Material compatibility, rate and uniformity, contamination from the crucible, and why the poor step coverage of an evaporator is the reason lift-off works. Also the cases where neither evaporator is the right tool.
A group buys an evaporator to put gold contacts on their devices, and for the first year it does exactly that. Then a student needs a titanium adhesion layer, another needs platinum, and somebody wants a silicon dioxide passivation layer without buying a second tool. At that point the choice made at purchase becomes the limit on what the group can publish.
Thermal and electron beam evaporation look similar from outside the chamber. Both produce a vapour from a source at the bottom and condense it on a substrate at the top, in a line-of-sight geometry, at high vacuum. What differs is how the source gets hot, and that difference decides almost everything else.
In thermal evaporation the charge sits in a resistively heated boat, basket or coil, usually tungsten, molybdenum or tantalum, sometimes alumina coated. Current through the boat heats the boat, and the boat heats the charge. Everything is limited by what the boat itself can survive.
In electron beam evaporation a filament outside the line of sight emits electrons that are bent by a magnetic field into the surface of a charge sitting in a water-cooled copper hearth, often with a crucible liner. The energy goes into a small spot on the charge, so the charge can be far hotter than the hearth around it. That is the whole trick, and it is why an electron beam source reaches materials a boat never will.
Before comparing anything else, write down every material the group expects to deposit over the life of the tool, including the ones that are somebody's idea rather than a current project. Then check each against the source.
| Material group | Thermal boat | Electron beam |
|---|---|---|
| Gold, silver, aluminium, indium, tin | Straightforward, the standard use case | Works, but is usually more machine than the job needs |
| Chromium | Sublimes readily from a coated rod or basket | Works well |
| Titanium, nickel, palladium, platinum | Attacks or alloys with common boat materials, boat life is short and contamination is likely | The normal choice |
| Tungsten, molybdenum, tantalum | Not practical, the boat melts before the charge evaporates | Practical |
| Oxides and fluorides such as silicon dioxide, aluminium oxide, magnesium fluoride | Not practical for most | Practical with the right liner, and the usual route for optical coatings |
| Organics and low temperature materials | Well suited, gentle and controllable | Possible but the radiation load is a risk |
Aluminium deserves a specific warning. It wets and alloys with tungsten, so a plain tungsten boat has a short and unpredictable life and the film picks up tungsten. Use a coated boat, expect to replace it often, and treat the boat as a running cost rather than an accessory.
Both source types are controlled by a quartz crystal monitor, and the monitor is where most thickness errors come from rather than the source.
Uniformity across a substrate is set by geometry. The flux from a small source falls off with angle and with distance, so a substrate held closer gets a thicker film in the centre. Moving the substrate further away flattens the profile at the cost of rate and of material utilisation, which matters when the charge is platinum. Substrate rotation, a planetary fixture and a correction mask are the three tools available, and a planetary fixture is the one most often left out of a research tool and most often wanted later.
An evaporated film picks up whatever else is in the chamber, and the sources of that are predictable.
One useful effect works in your favour: titanium getters. Depositing titanium improves the chamber vacuum while it runs, which is part of why a titanium adhesion layer deposited immediately before gold gives a cleaner interface than the same two films deposited in separate runs.
Evaporation is directional. Atoms arrive from a small source along a straight path, so vertical sidewalls receive very little material and the film breaks at a step. For a conformal coating that is a defect. For lift-off it is the entire mechanism.
In a lift-off process the metal on top of the resist and the metal in the opened window need to be discontinuous, so that solvent can reach the resist. A directional source with an undercut resist profile gives exactly that. A conformal process coats the resist sidewall, seals the pattern and turns lift-off into an exercise in ultrasonic agitation and lost devices.
So the questions are: does the process need to cover topography, or does it need to avoid covering it? If contacts are patterned by lift-off, an evaporator is the right tool and its directionality is a feature. If films must coat a trench or a released structure, an evaporator is the wrong tool no matter how good it is.
There are cases where neither evaporator will do the job, and buying one anyway is a slow way to arrive at the same conclusion.
The honest summary for a general purpose materials group is that a single evaporator does not cover the field. If the work is lift-off of noble metal contacts, a thermal system is enough and the money saved is real. If the list includes refractory metals or dielectrics, the source has to be electron beam. If the list includes a nitride, an alloy or a stoichiometric oxide, that is a sputter tool and no evaporator specification will change it.
Where the budget only stretches to one chamber, buy the chamber with spare ports and a feedthrough count you do not need yet. Adding a second source, a substrate heater, an ion source or a load lock to a chamber that was specified with room for them is a straightforward job. Adding them to a chamber that was bought to a minimum specification usually means buying the chamber again.