Unveiling surfaces for advanced materials characterisation with large-area electrochemical jet machining

Surface preparation for advanced materials inspection methods like electron backscatter diffraction (EBSD) generally involve laborious and destructive material sectioning and sequential polishing steps, as EBSD is sensitive to both sample topography and microstrain within the near-surface. While new methodologies, like focussed ion beam and femtosecond laser milling are capable of removing material in a layer-by-layer manner to enable the construction of tomographic datasets within the electron microscope, such techniques incur high initial capital cost for slow removal and reconstruction rates. In this study, ambient condition electrochemical slot jets are applied to rapidly etch (e.g. 31 s) large surface areas (e.g. 160 mm 2 ) at controlled depths (e.g. 20 µm) with no in-process monitoring. Unveiled surfaces are conducive to measurement by EBSD (raw index rates between 75-95%), despite topographic anisotropy arising both from the process and the material. The mechanisms of topography formation during dissolution under the slot jet are analysed and understood. It is proposed that this slot jet method can be applied to create measurement surfaces for analysis with optical-based microstructural measurement routines reliant on topography and directional reflectance, at a significantly lower cost and time intervention than electron beam-based analysis methods.


Introduction
In high-value manufacturing, there is often a requirement to understand in detail the microstructure and the metallurgy of substrates, specifically at the near-surface, which dictates the safe loading conditions and service lifetimes [1,2]. This is increasingly important considering the drive towards more aggressive operating conditions to maximise component performance in critical applications like turbomachinery for aerospace and land-based power generation [3]. Currently, high-level microstructural analysis is limited by the intervention time and cost penalty, and material wastage. Reducing these factors is necessary to translate highquality analysis methods from the laboratory to the factory floor, where increases to processing times are unacceptable to industrial operators [4].
Vacuum-based characterisation methods, like electron backscatter diffraction (EBSD), rely on surface preparation steps in order to obtain adequate measurement response. Conventionally, this is achieved through laborious sequential grinding and polishing steps, which are not sitespecific and cannot totally eliminate microstrain from the near-surface electron interaction volumes. In practice, microstrain and roughness are often removed by subsequent chemical or electropolishing treatments, enhancing diffraction-based imaging in both EBSD [5], and transmission electron microscopy [6], however it remains challenging to control material removal over a surface. Therefore, methods to prepare surfaces in a more controlled, faster, and less wasteful manner are sought. While sample preparation can in some cases be performed within the electron microscope, for example using Ga + and Xe + focussed ion beam (FIB) milling [7] and femtosecond laser (FSL) ablation [8], widespread application of these techniques is hampered by high capital and maintenance costs and expert user requirement.
When combined with techniques such as EBSD, FIB and FSL preparation methods are ideally suited to singular analysis and can generate intimate 3D tomographic datasets of complex materials [9], however they are largely incompatible with high throughput and low intervention scenarios.
To this end, ambient condition surface characterisation methods, especially those relying on optical [10] and laser ultrasonic inspection [11], are receiving increasing attention due to their potential to increase measurement throughput and as such, be applied throughout multiple areas of manufacturing. As a potential alternative, electrochemical surface preparation enables a high degree of control as material removal is directly proportional to charge transfer, which is readily modulated [12]. Electrochemical preparation is carried out under ambient conditions, where the application of high currents can increase material volumetric removal rates (e.g. 10 6 -10 8 µm 3 /s) beyond both FIB and FSL. In addition, electrochemical methods excel at rapidly removing small volumes and the atom-by-atom removal mechanism is understood to impart negligible near-surface stresses upon processing [13,14], for example in comparison to polishing with diamond slurry [15]. These factors make such methods ideal for preparing metallic specimens for further inspection. Surface preparation techniques, such as electrochemical jet machining (EJM), are showing promise towards enabling rapid, highthroughput, and location-specific measurement of microstructural texture [16] and functionalised surface preparation [17].
In the present study, we apply slot jet EJM (Figure 1a-b) to etch large surface areas (e.g. >100 mm 2 ) for rapid preparation (≈30 s) and analysis. Here, EBSD inspection was chosen as i) it returns a wealth of microstructural information, and ii) EBSD measurement quality is sensitive to material topography and near-surface microstrain, which challenges sample preparation [5]. Aspects of surface processing using slot jet apparatus have been explored, notably with regards to large-area electropolishing [18] and electrochemical turning [19]. However, as a tool to control microscale layer removal and enable the capture of microstructural aspects of the surface, the translational slot jet method is unexplored . Here, we uncover, discuss and address some of the challenges that affect the selective translational etching procedure, thus our present findings aim towards enabling the wider proliferation of the technique as part of an analytical tool.

Materials
Commercially pure rolled Ni (Grade 201, Unicorn Metals, UK) was used as the substrate in this investigation (ρ = 8.91 mg/mm 3 ). Ni was selected as the substrate material to aid conceptualisation, as it generally returns strong Kikuchi patterns in EBSD. Samples of asreceived Ni were annealed (1050 °C, 3 hr) to coarsen the grain size and modify the overall grain boundary character to appraise the effect this has on surface generation during etching.
The mean grain sizes before (13.2 µm) and after annealing (235.2 µm) were approximated using the linear intercept method from optical micrographs of etched surfaces taken from >200 discrete grains (Figure 2). of the apparent high density of low energy coherent twin boundaries (CTBs), examples of which are evident in the optical micrograph (Figure 2, inset). A high twinning density is expected in Ni after annealing, which has a relatively low stacking fault energy. Samples of Ni were sequentially ground and polished (to 1 µm diamond), before rinsing with deionised water prior to EJM. The electrolyte was made up with NaCl (Fisher, UK) and deionised water to the required concentration (2.3 M), to form an electrolyte of sufficient electrolytic conductivity to transfer the required charge [16].

Electrochemical jet machining approaches
Anodic dissolution is governed by Faraday's laws of electrolysis, which allow relative removal rate predictions, achieved by considering the electrochemical equivalent, eq, for a given material, the mass of a given material liberated by one coulomb of charge: where Ma is the atomic mass, F is the Faraday constant (the product of elementary charge and Avogadro's constant), and z the charge number of the reaction. Consideration of material density, ρ, enables prediction of volumetric removal, and etch depth, h: where Q is charge passed, and A is the processed area. Practically, Q is adjusted by control over current and feed rate, νf. In this study, EJM was performed on a custom-built 3-axis setup  Alternate EJM approaches were applied to characterise the resulting surface textures in the context of microstructural measurements using electron backscatter diffraction (EBSD).
Firstly, a single pass approach was applied to process an area (70 mm 2 ) at equivalent areal charge density (0.586 C/mm 2 ) to etch Layer 4 (Table 1), to etch a similar depth (≈20 µm), without sequential processing. This was achieved by applying identical current (2000 mA), but reducing feed rate (0.34 mm/s). Again, the slot jet was translated perpendicular to the nozzle.
Total etching time for this sample was 20.5 s, or 0.29 s/mm 2 .
Secondly, an orbital etching approach (tidally locked) was applied (Figure 3), wherein a circular motion with a radius of 0.25 mm and circumferential velocity of 2.14 mm/s was superimposed on the nozzle translation along the x-axis (feed rate: 0.34 mm/s, current 2000 mA). Orbital strategies will lead to discontinuous sinusoidal areal charge transfer along the toolpath that will influence local removal rates. Accordingly, the small orbital radius was selected to densely pack nozzle oscillations and homogenise local charge transfer. Areal charge density was maintained as above (0.588±0.012 C/mm 2 ).

Characterisation
Etch depths were determined from focus variation microscopy (FVM) datasets (Alicona G4 microscope, 10x objective). The depth of the central 8.5 mm of the 10.0 mm slot was interpreted to account for surface tension effects in slot jet EJM, which affect profile edges [19]. The depth was taken as the mean depth of all of the profiles (1,756) taken across the 1.51 mm width (field of view of the 10x objective), and the error bars are taken as the deviation between the mean depth profile and the best fit line (Supplementary Figure 1).
Surface roughness, Sq, and surface-texture aspect, Str, were appraised from datasets acquired using the aforementioned FVM (using the higher magnification 100x objective). An 80 μm cut-off filter was applied to the raw surfaces before analysis. Surface texture-aspect ratio is a measure of surface isotropy and is calculated as the ratio between the autocorrelation length and the horizontal distance of the slowest decay of the autocorrelation function to a set value (0.2) [20].
Scanning electron microscopy (SEM) was performed using a Philips XL-30 microscope, and EBSD was performed (Oxford Instruments HKL Advanced) with a field emission electron source (JEOL 7100F, 15 kV). EBSD data acquisition was performed over set areas (600 x 600 µm) and the step size was fixed as 10.00 µm for the coarsened Ni (4.3% of the mean grain diameter); exceptions are explicitly presented in text. EBSD patterns were acquired parallel to and perpendicular to the EJM feed direction (Figure 4a-b, respectively). EBSD data were processed using the MTEX toolbox add-on for Matlab (Figure 4c indicates map colouring [21]). No filling algorithms were applied to the datasets, and single grain areas were calculated

Scratching the surface: ambient planar depth profiling
Stepped layer profiling was applied to prepare large, but site-specific surface areas for inspection, retaining etch depth control. Samples of as-received and coarsened Ni were selectively etched (Table 1)

On the relationship between etch topography and EBSD index rate
Topographies of the etched surfaces for both material preconditions were appraised to further understand etch artefact formation and the effect they have on the response rates of further characterisation techniques, such as EBSD. The SE micrographs in Figure 6 show the development of parallel striations across the separate layers. These appear more pronounced upon successive passes (indicated by higher areal charge density), and occur in both the asreceived (Figure 6a-d) and coarsened (Figure 6e-h) specimens. In addition, the effective topographic wavelength, the separation between the discrete striation s, appears to be influenced by grain size. For the as-received specimen, the inter-striation distance is smaller than for the coarsened specimen, although the size difference does not appear to scale linearly with grain size in this case. For example, the micrographs in Figure 6b   EBSD measurement efficacy (index rate over sampling area) was appraised on these surfaces to understand the effect of the anisotropic etch artefacts. Here, the aforementioned areas of interest on the surface (Figure 5i-iv) were addressed orthogonally, that is to say identical areas were appraised parallel to (arbitrarily termed ϕ = 0°), and perpendicular to (ϕ = 90°) the translation direction. This was achieved through a rotation around the z axis (surface normal) in the electron microscope ( Figure 4). EBSD measurement quality is sensitive to topographic undulations. A corollary is that the orientation of the anisotropic surface relative to the detector will influence index rate. In this study, a 0.6 x 0.6 mm area was acquired at a step size of 10 μm (<5% of the mean grain diameter for the coarsened Ni).
The orthogonal EBSD results are presented in Figure 7a When etching fine grained samples at these parameters, there is likely to be a grain size limit below which the slot jet preparation method may struggle to generate surfaces conducive to EBSD measurement, especially where the striations approach the grain size. However, it should be noted that i) striation separation reduces with grain size, and ii) the Sq roughness is marginally lower for the finer grained sample (Figure 6i), both of which may reduce this dependence. These preparation times appear favourable when compared with conventional EBSD surface preparation methods, for example successive grinding, diamond polishing, colloidal silica polishing, and vibratory finishing steps [24]. Thus, it is considered that when applied appropriately, EJM can dramatically increase the throughput of what is generally considered a low-throughput-high-quality surface analysis technique in a site-specific manner.

On the origins and propagation of translational artefacts
The EJM etch artefacts are dependent on both the material microstructure and the translation of the electrolyte jet. The latter is evidenced by the anisotropy of the etch surfaces, with striations running parallel to the translation direction, crossing multiple grain boundaries.
Evidence to support the former is given in the correlative study Figure 8

a) SE micrograph of area of interest v, inset boxes showing acquistion areas of further microscopic study. b) Topography reconstruction (FVM, 100x objective) of area of interest v showing relatively large height differences across the striations. c ) Areas of EBSD zero solution (white pixels) are spatially correlated with topographic aberrations. d-f) Micrographs show striations are bounded by anisotropic etch facets, propagating these facets leads to striation development across multiple grains and triple points.
Determining the dominant factor in etch artefact formation is complicated by jet translation and electrolyte flow following the same axis (Figure 1b). To isolate these factors, it is necessary to observe features etched without translation, such as the termination regions, where the feed starts and stops, and static nozzle etching operations. The latter will return the footprint of the jet (jet analogue of a beam footprint [26]). The optical micrographs in Figure 9a show the starting and termination regions of a single pass toolpath in coarse-grained Ni (Section 2.3), with approximate locations of the toolpath start and end points marked by dashed boxes.
Striations occur beyond these areas. SE micrographs show the initiation of discrete pits within and beyond the termination region, which occur at defects like grain boundaries (Figure 9b).
The propagation of an artefact from a pitting event ahead of the jet is shown in Figure 9c Pit initiation here is consistent with corrosion studies, originating at surface defects, aligning with certain grain boundaries [27] and pre-existing grind marks [28]. In these pits, the surface is locally de-passivated, revealing anisotropic morphologies dictated by remnant slow etch surfaces. It is understood that increasing flow efficiency encourages the removal of electrochemical reactant products from the sample surface, preventing the accumulation of anodic films and the development of the mass transport limited condition [29]. These films can dissociate material removal from the underlying microstructure; conversely, efficient evacuation of these films will promote the anisotropic etching condition. Accordingly, it is intuitive that the striation artefacts are likely to channel the flow of electrolyte parallel to the surface, enhancing the local removal efficacy of reactant products (Figure 9i

Processing challenges: reducing repeats or orbital machining?
The processing approach required to overcome characteristic etch striations depends on the magnitude of the depth to be etched, and the surface inspection strategy to be used after EJM.
In this section, different approaches are explored to etch a given depth of material (≈20 μm), and the resulting surface quality in relation to the applicability to EBSD measurement is discussed. In the case that the striation amplitude is increased upon repeated passes, the facile route through which one might reduce these striations is by reducing the number of repeated passes to achieve a given depth. For a given current density condition, this is achieved by reducing nozzle feed rate, νf. As such, coarse Ni was processed at equivalent areal charge density to the multi-pass Layer 4 outlined in Table 1  The origin of these pitting events is likely to have a stochastic relationship to a number of factors, for example the distribution of defects in the material. Once a discrete pit has been initiated, propagated and halted ( Figure 8d) during a preceding machining pass, subsequent etching appears to preferentially attack these regions where the local corrosion resistance has been degraded and the electrolyte flow can be channelled, further focussing etching in these locations. The surface amplitude increases, and the anisotropy defined by the flow and the orientation-dependent etch rates is retained. Considering the orbital processing approach, the higher roughness likely results from the undulating charge density associated with the orbital jetting strategy, specifically where the retrograde nozzle motion will promote material removal in pre-existing channels, analogous to repeated passes, leading to deeper striations. This is likely due to the aforementioned retrograde motion that will lead to charge accumulation and promotion of material removal in pre-existing etch pits. Both datasets indicate an orientation dependence of index rate at these etching parameters. For example, grains 1 and 3 show index rates of 98% and 95% respectively, while within grain 2, a 59% index rate is achieved where the sample is 0° orientated. Index rates are lower for each of these grains in the 90° acquisition orientation.

On orientation-dependent index rates
Orientation-dependent index rates are intuitive given the anisotropic etch topographies and the surface sensitivity of EBSD. To quantify this dependency, a larger acquisition area (≈8 mm 2 , 3.29 x 2.42 mm) was measured by EBSD to sample a more statistically relevant number of grains at the same acquisition parameters (10 µm step size). In this study, the single pass approach was applied to remove ≈20 µm of material (0.586 C/mm 2 ) to minimise striation amplitude of and the EBSD dataset was acquired in the 0° sample orientation, parallel to the nozzle feed direction to maximise index rates. Figure 11a shows the resulting map with a global index rate of 92%, correlating with Figure   10h (etched at the same parameters), with the associated band contrast map of the data ( Figure 11b). Mean angular deviation (MAD) data indicates no strong correlation with material or topography (Supplementary Figure 2). The inverse pole figure (IPF, z direction) in Figure   11c shows the scatter of data from this set. Regions of high data density are indicative of discrete grains. To appraise the dependence of index rate on orientation, the data were separated into individual grains (4° misorientation threshold), where Figure 11a shows the resulting grain boundary overlay. Individual grains were plotted on the resulting IPF ( Figure   11d, marker area scaled to grain area). Local grain-averaged index rates were plot onto the IPF (Figure 11e), showing the spread in index rates (45-100%). As indicated in Figure 10h-i, grains oriented towards <001> directions tend towards lower index rates. This is intuitive, as after etching, orientations aligned towards <001> are typically characterised by inverted square pyramidal morphologies (Grain 2, Figure 10), rather than plateaued regions (Grains 1 and 3, Figure 10), which tend not to shadow backscattered electrons.
Both in-grain effects and the etching process drive differences in band contrast across the dataset. Data shows slight positive correlation (Pearson's R 2 = 0.56) between grain-averaged index rate and grain-averaged band contrast (Figure 11f), where it is known that band contrast is influenced by grain orientation. Differences in band contrast also arise over multiple grains, following the feed direction. Band contrast periodicity in the data is exemplified in AOI vi (red rectangle Figure 11a). Figure 11g shows the associated orientation data (IPF colouring),  Considering process-induced anisotropy, the flow-generated striations, predominantly caused by differential flow regimes over the surface, appear to be reduced by equalising electrolyte flow. Practically, this is achieved by increasing the jet velocity. Considering material -induced anisotropy, the scale of the etch facets can be reduced by increasing the applied current density at the surface, where the removal mechanism tends towards electropolishing.
However, it should be noted that for orientation measurement routines reliant on topography [16], or directional reflectance from etch facets [10,31], anisotropic etch textures may be beneficial.

Conclusions
In this study, we have introduced EJM as a route through which large -area (e.g. 160 mm 2 ) surfaces can be prepared in a depth-controlled manner for subsequent inspection using advanced characterisation methods such as EBSD. Surfaces were prepared rapidly (e.g. <0.3 s/mm 2 ) and were shown to be compatible with EBSD measurement. EBSD measurement quality (index rate) is shown to deteriorate upon multiple passes to etch a defined volume of material, although for a large sampling area (e.g. 0.36 mm 2 ), index rate does not deteriorate below 77% at the acquisition parameters applied in this study, and individual grains can be identified and studied. The surface texture created by EJM is shown to be anisotropic due to the propagation of discrete crystallographic pitting events, and this influences the EBSD index rate. Where surfaces are prepared by EJM using etching parameters, it is recommended that EBSD inspection is performed parallel to the feed direction (and therefore striation direction) when samples are prepared with a multi-pass approach.