I have a question about the chemical analysis of an alloyed metal, one
that had such elements as copper, antimony and arsenic added as
hardening agents. My understanding is that if one examines very small
samples, the grain boundaries that accumulate those elements (Cu, Sb,
etc.) could skew the results: in other words, using a small sample
would amplify your chances of mischaracterizing the chemical content of
the overall piece of metal. This is a result of microsegregation.
My question is this: within each sample, would the "spike" in antimony
or copper correlate with each other? If there was a jump in copper
in one sample, should there be a close/promixate jump in antimony in
the same sample, or would their relative accumulation in the grain
boundaries have nothing to do with each other?
Some of this will depend on the elements added, but sampling is and has been
a problem for many years.
There are a number of approaches to gaining a representative sample, some of
which are more effective than others. There are methods that allow for
cleaning the surface of the furnace prior to dipping a (maybe coated)
sampling ladle or a vacuum operated sampler, different approaches to
moulding, and different ways to cut or prepare a face to analyse. These
change depending on the type of metal you are trying to test. Hard metals
like iron and steel are normally prepared with a grinding operation, while
softer metals are often cut on a lathe or milling machine to prevent
smearing of "stuff" (*technical term) from one part of the sample to
There are also a range of matrix effects that occur simply because of the
way the sample has been taken or prepared. For example, MnS inclusions will
occur at varying depths from the face of a sample simply due to the way the
sample has cooled. Reading this sample with say an optical emission
spectrometer (spark machine or spectrograph depending on how old you are!)
will give different readings if varying amounts of metal are removed from
the face of the sample.
For spectrometers, which are the principal machine used in typical foundries
for metal analysis, the samples are normally taken such that a thin sample
that cools rapidly on a copper chill block or water cooled mould allows for
very little crystal or plate or grain or whatever growth, and the sample is
as homogonous as practically possible.
Some additions are particularly problematic, such as lead in free cutting
steel. The Pb form into little balls that provide a chip breaker function
for machining, like on a lathe. Of course, this makes it difficult to obtain
Hope this helps,
Thank you for your response. This is more of an after-the-fact
assessment. What I'm trying to get at is this: in trying to assess
whether or not the "damage has been done", would it be fair to look
for correlations between the elements added (in this specific case,
antimony, copper) in specific grain boundaries? Or am I making an
unwarranted assumption: could antimony accumulate in the grain
boundaries in levels completely unrelated to the accumulation of
copper? If someone came to me and said-- "since there is no
correlation between the alleged spike in antimony and the alleged spike
in copper in each of the given samples, we can place faith in our
analysis"--- would that person be making a fatal assumption?
Bushy Pete wrote:
How has your analysis been done? And how have the samples been taken?
I'm not up to date with SEM or similar techniques, the type of instruments
I'm used to are OES/spark spectrometers which will analyse an area of a
sample some 10mm in diameter and up to about 0.1mm deep, so testing a
particular grain boundary is out of the question.
If you are testing just a particular grain boundary, then a number of test
sites across the sample and averaging the results would seem to be in order.
It's not so much testing the grain boundary as testing small samples.
There aren't a tremendous amount per piece of metal from the same lot.
There are some discrepancies in the readings. There has been a
thought that if a one looks at the key elements over a set of samples,
IF there is no correlation between a rise in antimony and a rise in
copper, then any problems with the readings are not due to
microsegregation but something else. I'm wondering if that premise
is accurate: that mischaracterization due to microsegregation must
imply that the added hardeners (antimony, copper) accumulated in the
grain boundaries in some relation to each other... any increase in
copper must also see an increase in antimony. What does your
experience tell you in that regard? Thanks.
Bushy Pete wrote:
I'll try to stick to my area of "expertise"? I fix spectrometers to stop the
bank manager sending me dirty letters, so I'm not up to date with the
metallurgy, but can give you a bit of knowledge based on my experience.
Samples that are tested by spectrometer in a range of labs are normally
samples with a single face surface larger that 16mm, if you think of a coin
shaped sample, it would normally want to be at least 1 mm thick. This sample
size will cover the hole in the sample table where it is analysed, and
provide ample thickness that will prevent the sample from melting from the
current used to spark the sample such that up to about 10mm by 0,1mm becomes
the sample section that is actually melted and vaporised. (and tested!)
Samples prepared from molten metal are normally chilled as poured so there
is not much chance for the metal to for a complex grain structure, and to
try to make the sample as homogonous as possible. Samples are routinely
tested that have been cut from a larger casting and there may be quite some
difference through the casting, but the sample is taken from a part that may
be higher in alloying constituents than other sections of the casting.
For example, out of a continuous cast 7" diameter 6063 alloy aluminium log,
a 2" thick piece was cut. Then with a hole saw several 2" diameter pieces
were cut and the pieces tested to see if they could be used as setup
samples. The levels of several elements were higher in the pieces cut from
the outer area of the log, and different levels were found in the piece cut
from the centre of the log. There was also difference from one side of the
sample to the other side of the sample, which corresponds to a greater
distance from either the centre (slowest cooling!) and the outer edge
Another example, with a manganese steel sample poured into a paper cup
shaped sand mould, with a bottom diameter around 40mm and a top diameter of
about 60mm and a height of about 70mm, there is significant difference in
reading for both manganese and sulphur from the bottom surface through the
next 20 mm as the sample is re-prepared by grinding more metal off the
bottom test face. This could be referenced back to tests reported in Karl
Slicker's book on Spectrometry. It has a chapter on sampling of metal that
describes this in some detail. This once again comes from the cooling rate
allowing MnS inclusions to occur at different depths. Also note that as some
elements are solidifying at different depths, there is a corresponding
difference in the concentration in other elements that do not exhibit this
behaviour, due to dilution from the elements that do. It's sort of like a
can of worms, if all the big ones fight their way to the top, the little
ones are pushed down to the bottom!
I believe that there is quite some difference in pieces cut from larger
chunks of metal, compared to samples taken from molten metal and cooled such
that this segregation is kept to a minimum. The design of the sample mould
has quite a bearing on the stability of results, and varies from one metal
type to another. Maybe we are talking on different scales, and the
segregation that I can refer to is on a larger scale that the
microsegregation you are talking about, and I am thinking in terms of the
larger levels obtained by averaging several hundred grain boundaries that
are melted and vaporised in my machines and you are thinking on the levels
between any two adjacent grains. The SEM (scanning electron microscope) can
analyse down towards this sort of sample size.
Smaller samples that me be tested regularly range from turnings, filings or
drillings, which are often remelted and recast, ideally under an inert
atmosphere, there are sample remelt machines that can do this. Wires down to
3mm diameter can be held in a chuck and held over the sample stage, there is
reduced stability in the results when thinner wires are tested, but wires
over about 10mm can give quite reasonable results. Foils can be folded a
number of time (and compressed?) to obtain a thick enough sample. Other
chunks can be cast in epoxy, and as long as a sample clamp can make
electrical contact with the sample, normally by leaving part of the sample
exposed at the rear of the epoxy casting, the epoxy will seal the sample
table's hole and keep the argon in. With the exception of the smaller
diameter wires, the same cross-section of the sample is exposed to the
spark, and similar results can be obtained. There may be a bit of playing
around with the spark's electrical conditions, and the length of time taken
to flush the sample chamber before sparking, before good consistent results
are obtained, but this method development helps to keep my bank manager
What sort of metal, and how was it made, are you having problems with, and
how large a piece are you able to obtain from each lot? Is each "lot" from
the same melt, plant, company, or are these "scrap" samples?
There are a lot of reasons for segregation, and I'm happy to let someone
else jump in with their experience, but remember, there are lots times that
you will wish you could get a representative sample when you won't be able
to. However, there are ways that are accepted as suitable to get a
representative sample, that the powers that be, deem to be reasonable
methods, and some of these, well, there are three types of lies: lies, damn
lies and statistics!
Hope this helps,
This question is addressed to you. Since you say you are associated
with OES and more particularly in servicing them. Do you also deal in
used spectrometrs or refurbished ones. I am looking for a very basic
model around 10kUSD to analyse steel and stainless steels inhouse.
Bushy Pete wrote:
Suspect you are stateside! I would love to help you, but I am based near
Brisbane, Australia. I fear the shipping costs, and the cost of getting me
over to install it for you may blow the budget! If I heard the news right
this morning, the "Aussie Peso" is not worth much......
I do aftermarket service, and have worked with ARL spectrometers for over
twenty years, and a number of years before as a tech in the Aussie Army. I
now live on a small farm and have cut a hole in the side of my two bedroom
garage where I currently have two spectrometers in my modified second
bedroom. Out of the two, I could get one (a 34000) set up for steels, but
this could take some time as one is currently set up for aluminium base, and
the other (a 3580) is an ICP for liquid samples which has been substantially
raided for parts. The 34000 is about 600Kg and the 3580 is 860Kg, both are
top heavy machines that want a substantial shipping pallet, and are
regularly dropped in transport as some fork lift drivers don't allow for the
Keep an eye out for a "local" second hand ARL 34000, or 3460 or 3560 model
in "working?" order, and I can help you and your local sparky pull it down
for transport, set it up, and advise on it's use and servicing for the next
twenty years. If it needs extra channels fitted, I can do this, and have a
good range of parts. The earlier (and much cheaper) 34000 model is over
twenty years old, but can still be maintained, the later 3460 and 3560 are
still current instruments and most parts are available for both, direct from
Thermo, which is the company that owns ARL, or from local electrical supply
outlets. There is also a 3360 model that was made as a low budget foundry
machine that would do a good job as well, these could be around your budget,
as would a 34000. The 34xx instruments will go for more at foundry auctions
and so on, depending on if they have new or old series electronics, and what
channel set up they have. A new machine would set you back about US$80,000
to US$100,000, depending on options.
Laboratory instruments in general, are not highly desirable to the majority
of second hand dealers, due to the high "skill levels?" that can be required
to remove and reinstall them. Consequently, they can often go for
ridiculously low prices. If you have the time to wait, you can get bargains.
Try to obtain the paperwork from the lab as well, as the instruments had a
heap of circuit diagrams shipped with them, and there would be notes about
the calibration that may make later work easier. Many will also have a range
of spare parts on-site, and these box lots will also go for a song. If you
are dealing with a foundry that has upgraded with a replacement machine, the
value would depend on their perceived need for a back-up machine, and the
amount of floor space they have.
Your requirement for ferrous metals is common, and there are more
instruments out there for steel base than any other type. Most of these
would not require much modification, if any, to suit your steel and
Problems you may find with the older 34000 machines are often related to
temperature stability, and these can be addressed by having a better room
airconditioner and a bit of insulation, the later ones are more stable. The
room that you use can be cheaply set up, and also makes for a nice warm in
winter and cool in summer office. I have a couple of insulated 20 foot
shipping containers, that will in future, make ideal onsite labs with the
only external requirement being power. They are easy to keep clean and seal
against foundry dust quite well.
Maybe there is someone else that would like to share travel expenses, for a
bit of a look over their lab?
Hope this helps,