Skip to main content

Reactive ion etching and ICP-RIE: matching the etch tool to the device you are making

Physical and chemical etching, where anisotropy and selectivity trade against each other, why an inductively coupled source is added, and how to pick the right plasma etcher for the devices your group actually builds.

The most common mistake in a first plasma etch purchase is asking for "an RIE" without saying what will be etched. An etcher is not a general purpose machine. The gases it can run, the corrosion resistance of its wetted parts, whether it can clamp and cool a wafer, and whether it has one RF source or two, all follow from the materials on your list. Get that list wrong and the tool arrives, passes its acceptance test on a silicon dioxide coupon, and then cannot touch the material the group actually wants to work on.

This article explains what the plasma is doing, why anisotropy and selectivity pull against each other, what an inductively coupled source adds and what it costs, and how to match the tool to the device.

What is happening inside the chamber

Radio frequency power, conventionally at 13.56 MHz, is applied between two electrodes in a low pressure gas. Electrons are far more mobile than ions, so the powered electrode carrying the wafer charges negative relative to the plasma and develops a DC self-bias. Positive ions are accelerated across that sheath and arrive at the surface travelling nearly straight down. At the same time the plasma cracks the feed gas into reactive neutral radicals, fluorine from SF6 or CF4, chlorine from Cl2 or BCl3, oxygen from O2, which drift to the surface and react chemically.

So two things are happening at once: a directional physical bombardment and a non-directional chemical attack. Everything useful about reactive ion etching comes from the interaction between them.

Physical, chemical, and why you need both

MechanismDirectionalitySelectivityDamageWhere it is used
Pure physical, argon sputter etchStrongly directionalAlmost none, removes everythingHigh, plus redepositionSurface cleaning, hard materials, ion milling
Pure chemical, downstream or barrel plasmaIsotropic, undercuts the maskVery highLowResist stripping, descum, gentle removal
Reactive ion etching, both togetherControllable, near vertical achievableModerate to high, recipe dependentModeratePattern transfer into oxide, nitride, silicon, metals

The synergy is the point. Ion bombardment breaks bonds and clears reaction products and passivation from horizontal surfaces, so radicals attack the floor of a feature much faster than the sidewalls, which are shielded from the vertical ion flux. That difference in rate is what produces a vertical wall. Remove the ions and you get an isotropic chemical etch that undercuts. Remove the chemistry and you get slow, dirty sputtering.

Anisotropy and selectivity pull in opposite directions

Raise the bias to make the walls more vertical and the same energetic ions start sputtering the mask, so selectivity to photoresist falls and the mask erodes before the etch finishes. Lower the pressure and the ions arrive more directionally, because they suffer fewer collisions, but radical density drops and so does the rate. Add oxygen to a fluorocarbon chemistry and you consume the polymer that was protecting the sidewalls, which speeds the etch and rounds the profile.

Three effects that surprise people running their first deep features:

Endpoint detection is what turns a recipe into a process. Optical emission spectroscopy watches a spectral line associated with the film or the product and tells you when it disappears. Laser interferometry counts fringes as the film thins. For a university tool, specify at least optical emission with a usable software interface, because a timed etch on a variable film thickness will over-etch somebody's device sooner or later.

Why an inductively coupled source is added

In a simple capacitively coupled RIE, one generator sets both the plasma density and the ion energy. You cannot make the plasma denser, and therefore the etch faster, without also making the ions more energetic and therefore more damaging. That single knob is the limitation of the whole architecture.

An inductively coupled plasma source solves it by adding a coil that couples power into the gas to generate a dense plasma, while a separate RF supply on the platen sets the wafer bias independently. Two knobs instead of one. You can run a high density plasma with low ion energy, which means a fast etch with little damage, and you can operate at lower pressure, which improves directionality.

What that buys in practice: deep silicon etching and the Bosch process for MEMS, chlorine based etching of gallium arsenide, gallium nitride and indium phosphide, smooth sidewalls for photonic waveguides where roughness turns directly into optical loss, and gentle processing of thin layers and two-dimensional materials. What it costs: a second RF generator and matching network, a more complex gas panel, more cooling, and almost always helium backside cooling with mechanical or electrostatic clamping to keep the wafer from overheating during a long etch.

Matching the tool to the device

What the group makesEtch requirementTool that fits
Resist stripping, descum, surface activationIsotropic, gentle, high selectivityBarrel asher or a basic RIE running oxygen
Patterning oxide and nitride, sub-micron to a few microns deepModerate anisotropy, good mask selectivityCapacitively coupled RIE, fluorocarbon chemistry
MEMS structures, deep trenches, through-wafer featuresHigh aspect ratio, vertical walls, wafer coolingICP-RIE with Bosch or cryogenic capability
Compound semiconductors, GaN, GaAs, InPChlorine chemistry, heated chamber, corrosion resistant hardwareICP-RIE specified for chlorine from the outset
Photonic waveguides in silicon or silicon nitrideVery smooth sidewalls, low damage, low biasICP-RIE with fine bias control and endpoint
Graphene and thin two-dimensional layersVery low energy, short controlled exposureDownstream or remote plasma, or ICP at minimum bias
Patterned metals such as aluminium, titanium, chromiumChlorine chemistry and post-etch passivationICP with chlorine, or redesign the process for lift-off

One decision on that table is far more consequential than the rest. Chlorine capability is a building decision, not an option box. A chlorine tool needs corrosion resistant wetted parts, a heated chamber and pump line to stop involatile products condensing, a toxic gas cabinet with a purge panel, dedicated detection, and abatement rated for the chemistry. Retrofitting chlorine onto a fluorine-only system, in a room that was never designed for it, is not a small job and is often refused by the manufacturer. If compound semiconductors or metal etching are anywhere in the department's five year plan, decide now.

Reproducibility: what students discover in month three

Etch rate in a shared tool is governed as much by the condition of the chamber walls as by the recipe. Polymer deposited by the previous user changes the radical balance for the next one. Well run university labs deal with this by process, not by hope.

  1. Run a chamber clean between users, typically an oxygen plasma for fluorocarbon polymer and a fluorine plasma for silicon residue, and log it.
  2. Run a conditioning or seasoning step before a real sample so the walls reach a repeatable state.
  3. Etch a monitor coupon of known film thickness alongside anything that matters, and measure it.
  4. Standardise the carrier wafer and the exposed area, so loading is constant.
  5. Keep a chamber logbook with user, recipe, gases, duration and any unusual behaviour. In a multi-user lab, the single largest cause of "the tool is broken" is the chemistry left behind by the person before.

Buying and running one in Bangladesh

The Oxford PlasmaPro reactive ion etcher is the platform most Bangladeshi groups will encounter first, and Vvon supplies it here along with Oxford's sputter and PECVD systems. Whatever the make, the same site issues decide whether the tool runs.

Back to all Insights