Electron%20Microprobe%20Techniques%20for%20use%20in%20Tephrochronological%20Analyses

1
Electron Microprobe Techniques for use in
Tephrochronological Analyses

• J. Fournelle, K. Severin, K.Wallace, J.Beget, J.
  Larsen

Material in Poster V33B-0663 presented at AGU
December 2006 in Tephra Dispersal and
Sedimentation Field Studies, Modeling, and
Hazard Assessment III session
2

• Tephrochronology generally assumes that a layer
  of volcanic ash represents a snapshot of
  eruption/deposition and of a region within the
  subvolcanic magma chamber. Correlation of tephra
  deposits over long distances helps establish age
  control for other deposits (volcanic and
  nonvolcanic). Reliable correlations depend on
  establishing similarity among tephra deposits.
• Although multi-parameter characterization of a
  tephra enhances long-distance correlations,
  identification and correlation of unknown tephras
  is often done using only geochemical analyses.
  Techniques vary but generally deal with
  chemically characterizing all (bulk) or portions
  (glass, crystals) of the tephra layer, with
  various geochemical techniques at various spatial
  scales.
• Electron probe microanalysis (EPMA) is the most
  commonly used analytical tool for geochemical
  analysis and imaging of micron-size volumes of
  glass and crystals. However, a standard method
  for collecting, reducing, and reporting tephra
  data among and within laboratories is not common
  practice, making comparison of data sets
  problematic. There is also a range of
  proficiencies, from advanced to introductory
  some EMP and SEM labs have mature established
  procedures, whereas other labs may have little
  experience with this somewhat specialized work
  (hey, someone comes and asks you if your probe
  will do glass analysis...of course it will).

3
INQUA/Froggatt (1992) Recommendations

• Froggatt (1992), reporting on the 1990 INQUA's
  Committee on Tephro- chronology, listed
  recommendations related to quality control of
  EPMA data
• (3) All EMP data should be accompanied by a
  statement of probe make and model, whether WDS or
  EDS, and accelerating voltage, beam current
  (Faraday or absorbed on some standard), and beam
  diameter. If referenced to earlier work, that
  should be readily available.
• (4) All tephra analyses should be based on common
  working standards, similar to the unknowns. At
  present the best are KN18 and VG99. Also suitable
  minerals are feldspar OR1A, Kakanui Augite, and
  Engels amphibole.
• (5) All published EMP analyses should be
  accompanied by typical analyses, or the mean, and
  other statistics on one or more standards,
  especially for glass data, or reference to where
  these standard data are available
• (6) All EMP analyses of volcanic glass should use
  an electron beam defocussed to at least 10 um
  diameter to reduce potential loss of Na and other
  elements.
• (continued on next slide)

4
INQUA/Froggatt (1992) Recommendations-continued

• (7) Give single shard analysis, rather than
  averaging out heterogeneity.
• (8) All glass EPMA should be of freshly exposed
  (polished) internal surfaces.
• (9) EMP data on glass should be normalized to
  100 to facilitate proper comparison. The
  deficiency from 100 can be recorded as 'water by
  difference' or the original analysis totals
  listed.
• (10) A minimum of 10 analyses on different shards
  should be collected.
• (11) All data should be examined for homogeneity,
  and unusual standard deviations explained.
• (12) Minimally publish mean and std dev of each
  element, after normalization 'water by
  difference' or original analysis totals number
  of separate shard analyses in the mean. Also (3)
  and (5) should be noted.

5
Survey of Recent Tephrochronology Publications
Reporting EPMA Glass Chemical Data
To better understand the extent of EPMA use in
tephrochronology, and how EPMA operation was
documented, a literature survey was done
recently by searching bibliographic databases
with the terms tephra chronology glass.
Around 200 articles were then checked to find
those with published glass analyses by EPMA. 52
articles were found with EPMA data, 39 published
since 2000, and 46 published since 1996 (these
figures are weighed toward those journals with
online pdf files). There are about 2 dozen
laboratories represented, of which 13 are
recognized by name as established EPMA
facilities with dedicated WDS electron probes
(Cameca camebax, sx50/51, sx100 Jeol 733, 8600,
8900 Microscan V), the balance SEM facilities
plus 6 "mystery" labs (no details given in the
publications about their equipment).
6
Journals used for this survey
7
Reporting basic conditions. some well done,
others not so well done
8
Reporting conditions related to element mobility
(e.g. Na) in glasses. some well done, others not
so well done
9
Reporting standards, important for quality control
10
Normalization of tephra glass is somewhat
controversial a good argument has been made that
even if it is done, the original values should be
able to be recovered.
11
Suggestions by Hunt and Hill J.B. Hunt and P.G.
Hill in a series of papers (1993, 1996, 1998)
followed up Froggatt's challenge to the
tephrochrology and EPMA communities. In Tephra
geochemistry a discussion of some persistent
analytical problems (1993, The Holocene 3, p.
271) they agreed with many of Froggatt's
suggestions, but questioned others, e.g., the
acceptance of "low totals" via data
normalization. They also pointed out that it is
not just Na (and K) that are "lost" in EPMA, but
other elements (Al, Si) increase under beam
irradiation. They argued that some North Atlantic
tephra EPMA data are of poor quality and
"normalization to 100" permits erroneous
conclusions as errors in analysis are
allowed. They suggested not using any analysis
with a total lt95, that gt2 wt H2O implies
alteration of a glass, inappropriate operating
conditions, poor sample selection, or poor sample
preparation. (They did not address the question
of the situation where the only glass available
is truly hydrated). They gave a list of 8 EPMA
parameters that should be included in any
publication. They repeat many of those listed by
Froggatt (1992), but add counting time (peak and
background), total analysis time (x-ray
counting), and exposure time (beam on, on the
sample).
12
From Hunt and Hill, 1998
13
Suggestions by Hunt and Hill - continued
In the 1996 paper "An inter-laboratory comparison
of the EPMA of glass geochemistry" (Quaternary
International 34-36, p. 229), they report on a
round robin of 7 northern European EPMA labs,
where a rhyolitic glass (Lipari obsidian) was
analyzed. There is a large amount of scatter of
the plotted data (Na2O vs SiO2). Only 4 probes (2
of them theirs) at 3 labs submitted precise and
accurate data. (The labs were not identified, as
some of the results were clearly erroneous, with
several operator and instrumental issues later
identified, e.g. failure to carbon coat causing
extreme Na enrichment and Si loss, and analysis
of the edge of the material, and standarization
and analytical procedure problems. They observed
a difference between the results of using a ZAF
vs PAP (phi-rho-Z) matrix correction (the ZAF
gives 1 wt lower SiO2). "The magnitude of the
discrepancies seen in the inter-laboratory
programme, if repeated in tephra analysis, may be
sufficient to result in miscorrelation to
volcanic source, in a failure to distinguish
discrete tephras, or in faulty correlation
between tephras, all of which could have
important tephrochronological consequences." (p.
237)
14
Suggestions by Hunt and Hill - continued
"The reliability of published tephra geochemical
data can be gauged by the wider tephra communigty
only by the inclusion of representative
geochemical analyses of an secondary glass
standard acquired during the same analytical
session in which the tephra was analyzed. At the
3rd U.K. tephra meeting in Cheltenham in 1994,
the U.K. tephrochronology community agreed to a
publishing protocol using the Lipari obsidian as
a secondary standard." (p. 237) In
Interlaboratory comparison of electron probe
microanalysis of glass geochemistry (Hunt et al,
1998, ODP Report), the results of another round
robin are given, this time with 5 labs, 3 from
the U.S. (and they are identified). All but one
lab show similar data. One feature, not
discussed, is the slight but clearly evident
shift to higher Na2O (0.3 wt) and lower SiO2
(0.4 wt) by the one lab that used the Nielsen
and Sigurdsson (1981) time series t0
extrapolation (figure in following slide).
15
From Hunt and Hill, 1998
16
Volatile or Mobile Elements... beam size matters
Six papers in over the past 25 years have
considered the problems of mobile elements in
volcanic or experimental glasses under the
electron beam, and suggested procedures for
either reducing/eliminating the need for
correction, or a time series method for
correction.
1981 Nielsen and Sigurdsson (Am. Min.)
demonstrated the exponential decay of Na under a
focused 20 nA beam on anhydrous rhyolite (KN18,
see figure to right), and showed that a time
series of 2 second counts count be modelled to
yield the Na value at t0. They concluded that
this reduced the errors greatly, but not
totally--the result was 1-3 low (relative).
17

• 1995 Devine et al (Am. Min.) evaluated hydrous
  Si-rich glasses using this time series technique
  and showed that Na and K could be corrected well
  and yielded totals whose deficiency were very
  close to known H2O contents (up to 6 wt). They
  noted that defocussing the beam to 15 microns
  eliminated almost all of the Na loss. Also they
  noted that Na loss in basaltic glass under the
  (defocussed?) beam is neglible.
• Re this time series extrapolation method neither
  Jeol or Cameca offer this as a software option
  currently the only commercially available option
  is in the Probe for Windows software (where it is
  applied before the matrix correction.) Thus
  almost everyone uses the defocused beam method.

18
1995 Spray and Rae (Can. Min.) presented a
review of the physics of the phenomenon (i.e.,
heating and development of a space-charge layer),
and pointed out a study (unclear what the glass
composition was) where low current (0.2 nA),
defocussed beam and high keV (esp. 30 keV)
yielded virtually no drop in Na counts. They
showed Na drop in basaltic glass VG2 with a fixed
beam at 2.5 nA (15 keV), but none when defocussed
to 20 um.
19
1996 Morgan and London (Am. Min.) also
considered hydrous Si-rich glasses, quantifying
the effect on Na, K, Al and Si. Showed that for
currents gt5 nA, the time series extrapolation to
t0 does not retrieve counts lost in the first
instants of irradiation. They concluded that
optimal EPMA acquisition of Na, K, Al and Si was
with as low current (e.g. 2 nA) and beam diameter
the largest possible (15-20 um) with long
counting times (20-40 sec) to improve statistics.
The remaining elements could be measured with a
second pass at 20 nA beam. For hydrous glasses
where less than a 15 um beam is used, similar
treatment will be calibrated on a glass standard
of similar H2O content, and a correction factor
determined and applied.
20
2001 Hunt and Hill (J. Quat. Sci.) (figure
below) evaluated different beam sizes and
concluded square beam rastering (10-15 um width)
was their preferred technique, using 20 keV and
15 or 20 nA to probe dacitic glasses.
21
2005 Morgan and London (Am. Min.) addressed
the problem of small glass inclusions, evaluating
2-50 nA currents, 2-20 um beam diameters, at 20
keV. They recommended using 2 nA and 5 um
beam size, counting for 30 seconds for improved
statistics. For lt5 um, correction factors (on
standards) are determined and applied. They
compared element migration effects per current
density (nA/microns squared) as shown in their
Table 2 below, and Figure 3C. They also addressed
the issue of mixing various beam size conditions
in intra- and inter-sample comparisons,
recommending that only one size, the largest spot
size that can be accommodated by all
samples/inclusions, so that all are treated
similarly.
22
Glass Standards in EPMA
A critical component of a good tephra electron
probe analysis is a good standard. What makes a
standard good? - It is homogeneous and
accurately characterized -- and optimally similar
in composition to the unknown. One concern
must always be the presence of crystals within
the standards analytical volume (upper 2
microns of sample) isolated crystals can be
bypassed, but pervasive devitrification can be
present in materials distributed as official
standards and need to be checked for.
Above some commonly distributed natural glass
standards not saying that all fragments are
like these, but you dont know until you look
23
There is now an additional source of glass
standards for EPMA the USGS for the past 3
years has been developing and testing several
remelted natural basalts. At least one glass
(BIR-1G) is available from Stephen Wilson, USGS,
Denver (swilson_at_usgs.gov) An added benefit of
these compared to the original USNM glasses is
that they are large in size. Some (many?)
EPMA users assume that a chemical analysis
distributed with a standard is the actual
composition of the grains they have. While this
may be true for a percentage of the cases, in
general, one should assume this is the mean value
of a subset of the standards distributed -- a
good initial 1st approximation -- but then the
lab needs to verify that, usually using a suite
of other standards and checking for consistent
results. Jarosewich et al (1980) state that
any particular grain may not be the same
composition as the published value. There are
homogeneity index (HI) values listed (measured
standard deviations divided by 1 sigma of
counts), and most major elements are low (lt1.5)
where they say gt3 is hetereogeneous. The most
heterogeneous of the Yellowstone VG568 glass had
an HI for Na of 3.45.
24
Evaluating EPMA Standards
Part of general EPMA quality control is the
verification that the mineral and glass
(particularly natural) standards are actually the
composition that they are supposed to be (they
are not always!). The only real way to do this is
to check for internal consistency amongst a suite
of standards. One can use one to measure another
to measure another etc. An alternative is to plot
them all up together, with their nomination
compositions vs their actual x-ray intensity
times the ZAF of that element in that matrix. If
all the standards were perfect, they would all
line up on a straight line. Those falling off the
line have a problem, and probably are of a
somewhat different composition than the
official composition. In a few cases (e.g. Al,
Mg) peak shifts may also be involved if minerals
are also used (and not peak centered
individually). To the right are plots from
running the program Evaluate (part of the Probe
for Windows-Enterprise software). Std 358 needs
attention.
25
Other EPMA quality control issues

• WDS
• P overlap by 2nd order Ca K?
• EDS
• This deserves a more systematic treatment than
  can be done here. While EDS systems are easy to
  use, there are many pitfalls particularly to the
  unwary. And never standardless.

26
In basalts this can be a significant amount of
measured P! It matters whether you measure with
TAP or with PET. And it is rarely mentioned in
EPMA result writeups.
Above left, with TAP, on P-free K412 glass (with
15.3 wt CaO), the apparent P2O5 from the 2nd
order Ca K? overlap is 0.3 wt in ordinary
differential mode PHA. Differential mode PHA
removes it. With PET (right) there is sufficient
separation from the P Ka peak so no PHA is
necessary.
27
Proficiency Testing
The International Association of Geoanalysts has
been running a series of EPMA proficiency tests
over the past 3 years, using a series of new
basaltic glass standards being developed by the
USGS. Apparently the first one was a pay to
play operation, but for the past 2 tests, they
are open to anyone. The most recent one, G3, has
just been completed. These are an excellent
opportunity to evaluate ones EPMA operation. If
you are interested in being included in the next
round, contact Stephen Wilson (swilson_at_usgs.gov)
and Phil Potts (p.j.potts_at_open.ac.uk)
28
Communication
One general observation that developed in the
course of studying the literature on
tephrochronology and glass EPMA is the uneven
level of communication between various players
in the game. It is understandable, people move in
different circles and publish in different
journals, but for a field as small as EPMA, one
would think there would be more exchange. For
example, Hunt and Hill (2001) apparently did not
know of Morgan and London (1996), and Morgan and
London (2005) did not know of Hunt and Hill
(2001) -- all which dealt directly with the same
topic. And Hunt and Hill (2001) apparently were
unaware of the physics modelled in Spray and
Raes 1995 paper that could have improved their
paper.
29

• Points for Further Discussion
• 1. Standard composition verification Need for
  critical evaluation of ones own specific glass
  standards used for EPMA. Cant assume ones grains
  are necessarily the same as published official
  values
• 2. Publish standards EPMA as a check on
  procedure Need for including EPMA measurements
  of secondary glass EPMA standards together with
  the unknown tephra analyses
• 3. Volatile/mobile elements
• Can some optimal settings be agreed upon (keV,
  nA, defocussed beam diameter) for tephra glass?
  Basaltic vs rhyo-dacitic?
• How best to deal with situations where only thin
  glass walls are available for probing, so beam
  defocusing is not possible, and software for time
  series extrapolation is not available.
• 4. Low total validation Possibility of verifying
  that low totals are due to hydration and not due
  to analytical error by measuring oxygen, where 45
  or 60 LSM (pseudocrystals) are available. See
  Nash (1992)
• 5. Proficiency testing/round robins
  Participation needs to be encouraged,
  particularly of those organized by the
  International Association of Geoanalysts (free!)
• 6. Consider implementation of interference
  correction on P K? in basaltic glasses by 2nd
  order Ca K? peak (can be a significant amount of
  the P)

30
To be continued .