Cornerstone guide · Technique

Sputter Coating: Metal, Carbon, and How Thick Is Too Thick

The coater is the one instrument in the lab where everyone has a private recipe and nobody has a thickness monitor. Here is the reasoning behind the recipe.

1Why coat — and why not

A conductive coat does three jobs. It gives surface charge a path to ground, so insulators stop deflecting the beam and glowing at the edges. It raises secondary-electron yield, so the image gets brighter and cleaner at the same beam current. And it takes the beam's heat and dose in the top few nanometers, which spares delicate polymers and biological material some of the damage they would otherwise absorb directly.

It also costs you three things. A coating is a layer of foreign material sitting on top of the surface you wanted to see; at high enough magnification you image the coat, not the sample. It adds its own elements to every EDS spectrum taken through it. And it is, for practical purposes, permanent — an archival specimen that is sputtered has been altered. The coat-or-don't-coat decision is therefore made per sample, not per lab: what do you need to learn, and which of those costs can you afford? The sample-preparation guide walks through that decision; this page assumes you've made it and now need to coat well.

2What a sputter coater actually does

A benchtop sputter coater is a small vacuum chamber with a metal target at the top and your stubs on a stage below. Pumped to a few pascals and backfilled with a trickle of argon, it applies a negative potential to the target; the resulting glow discharge accelerates argon ions into it, and each ion knocks loose a handful of target atoms. Those atoms fly off, collide repeatedly with the argon gas on the way down, and arrive at the sample from every direction at thermal energies — which is exactly why sputtered coats wrap around edges and into crevices so well. Most modern coaters use a ring magnet behind the target (a magnetron) to trap electrons near the target surface, which makes the plasma denser and the deposition faster at lower sample heating.

Carbon is the exception: it is almost always evaporated rather than sputtered — a carbon thread or sharpened rod is resistively heated until it flashes — because sputtering carbon is slow and produces a poor-quality film. Evaporation is line-of-sight: atoms travel straight from source to sample, shadows and all, which is why carbon-coating a rough specimen without tilting and rotating it leaves the lee side of every feature uncoated.

3Choosing the metal (or the carbon)

Every coating material trades among four things: how fine its grain is (finer grain stays invisible to higher magnification), how many secondary electrons it throws off (more means brighter images for less dose), how fast and easily it deposits in an ordinary benchtop coater, and which X-ray lines it adds to your EDS spectrum. No material wins all four.

4How thick is too thick

The right thickness is the thinnest film that stops the charging. For most work that lands between 2 and 10 nm of metal; biological specimens and rough topography often want the upper end, polished or flat samples imaged at high magnification the lower. Above about 10–15 nm two things happen: the film's own grain structure becomes the dominant fine texture in the image, and genuinely small features — sub-50 nm pores, fibrils, cracks — are rounded over and filled in. You end up publishing a micrograph of your gold.

The honest problem is that very few benchtop coaters tell you what you deposited. A quartz-crystal thickness monitor is the right answer; without one you are working from a recipe — plasma current, time, pressure, and source-to-sample distance — that was calibrated once, by someone, possibly on a different target. Recipes drift as targets erode and as the argon line acquires leaks. Two sanity checks cost nothing: first, put a scrap of clean glass slide in with every run and look at it against a white page — a film you can see clearly through is thin; one that looks metallic is thick; a gold film that reads blue-green in transmission is in the useful range. Second, image a known feature — a standard, or your own well-characterized sample — at the magnification you care about, and look for grain. If you can see grain, you coated too thick for that magnification (or chose the wrong metal).

Thickness also interacts with kV. At 15–20 kV the primary beam passes through a 5 nm coat almost unimpeded and your secondary signal comes from the sample; at 1–2 kV a 5 nm gold film is a significant fraction of the interaction depth, and the "sample" you image is increasingly the coating. Low-kV work wants thinner, finer-grained coats — iridium, chromium, or a whisper of platinum — or no coat at all.

5Getting an even coat

Sputtering is forgiving but not magical. Tall features still shadow their downstream side, deep holes stay thin at the bottom, and a stub loaded at the edge of the stage sees a different flux than one dead center. The habits that help, in rough order of payoff:

• Tilt and rotate. If the coater has a rotating or planetary stage, use it, every time. If it only tilts, do two half-length runs at opposite tilts rather than one long run flat.
• Pulse rather than bake. Several short bursts with a pause between them deposit the same film with less heating — relevant for polymers, waxes, and anything that melts below a cup of coffee.
• Mind the pressure. Higher argon pressure scatters more and coats more conformally, at the cost of coarser grain; lower pressure gives finer films that shadow more. For rough samples err higher; for flat high-resolution work err lower.
• Keep the argon honest. A leaky line or a nearly empty cylinder admits air; oxygen in the plasma turns chromium purple and makes every metal's film coarser and duller. If coats have quietly gotten worse over months, suspect the gas before the target.
• Ground the sample, not just the stub. A perfect 5 nm coat over an insulating puck still charges if the coat has no path to the stub. Bridge coat to stub with a stripe of paint, or coat with the stub tilted so film runs down the side.

6Carbon for analysis

Carbon is the analytical coat: nearly transparent to X-rays, essentially invisible to backscatter contrast, and a reasonable conductor at 10–25 nm. It is applied by evaporation, so tilt and rotate. Thickness control on an evaporator is by colour: a piece of polished brass placed beside the sample goes from orange to red to indigo to blue as carbon builds, and a blue-indigo brass is around 15–20 nm — enough for most EDS work. Thicker than that and the carbon layer itself starts absorbing the soft X-rays (oxygen, nitrogen, carbon itself) you may be trying to quantify; thinner and rough samples keep charging. For EBSD, where the pattern comes from the top few tens of nanometers, keep it thin — 2–5 nm — or use low vacuum instead. Low-vacuum mode is, in fact, the main reason many labs carbon-coat far less than they used to; it is the subject of an upcoming guide.

7Troubleshooting

8Quick reference

9Further reading

Echlin's Handbook of Sample Preparation for Scanning Electron Microscopy and X-Ray Microanalysis (Springer, 2009) is the standard reference on coating and its alternatives, and is far more practical than its title suggests. Goldstein et al., Scanning Electron Microscopy and X-Ray Microanalysis (4th ed.) covers coating's effect on both imaging and microanalysis. myScope (Microscopy Australia) has a free, structured module on specimen coating that is a good first read for new operators. And the sample-preparation guide on this site is the upstream decision — whether to coat at all.