1
How deep EBSD looks, and why that is the whole problem
The electrons that carry Kikuchi contrast escape from a very thin skin. Anything deformed, amorphous or smeared that is thicker than that skin replaces your pattern with a grey blur, and the blur looks exactly like a bad detector setting.
The usual figure
10 to 40 nm at 20 kV, with the lower end for denser materials (Winkelmann 2010). For molybdenum at 20 kV, 80% of the diffracted intensity comes from the first 10 nm.
Density sets the depth
At 40 kV and 20° incidence: about 100 nm for aluminium, 20 nm for nickel and 10 nm for gold (figures attributed to Michael and Goehner).
Two answers, seventy times apart
Orientations were still measured through about 116 nm of amorphous silicon at 30 kV. Copper patterns died under 1.6 nm at 5 kV using a Hough-peak threshold.
Which one you need
The first asks whether any solution can be indexed. The second asks whether the pattern is still sharp. Phase mapping needs the first; measuring a two-degree misorientation needs the second.
More detail: the published depth figures and why they disagree
The number usually quoted is 10 to 40 nm at 20 kV, with the lower end for denser materials (Winkelmann 2010). Simulation puts a sharper point on it: for molybdenum at 20 kV, 80% of the diffracted intensity is already collected in the first 10 nm. At 40 kV and 20° incidence the figures attributed to Michael and Goehner are about 100 nm for aluminium, 20 nm for nickel and 10 nm for gold, which is the density dependence in three numbers.
That consensus is not the whole story, and the page would be dishonest if it pretended otherwise. Wisniewski and co-workers milled a wedge of amorphous silicon over a crystal and could still measure orientations through about 116 nm at 30 kV. Chen and co-workers, using a threshold on the Hough peak instead, found copper patterns dying under only 1.6 nm of amorphous overlayer at 5 kV. Those two numbers differ by seventy times and both are correct: they answer different questions. The first asks whether any solution can be indexed, the second asks whether the pattern is still sharp. Which one you need depends on whether you are mapping phases or measuring a two-degree misorientation.
Try this: switch the density chips and the two "what counts as working" chips, and watch the EBSD band move against the damage bars of each method.
Grinding and diamond polishing both leave damage measured in hundreds of nanometres to micrometres. EBSD sees tens of nanometres. So the last step always has to be something that is not grinding or diamond: colloidal silica, electropolishing, or an ion beam. Everything before it exists only to get you close enough for that last step to finish the job in a sensible time.
2
What each method leaves behind
Six routes to a surface. They are not ranked, because each one fails on a material class the others handle, and picking wrongly wastes a week.
| Method | What it removes | What it leaves | Fails on |
|---|---|---|---|
| Grinding, SiC paper | Sectioning damage, and the previous grit's damage | A fragmented layer of its own. Measured at about 0.5 µm on tungsten after P400 to P4000. On hardened steel, 1.5 to 2 µm from 320 grit down to about 0.3 µm from 20,000 grit. | Nothing, but it can never be the last step. A 600 grit diamond plate left 4 to 6 µm, deeper than 320 grit paper, so grit number alone does not predict damage. |
| Diamond polishing | Grinding scratches, quickly | Deformation and residual stress. On titanium, diamond slurry left up to 300 MPa compressive stress; colloidal silica and electropolishing left none. On tungsten, heavy deformation survived 1 µm diamond at 35 N for 10 minutes. | Soft metals, where it smears rather than cuts. Struers say plainly that going to finer and finer diamond keeps adding deformation rather than removing it. |
| Colloidal silica, chemo-mechanical | The deformed layer, by growing a brittle reaction layer and then abrading it away with 20 to 60 nm particles at pH 9 to 10 | Essentially nothing. No measurable residual stress. Some relief between phases of different hardness if run too long. | Nothing, but it is slow: about 1 to 2 µm per hour. It cannot rescue a badly ground surface in a few minutes. |
| Vibratory polishing | The last traces of deformation, with gravity as the only load | Nothing. A stress-free surface, and less edge rounding than a wheel. | Serious damage from earlier steps. MetPrep put it exactly: it “will not remove serious damage”, it only improves a reasonably prepared surface. |
| Electropolishing | Everything mechanical, in seconds | No mechanical damage at all. But rounded edges, exaggerated pores and inclusions, and a surface that is only flat if the material is one phase. | Multiphase alloys, inclusions, porosity, composites, ceramics, anything non-conductive. It is a single-phase-metal technique. |
| Ion milling, broad beam or FIB | Whatever you point it at, including damage nothing else reaches | An amorphous or damaged layer set by the ion energy. In silicon: 22.5 nm at 30 kV Ga, 2.5 nm at 5 kV, 1.0 nm at 2 kV; argon gives 8 nm at 5 kV and 3 nm at 2 kV. Below 1 kV at a 10° angle, under 1 nm. | Nothing technically, but it is slow for large areas, and gallium attacks aluminium grain boundaries chemically. |
Why colloidal silica is the workhorse, with the numbers
It is not just a very fine abrasive. Struers describe the mechanism plainly: the suspension reacts with the surface to build a relatively brittle reaction layer, and the small particles then remove that layer. So it cuts chemically and carries away mechanically, which is why it takes material off without pushing new dislocations in.
The measurement that settles it is Everaerts and co-workers, who used focused ion beam ring-core milling with digital image correlation to read residual stress directly on titanium. Diamond slurry left up to 300 MPa of compressive stress, and the bigger the diamond the deeper it reached. Colloidal silica left no significant residual stress, and neither did electropolishing. That is the whole argument in one experiment.
The payoff shows up as indexing rate. On INCONEL 600, EDAX measured 4.5% of points correctly indexed straight off 1200 grit paper, rising to 99.9% after an hour or more of colloidal silica on a vibratory polisher. On dual-phase steel, Sandia took 87% to 97% by adding four hours of vibratory alumina and silica to an already good diamond sequence. On nickel, image quality more than doubled between fifteen minutes and two hours.
Buehler measured band contrast before and after twenty minutes of vibratory polishing across seven materials: nickel gained 20.9%, silicon 13.2%, titanium 9.1%, magnesium 8.7%, niobium only 3.8%. Lead went from no pattern at all to an indexable one, and needed sixty minutes. The pattern in those numbers is that soft, smearing metals gain most and refractory metals gain least, which matches the tungsten result that refractory metals need long chemo-mechanical polishing rather than a vibratory finish.
Three practical points that cost people whole samples. It crystallises when it dries, and crystalline silica scratches, so filter it and never let the cloth dry out. Rinse with water for the last 20 to 30 seconds of the step, with the suspension turned off. And dry with alcohol rather than acetone, because acetone causes condensation on the surface.
3
The recipes, by material
Every recipe below is published by a named source, and the source is on each one. Loads are per specimen unless the note says otherwise, which matters: a Struers table quoting 150 N usually means the total central force for six 30 mm specimens, while Buehler quotes force per specimen.
Pick your material. The others stay closed so the page stays short.
Aluminium and soft face-centred-cubic metals
Pure aluminium is one of the hardest things on this page to prepare, because it is soft enough to smear and soft enough to swallow abrasive whole. The two fixes are a wax-rubbed grinding paper and a long final polish.
Buehler TECHNotes Vol 5 Issue 2, high-purity aluminium, written for EBSD. Loads per specimen, contra rotation throughout.
| Step | Surface | Abrasive | Size | Load | Speed | Time |
|---|---|---|---|---|---|---|
| 1 | CarbiMet | SiC, water | 240 grit | 22 N | 240 rpm | 1 per sheet |
| 2 | UltraPol silk | MetaDi diamond | 9 µm | 22 N | 150 rpm | 5 min |
| 3 | TriDent polyester | MetaDi diamond | 3 µm | 22 N | 150 rpm | 5 min |
| 4 | TriDent polyester | MetaDi diamond | 1 µm | 22 N | 150 rpm | 3 min |
| 5 | MicroCloth | MasterMet silica | 0.05 µm | 22 N | 150 rpm | 3 min |
| 6 | MicroCloth on VibroMet 2 | MasterMet silica | 0.05 µm | gravity | vibratory | 20 min or more |
Struers Application Note, pure aluminium. General metallography rather than EBSD-specific, so treat the oxide step as the minimum.
| Step | Surface | Abrasive | Speed | Load | Time |
|---|---|---|---|---|---|
| Plane grind | SiC foil or paper | #320 or #500, water | 300 rpm | 25 N | until flat |
| Fine grind | SiC | #800, #1200, #4000, water | 300 rpm | 25 N | 0.5 min each |
| Diamond | MD-Mol | 3 µm DiaPro Mol R3 | 150 rpm | 25 N | 4 min |
| Oxide | MD-Chem | 0.04 µm OP-U or OP-S | 150 rpm | 15 N | 2 to 5 min |
For cast aluminium and silicon, Struers shorten the oxide step to one minute on purpose, because a longer one carves relief around the silicon particles.
LightForm, University of Manchester, written for EBSD: grind at SiC 180 and 320 to shape, then jump straight to 4000. Diamond pad 3 µm for 2 min, 1 µm for 2 min, then neat colloidal silica at about 0.25 µm for 1 to 2 minutes. No water at all for corrosion-sensitive alloys; rinse in ethanol and air dry.
MetPrep Application Note 004, AA1070 at about 20 HV: a Saphir Vibro with a low-nap cloth and 0.04 µm colloidal silica for fourteen hours. That is what commercially pure aluminium costs if you want it perfect.
Electropolishing. Struers A2 electrolyte on a LectroPol-5, 39 to 48 V, 1 cm² mask, 15 to 35 s at 22 °C. Struers' own aluminium note gives 39 V with a 2 cm² mask for 20 s, so treat the voltage as settled and the mask, flow and time as a range to tune. Manchester's manual cell route is 30% nitric acid in methanol at −40 °C and 9 V in three-second bursts, and it must stay below −20 °C; read section 5 before mixing that one.
The traps. Abrasive embeds itself in pure aluminium. Struers say to rub the SiC paper with wax first; if diamond has already been pressed in, the answer is a longer oxide polish, not a finer diamond. Gallium in a FIB attacks aluminium grain boundaries, so if you are ion polishing aluminium use a xenon plasma source, argon broad beam, or accept a worse hit rate.
Carbon, stainless and tool steels
Steels are the easy case, and the standard four-step sequence works. The interest is in the multiphase ones, where relief around carbides and inclusions is the thing that spoils a map.
Buehler SumMet contemporary four-step method, steels. Loads per specimen.
| Surface | Abrasive | Load | Speed | Time |
|---|---|---|---|---|
| CarbiMet | SiC 120 to 280 grit, water | 27 N | 300 rpm | until flat |
| UltraPad | 9 µm diamond | 27 N | 150 rpm | 5 min |
| TriDent | 3 µm diamond | 27 N | 150 rpm | 4 min |
| ChemoMet | 0.02 to 0.06 µm colloidal silica | 27 N | 150 rpm | 2 min |
Buehler TECHNote Vol 5 Issue 3, tool steels. Note the alternating rotation, which is deliberate.
| Surface | Abrasive | Load | Speed | Time | Rotation |
|---|---|---|---|---|---|
| CarbiMet | P400 SiC, water | 27 N | 300 rpm | until flat | complementary |
| UltraPad | 9 µm MetaDi | 27 N | 150 rpm | 5 min | contra |
| TriDent | 3 µm MetaDi | 27 N | 150 rpm | 3 min | complementary |
| MicroCloth | 0.05 µm MasterPrep alumina | 27 N | 150 rpm | 2 min | contra |
Buehler give a fix worth memorising: if relief appears around oxides or sulfides after the last step, repeat that step with complementary rotation instead and it goes away. Use a finer plane-grinding grit when carbide pull-out is the problem.
Sandia, dual-phase steel, prepared for EBSD, with the result measured.
| Surface | Abrasive | Load | Time |
|---|---|---|---|
| MD-Largo | 9 µm | flatten | 5 min |
| MD-Pan | 9 µm | 30 N | 5 min |
| MD-Pan | 6 µm | 35 N | 5 min |
| MD-Mol | 3 µm, red lubricant | 35 N | 5 min |
| MD-Nap | 1 µm | 35 N | 5 min |
| DP-Nap, vibratory | 0.3 µm alumina | gravity | 2 h |
| DP-Nap, vibratory | 0.02 µm OP-U | gravity | 2 h |
Indexed fraction went from 87% to 97% when the two vibratory steps were added. The four hours bought ten points of indexing on a sequence that was already good.
Electropolishing. Struers presets at 22 °C, 1 cm²: low carbon steel A2 at 40 V for 12 s; medium carbon AC2 at 53 V for 20 s; hardened steel A3 at 63 V for 18 s; stainless A3 at 35 V for 25 s. Struers report that it works well for duplex and ferritic steels in general, but that ferritic steels with non-metallic inclusions are a problem: the inclusions separate from the matrix. If you are counting inclusions, do not electropolish.
Titanium alloys
Titanium is where the usual instinct is wrong. Do not walk down through finer and finer diamond. Stop diamond early and let peroxide-boosted colloidal silica do the rest, for a long time.
Struers Application Note, Titanium. Unmounted, 30 mm diameter, loads as printed.
| Step | Surface | Abrasive | Speed | Load | Time |
|---|---|---|---|---|---|
| Plane grind | MD-Mezzo | #220 diamond, water | 300 rpm | 40 N | until flat |
| Fine grind | MD-Largo | 9 µm DiaPro | 150 rpm | 30 N | 5 min |
| Oxide | MD-Chem | 90% OP-S + 10 to 30% hydrogen peroxide (30%) | 150 rpm | 30 N | 5 min, up to 45 min for commercially pure Ti |
For commercially pure grades 1 to 4 Struers drop the loads to 15 to 20 N, start at SiC #320, and use 80% OP-S with 10% peroxide and 10% ammonium hydroxide for 25 minutes or more.
Struers, in their own words, on why the diamond ladder stops: “Contrary to the usual procedure of using finer and finer diamond for polishing, diamond polishing actually introduces continuously mechanical deformation which leaves scratches and smearing on the surface.”
LightForm, Manchester, tuned by alloy hardness. Mixtures are water : colloidal silica : peroxide.
| Alloy | Mixture | Time and pressure |
|---|---|---|
| Harder than about 350 HV | 4 : 1 : 1 | 1 to 2 min hard, then about 30 s light |
| Softer than about 350 HV | 4 : 1 : 1 | 15 min, light pressure |
| Bimodal Ti-6Al-4V | 4 peroxide : 1 silica | 25 min or more, then 5 min under running water |
| Fully lamellar Ti-6Al-4V | 1 : 1 peroxide and silica | 5 min, then 1 : 1 : 4 until scratch free |
Electropolishing. Struers A3 on a LectroPol at 18 to 20 °C, 35 to 45 V, flow 10 to 15, 20 to 30 s, starting from a #1200 or finer ground surface. Manchester's manual route for soft titanium is 5% perchloric acid in methanol at −40 °C and 12 V in three-second bursts; see section 5. There is a published non-acid alcohol-based route specifically validated for EBSD of pure Ti and Ti-6Al-4V (Materials Characterization, 2020) that is worth pulling if your lab is trying to get perchloric acid out of the building.
Vibratory. About an hour with colloidal silica plus 20% peroxide finishes Ti-6Al-4V. Buehler measured a 9.1% gain in band contrast from twenty minutes on high-purity titanium.
Magnesium alloys
The rule that governs everything here is no water. Magnesium corrodes while you watch, and the corrosion product is what stops the pattern.
LightForm, Manchester, written for EBSD.
- SiC 180 and 320 to shape, then jump straight to 4000.
- Diamond paste pad 3 µm, 2 min, oil-based lubricant.
- Diamond paste pad 1 µm, 2 min, oil-based lubricant.
- Fumed silica at about 0.25 µm, 4 min, no water.
- Finish with ethanol and ethylene glycol at 10 : 1, then ethanol, then air dry.
Buehler TECHNotes Vol 4 Issue 2, optical metallography rather than EBSD: 320 grit SiC with water, then 9, 3 and 1 µm diamond in oil-based suspensions, then 0.05 µm MasterPolish on ChemoMet. Contra rotation for the diamond steps, complementary for grinding. Buehler note that colloidal silica etches magnesium alloys, though it is acceptable on pure magnesium.
Electropolishing. Manchester use 30% nitric acid in methanol at −40 °C and 9 V in three-second bursts, with a hard limit at −20 °C. Read section 5 first: this combination has a serious published warning attached and the temperature is not a comfort setting, it is the safety interlock. Struers list magnesium under AC2 but report electropolishing as a limitation for magnesium alloys rather than a strength, and recommend deep cooling of the electrolyte.
The trap. Magnesium swarf is pyrophoric. Do not use water on a grinding wheel that is generating magnesium dust, and know where the class D extinguisher is. Buehler also warn that magnesium alloys carry precipitates far harder than the matrix, which makes relief hard to control, so avoid napped cloths.
Nickel superalloys
The best-documented EBSD preparation study of any material is on INCONEL 600, and its headline is the clearest number on this page.
EDAX, Nowell, Wright and Carpenter, "EBSD Sample Preparation: Techniques, Tips, and Tricks", INCONEL 600.
- Section on a low-speed diamond saw at about 200 rpm with cutting fluid.
- Mount in copper-filled conductive powder, 3800 psi, 175 °C.
- Grind 240, 320, 400, 600, 800 and 1200 SiC with water, 45 N, 150 rpm, 30 s each, fresh paper every time.
- Polish on Imperial cloth with 1.0 then 0.3 µm alumina, 10 min each, 40 N, 130 rpm, water.
- Vibratory with 0.05 µm colloidal silica at pH 9.8, under 5 N.
Correctly indexed points: 4.5% straight off the 1200 grit paper, and 99.9% after an hour or more of vibratory colloidal silica. The same study tested 15, 30, 60, 120 and 240 minutes; the curve is essentially flat after 60. Buehler separately measured a 20.9% gain in band contrast from just twenty minutes of vibratory on high-purity nickel, the largest gain of any material they tested.
Electropolishing. Honest answer: I could not find a published electropolishing recipe validated for EBSD of a nickel superalloy. Struers list nickel under A2, A8 and AC2 but publish no preset method for it. A deep eutectic solvent route (1 choline chloride : 2 ethylene glycol, 10 to 50 mA/cm², 20 °C, 30 min) takes nickel to under 5 nm of roughness, but it has not been shown to produce EBSD-ready surfaces. The defensible route for superalloys is mechanical plus a long vibratory polish, which is what the numbers above say anyway.
Copper and brass
Copper is the one material family where the perchloric-free electropolishing route is fully documented and routine, so use it.
Buehler TECHNotes Vol 5 Issue 2, high-purity copper, written for EBSD. Contra rotation throughout.
| Surface | Abrasive | Load | Speed | Time |
|---|---|---|---|---|
| CarbiMet | 240 grit SiC, water | 27 N | 240 rpm | 1 per sheet |
| UltraPol or TriDent | 9 µm MetaDi | 27 N | 150 rpm | 5 min |
| TriDent or TexMet | 3 µm MetaDi | 27 N | 150 rpm | 5 min |
| TriDent or TexMet | 1 µm MetaDi | 27 N | 150 rpm | 4 min |
| MicroCloth or ChemoMet | 0.05 µm MasterMet | 27 to 31 N | 150 rpm | 3 min |
| MicroCloth on VibroMet 2 | 0.05 µm MasterMet | gravity | vibratory | 20 min or more |
Oxford Instruments and Kemet application note AN012 cover aluminium, titanium and brass with one table. SiC 320, 600, 800 and 1200 with water at 22 N, 300 rpm platen and 150 rpm specimen, 1 min each, complementary. Then 3 µm polycrystalline diamond on a Gold Label cloth for 10 min, 1 µm on Kempad for 10 min, both at 22 N and 150/150 rpm, then colloidal silica. The 1 µm diamond step is there specifically to work out the 3 µm particles that embedded themselves.
Electropolishing. Struers D2, 24 V, 20 s at 22 °C for copper, and E5 at 56 V for 18 s for brass and bronze. D2 is phosphoric acid, ethanol and propanol: no perchloric acid at all, confirmed from the safety data sheet. If your lab is trying to eliminate perchloric acid, copper and brass are the easy win.
Tungsten and refractory metals
Refractory metals gain least from vibratory polishing and most from long chemo-mechanical polishing. The Max Planck study on tungsten is the one paper that watched each step and measured what was left.
Max Planck Institute, "A step-by-step analysis of the polishing process for tungsten specimens", with the damage measured at each stage.
| Step | Conditions | What was left |
|---|---|---|
| Grinding | SiC P400 to P4000, 20 N, 1 min each | About 0.5 µm of fragmented grains |
| Diamond | 1 µm suspension on a textile pad, 35 N, 10 min | Heavy deformation still present |
| Chemo-mechanical | Alkaline colloidal silica on synthetic felt, 30 N, 10, 20 and 30 min | 30 minutes needed for complete removal of the deformed layer |
| Optional electropolish | 2.5% sodium hydroxide, 25 V, 20 °C, 2 to 3 min | Then vacuum anneal at 1200 K |
Electropolishing, perchloric-free. A published aqueous route for tungsten: sodium hydroxide at 0.27 mol/L, 8 V, 1.0 mm electrode gap, 667 rpm stirring, copper cathode, 400 s, reaching 7.5 nm roughness. The same paper found the conventional concentrated phosphoric acid and glycerol electrolyte ineffective on tungsten, which is a useful negative result.
Zirconium, antimony and vanadium. Buehler report that all three needed SiC paper coated with paraffin wax to stop abrasive embedding. Their zirconium method is SiC 240 then wax-coated 320, then 9, 3 and 1 µm MetaDi at 27 N and 200 rpm for 10, 7 and 5 min, then 0.05 µm MasterMet for 7 min, then 20 min or more of vibratory. Zirconium also benefits from a 5 : 1 colloidal silica to peroxide final polish. Niobium gained the least of any material from vibratory, only 3.8% in band contrast.
Ceramics and cemented carbides
The problem is relief. Tungsten carbide is 1800 to 3200 HV and its cobalt binder is about 200 HV, so the binder disappears under the carbide and the surface stops being a plane.
Rigid discs, not woven cloths, for the intermediate steps. That is the single fix that matters: a rigid composite disc keeps the hard phase flush with the soft one, while a woven cloth lets the binder recede and leaves carbide grains standing proud. The same applies in silicon nitride, where the glassy grain-boundary phase polishes faster than the nitride grains.
A workable sequence for cermets and cemented carbides, from PACE Technologies. Their numeric tables are not attributed to a named author, so treat these as a starting point rather than a specification.
- Section with a diamond wafering blade, very slow feed, rubber-jaw clamping.
- Mount in castable epoxy with vacuum impregnation. Never compression mount: thermal shock cracks glasses and carbide interfaces.
- Grind on 70, 45, 30 and 15 µm diamond discs at 10 to 15 N, very light, with oil-based lubricants to stop the cobalt or nickel binder oxidising and staining.
- Polish 9 µm then 3 µm diamond on rigid composite discs, then 1 µm on a hard pad, all at 5 to 10 lb and 100 to 200 rpm.
- Finish with 0.05 µm colloidal silica or alumina on a napless cloth. Napless is mandatory or the binder smears.
Sandia, on ceramics for EBSD, and the most useful sentence in this whole section:
“Fracture surfaces generally produce excellent patterns in ceramics.” For a brittle ceramic you can sometimes skip preparation altogether and index a fresh fracture surface. Sandia also report that mechanical polishing followed by a light chemical etch is often enough, that vibratory polishing with colloidal silica is a good last step, and that ion beam preparation works for some oxides: they used it on alumina and zinc oxide. Their alumina route is mechanical polishing followed by phosphoric acid at 250 °C.
For hard covalent ceramics such as silicon carbide, broad ion beam polishing is the practical last step, because mechanical polishing cannot reach the subsurface damage. Electropolishing is not available to you here at all: it needs a conductor.
Silicon and semiconductors
Silicon is the material with the best-measured ion damage numbers, so this is where the low-kV finishing rule comes from.
Sandia, residual amorphous layer on silicon after ion polishing, measured.
| Ion and energy | Amorphous layer left |
|---|---|
| Gallium, 30 kV | 22.5 nm |
| Gallium, 5 kV | 2.5 nm |
| Gallium, 2 kV | 1.0 nm |
| Argon, 5 kV | 8 nm |
| Argon, 2 kV | 3 nm |
Oxford and Kemet add that below 1 kV at a 10° angle the residual amorphous layer is under 1 nm. Put those numbers beside the 10 to 40 nm information depth from section 1 and the whole low-kV finishing rule follows: a 30 kV gallium surface is damaged right through the depth EBSD reads, and a 2 kV finish is not.
Mechanical. Struers advise fine grinding with diamond on a rigid disc rather than SiC foil, then diamond on silk, then a brief colloidal silica polish to avoid relief. Buehler measured single-crystal silicon band contrast rising 13.2% from twenty minutes of vibratory polishing, so the mechanical plus vibratory route does work; they just do not publish the full table.
Thin films on semiconductors belong in the thin films section below: if the film is thinner than the interaction volume, polishing is the wrong tool entirely.
Additively manufactured metal with porosity
Porosity changes the mechanical preparation and rules out electropolishing, for the same reason in both cases: a pore is an edge, and edges get preferential treatment.
Impregnate first. Struers give the trick that matters: heat the epoxy to 50 to 60 °C so it flows into the pores, add fluorescent powder so you can see where it went, and cool actively while it cures. Skip this and the pores fill with polishing slurry, which then bleeds out during the next step and leaves comet tails and false porosity.
- Cast-mount under vacuum. Do not compression mount porous material: the pressure collapses pores.
- Grind 120, 240, 400 and 600 for 30 to 60 s each with lighter pressure than you would use on dense material, to avoid collapsing and smearing pores shut.
- Clean thoroughly between every step to get trapped abrasive out of the pores.
- Diamond 9 µm for 3 to 5 min on a firm low-nap cloth, 3 µm for 3 to 5 min, 1 µm for 2 to 3 min on a soft cloth, all light.
- Colloidal silica 0.05 µm for 1 to 2 min, then rinse in water and then alcohol, then ultrasonic clean to get the compound out of the pores.
Struers route additively manufactured material by alloy: rigid plane grinding for aluminium to avoid edge rounding; avoid very coarse grinding and high pressure on stainless and nickel; and for titanium, avoid diamond polishing and electropolish the less-alloyed grades.
Reading the defects. Compound trapped in pores means you need the ultrasonic step. Poor edge retention means a harder mount or less polishing time. Relief around pores means you polished too long, not too little. And grain pull-out in a loose mount reads as porosity, which in an additively manufactured part is exactly the measurement you were trying to make.
Why not electropolish. ASTM E1558 states that the areas around non-metallic inclusions and voids polish faster than the matrix, exaggerating their apparent size, and that preferential attack rounds specimen edges. A porous additively manufactured sample is that situation repeated thousands of times, so electropolishing systematically overstates the porosity. Struers did report success electropolishing laser-sintered Ti-6Al-4V, so for titanium it is defensible; for aluminium with gas porosity it is not.
There is a dedicated open-access study, "Developing a best practice for sample preparation of additive manufactured AlSi10Mg for electron backscatter diffraction analysis" (Applications in Engineering Science, 2023). It is the only paper written specifically on this question and it is worth reading before you commit a batch.
Thin films and coatings
If the film is thinner than the interaction volume, do not polish it. There is no polishing recipe that stops at the right depth.
Three routes, in order of preference.
- Transmission Kikuchi diffraction. Thin the film to 50 to 100 nm and work in transmission. Spatial resolution goes from 25 to 100 nm in conventional EBSD down to 2 to 10 nm, which is the only way to measure a genuinely nanostructured film. Prepare by electropolishing a 3 mm disc for a conductive metal, or by focused ion beam lift-out for anything site-specific or non-conductive. This also rescues heavily deformed material, where the dislocation density defeats conventional EBSD.
- Drop the accelerating voltage. For a plan-view map on a film a few hundred nanometres thick, lowering the kV shrinks the interaction volume so the signal comes from the film rather than the substrate. There is no published “polish for N seconds” answer here because polishing is not the variable.
- Low-angle plasma FIB cross-section. For a cross-section through a coating, 30 keV down to 10 keV xenon at 30 to 100 nA is the published envelope, with EBSD then collected at 20 keV. The angle is the point of the technique and the public documentation does not state it, so ask your supplier.
A broad-beam recipe that is published, from Manchester, for titanium: 7 kV at a 12.5° beam angle for 30 min, then 7 kV at 5° for 30 min. Halve both times for magnesium and aluminium.
Geological and mineral samples
Two independent university facilities publish almost the same recipe, which is reassuring. It is long, and it starts from a thin section rather than a mount.
Carleton SERC and Bowdoin College, both working from standard polished thin sections.
- Start from a microprobe-quality polished thin section, 30 µm thick, mounted to a brass holder with sheet adhesive.
- Vibratory polish with non-crystallising colloidal silica on a 12 inch cloth, pad dampened to 50 to 75% coverage, rock side down.
- Three to six hours. Carleton run two thin sections at a time for six hours; Bowdoin use three. Removal is 1 to 2 µm per hour, and quartz-rich samples take four to five.
- Rinse thoroughly with water so the silica does not crystallise, ultrasonic for about ten minutes, then an ethanol wipe or hot-air dry.
Earlier steps, where a thin section is not already available, are diamond paste at 3, 1 and then 0.5 or 0.25 µm before the chemo-mechanical step. Colloidal silica has rounded particles and a pH above 9; colloidal alumina has platy particles and a pH near 7.
Charging, which is the real problem here. Two published routes, and you pick one:
| Route | Setting | Cost |
|---|---|---|
| Carbon coat and stay in high vacuum | 2.5 to 5 nm of carbon. One study found 2.5 nm sufficient for conductivity and 5 nm or more measurably degrading the diffraction intensity. | Above 5 nm you lose indexing fraction before you lose the image, so there is no visual warning. |
| Do not coat, run in low vacuum | 10 to 15 Pa in practice; Oxford give 10 to 50 Pa as the range and say that any higher and the pattern is too weak. | Gas scatters both the incident beam and the diffracted electrons, so contrast falls monotonically with pressure. Nobody has published the indexing penalty as a number. |
Both facilities run 20 kV at 25 mm working distance with 1 to 3 nA. If you do coat, carbon is the right element because it is light; gold or tungsten cost far more pattern contrast per nanometre. Before reaching for a coater, try the cheaper fixes: a better polish, tilting to 70° before turning the beam on, faster scanning, lower probe current, and conductive paint around the edges. Silver paint conducts better than carbon paint, and carbon paint better than carbon tape. Carbonate biominerals give better results when embedded in epoxy.
Polymers and polymer composites
The honest answer is that EBSD is not the tool for bulk polymers, and no published preparation recipe exists because the technique does not apply.
EBSD needs a crystalline, reasonably conductive, flat surface. Semicrystalline bulk polymers do not give usable Kikuchi patterns under normal conditions, and the crystallography of polymers is done by X-ray scattering, by electron diffraction in a transmission microscope, and increasingly by four dimensional scanning transmission microscopy.
The interesting exception. It has been done on crystalline organic molecular thin films. A 2018 study collected EBSD maps from rubrene, described by the authors as the extreme example. Two things make it hard: the Kikuchi patterns are weak because the atomic numbers are low, and the patterns fade and vanish once a critical electron dose is exceeded. So it is a specialist, dose-limited measurement rather than routine work, and it opens a door for crystalline organics rather than for polymers in general.
If your composite has a crystalline reinforcement, the EBSD target is the reinforcement and the preparation problem is metallographic. Cast mount at room temperature with a two-part epoxy and vacuum impregnation; section with a precision diamond wafering blade at 0.5 to 1.0 mm per minute with flood cooling, letting the blade do the work, because force delaminates rather than cuts; grind 240 through 1200 at very light 5 to 15 N; then polycrystalline diamond at 9, 3 and 1 µm with the pad getting softer as the abrasive gets finer, and a short colloidal silica finish.
Aim for relief under 100 nm. If pull-out persists after you have reduced the force and firmed the pad, the cause is subsurface damage from the sectioning, not the polishing.
4
Reading a bad map backwards
A preparation fault does not announce itself. It arrives as a detector problem, a settings problem or a material result. This table runs the other way: the symptom first, then the step that caused it.
| What you see | Usually caused by | What to change |
|---|---|---|
| Uniformly weak or absent patterns over the whole map, on a surface that looks mirror-bright | A deformed layer thicker than the information depth. Optical brightness only tells you about micrometres. | Add or lengthen the last non-mechanical step. On nickel this took indexing from 4.5% to 99.9%. |
| Good patterns in some grains, none in others, on a single-phase sample | Smearing, usually from diamond on a soft metal, or embedded abrasive | Stop the diamond ladder earlier and lengthen the oxide polish. Wax the SiC paper if the abrasive is embedding. |
| Indexing falls off near one edge or one phase | Relief. The local surface normal is tilted, so the pattern geometry is wrong even though the pattern is sharp. | Rigid discs instead of woven cloths for the intermediate steps; shorten the oxide step; try the same step again with the opposite rotation. |
| Parallel bands of poor indexing across the map | Scratches that survived, or the grinding direction | Rotate 90° between grinding steps so you can see when the previous step's scratches are gone. |
| The map drifts or the pattern moves during a long scan | Charging. Either no conductive path, or a coating that is too thin. | Conductive paint to the stage, or 2.5 to 5 nm of carbon, or 10 to 15 Pa of chamber gas. Not all three. |
| Patterns get weaker as you add carbon coat | The coating is over 5 nm | Coat thinner. Above 5 nm the indexing fraction falls before the image looks any different. |
| Pores and inclusions look larger than they do in the optical image | Electropolishing. Voids and inclusions dissolve faster than the matrix. | Do not electropolish anything you intend to measure quantitatively for porosity or inclusion content. |
| Pits scattered across an electropolished surface | Voltage above the polishing plateau, so oxygen is evolving at the surface | Run the current against voltage scan and work in the plateau. Too low a voltage etches instead; too high pits. |
| Grains missing at the edge of the field, or the sample edge rounded | Electropolishing, or a soft mount | Harder mounting compound, or switch to vibratory, which rounds edges far less. |
| A gradient in indexing quality across a focused ion beam cross-section | The final milling energy was too high, or the polish was done at one angle | Finish at 2 kV or below. At 30 kV gallium the damage is 22 nm in silicon, which is the whole information depth. |
| Aluminium indexes far worse than it should after focused ion beam work | Gallium. It dissolves into aluminium and concentrates at grain boundaries. | Use xenon plasma or argon broad beam. One vendor comparison put the hit rate at 51% for gallium against 76% for xenon on the same material. |
| Scratches appear that are coarser than anything you used | Colloidal silica that dried and crystallised | Filter it, keep the cloth wet, and rinse with water for the last 20 to 30 seconds of the step. |
| Holes where a hard particle used to be | Pull-out, from too coarse a plane grind or a loose mount | Finer plane grinding grit, and matched platen and holder speeds at 150 and 150 rpm rather than a big velocity difference. |
Prepare a second sample of a material you know indexes well, in the same session on the same instrument, and map it. If the known sample indexes, the problem is your preparation. If it does not, the problem is the microscope or the detector, and no amount of polishing will help. This costs one afternoon and it is the only test that cleanly separates the two.
5
Electropolishing safety
Most electropolishing electrolytes for metallography contain perchloric acid. Used the way the vendors specify, they are routine. Used outside that envelope they have killed people.
Perchloric acid and acetic anhydride. The standard metallography safety guidance is blunt about it: the mixture is “difficult to prepare and highly explosive”, electrolytes made from it “are not recommended”, and many companies forbid it while some cities have banned it. In 1947 a plating works in Los Angeles was using a perchloric and acetic anhydride bath to polish aluminium. The bath was boiling, the concentration had drifted far above normal, and a plastic rack went into it. Seventeen people died, more than a hundred were injured and eleven buildings were destroyed.
Three things were true at once: the concentration had climbed, the temperature control had failed, and an organic material touched a hot concentrated bath. Any one of those alone is survivable. That is the lesson, not the specific chemistry.
The published safety guidance, in the sources' own terms:
- Acid concentrations of 30% or less were judged safe to use as long as the operating temperature does not exceed 20 °C.
- Perchloric acid with alcohol, with or without organic additions, and perchloric acid with glacial acetic acid, are described as safe to use. Mixtures of acetic acid with 5 to 10% perchloric acid have been used routinely on iron-based alloys and are called reasonably safe.
- Mixtures of perchloric acid with 2-butoxyethanol at 60 to 90% acid were explosive at room temperature. This one matters because 2-butoxyethanol is the additive in the Struers A2 and A3 solvents. The margin is entirely in the dilution, which is why you mix to the vendor ratio and never strengthen a bath.
- Never electropolish a sample mounted in phenolic or other plastic with a perchloric solution. Explosions have resulted.
- Do not use these solutions on bismuth, arsenic or tin. Explosions have occurred.
- Do not store the electrolyte for more than a few days, and discard it once dissolved metal has coloured it.
- Keep it cool. Raising the temperature raises the oxidising power of perchloric acid.
- Regular use needs ventilation designed for perchloric acid, with a washdown fume hood, because perchloric vapour condensing in ductwork forms metal perchlorates, which are shock sensitive. Heated perchloric acid must never go in an ordinary fume hood. Cold electropolishing in a standard hood is the normal metallography case; heating is what triggers the special hood.
The same safety guidance says that mixtures of methanol with more than 5% nitric acid are subject to violent decomposition if heated, and that 33% nitric acid in methanol has decomposed suddenly and violently. Yet a university facility publishes 30% nitric acid in methanol as a routine electropolishing recipe for aluminium and magnesium. Both are real sources and neither is wrong.
The reconciliation is temperature. That recipe runs at −40 °C with a hard limit at −20 °C, and the warning is about heating. In electropolishing the chemistry and the thermal interlock are one inseparable specification. A recipe quoted without its temperature is not a recipe. Note also that a bench electropolisher may not accept electrolytes below 0 °C at all, so the very cold routes are manual-cell work.
Routes that avoid perchloric acid entirely. Copper, brass and gold have the cleanest answer: the Struers D2 electrolyte is phosphoric acid, ethanol and propanol, with no perchloric acid and no oxidiser hazard on its safety data sheet. Tungsten has a published aqueous sodium hydroxide route. Nickel and cobalt have a deep eutectic solvent route. Titanium has a published non-acid alcohol-based route validated specifically for EBSD. And for anything at all, the two escape routes are a long vibratory colloidal silica polish or broad ion beam milling, both of which give a stress-free surface with no dangerous chemistry at all.
This section summarises published safety guidance so you know which questions to ask. It is not a substitute for your own institution's chemical safety approval, your supplier's safety data sheets, or training from someone who has run the bath before. Buy pre-mixed electrolytes where you can: most labs do, and it removes the mixing step, which is where the 1947 accident began.
Sources
Every number on this page comes from one of these. Where sources disagree the page gives the range and says so rather than picking a winner. Vendor application notes are marked, because they are reliable on their own consumables and not peer reviewed.
Show the sources
Information depth
- A. Winkelmann, "Dynamical simulation of electron backscatter diffraction patterns", J. Microsc. 239, 32 (2010): the 10 to 40 nm range at 20 kV, and the molybdenum simulation where 80% of the intensity comes from the first 10 nm.
- A. Winkelmann, T. B. Britton, G. Nolze, Phys. Rev. B 99, 064115 (2019): mean excitation depths of 13 nm in silicon at 15 keV and 6 nm in BaFe2As2 at 20 keV.
- R. A. Schwarzer, "Backscatter and transmission Kikuchi diffraction for materials science", after J. R. Michael and R. P. Goehner (1994): about 100 nm for aluminium, 20 nm for nickel and 10 nm for gold at 40 kV and 20° incidence.
- M. Wisniewski and co-workers, Ultramicroscopy 173, 1 (2017): orientations still measurable through about 116 nm of amorphous silicon at 30 kV, and the distinction between a core information depth and a maximum one.
- S.-H. Chen, J.-C. Kuo, C.-L. Wu, Ultramicroscopy 111, 1488 (2011): copper patterns lost under 1.6 nm of amorphous overlayer at 5 kV and 7.9 nm at 30 kV, on a Hough-peak threshold.
- Oxford Instruments EBSD sample preparation pages: the general statement that diffracted electrons escape from only a few tens of nanometres, the 10 to 50 Pa low-vacuum range, and the coating guidance.
Damage and its removal
- Max Planck Institute, "A step-by-step analysis of the polishing process for tungsten specimens": the 0.5 µm fragmented layer after grinding, the deformation surviving 1 µm diamond, and the 30 minutes of chemo-mechanical polishing needed to clear it.
- T. Everaerts and co-workers, Adv. Mater. Interfaces (2019): up to 300 MPa of residual stress from diamond slurry on titanium, and none from colloidal silica or electropolishing, measured by focused ion beam ring-core milling.
- J. R. Michael, Sandia National Laboratories, "EBSD Sample Preparation": the ion damage table for silicon, the dual-phase steel sequence, the 87% to 97% result, and the note that fracture surfaces give excellent patterns in ceramics.
- M. Nowell, S. Wright, J. Carpenter (EDAX), "EBSD Sample Preparation: Techniques, Tips, and Tricks": the INCONEL 600 sequence and the 4.5% to 99.9% indexing result.
- Buehler TECHNotes Vol 5 Issue 2: band contrast before and after twenty minutes of vibratory polishing across seven materials, and the wax-coated paper trick for zirconium, antimony and vanadium.
- Vendor application note: Thermo Fisher AN0088, aluminium with gallium against xenon, the 1.8 nm damage depth at 2 kV and the 51% against 76% hit rate.
Recipes
- Vendor: Buehler SumMet and TECHNotes (steels, tool steels, aluminium, copper, magnesium, zirconium); Struers application notes and the Metalog Guide (aluminium, titanium, microelectronics, additive manufacturing); Oxford Instruments and Kemet AN012 (aluminium, titanium, brass).
- Facility: LightForm at the University of Manchester (aluminium, magnesium, titanium, and the broad ion beam recipes); Carleton SERC and Bowdoin College (thin sections and charging); MetPrep Application Note 004 (commercially pure aluminium).
- PACE Technologies preparation guides (cermets, ceramics, composites, additive manufacturing): used for sequences where nothing better is published, and marked as such, because their numeric tables are not attributed to a named author and two of their sibling sites publish different values for the same rule.
Electropolishing and safety
- Struers LectroPol-5 manual and application notes for the preset methods; Struers safety data sheets for the A2, A3 and D2 compositions.
- Laboratory Safety in Metallography (University of Waterloo, reproducing the standard ASM guidance): the perchloric acid rules quoted in section 5.
- Cornell University Environment, Health and Safety laboratory safety manual: the washdown fume hood requirement for heated perchloric acid.
- ASTM E1558, Standard Guide for Electrolytic Polishing of Metallographic Specimens: the statements on multiphase surfaces, exaggerated voids and inclusions, and edge rounding.
- Published perchloric-free routes: sodium hydroxide for tungsten (Advances in Manufacturing, 2020); deep eutectic solvents for nickel and cobalt; a non-acid alcohol-based route for titanium validated for EBSD (Materials Characterization, 2020).
What is deliberately not on this page
- Gallium damage layer thicknesses for copper, titanium and nickel superalloys. They are cited in the literature but I could not reach a primary measurement, so no number is given rather than a plausible one.
- A xenon damage layer thickness in nanometres. Only the qualitative result on aluminium at 2 kV is published.
- A single “damage equals N times the abrasive size” rule. Four vendor sources give four different multipliers, and two of them are published by the same company. The measured depths in section 2 are given instead. The book that would settle it is L. E. Samuels, Metallographic Polishing by Mechanical Methods, chapter 5.
- A damage depth in micrometres for an abrasive cut-off wheel or a precision saw. Nobody publishes one. Wire electrical discharge machining is the only cutting method with a measured recast layer, at 3.6 to 12.8 µm on one nickel alloy.