a) why you don't use less kV?

b) if it is a single crystalline Si....can you use the static background subtraction?

Like Gert said, my first suggestion would be to reduce the voltage down to 10-13 kV in order to reduce the interaction pear.

I am not use that static background will help on several points since (I guess) you will get intensity variations of your backgroung.

Next suggestion is to try to change your substrate with an amorphous one.

One question comes to my mind : how large and thick are these micro- and nanocrystals ? If they are smaller than 50 nm, it will become more and more difficult to do EBSD on them.

Unfortunately, less kV reduces the signal in such a large amount that I become problems with drift due to slower scanning. The electron current is also limited to a maximum of 1nA. Above that I lose the important high resolution.

Thank you for mentioning the background correction. I will try that.

The gold crystals have a size of 1-10µm an initial thickness of 50nm, but I thin them down to 10-20nm by using FIB polishing. Even at 10nm I get very nice patterns.

The silver nanowires have a thickness of 50-100nm and I want to analyze grain boundaries inside them. This becomes really difficult with the constant Si signal.

Another problem is the similar space group of silicon, silver and gold. The EBSD software cannot differentiate between the two phases. This is why I am searching for a better substrate.

The space group is not deciding whether a structure has a similar signal or not. For ni-base superalloys the gamma'-Phase is Pm3m, the gamma-Phase Fm3m. They are so similar that even using XRD they are hard to differenciate. But I understand your point although the argumentation is not correct. Anyway...

I have strong doubts that you still see Si through a 10 or 20nm Au layer. Dingley showed experimental maps using 2..4nm step width for Pt and I believe him. Is your Au or Ag layer perhaps porous so that you observe the signal coming from scattered electrons? Si has a much smaller backscattering coefficient. It would surprize me when this signal can penetrate undestroyed the Au layer. Vice versa it would be more understandable....

regading the background: I am not sure whether it works, but with a scratch on a screen it is similar. There is a constant signal in the background . I admit it only would work if your layer is very homogeneous.

I would try to reduce the kV for single pattern and would look when Si disappears. If it doesn't something else happens.

By the way....the scattering volume of electrons in Si is quite big. Taking into account the investigations of Stefan Zaefferer (2 years ago?) who is claiming that even in steel the physical resolution is not better than 100nm using MC simulation (experimental resolution is perhaps 30..40nm since EBSD systems still can interprete pattern overlaying) for Si the spatial resolution I wouldn't expect better than 60...100nm.

Dear Gert,

thank you for your help! I know my knowledge about EBSD is not so extensive like yours, so I apoligize for mixing things up or being wrong with details.

I think I need to clear some things up, or rephrase my question. It is right that the Si signal doesn't come though a very thin gold layer. My problem is, that I write fine nanoantennas into these layers and perform EBSD on them. In the holes, where all Gold is etched away, I have see the Si pattern of course. When I scan several Ag nanowires, I also want to perform statistics on the grains and on the grain boundaries inside the wires. Unfortunately, the software interprets the Si signal beside the wires also as Silver and thus I get a large grain everywhere where no wires are lying. Furthermore I get fake grain boundaries which ruin these statistics.

With the gold crystals I can scan so fast, that it indexes the gold, but not the silicon. With the silver wires it is not so easy.

And this brings me back to my initial question: Is there an alternative substrate known or do you know some methods to have the crystalline material just lying on an conductive, amorphous layer? If that is not an option, I accept to rethink my measurement methods. ;)

Dear Björn,

there is still another option one could try: using the BSE contrast. You should be able to differenciate between Si, Ag and Au very easily by the raw signal, i.e. how many BSE will be "reflected". It is quite challanging in your case because of the topography, so that I would need a test map to process the data, i.e. the original patterns saved. I don't need any other information, only the saved bcf file. Write it on a DVD and send it to me.

I still had another idea yesterday. Did you used already 80x60 pixel images or the 160x120? If not try this! So you can measure clearly faster (factor of 4) or your signal becomes 4 times more intense.

Hi,

My two suggestions: (a) to oxidize the silicon (b) silicon pattern is constant while the grains change orientation. You may try pattern substraction. Let me know when you are successful.

Best,

Zahava Barkay

Hi,

I had some time the last days to try some suggestions. First the negative results: Background (or pattern) subtraction does not work, because I will get some negative patterns in the areas with thicker metal particles. The reduction of voltage also did not help as the scanning speed reduced so much, that drift became a serious problem.

The solution was quite surprising: increasing the voltage! I used 25kV which allowed such a high acquisition (>200fps) that the nanostructures were indexed well, but the silicon pattern was not seen. Furthermore the drift was no problem anymore.

Regarding a new substrate type, Zahava Barkay gave me an idea. Oxidized Si may be not a good choice in terms of conductivity, but highly doped amorphous Si could be an option. I will try that soon.

I attached one of the measurements, if you are interested.

The map looks impressive...and I see, you still have a eFlash1000 and no HR.

Maybe it is a good idea to present it in Freiberg?