CL spectrum acquisition on Cameca

[Probeman]

My instrument engineer recently finished an optical adapter on our Cameca Sx100 which allows for normal panchromatic CL imaging using the PMT, but also allows the user to slide it to one side (there are spring detents for proper alignment), for an optical fiber coupling to a inexpensive Ocean Optics spectrometer as seen here:



This visible light spectrometer (USB4000-FL) is not the high sensitivity "back thinned" CCD type, but then again it only costs around $4K compared to around $15K for the back thinned model.  So while the sensitivity isn't as good as it could be, we can now acquire CL spectra along with the WDS and EDS in Probe for EPMA.

Attached below (remember to login to see attachments) are some examples of a number of CL emitters in my Mac standard block.  Note that these are net spectra which means the dark noise has been subtracted.  Also I specified a "boxcar" width of 5 channels to smooth out noisy pixels in the spectrometer CCD.

I think the infrared peaks in the InP and GaAs are pretty nice but I have no experience with CL spectra and how the sensitivity here compares to other CL spectrum acquisition systems which utilize the EPMA light optics.  I would appreciate any suggestions or feedback...
john

[Probeman]

This is kind of cool as it demonstrates the necessity of performing a dark spectrum acquisition.

Here is a raw (not dark spectrum corrected) CL spectrum from the mineral Alamosite (PbSiO3):



Nothing visible right?  Now here is the same spectrum but this time corrected for the dark spectrum:



The conditions were 15 keV, 100 nA and a focused beam.

[Probeman]

Maybe someone can explain something to me.  Attached below (be sure to login to see) are 5 CL net intensity acquisitions (dark spectrum corrected) on InP in which the dark and CL acquisition time together ranged from 5 to 80 seconds (15 keV and 100 nA).

It surprises me that although the background seems to improve somewhat as the integration times increases from 5 to 80 seconds, the emission peak is relatively unchanged. 

I will try the same test on a weaker emission line (for example the PbSiO3), but I wonder what sort of integration times other people use and how exactly does CL integration time affect the spectrum "noise".  I realize that it's a very different animal than x-ray counting but I am also surprised that one can acquire a quite decent CL spectrum in 5 seconds.

I guess I should also try some lower beam currents.

[Probeman]

I measured a number of REE (Smithsonian) phosphates at 15 keV and 30 nA at 5, 20 and 80 sec counting times.  I figured that because the SmPO4 signal is less intense (and using a lower beam current), it might be a better sensitivity test for our "homemade" optical coupling to the fiber. See below (remember to login to see attachments).

So does anyone have CL spectra from this standard to compare?   It seems to me that for materials such as this, the longer integration times do seem to improve the finer spectral details.
john

[qEd]

Because the relaxation times are fast for CL transitions, one would expect a linear increase in response to increasing probe current. Therefore, you may be exciting and piling more charge into a beam induced defect and lowering the probe current is an excellent idea.

On the SEM I typically collect parallel spectra (using a CCD, not a serial spec) with 100s of pA to 20 nA  with acquisition times of 10s - 100s of milliseconds. NB: The dwell time is heavily dependent upon your collection optics, detector, and of course the sample.

Ed V.

[John Donovan]

Because the relaxation times are fast for CL transitions, one would expect a linear increase in response to increasing probe current. Therefore, you may be exciting and piling more charge into a beam induced defect and lowering the probe current is an excellent idea.

On the SEM I typically collect parallel spectra (using a CCD, not a serial spec) with 100s of pA to 20 nA  with acquisition times of 10s - 100s of milliseconds. NB: The dwell time is heavily dependent upon your collection optics, detector, and of course the sample.

Ed V.

Hi Ed,
Yes, I've seen that increasing the scan rate can increase CL output on our SEM. But I wonder if anyone with a CL detector on their EPMA has acquired point spectra of any of these materials to compare the absolute intensities...  I'm mainly interested to know how efficient our "homemade" optical fiber coupling is compared to other systems.
john

[Probeman]

I finally got around to taking a new snapshot of our recently modified CL to fiber optic coupling:



We had to extend the fiber out a bit because it turns out that the cone of light coming out the front of the Cameca instrument isn't quite parallel. So in order to use the coupling lens we already bought, the Al tube had to be lengthened somewhat.

If anyone is interested in getting the drawings of this fiber optic coupling, please let me know and I will scan them and post them here.
john

[Probeman]

Below are the CL spectra that was acquired on a couple more materials (EuPO4 and a natural anhydrite or CaSO4).





It surprises me how complex some of these spectra are in such simple compounds...

[Probeman]

Paul Edwards at Strathclyde University wrote me the following and I post here, with his permission, a comparison measurement on his SX100 system on a sample of GaAs:

Dear John,

Attached is a CL spectrum acquired from a piece of GaAs at 15 kV and 30 nA. The acquisition time was 10 s. (I tried 80 s, but the number of "cosmic ray" spikes during such long acquisitions makes this data pretty unusable.) The spectrum has been background (dark count) corrected, and the intensity is in counts/second/nanometre, to account for the varying spectral width of each channel. I think this should make it more comparable with yours, as I seem to remember that the Ocean Optics detectors give their output in exact 1 nm intervals. I used a 50 µm spectrometer entrance slit with 2× binning of the 1024 pixel detector to get 512 spectral channels, and chose a grating to cover a range of around 440-950 nm, again to try to more closely match your spectrometer.

This sample has been hanging around for a decade or so, so I should probably find a fresh bit to try it on before we make any firm comparison. The same is true for the InP which I looked at, which had extremely bright luminescence (saturating the detector at < 500 ms) but an anomolously short wavelength emission (above the InP band gap!). We will try to source some new material and try it again.

I have a question regarding the "counts per second" scale on your spectra. If you are measuring for 80 s and have a peak of 12000 cps, this means you have an absolute peak of around a million counts. Most detectors (or the ADC attached to them) are 16-bit and would therefore saturate around 65.5 kcps. Are you (or is the Ocean Optics driver) instead integrating a number of shorter acquisitions rather than doing one long exposure?

By the way, I think the broad shorter wavelength band in my spectrum is the filament incandescence reflected off the surface of the sample. We tend to see this with such long acquisition times and when the sample is highly reflective.

Regards,
Paul

See his attached GaAs spectrum (remember to login to see attachments).  We have quite different CL spectrometer hardware so the comparison is not exact, but overall it seems to be in agreement with the GaAs spectrum attached to the first post in this topic.
john

[Changkun]

Hi,

This "homemade" system looks very impressive and much more better than JEOL monoCL system. JEOL monoCL system is "completely" separated from EPMA system.

Since I have never used Cameca EPMA, I can't figure out how you can get both panchromatic CL image and CL spectrum. Do you need to retract the optical microscope (OM) tube in order to put your sliding interface? (This is the way JEOL does for their own monoCL system so that it is impossible to focus using OM on the spot where you want to get CL.)

If not, is it possible to apply your "homemade" CL system (both for imaging and spectrum) to JEOL EPMA (JXA-8530F) possibly into a port on the right of OM port as seen an attached photo)?

Thanks,
Changkun

[John Donovan]

This "homemade" system looks very impressive and much more better than JEOL monoCL system. JEOL monoCL system is "completely" separated from EPMA system.

Since I have never used Cameca EPMA, I can't figure out how you can get both panchromatic CL image and CL spectrum. Do you need to retract the optical microscope (OM) tube in order to put your sliding interface? (This is the way JEOL does for their own monoCL system so that it is impossible to focus using OM on the spot where you want to get CL.)

If not, is it possible to apply your "homemade" CL system (both for imaging and spectrum) to JEOL EPMA (JXA-8530F) possibly into a port on the right of OM port as seen an attached photo)?

Thanks,
Changkun

Hi Changkun,

Great questions. 

On our Cameca instrument seen here:

http://probesoftware.com/smf/index.php?topic=883.msg5626#msg5626

you will note that the CL detectors have a sliding mechanical interface with two detents. One position for normal "mono" or panchromatic analog signal CL imaging using the OEM PMT detector, and another 2nd detent position that allows light from the light optics to be focused to the tip of a fiber optic that connects to our optical spectrometer from Ocean Optics for collecting CL spectra from 350 nm to 1000 nm.  Obviously not at the same time though!   

One could also add a mechanical connection that inserts a parabolic mirror in the side port of the Cameca EPMA instrument (as we have done in our homemade CL system for our SEM), but if we did that on the EPMA then one cannot collect WDS and optical signals at the same time...

http://probesoftware.com/smf/index.php?topic=201.msg3720#msg3720

You might want to see what they did in Singapore on their 8530:

http://probesoftware.com/smf/index.php?topic=517.msg2883#msg2883

Basically it has a separate optical tube which is focused on the beam incident spot, but located as close to the sample as it can be for optical efficiency.  Currently I only have drivers for Ocean Optics spectrometers, so I would suggest that you fabricate (or purchase) an optical unit that can be inserted into the low power optics port (I think that's what your photo shows), and then run a fiber optic from there to a spectrometer.  I bought an Ocean Optics USB-4000 for about $5K a while back, but better spectrometers (back thinned CCD) are available from them also, for more money.

By the way, what the xCLent system does is replace some elements of the JEOL light optics (for better IR transmission) and collect the light for a fiber connection to an Ocean Optics spectrometer. But it is rather expensive...  so if you only need CL spectra from points (no CL spectrum imaging), then an optical unit, fiber and Ocean Optics spectrometer is all you need to automate CL spectrum acquisition with your WDS and EDS in Probe for EPMA.

john

[Probeman]

By the way, these are very crude drawings from my electronics engineer (see pdf attached below), but they might help if anyone wanted to try and replicate our "homemade" dual PMT-fiber optic mechanical mounting shown on our SX100 as seen in the above post:

http://probesoftware.com/smf/index.php?topic=883.msg5626#msg5626

on their own Sx100 or SXFIVE instrument...  let me know if you have further questions on the details.  For example, I do know that x-y stage for adjusting the focused light spot to the fiber tip, is just a commercial product for optical benches. It also has some Z adjustment for the best focus.

We are using this fiber coupled CL spectrometer on several recent probe runs, e.g., apatites and getting nice spectra.  A back-thinned CCD should be even better.
john