition and solute atom concentration influence the optimal polishing voltage or the material removal efficiency under otherwise identical polishing conditions (Tables 1 and 2).

FIGURE 3.

STEM‐BF images of flash‐polished W‐based materials: (a) single crystal W irradiated with 10.8 MeV W ions at 17°C to 0.008 dpa (Z = <100>); (b) polycrystal W neutron‐irradiated at 567°C to 0.33 dpa (Z = <110>); (c) polycrystal W neutron‐irradiated at 850°C to 0.42 dpa (Z = <111>); (d) non‐irradiated W–3Re (Z = <110>); (e) non‐irradiated polycrystal K‐doped W (Z = <111>); and (f) polycrystal K‐doped W neutron‐irradiated at 850°C to 0.42 dpa (Z = <110>).

By measuring the TEM foil thickness before flash polishing (using SEM) and after flash polishing (using the STEM–EELS log‐ratio method), the thickness removed as a function of flash polishing time for polycrystalline W was determined, as shown in Figure 4a. With the polishing voltage fixed at 18 V, increasing the polishing time did not result in a linear increase in removed thickness. Specifically, the removed thickness increased from approximately 150 to 325 nm as the polishing time increased from 1 to 5 ms, and then reached a saturation regime between 5 and 8 ms. For neutron‐irradiated W and K‐doped W samples irradiated over a temperature range of 567°C–850°C (Figure 4b), increasing irradiation dose exhibited a trend of slightly reduced thickness removal under applied voltages of 18–20 V and a fixed polishing time of 5 ms. Under neutron irradiation (especially thermal neutron dominant spectra), W is well known to undergo nuclear  transmutation, producing Re and Os. ³⁸ , ³⁹ , ⁴⁵ The increased concentrations of Re and Os appear to also influence the flash polishing results, either by requiring higher applied voltages to initiate effective polishing or by reducing the overall material removal rate, as shown in Table 1.

FIGURE 4.

TEM foil thickness removed by flash polishing as a function of (a) polishing time and (b) neutron dose (Tirr: irradiated temperature). Flash polishing was performed at 18–20 V using a room‐temperature electrolyte (1 wt.% NaOH in deionised water).

4. DISCUSSION

4.1. FIB‐induced surface amorphisation and ‘black spot’ defects

The formation of ‘black spot’ features (or ‘black spot’ defects) after flash polishing followed by low‐energy Ga (2–5 keV) and 600 eV Ar ion cleaning – particularly after 600 eV Ar nanomilling – is noteworthy (Figure 2c and d). Low‐energy ion cleaning is generally considered to improve sample quality or, at least, only slightly modify the microstructure. These results demonstrate that, regardless of the FIB operator's skill (including the use of protective layers, optimised FIB voltages and currents, and appropriate sample tilting to minimise ion damage) ‘black spot’ defects can still form in W even under 600 eV Ar cleaning conditions. This finding is critical for studies of irradiation‐induced defects in W, as FIB‐induced black dot defects (Figure 2) are indistinguishable from those produced at low to intermediate temperatures by high‐energy ion irradiation (excluding FIB) and neutron irradiation (Figure 3a and b).

Previous studies have shown that FIB damage commonly includes irradiation‐induced ‘black spot’ defects, amorphous surface layers (often more pronounced in ceramics and intermetallics), and compositional artefacts associated with implantation and redeposition. ¹² The formation of amorphous layers and ‘black spot’ defects during FIB primarily arises from the combined effects of displacement damage, sputtering, and localised ion energy deposition. ²² Incoming Ga ions generate dense near‐surface displacement cascades dominated by nuclear stopping (Figure 5a), producing high concentrations of point defects (vacancies and interstitials) and small point defect clusters within only a few nanometres of the surface (∼10 nm for 30 keV W, as simulated by SRIM ⁴⁶ , ⁴⁷ in Figure 5a), which are likely responsible for the observed ‘black spot’ contrast. Simultaneously, continuous sputtering removes partially recovered lattice regions and exposes newly damaged material, resulting in a near‐surface layer with persistently high defect density.

FIGURE 5.

Calculation of stopping power and formation of point defect clusters: (a) SRIM‐calculated electronic and nuclear stopping powers of W self‐ions in a W target, with a schematic illustration of ion–target interactions (PKA: primary knock‐on atom, VAC: vacancy, SIA: self‐interstitial cluster); (b) ratio of nuclear to electronic stopping power as a function of ion energy; (c, d) molecular dynamics (MD) simulations ⁴⁸ of displacement cascades for (c) 40 eV Fe ions in Fe and (d) 80 eV W ions in W. The displacement energies of Fe and W are assumed to be 40 and 80 eV, respectively.

Although low‐energy ion milling (e.g., 600 eV Ar) is widely regarded as an effective cleaning method for materials such as Si in semiconductor applications, ⁴⁹ the forementioned coupled processes still lead to the formation of ‘black spot’ defects in W. Also, as shown in Figure 2, after flash polishing followed by low‐energy Ga or Ar ion cleaning, a distinct ‘damaged’ surface layer is no longer evident; however, the presence of small ‘black spot’ defects remains clear. The strong ‘black spot’ contrast observed in W is likely attributable to several factors: (1) the relatively high nuclear‐to‐electronic stopping power ratio of W at FIB‐relevant energies (compared with Fe and Si, Figure 5b), which favours displacement damage and point‐defect clustering; ⁵⁰ (2) the formation of fewer but larger defect clusters in W (Figure 5d), as suggested by molecular dynamics simulations, ⁴⁸ in contrast to the higher density of smaller point defect clusters in Fe (Figure 5c) that may fall below TEM contrast limits; and (3) the shorter electron mean free path in high‐Z W materials, which enhances electron scattering and diffraction contrast, thereby amplifying the visibility of ‘black spot’ defects in W. ⁵¹

Additionally, ion energy and ion species strongly influence the implanted ion concentration and displacement damage depth profiles, and thus the resulting FIB damage layer. ⁸ As shown in Figure 6, SRIM simulations provide ion depth profiles and corresponding displacement levels for Ga, Xe, and Ar ions implanted into a pure W target at various energies. The SRIM simulations were performed in quick calculation mode using 10,000 ions, a total fluence of 1 × 10¹⁶ cm⁻², and a displacement energy of 90 eV for W. ⁴⁷ In general, higher ion energies produce thicker FIB‐induced damage layers, whereas lower‐energy ions tend to yield a higher concentration of implanted ions near the surface. It should also be noted that typical FIB milling angles for lamella fabrication range from 2° to 15°, with angles up to 25° used in certain steps. ⁵² Therefore, the actual penetration depth is often smaller than the simulated values obtained at normal incidence (90°), which represent the maximum depth. The black spots in Figure 2c appear significantly denser than those in Figure 2d, likely reflecting differences in ion–matter interactions between 2 and 5 keV Ga ions and 600 eV Ar ions. Under comparable ion flux, 2–5 keV Ga ions generate peak displacement damage at greater depths, resulting in a thicker damage layer. In contrast, 600 eV Ar ions induce damage that is more localised near the surface (<3 nm). Due to surface effects, many of the displacement‐induced defects in the Ar case can diffuse to the surface and recombine via enhanced interstitial–vacancy annihilation, leading to fewer point defect clusters. Consequently, fewer defects are formed and remain stable at the surface, leading to reduced contrast (i.e., fewer black spots) in STEM images compared to the 2–5 keV Ga ion case.

FIGURE 6.

SRIM simulation of depth profiles of (a) ion concentration and (b) displacement damage for Ga, Xe, and Ar ions at a fixed fluence of 1 × 10¹⁶ cm⁻² and various ion energies, normal to a W target (90° incidence and E_d = 90 eV).

4.2. Determination of optimised flash polishing parameters

Selecting an appropriate intermediate polishing plateau voltage for flash polishing is analogous to twin‐jet electropolishing of 3 mm TEM discs and surface electropolishing of bulk W materials, where insufficient voltage typically leads to etching whereas excessive voltage promotes pitting. Jacquet first employed the current–voltage (I–V) curve to analyse the anodic dissolution behaviour during the electropolishing process. ⁵³  Figure 8 illustrates a typical I–V curve for electropolishing, which can be divided into four characteristic regions: etching, passivation, limiting current plateau, and pitting. In the low‐voltage region, direct anodic dissolution dominates, resulting in surface etching. In the passivation region, the current density slightly decreases with increasing voltage due to the formation of a viscous electrolyte layer (or passive oxide film) on the anodic surface. In the limiting current plateau region, the current remains nearly constant despite increasing voltage, and this region is generally considered the optimal voltage range for achieving uniform electropolishing. At higher voltages, the passive oxide layer breaks down, and anodic dissolution is accompanied by gas evolution (typically oxygen). ⁵⁴ Gas bubbles trapped on the material surface led to localised dissolution, resulting in pitting; therefore, this regime is referred to as the pitting region. It should be noted that a later study by DuBose and Stiegler reported that, under short electropolishing durations, etching and pitting effects can be largely suppressed, and a well‐defined limiting current plateau was not observed. ³⁰  Instead, the polishing current increased monotonically with increasing applied voltage. Nevertheless, they emphasised that an intermediate voltage is still required to achieve uniform electropolishing of TEM foils, as both insufficient and excessive voltages lead to non‐uniform polishing.

FIGURE 8.

Representative features and potential artefacts observed in unsuccessfully flash‐polished samples. (a) Non‐uniform etching caused by insufficient voltage (13 V). (b) Selected area electron diffraction (SAED) patterns acquired from the yellow‐highlighted regions in (a) and (c). (c) Pitting and surface oxidation (white arrows in c) induced by excessive voltage (24 V). (d) Surface morphology changes resulting from reduced mesh size (aperture 1 mm; wire diameter 0.4 mm) at 20 V. (e) Residual surface debris due to insufficient cleaning time. (f) Preferential polishing originating from locally thinner regions. (g) Void‐like features (white arrows in g) associated with detachment of Y–Ti–O particles in FeCrAl ODS alloys, confirmed by (h) STEM‐BF and STEM‐EDX spectrum. ⁴⁶ For comparison, TEM specimens of CuCrZr alloys prepared by different methods are shown: (i) 2 keV Ga‐ion FIB, (j) 1 keV Xe‐ion PFIB, and (k) flash polishing. Dislocation‐like features (white arrows in i and j) are evident in the FIB‐ and PFIB‐prepared samples; (l) Cr clusters in CuCrZr, confirmed by on‐zone STEM‐BF and STEM–EDX spectrum. ⁵⁵

However, flash‐polishing voltages cannot be directly transferred from conventional twin‐jet electropolishing recipes. The applied voltage (or corresponding current density) is the most sensitive electrochemical variable, and uniform, clean polishing occurs only within a narrow intermediate voltage window corresponding to the limiting current plateau. Importantly, the optimal voltage strongly depends on material composition. In addition, the presence of irradiation‐induced defects, which are well known to affect the electrical resistivity of materials, appears to further increase the voltage or polishing time required, necessitating material‐specific optimisation (Figure 4). It should also be noted that cathode materials and geometry (e.g., mesh size and cathode basket design) can influence the effective current density and, consequently, the polishing behaviour. As a result, a flash polishing recipe developed on one material or electropolishing system cannot always be directly transferred to other systems.

The starting lamella thickness is another critical parameter. FIB‐induced damage commonly extends tens of nanometres beneath each foil surface; therefore, to avoid inadvertently introducing damage into the central region of the foil, the lamella must be sufficiently thick to allow effective removal of the damaged layers. This consideration motivated the selection of a relatively large initial thickness (∼400 nm) for the W flash polishing recipe in the present work. In addition, the quality of the FIB lift‐out foil – such as the uniformity of thinning within the window region and the absence of strong thickness gradients or pre‐existing holes – also affects the flash polishing outcome. We also observed that when an additional round of flash polishing is required, the FIB‐prepared sample surface must first be cleaned using 2–5 keV Ga ions to achieve improved polishing performance. This behaviour is likely associated with the formation of a passive surface layer during the initial flash polishing step. The development of such a passive layer is also believed to be responsible for the saturation in removed thickness as a function of polishing time, as shown in Figure 4a.

Other parameters, including electrolyte composition and polishing temperature, should be fixed at the early stage of recipe development, typically based on successful twin‐jet electropolishing conditions and subsequently refined if necessary. The choice of grid or mounting materials further influences polishing behaviour; inert Au‐coated grids are preferred to minimise galvanic electrochemical reactions and ensure stable electrical contact during polishing and rinsing. In addition, the choice of rinses is totally dependent on the specific electrolyte used for each material. For steels, one would never rinse in H₂O, but only in either absolute alcohol or methanol to avoid any water that will promote oxidation. Overall, optimal flash electropolishing is achieved through careful control of lamella thickness, material‐dependent voltage selection, electrolyte chemistry, temperature, and polishing timing, thus enabling uniform dissolution within the limiting current plateau.

4.3. Potential artefacts and challenges of flash polishing

Unsuccessfully flash‐polished W samples may contain flash‐polishing‐induced artefacts and should not be interpreted as intrinsic microstructural features of the material. As summarised in Figure 8, observed flash‐polishing artefacts include: (1) non‐uniform etching caused by insufficient voltage; (2) pitting and surface contamination induced by excessive voltage; (3) surface morphology changes resulting from reduced mesh size; (4) residual surface debris due to insufficient cleaning time; (5) preferential polishing originating from locally thinner regions; and (6) void‐like features associated with detachment of fine particles.

Assuming that flash polishing follows the typical electropolishing I–V curve, non‐uniform etching was observed when an insufficient voltage of 13 V was applied. As shown in Figure 8a, the etching effect produced a crater‐like surface morphology, resulting in significant foil thickness nonuniformity. Although dislocation contrast could still be observed, the pronounced surface morphology variations substantially degraded TEM sample quality and limited reliable microstructural data. In contrast, when an excessive voltage of 24 V was applied, pitting‐related artefacts dominated the foil surface (Figure 8c). The TEM lamella exhibited anomalous light–dark oval features, and bright contamination contrast was frequently observed near grain boundaries close to the surface. These features are likely associated with chemical reactions (e.g., pitting corrosion) between the electrolyte and the FIB‐prepared foil, potentially involving the Pt protective layer, implanted ions, or the W matrix itself. Similar to the low‐voltage condition, although dislocation contrast remained visible, detailed microstructural features were obscured by flash‐polishing‐induced surface artefacts. It is worth noting that the selected area electron diffraction (SAED) patterns showed differences between the two polishing conditions (Figure 8b). Samples polished at insufficient voltage show no additional diffraction spots beyond those of the W matrix, whereas the excessive‐voltage case exhibits extra reflections originating from the white‐contrast features in Figure 8c. With a selected area diffraction aperture covering a ∼500 nm diameter region, the additional reflections are consistent with the most common monoclinic WO₃ structure (space group P2₁/n). During electropolishing of W in aqueous NaOH, Han and Fang reported that hydrogen evolution occurs at the cathode (6H₂O + 6e⁻ → 3H₂ + 6OH⁻) while anodic oxidation forms WO₃ (W + 6OH⁻ → WO₃ + 3H₂O + 6e⁻). ⁵⁶ However, amorphous rings are also present in Figure 8b (bottom image), and potential grain overlap may occur, which complicates the identification of the white‐contrast features as WO₃.

In an effort to further refine the flash‐polishing method and reduce the time required for the post‐polishing rinsing step, flash polishing was performed using a smaller cathode basket mesh size (aperture 1 mm; wire diameter 0.4 mm), while the rinsing time was reduced by half. Contrary to expectations, the reduced mesh size did not improve polishing uniformity; instead, pronounced surface morphology changes were observed, characterised by canyon‐like height gradients across the TEM foil surface (Figure 8d). These height‐gradient features obscured dislocation contrast and significantly hindered reliable microstructural analysis. It is likely that changing the cathode basket mesh aperture size (and thus the mesh density) altered the effective current density during flash polishing (Figure 7), suggesting that different combinations of applied voltage and polishing time may be required when using different mesh sizes. For the post‐polishing rinsing procedure, in which the sample is sequentially transferred through multiple cleaning solutions to remove residual electrolyte, reducing the immersion time from 1 min to 30 sec per solution resulted in the accumulation of debris particles on the foil surface (Figure 8e), indicating insufficient removal of polishing residues. In addition, preferential thinning during flash polishing was observed at locally thinner regions of the FIB lift‐out sample, which in some cases resulted in the formation of holes near the Pt protective layer on the W surface (Figure 8f). This behaviour is particularly unfavourable for ion‐irradiated samples, as the most critical microstructural information is typically located in the near‐surface region adjacent to the Pt protection layer.

FIGURE 7.

Schematic of a typical current–voltage (I–V) response during electropolishing, illustrating the relationship between the current density and the applied voltage between the specimen and the cathode and the corresponding electrochemical reaction regions.

Another potential flash polishing artefact observed in FeCrAl ODS alloys is the presence of void‐like features(Figure 8g), confirmed by Fresnel‐contrast method, ⁵⁷ associated with the detachment of Y–Ti–O particles (Figure 8h) in very thin regions after flash polishing. ⁵⁵ These void‐like features closely resemble irradiation‐induced cavities (e.g., voids or bubbles) typically observed in materials irradiated at relatively high temperatures or in the presence of gas atoms. Although this artefact was identified in FeCrAl ODS alloys, similar behaviour is expected in other material systems containing second‐phase particles, as the electrolyte primarily interacts with the matrix material rather than with embedded particles or precipitates. Consequently, in materials with larger particles, artefactual void‐like structures resulting from particle detachment may occur even in relatively thicker foils. However, from another perspective, flash polishing can be advantageous for analysing precipitates that remain stable during the flash polishing process. For example, in CuCrZr alloys containing fine Cr‐enriched precipitates, FIB foils subjected to 2 keV Ga FIB (Figure 8i) and 1 keV Xe PFIB (Figure 8j) cleaning exhibit dislocation‐like features, ⁵⁸ and the precipitate contrast were strongly suppressed and obscured by FIB‐induced background contrasts (all STEM‐BF on‐zone images). After flash polishing (Figure 8k), the strain‐field contrast in STEM‐BF associated with the Cr‐enriched precipitates becomes significantly clearer (Figure 8l), likely because flash polishing preferentially removes the Cu‐based matrix while leaving the Cr‐rich precipitates largely intact. It is worth noting that the dislocation density in the CuCrZr alloy is expected to be very low. The origin of the dislocation‐like features remains unclear, whether they are directly related to FIB‐induced damage or foil bending during FIB thinning, but they are clearly absent in the flash‐polished specimen. Additionally, as discussed in Section 4.2, FIB‐induced damage depends on multiple factors, including sputtering, surface effects, ion species and energy, target material, and beam quality. Since Figure 8i and j uses different cleaning conditions (2 keV Ga vs. 1 keV Xe), it does not support a direct comparison or the conclusion that Ga ions (Ga FIB) produce less damage than Xe ions (PFIB).

As a direction for future work, performing cross‐sectional TEM analysis on samples before and after flash electropolishing (using a greater initial thickness or shorter polishing times to retain sufficiently thick regions for cross‐sectional preparation) would provide valuable insight. This approach would enable direct characterisation of surface and damage‐layer features, as well as more accurate determination of their thickness, including amorphisation, dislocation‐like features, oxides, and other electropolishing‐related effects. Additionally, compositional artefacts arising from differential implantation of source ions may influence FIB‐induced defect evolution and could help clarify the origin of the observed dislocation‐like features.

4.4. When flash polishing is necessary for reliable microstructural analysis

Although flash polishing requires well‐controlled FIB preparation and demands considerable time and financial investment, especially during development of a new polishing recipe, it becomes necessary when FIB/PFIB preparation (or any ion‐beam‐based thinning or cleaning step) introduces artefacts that can mimic, mask, or bias the true microstructure, particularly in radiation effects studies. Two representative cases are discussed below.

4.4.1. Quantitative analysis of fine irradiation‐induced defects

For irradiated materials, particularly those exposed at low homologous temperatures or low doses, the dominant microstructural features often consist of extremely small defect clusters (dislocation loops or ‘black spot’ defects) with characteristic sizes of only a few nanometres. In such cases, FIB‐induced damage can produce defect structures of comparable size and contrast, which may substantially mask the true irradiation microstructure. Without removal of the FIB‐damaged surface layers, it becomes difficult to distinguish preparation artefacts from irradiation‐induced defects, leading to significant uncertainty in defect identification, size measurement, and number–density quantification. Under these conditions, flash polishing is strongly recommended to ensure reliable and defensible quantitative analysis.

4.4.2. When ion‐beam artefacts extend beyond ‘black spot’ defects

In addition to ‘black spot’ defects, FIB‐prepared lamellae may exhibit a range of other artefacts, including amorphous layers, phase transformations, and ion‐beam‐induced precipitates. These artefacts can significantly complicate interpretation of diffraction contrast, defect morphology, and phase identification, particularly in neutron‐ or ion‐irradiated materials where subtle microstructural features are often of primary interest. In such cases, flash polishing provides an effective means to remove these ion‐beam‐induced artefacts and restore a microstructure more representative of the bulk material.

4.5. Recommended best practices for implementing flash polishing on a new material system

Implementing flash polishing on a new material system requires a systematic and controlled workflow to ensure reproducible flash polishing results while efficiently minimising FIB‐induced artefacts. The recommended best practices for applying flash polishing to a new material system are summarised in Figure 9. First, the electrolyte composition and concentration should be selected based on previously successful twin‐jet electropolishing conditions for the same or closely related materials. The flash polishing tools, including the cathode basket, tweezers, and grid materials should then be maintained constantly to ensure consistent electrochemical conditions. In the present work, the cathode basket was fabricated from 304 stainless steel, while gold tweezers and gold grids were used to minimise galvanic electrochemical reactions and contamination. Although not yet experimentally evaluated, platinum mesh is expected to be a promising substitute for 304 stainless steels because of its superior electrical conductivity and greater electrochemical stability. The cathode basket dimensions (4.5 cm × 3.5 cm × 6 cm) and mesh geometry (4.5 mm aperture; 0.7 mm wire diameter) also influence current distribution and polishing uniformity; therefore, these parameters should be fixed prior to subsequent optimisation steps.

FIGURE 9.

Workflow outlining key steps for establishing a flash polishing protocol for a new material system.

A polishing strategy must be decided based on the initial FIB lamella thickness, as this strongly influences TEM foil uniformity and required flash‐polishing time. Our experience indicates that, under otherwise constant conditions, longer polishing times tend to promote preferential removal in the thinnest regions. Consequently, starting from a relatively thick window (∼400–500 nm) often requires extended flash polishing, which enhances preferential thinning at locally thin regions and can lead to hole formation near the Pt protective layer (Figure 8f). Because this region is often thinner after FIB processing, it is especially susceptible to over‐flash‐polishing. This is particularly problematic for ion‐irradiated samples or studies focused on the surface region, where the most critical microstructural information typically resides beneath the Pt layer. One mitigation method is to tilt the lamella during FIB thinning and preferentially thin the bottom portion of the foil (away from the Pt layer), ensuring that subsequent flash‐polishing removal occurs farther from the surface of interest. Beginning with a thicker lamella can also help electropolishing remove more of the FIB‐damaged layer, which is advantageous when thinning to 100–200 nm is difficult without introducing central‐foil damage, especially for less experienced FIB operators. Thickness gradients in the lamella may additionally benefit studies targeting microstructures across multiple length scales. Alternatively, starting with a thinner window (∼100–200 nm) reduces flash polishing time and improves analysis‐area uniformity but requires more extensive FIB thinning and therefore increases the risk of introducing FIB‐induced damage into the central region. Moreover, if the lamella is uniformly thin rather than wedge‐shaped, excessive flash polishing may remove the entire electron‐transparent region.

Flash‐polishing parameters (including voltage, temperature, and polishing duration) must be iteratively optimised, noting that effective flash polishing often requires higher voltage and lower temperature than conventional twin‐jet electropolishing. To minimise the formation of surface oxides that hinder material removal, flash polishing should be performed as soon as possible after FIB thinning, or the specimen should be temporarily stored in a vacuum desiccator prior to processing. If re‐flash‐polishing is required, the specimen surface should first be cleaned by low‐voltage ion milling to remove contamination or passive oxide layers. During early recipe development, individual lift‐outs may be subjected to multiple re‐flash‐polishing cycles. Because polishing time must be limited to avoid removing the entire foil, the FIB‐damaged layer may not always be fully eliminated. This observation also indicates that, once an optimised recipe is established, a single flash‐polishing step is preferable to repeated cycles for minimising residual FIB damage. Together, these considerations provide a practical framework for establishing a reliable flash‐polishing protocol for new material systems.

5. CONCLUSIONS

This study clarifies the role of flash polishing in improving the reliability of TEM specimen preparation for W materials. Owing to tungsten's strong sensitivity to ion‐beam interactions, preparation‐induced artefacts can readily obscure or distort intrinsic microstructural features. Through systematic evaluation of flash polishing behaviour and outcomes, this work establishes practical guidelines and mechanistic insight for applying flash polishing in a controlled and reproducible manner. The principal conclusions are summarised as follows:

1. When properly optimised, flash polishing effectively removes FIB‐damaged layers and produces TEM foils with substantially improved microstructure images. Compared with FIB‐only or ion‐cleaned specimens, flash‐polished lamellae exhibit substantial suppression of ‘black spot’ defects, elimination of amorphous/damaged surface layers, and significantly enhanced dislocation contrast, enabling reliable observation of microstructures in tungsten.

2. Successful flash polishing requires careful control of several interdependent parameters, including starting lamella thickness, polishing voltage, polishing duration, electrolyte composition, temperature, and cathode geometry. In particular, operation within a narrow intermediate voltage window corresponding to the limiting current plateau is essential for achieving uniform dissolution while avoiding etching or pitting. The results further demonstrate that alloying elements and irradiation‐induced crystalline defects can reduce material removal efficiency and may require higher applied polishing voltages.

3. Flash polishing is strongly recommended when quantitative analysis of nanoscale irradiation‐induced defects is required or when ion‐beam artefacts extend beyond ‘black spot’ defects. The results show that conventional low‐energy Ga or Ar ion cleaning, even at Ar ion energies as low as 600 eV, is insufficient to fully eliminate these FIB artefacts in tungsten, underscoring the necessity of flash polishing for reliable microstructural interpretation.

Lin, Y.‐R. , Zhong, W. , Calzada, S. E. , Lach, T. G. , Nuckols, L. J. , Parish, C. M. , Zinkle, S. J. , & Burke, M. G. (2026). Flash electropolishing for TEM: Reducing FIB‐induced defects in tungsten with protocols for new materials. Journal of Microscopy, 303, 154–170. 10.1111/jmi.70122