Cosmogenic isotope
measurement inter-comparison
Marian Scott, Tim Jull
University of Glasgow, University of
Arizona
July 2008
The CRONUS inter-comparison
• To assess comparability of measurements
made by the different laboratories
• To assist in the definition of overall
uncertainty when results from different
laboratories used.
• Carried out through a programme of
comparisons, small number of samples
distributed to the laboratories.
•The quality of the measurement is determined
by the laboratory.
• Sound, reliable, precise and accurate
measurement requires trace-ability to
community agreed reference materials and
standards
Reference materials
• Basic purpose: improvement of
comparability of measurement results
• Possible uses
– for calibration, to demonstrate trace-ability
– for quality control, to verify the performance of
a method
Reference materials
• For calibration, material often artificially
produced so its properties are known with
low uncertainty
• for quality control, material is often ‘real
world’, so that it behaves as similarly as
possible to the samples being measured.
Objectives of TCN within
CRONUS
• To explore the comparability of results from
the different laboratories
– generate consensus values for a range of
reference materials
• to assist laboratories in independently
assessing quality and
• to quantify precision and accuracy
What information can be quantified from an
inter-comparison?
• Accuracy (from known activity samples)
• Laboratory precision (from duplicate samples)
leading ultimately to quantification of
• Measurement uncertainty
• Quality Assurance (QA)
What is QA?
 QA is an early warning system - it is retrospective
and dynamic, based on judgement of the
measurements on backgrounds, standards and
reference samples (including internal laboratory
materials).
 It is the laboratory paper trail, and in-house
checks
Quality issues
 User concerns
 How good are my results? what does the quoted
uncertainty represent?
 Laboratory concerns
 Is my system stable? Are there any sources of
contamination in the laboratory? How do the results
compare to those expected?, ……How good are my
measurements?
QA involves
 Internal checking
 Measurements made on a series of replicate
samples (Polach (1989) noted ‘internal
checking needs suitable quality control and
reference materials’.)
 Monitoring of background, standards, knownage and reference materials
 Independent (external) checking
 Laboratory inter-comparisons
Time line of activities within
CRONUS
• Design the inter-comparison
– Identify suitable samples (and criteria) done
– Agree timescale for results- done- phase one
results due in April 2008
– Define format for reporting results done
• Inform laboratories and ask for expression
of interest to participate (done)
• Distribute samples- phase 1: done
Timeline
• For phase 1
– Distribution July 2007, results returned April
2008
– Archive material for future use
• Location of archive- currently in Arizonasufficient material for at least 10 more
intercomparisons of the same size
– 23 laboratories sent samples
identified inter-calibration
standards
•
Noble gases
there are 2 potential “standards” – the pyroxene
of Schaefer and an EU sample (from Wieler) .
– pyroxene sample distributed
•
AMS standards
prepared and available from Nishizumi- not
distributed as part of CRONUS
identified inter-calibration
samples-distributed July 2007
• (A)ntarctic sample: high in Al-26 and Be-10.
Quartz was separated at U of Vermont, etched 3
times in HF and washed.
Recommended 5g be used. Approx 37g provided
• (N)amibia: a low latitude sample,
recommend 20g be used. approx 75g provided
• For in-situ C-14, same samples provided in glass
vials.
typical format for reporting
results eg Al-26
Mass of sample (quartz) (g) used in the measurement:
AMS Standard used in the measurement
Half-life used
Background material used
measured 26Al/27Al ratio (1 uncertainty) (specify units)
mass (number of atoms of 27Al) in sample
number of atoms 26Al in lab process blank
Participating laboratories
Al, Be, C- USA, UK, France, Switzerland, Germany,
The Netherlands, Sweden, Australia, Canada (24 in
total)
Noble gases: USA, UK, Switzerland, France, Germany
(14 in total)
Results so far in CRONUS
• So far, results from 7 laboratories for samples A
and N, from 2 laboratories for sample P:
laboratories have often reported replicate results
• There is general consensus on half-life used for
Al, but some variability for Be (1.5,1.51, 1.36,
1.37 x 106)
• Standards used
• Be:- NIST SRM4325, Nishiizumi SYD Be01-5-4
• Al:- Z92-0222(PRIME, Purdue), Nishiizumi STD
Al 0143
Results so far in CRONUS
• Sample A,
– Be analysis
– 13 results, coefficient of variation (CV)
(stdev/mean*100%) is 4.92
– Al analysis
– 6 results, CV 7.92%
Results so far in CRONUS
• Sample N (lower Be and Al by approx
factor of 100 than sample A),
– Be analysis
– 15 results, coefficient of variation (CV)
(stdev/mean*100%) is 9.34%
– Al analysis
– 6 results, CV 9.2%%
Potential analysis
• Similar to that used commonly for the C-14 intercomparisons, defining reproducibility but will use
z-scores, defined as standardised deviations from
a consensus value
• for each sample, we define an agreed value
(usually a robust estimate based on all results)
• the z-score is defined as the difference between an
individual result and the robust value standardised
to account for the uncertainty (also based on a
robust estimate).
• properties of Z-scores well understood- used
internationally in proficiency trials.
The error in a measurement
• A single value, which represents the
difference between the measured value and
the true value
• However, for these samples, we do not have
the true or real Al/Be atoms/g, so we use the
inter-comparison to define this value (as a
consensus from the participating
laboratories)
Key properties of measurement
• Accuracy of the measurement refers to the
deviation (difference) from the true value (or
sometimes expected or consensus value)
• Precision refers to the variation (expected or
observed) in a series of replicate measurements
(obtained under identical conditions).
High
precision, low uncertainty
Accuracy and precision
Accurate and
inaccurate and
precise
Accurate and
inaccurate and
imprecise
Offset (years BP)
Evaluation of accuracy
500
400
300
200
100
0
-100
-200
-300
-400
-500
-600
0
10
20
30
40
50
60
70
laboratory identifier
80
90 100
• In FIRI and VIRI,
known-age material is
used to define the
‘true’ age
• The figure over shows
a measure of accuracy
for individual
laboratories
Between laboratory variation
• reproducibility –identical samples, different
laboratories
Reliability and reproducibility
• Repeatability (r) refers to measurements made
under identical conditions in one laboratory,
• Reproducibility (R) refers to measurements made
in different laboratories, under different
conditions.
• Reproducibility is the closeness of agreement
between test results under conditions where the
same method is used in different laboratories.
Reliability and reproducibility
• The reproducibility value R is the value below
which the absolute difference between two single
results obtained under reproducibility conditions
may be expected to lie with probability 0.95.
• A difference larger than R cannot be ascribed to
random fluctuations and would warrant
investigation of possible sources of systematic
differences.
Reproducibility (VIRI phase 1)
100
~95% Confidence Intervals for Age, for Sample D
51
59
37
20
40
60
80
70
66
22
47
52
49
66.1
69
41.1
12
0
18
19
2
1
17
35
43.2
25
7
62.2
46.1
8
24
26
67.1
46.2
43
41
21
65
25.2
9
22.1
27
25.4
16.1
16
25.3
56
23
10
3
48
46
6
62.4
62
46.4
53
4
58
45.1
57
70.1
55
13.1
25.1
67
20
62.1
43.1
31
15
63
46.3
36
45
5
13
44
70.2
69.1
62.3
50
37.2
63.1
54
33
64
71
42
30
37.1
34
32
29
39
41.2
40
11
14
Archaeological Age = 2850-2900
61.2
61
38
28
37.3
2500
3000
Age (Years)
3500
4000
Estimation of r and R
• Model: Y = m + B + e
where Y is the measurement, m is the average activity, B is
the between-laboratory variation and e is the random error.
• B is assumed random and var (B) = 2L
• e is assumed random and for a single laboratory
var(e) = 2W.
• 2W assumed constant for all laboratories, with
average value 2r.
r and R
• The repeatability value r is 2.8 r
• The reproducibility value R is 2.8 R ,
where R = (2L + 2W)
•
2L , 2W and r must all be estimated.
Conclusions
• All measurement is subject to uncertainty, the test
of which is to make replicate measurements
• Inter-laboratory trials provide generic measures of
reproducibility and assessment of laboratory
comparability
• For cosmogenic isotope work, this is still at an
early stage
• CRONUS inter-comparison has archived material
for future use, but for satisfactory characterisation,
we need more results
Actions- for 2008
• Finalise acquisition and preparation of the 2nd
suite of samples.
• Agree a timescale and distribute the samples
(ideally distribute shortly after results of phase 1)
• await results- for phase 2, assuming distributed
September 2008, deadline for results- Feb 2009.
• analysis of results from phase 1- reported by
August 2008.
other potential inter-calibration
samples
• Antarctic sample: 14C, 10Be, 26Al, 21Ne.
• Namibia: 14C, 10Be, 26Al, 21Ne.
• Maine: 14C, 10Be, 26Al, 21Ne.
• NMT basalt: 36Cl.
• Carbonate/sandstone: 36Cl, 10Be, 26Al.
Some other potential materials are:
• Lake Bonneville basalt sample.
• Promontory Point quartzite
• Blank quartz
acknowledgements
• All the participating laboratories, the
sample providers (Paul Bierman and Joerg
Schaefer), NSF for funding.