[meteorite-list] XRF issues response to Dan Wray

[Impactika at aol.com]

On behalf of Blaine Reed who is not (yet!) a member of the MetList.
Anne M. Black
_http://www.impactika.com/_ (http://www.impactika.com/) 
_IMPACTIKA at aol.com_ (mailto:IMPACTIKA at aol.com) 
President, I.M.C.A. Inc.
_http://www.imca.cc/_ (http://www.imca.cc/) 
 
 
In a message dated 3/15/2011 2:00:19 PM Mountain Daylight Time, 
brmeteorites at yahoo.com writes:
Dan,

I wanted to reach you to talk with you a bit about this, but I have no 
phone number for you, only a mailing address. As I am leaving (tomorrow morning, 
supposedly) to deal with the aftermath of yet another death in the family 
(the third in the past year now), I don’t have time to take to research 
contact info. Also, your comments could be considered rather disparging towards 
my efforts in this field by some (Carl definitely though so, and was the one 
to drop this bomb in my lap) – so I need to state my case publicly anyway. I 
am having Anne send this out for me for me.   

Some of what you say is indeed true (I’ll go into that here in a bit), but 
much of your comment leads me to believe that the machine you are running 
does not have the proper programming in it. 

These things come with a number of analysis “program” options. I believe 
that the “Alloy”(indeed mostly for analysis work on man-made metals, but I 
have had quite good results on iron meteorites using this – more on that 
later) comes from the factory on all of them. The “extra” programs often cost 
substantial extra dollars and many buyers fail to add the ones that are most 
useful for the job they really want to do. The company I got my unit from 
(Innov-X) allows you one free additional program. Any additional programs (of 
which we hope to add a “meteorite identification” program to the list of 
potentials at some point) run $2500 to $3500 each, I believe. The devil is in 
the detail of what extra programs are on the machine being used.

I really hate to make public just what is the proper programming at this 
point, but I really don’t see that I have much choice while addressing your 
comments. 

For irons: “Alloy” is fine! However, it is probably NOT a good idea to let 
the machine go over to the second beam (the one that sees the light 
elements like Aluminum, Magnesium) as this often (but not always) really screws up 
the results. That second beam is really meant to run when you already know 
that the material you are looking at is indeed mostly Al and Mg. These light 
elements are indeed VERY hard for these machines to nail down accurately (in 
ANY programming mode) unless you are indeed looking at a man-made material. 
MANY other elements do indeed interfere with the detection of these! This 
has made my job A LOT harder as I thought that by getting some kind of 
accurate Mg amount might allow me to make an estimate of a faylite number (but I 
would have to have also found a way to “back out” the fresh metallic iron 
from the sample analysis results first). Unfortunately, I will not be able to 
do this.

I don’t understand how you are not getting anything remotely accurate for 
Ni content of irons with your machine (assuming you have an “Alloy” mode). I 
have gotten quite good results from mine. I have consistently (on cut 
surfaces, weathered surfaces give somewhat skewed results  – a bit on that in a 
later note) gotten results shockingly close to actual published results. In 
fact, my results where so accurate (down to even seeing the gallium and 
germanium on the graph in proper peak height relation – those were not turned on 
in the particular alloy program I was using) on a Canyon Diablo end piece I 
had that it prompted Dr. Carleton Moore to come back to the show a few days 
later with a couple other sample to run for him!

I do indeed “look at the graph” from time to time (mostly when I get a 
real screwy number on something). I had Rob Wesel bring me a chunk of slag (it 
had crystal structure in it superficially resembling Widmanstatten lines on 
the cut face). My initial analysis tried to tell me that it had some 24% 
Iridium! (at around $1000/ ounce, that would make it one valuable rock!). 
Knowing this to be highly unlikely, I went “to the graph”. I quickly saw that 
what I really had was arsenic. This element is not turned on for detection in 
the alloy mode, so the thing saw it as iridium, as those two elements do 
share one of their peaks (and the program will automatically assume it is 
looking at material that is consistent for the program being run).

This definitely proves your point of needing to look at the graph if 
something seems amiss (and it is not a bad idea to have a look at it from time to 
time even when things seem to be proper, just to be sure).

Now, stones are another animal. Running one of these (as I ran, for grins, 
a couple samples of Murchison, and Ls) in this mode gives COMPLETELY screwy 
results. Running any rocky material in Alloy mode will indeed leave you with 
EXCTLY what you describe (can merely tell Ni, for example is present but 
the numbers will be meaningless). The program will tend to “fill in” the 
report for the elements it can’t see by drastically increasing the ones it can 
see (so all the metallic elements, even if they only make up a small portion 
of the sample, will add up to 100% in the final report).  This is why I got 
messed up results like Murchison having 60+% iron.

There is also (really commonly on these things from the factory) what is 
called “mining mode”. In this mode the machine will report the elements seen 
(at least the ones “turned on” for reporting) and a number roughly 
coinciding with the size of the graph peak. The results in this mode is often (from 
my goofing around with this mode anyway) relatively close to what the proper 
mode would give and can be used for comparing one type meteorite to another 
(for example, a CK might consistently give a higher reported result or a 
certain element than a CV and both of those may be higher than a CM).  As 
such, you can gain some headway towards identifying different types using this 
mode. HOWEVER – THE ACTUAL NUMBERS GIVEN IN THIS MODE ARE NOT TO BE 
CONSIDERED ACCURATE!!! To get the accurate numbers from those reported in this mode 
would indeed require (just as you describe) a detailed certified lab analysis 
to tweak them to accuracy. For some unknown reason, this mode is OFTEN the “
second program” on these things from the factory (the Colorado Geologic 
Survey had only this and alloy on their machine. I am really thinking that this 
may be how yours was set up as well).

Stone meteorites should be run in what is referred to as “SOILS MODE”. 

It is true that the numbers for elements in concentrations well above 10% 
may not be super accurate though. This mode is really for getting data on the 
things in ppm concentrations. 10ppm = .0001% folks! There are indeed some 
interferences that one (always) has to keep in mind, but these machines 
should give quite accurate results for most stone (or soil) type samples (at 
least for the majority of elements). This MUST be the case, or these things 
would not be allowed to be used for such things as soil contaminate analysis by 
the EPA (a MAJOR uses for these devices). I think the acceptable “tolerances”
 for these uses is that it must be accurate to within 10 or 20% (in other 
words, if it tells you you have .147% Ti in a sample, the certified lab 
results must be between .125% and .170% or your XRF is considered to be out of 
calibration – assuming that is, that you have not overlooked an interference 
issue). 

The MANY meteorites I have run with this thing all come out quite close to 
official lab analysis (for ones I have other lab results for) for most 
elements. I was VERY surprised to see how close the data I have for the dozen or 
so different Lunar anorthosites matched the data the Carl gave me for 
Calcalong Creek (I can only compare to “bulk” analysis data as that is what this 
machine really does. Lots of lab data is for only particular minerals in a 
meteorite and not the bulk). The numbers I am getting for most of the 
elements (I had my machine custom programmed to be looking for elements in 
meteorites that may allow me to sort them by type) have all been, for the most part, 
very consistent and close to what little “bulk analysis” data I have for 
the same type meteorites.

Now, that being said, I do have some problems with crust and weathering. 
The elements in crust have been oxygenated (and some volatilized and removed, 
compared to an internal cut surface), so the numbers get a bit skewed. The 
same goes for a natural weathered surface (those have the additional problem 
that often other things have attached or leached in – like Al, Ca, etc.) I 
am trying to learn (by running hundreds of different samples) how to 
compensate for these issues (and may, ultimately, not be successful in this, 
unfortunately).  

Any way, I have spent a couple valuable hours typing all of this (I have 
dozens of L and H chondrites waiting for analysis, and I need to pack for my 
trip) and need to move on. Some of what you say is indeed true, but much 
leads me to believe that you are attempting to use a machine that is NOT set up 
correctly for the job. One of these days (Denver spring show time perhaps) I 
would love to see what the School of Mines got for an XRF and have a look 
at how is was set for them.

Thanks (everybody) for taking the time to read all of this and give it some 
thought!

Blaine