Non conductive, dry, specimens require to be coated with a thin (5-20nm) layer of metal in order to avoid charge-up image defects and to be correctly visualized.
Metallizers from lab suppliers are very expensive (10.000/25.000€) and out of an enthusiast budget. They are also hard to find second hand, at least within Europe, and still price would be over ~2000€.
After studying plasma deposition process I realized that building one coater by my own would be quite easy and wouldn’t cost too much. On Youtube many videos showing home made magnetron sputters can be found.
Types of sputtering methods.
Two different types of sputtering processes are commonly used for SEM samples:
1) Plasma sputtering
Plasma is generated in a chamber under vacuum by DC or RF power and concentrated onto the target metal to be sputtered by a strong magnetic field. Highly dense and energized gas ions hit the target, causing its atoms/molecules to physically knocked off and sputtered in all directions. A relatively “low” vacuum in the range of 10-100 millitorr is required for this process and Argon is used as heavy gas for target bombardment. Oxigen and Nitrogen work too, but they may react with target material, forming oxides and other compounds that we don’t want to be on a SEM sample. The optimal is to produce a very high vacuum with a diffusion/turbomolecular pump and then fill the chamber with Argon to reach a final pressure of a few millitorrs. Using gold as a target, excludes the chance of reacting with gas ions. Thickness of the sputtered layer can be monitored by using a quartz-oscillator microbalance monitor, which can be found second hand for good deals on Ebay every now and then. If the power supply of the magnetron gun is let through the thickness monitor relay, a desired thickness can be set and the power of the gun will be shut as soon as it’s reached.
This process requires a very high vacuum and thus a high vacuum pump. Oil diffusion pumps have low maintenance, but require a water cooling system and have long starting and shut-down times. Turbomolecular pumps are more delicate and more expensive, but offer low downtimes. In both cases the vacuum system requires more projecting effort than a low vacuum plasma chamber.
The metal we want to cover the sample with is put into a tungsten/molybdenum boat which is held by two electrodes. Once a high enough vacuum is reached inside the chamber, a high current is let through the electrodes and the metal in the boat will soon evaporate, covering everything in the chamber. This method is the way to go if you want to cover a sample with carbon/graphite for EDX analysis or prior to gold sputtering. My SEM can only capture morphological information by seconday electrons, so there is no reason for me to use carbon coatings. This method doesn’t allow as much precise control of the deposited layer thickness as plasma sputtering does.
The vacuum chamber
A glass bell jar of 20dx20h, priced 25€, was used as chamber for the system. A dual stage rotary vacuum pump able to take pressure down to a few millitorr was bought second hand for around 350€ and can be considered the most expensive part of the project.
An oil mist filter was also put on the pump output to reduce dangerous vapours diffusion.
The basement of the chamber was made from a square 30cm, 10mm thick, aluminium sheet. Two motor spark plugs have been used as cheap high voltage feedthroughs and a couple of needle valves allow slow venting and Argon inlet.
A 5lt Argon cylinder was bought for the purpose off Ebay for 120€.
Venting and Argon inlet will be regulated by solenoid valves in the final setup.
The initial setup was composed by power sopply and two aluminum square electrodes.
The purpose was to check power supply and see if I could generate plasma before developing the magnetron gun. Edges and uneven surfaces cause a lot of sparkling in the plasma and need to be limited or eliminated. Check this video.
Powering up the plasma
Two electrodes, anode and cathode, are required to ignite plasma in a vacuum chamber.
Depending on pressure, their shape and distance, gas composition and other factors, a variable high voltage potential is needed to start the plasma. Once it’s turned up, voltage (and current) can be lowered down as sustaining power is lower than starting power.
Building a variable high voltage DC power supply isn’t really hard. Salvaging a transformer and condenser from a microwave oven will give us all we need for the task.
So I bought a second hand microwave oven for 30€ and got all I needed. I just had to buy 4 high voltage diodes for building the diode bridge.
A 3Amp variac from Ebay (around 50€) is all you need for output voltage regulation. Ammeter and voltmeter on the output of the diode bridge are a plus and not really needed. Ammeter might be more helpful than voltmeter in order to monitor and set the current.
The operating voltage ranges from 600V to 1200V and current from 40mA to 100mA.
The plasma magnetron gun
As last step I developed the plasma gun which was then machined on a lathe by Antonio Possemato, a member of my Microscopical Association, “Amici del Microscopio”.
In the last picture you can see the erosion ring on the copper target due to plasma concentration in the magnetic field generated by strong neodimium magnets.
The magnets setup is as follows (image from DirectVacuum.com)
You can check my plasma gun in action in this short video.
So far, I’ve been using a copper plate as target for tests. This will be replaced with a gold foil once the setup will be perfected.
Here you can see a deposited film of copper on a microscope slide.
I also own a thickness monitor microbalance (150€ second hand off Ebay), which will be implemented soon into the system. Also the samples will be held by a tilted, rotating stage under the gun in order to obtain an as even as possible layer of metal on them.
So far the cost of the machine has been around 800€. A lot lower than any other commercial system on the market.
These are the first images recorded with my DSLR of samples copper sputtered with my coater: diatoms/radiolarians and a butterfly.
I’ve finally got a gold target! A 12 micron sheet of 24K gold.
The gun is finally fully operative and first sputtering went extremely well. I’ve just realized that sputtering gold takes a lot lower current (aprox 40 mA) and shorter time, like 1-2 minutes. Still need to find out what’s the best sputtering time while waiting for the quartz microbalance to be ready.