Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/8494068

Analytical strategies to confirm Scotch whisky


authenticity. Part II: Mobile brand
authentication

Article in The Analyst · August 2004


DOI: 10.1039/b403068k · Source: PubMed

CITATIONS READS

38 413

2 authors, including:

Ross Aylott
Aylott Scientific
21 PUBLICATIONS 194 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ross Aylott on 17 June 2014.

The user has requested enhancement of the downloaded file.


Analytical strategies to confirm Scotch whisky authenticity.

Part II:† Mobile brand authentication


W. M. MacKenzie and R. I. Aylott
Diageo PLC, Brand Technical Centre, Menstrie, Clackmannanshire, Scotland, UK FK11 7ES

Received 1st March 2004, Accepted 5th May 2004


First published as an Advance Article on the web 8th June 2004

Whilst the authentication of Scotch whisky brands using laboratory based gas chromatography is well established, there
is a need for authentication methods for use under field test conditions. The UV/visible absorbance spectra of specific
brands of Scotch whisky were found to produce consistent absorbance ranges which allowed the development of a faster,
cheaper and more mobile test method. Test samples with spectra outside the acceptable ranges for genuine reference
samples were therefore deemed suspect, enabling their authenticity to be later confirmed in the laboratory by GC. This
novel application allowed brand authenticity analyses to be undertaken at remote sites where gas chromatography was
unavailable and has also allowed the introduction of field testing using a small portable hand held spectrophotometer.

Introduction Experimental
Absorbance spectra of test samples were collected using two
As part of its worldwide popularity, blended Scotch whiskies can be
different spectrophotometers, one laboratory based and one
subject to illegal counterfeiting and substitution. Consumer protec-
portable. Absorbance data at set wavelengths from both instru-
tion agencies and producers require analytical laboratories to
ments were transferred electronically via RS232 interfaces into
support their consumer and brand protection programmes with
Microsoft® Excel spreadsheets.
effective brand authenticity analyses.
Initial laboratory development was conducted on a WPA
Current authenticity analysis is based upon the gas chromato-
graphic determination of the major volatile congeners present in Lightwave® S2000 UV/visible diode-array spectrophotometer
(Walden Precision Apparatus Ltd., The Old Station, Linton,
Scotch whisky. Normal concentration ranges are first set for
Cambridge, UK CB1 6NW) fitted with deuterium and tungsten
genuine product, against which the results for suspect samples may
lamps. Samples were introduced by drawing liquid up into a 1 mm-
be compared. Suspect samples are then analysed for alcoholic
path length flow cell using a 5 ml plastic syringe. Absorbance data
strength and major volatile congeners. If the suspect sample’s
was collected at 1 nm intervals between 200 and 500 nm and
alcoholic strength and congener results fall within the normal
transferred into the spreadsheet using WPA Lightwave® software.
ranges for the genuine brand, then the sample is considered
authentic. If the results are outside the normal ranges, the sample is At the start of operation, the spectrophotometer was referenced
against 40% vol. ethanol in water.
not authentic. These analyses may then form part of the evidence
The development of mobile authentication was conducted on a
used in the prosecution of offenders.1,2
However, investigations are enhanced when supported by fast SAD Brand Authenticator® (Spectroscopic and Analytical De-
vices, West Grinstead, West Sussex, UK RH13 8LH. http://
and reliable forensic information, ideally based on field tests.
www.spectroscopic.co.uk). This small novel instrument was de-
Whilst the current gas chromatographic approach provides accurate
signed and built specifically by SAD for the purpose of enabling
and precise results which enable reliable conclusions, it can take
authenticity analysis to be conducted under field conditions. The
significant time and cost to locate and utilise local laboratories with
instrument measures 27 cm long 3 9 cm wide 3 10 cm deep,
appropriate equipment and expertise. Spectroscopic techniques
weighs 1.75 kg and is shown in Fig. 1. It is powered by a 20 V
were therefore considered as an alternative means of authenticating
whiskies. rechargeable battery that enables approximately 2 h continuous use.
UV and visible absorbency in Scotch whisky is primarily derived
from three sources. First, the Scotch whisky maturation process
results in the addition of a range of congeners, which are extracted
from the oak cask into liquid through the interaction of ethanol with
wood lignin. Secondly, volatile phenolic compounds are found in
whiskies where peat is used in the kiln drying of the malted barley
used in the fermentation. The spectroscopic properties of the
maturation and phenolic congeners are employed in their high
performance liquid chromatographic determination with UV and
fluorescence detection. Thirdly, batch to batch colour consistency
is often maintained through the addition of trace spirit caramel
(E150a). Indeed, the colour consistency of Scotch whisky is
DOI: 10.1039/b403068k

monitored at the blending and bottling stage by visible spectros-


copy typically around 500nm.3 Maturation congeners in brandies
have been shown to result in similar spectra.4

Fig. 1 Photograph of a test sample being injected into the SAD Brand
† For Part 1 see ref. 1. Authenticator®.

This journal is © The Royal Society of Chemistry 2004 Analyst, 2004, 129, 607–612 607
The Brand Authenticator® was fitted with a deuterium lamp and a represent blend compositions containing increasing proportions of
diode array detector. Samples were introduced by either injecting or malt whiskies. Compositions may be shown from their 2- and
drawing liquids into the 1 mm-path flow cell using a plastic syringe. 3-methylbutanol concentrations, congeners primarily associated
Absorbance data was collected at defined intervals at approx- with the malt whisky contents of blended whiskies.3
imately 1 nm resolution between 200 and 460 nm. Four patents The spectra for single samples taken from 10 different batches
have been applied for covering various features of this instrument, representing five different brands (labelled A to E) produced over
which is designed to be simple to operate in stand-alone mode a two-year period are shown in Fig. 2. There were clear separations
controlled by internal software. There are five control keys, a two- between the spectra of certain brands, particularly brands A, B, C
line text screen and three result lights. Reference data for genuine and E. There was partial overlap between others (such as brands C
brands is uploaded to the authenticator via its electronic interface. and D) particularly at higher wavelengths. Whiskies of known
Similarly, result data from field measurements may be downloaded authenticity were then analysed in a similar way. Fig. 3 shows that
to the result spreadsheet for onward transmission and critical a sample of genuine brand B (sample T1) fitted, as expected, within
examination at a central laboratory. the minimum and maximum ranges for that brand, whereas four
A process for drawing conclusions on sample authenticity was non-authentic samples (samples T2–T5), which were known
devised. An algorithm was constructed based on the sum of squared counterfeits, each had spectra outside these ranges.
scores (S z2) for absorbance results at 10 pre-selected wavelengths The principles of this spectroscopic method were then trans-
ranging from 230 to 500 nm for the WPA Lightwave® S2000 and ferred from the laboratory WPA Lightwave® spectrophotometer to
between 220 and 360 nm for the SAD Brand Authenticator®. The the portable SAD Brand Authenticator®. Brand range data were
absorbance of the test sample and the average and standard generated on the Brand Authenticator® in the same way as that
deviation of at least 10 reference samples were entered into the used for the laboratory instrument. Fig. 4 shows equivalent spectra
calculation. The z-score5 was calculated from z = (x 2 X)/s for for 10 samples of each of the five Scotch whisky brands A to E.
absorbance data at 10 wavelengths chosen to reflect changes in the Minor differences in the absorbencies at higher wavelengths
slope of the spectral curve, where x is the determined value, X is the between the two instruments (Figs. 2 and 4) were thought to arise
average of the reference values and s is the standard deviation of from secondary transmissions caused by the higher excitation
the reference values. For example, the Brand Authenticator® used energy used in the Brand Authenticator®. This highlighted the
wavelengths at 220, 230, 240, 250, 260, 280, 300, 320, 340 and necessity to establish brand ranges on each model of instrument
360nm. The formula for the z score at each given wavelength (l used at this stage of method development.
nm) was therefore Analytical precision was investigated on the Brand Authentica-
tor® over the range 220 to 360nm. The repeatability of 10
measurements on one sample of brand C (expressed as ±2s/mean 3
100%) ranged from ±0.32% at 260nm to ±1.43% at 360nm (Table
1). The equivalent confidence limits for brand authenticity analysis
by gas chromatography were approximated double the spectral
Finally, S z2 = Sum of squared scores (z(l)) at each monitored
limits and ranged from ±1.6% (for methanol analysis) to ±2.7% (for
wavelength.
2-methyl butanol analysis).1 The spectral confidence limits for 10
The same test samples were subjected to brand authenticity
production batches of brand C analysed once ranged from ±9.8 to
analysis by gas chromatographic methods previously described.1
±10.9%, indicating that variation in instrument repeatability was
Methanol, n-propanol, isobutanol and the combined value for 2-
very small compared to batch to batch variation.
and 3-methyl butanol were determined. For the gas chromato-
The authenticities of a series of samples were then assessed both
graphic process a similar algorithm was applied to each of the 4
by the spectroscopic method on the Brand Authenticator® and by
congeners.
gas chromatography. Reference samples of the same five brands
were analysed for their major volatile congeners. Minimum and
Results maximum range data, shown in Table 2, indicated that similar
Initial stages of this spectroscopic method development determined analytical overlap of certain congeners used in authenticity
the consistency and specificity of the spectral characteristics of a assessment occurs and therefore specific congeners, such as 2- and
range of Scotch whisky brands. The brands were selected to 3-methyl butanols, are particularly important in reaching au-

Fig. 2 Absorbance spectra of 10 samples from different production batches of five different brands of Scotch whisky collected using the WPA Lightwave®
S2000 spectrophotometer.

608 Analyst, 2004, 129, 607–612


Fig. 3 Absorbance spectra of five samples of known authenticity (one genuine brand B [T1] and four counterfeits [T2–T5]) shown against minima and
maxima of genuine brand B collected using the WPA Lightwave® S2000 spectrophotometer.

Fig. 4 Absorbance spectra of 10 samples from different production batches of five different brands of Scotch whisky collected using the SAD Brand
Authenticator®.

Table 1 Precision of spectral analyses for brand C using the Brand Authenticator®. The repeatability of ten measurements on one sample and the results
obtained from single analyses on ten production batches of brand C

Wavelength/ Confidence Confidence 2s/


nm Absorbance range Mean limit (2s) 2s/mean,% Absorbance range Mean limit (2s) mean,%

Ten measurements on one sample of brand C Single analyses on ten production batches of brand C.

220 1.141–1.150 1.145 ±0.0066 ±0.58% 1.109–1.302 1.205 ±0.118 ±9.8%


230 0.947–0.953 0.951 ±0.0038 ±0.40% 0.922–1.091 1.002 ±0.103 ±10.3%
240 0.819–0.825 0.822 ±0.0033 ±0.40% 0.782–0.934 0.853 ±0.090 ±10.6%
250 0.734–0.742 0.738 ±0.0047 ±0.64% 0.720–0.857 0.780 ±0.081 ±10.4%
260 0.740–0.744 0.743 ±0.0023 ±0.32% 0.723–0.855 0.781 ±0.079 ±10.1%
280 0.714–0.731 0.726 ±0.0103 ±1.43% 0.707–0.836 0.760 ±0.076 ±10.0%
300 0.507–0.513 0.509 ±0.0038 ±0.74% 0.491–0.584 0.525 ±0.056 ±10.7%
320 0.356–0.361 0.359 ±0.0030 ±0.82% 0.357–0.425 0.384 ±0.042 ±10.9%
340 0.279–0.282 0.280 ±0.0023 ±0.81% 0.279–0.328 0.297 ±0.029 ±9.8%
360 0.229–0.235 0.232 ±0.0033 ±1.43% 0.229–0.270 0.244 ±0.024 ±9.8%

Analyst, 2004, 129, 607–612 609


thenticity conclusions.1 Interestingly, brands B and E showed for authenticity analyses could also have been produced prior to the
similar congener ranges while their spectral characteristics were reference period or be derived from production batches other than
quite different (Fig. 2 and 4). In an opposite way, brands C and D those used for the reference data. The robust performance of this
exhibited overlapping spectra while their congener ranges were methodology was then demonstrated by authenticating other brands
quite distinct. in a similar way.
The spectra of twenty suspect samples purporting to be brand B An authenticity decision process was established based upon S z2
(labelled S1 to S20 in Table 2) and 15 other brands of Scotch values. Pass/fail limits were initially based upon values obtained
whisky (labelled X1 to X15) were collected and compared against from chi-squared tables. For 10 data points the limits were set at
the known spectral ranges of brand B. Their spectra, shown in Fig. 18.6 and 26.9 for 95.5% and 99.7% confidence limits, respectively.
5, indicated that suspect samples S2, S4, S7, S11, S12, S14 and S20 From experience and observation during testing of the WPA
each had spectra within the ranges for genuine brand B. Their Lightwave® instrument, these limits were found to fail genuine
resulting spectral S z2 scores are shown in Table 3 together with samples at a higher frequency than would have been expected. A
equivalent results, S z2 scores and conclusions by gas chromatog- decision was taken to increase these limits, as these would only
raphy. These results, plotted on Fig. 6, indicated that the have been appropriate if multiple reference samples from the same
spectroscopic and gas chromatographic methods came to the same batch/period as the suspect sample were available to use to establish
conclusions that these seven samples were authentic. The remain- the baseline figures for comparison. A pragmatic decision was
ing suspect samples each had spectra outside the ranges for genuine made to apply more usable limits, based on less rigorous
brand B indicating that they were not genuine. These conclusions application of statistical concepts, using the following calcula-
were confirmed by gas chromatography. tions:
The spectra for the 15 brands of other Scotch whiskies also
exhibited spectra outside the normal ranges for brand B, except for Lower limit = S(2la)2 + (2lb)2 + …
sample X14. However, the gas chromatographic results for this
sample indicated that it was not brand B but was a Scotch whisky Upper limit = S(3la)2 + (3lb)2 + …
with a higher malt whisky content than brand B. The spectral
conclusions for whiskies X4 and X12 indicated that they were not Therefore for 10 data points the limits and decision points were as
brand B, whereas gas chromatography found both samples to have follows: If S z2 was < 40, the sample was deemed authentic. If S z2
certain congener results only slightly out of range. These results was > 40 and < 90, the sample was deemed suspect and if S z2 was
demonstrated the complementary value of the two techniques > 90, the sample was deemed to be not authentic. S z2 values in
working together to highlight differences between brands based on these categories resulted in the Brand Authenticator® showing a
colour and/or congener concentrations. green (genuine), yellow (suspect) or red light (not authentic) at the
Whilst the reference spectra were collected from up to ten completion of each test.
production batches of each brand, many more production batches of A similar decision process was applied to the gas chromato-
these brands had been produced during the two year time frame in graphic results. If S z2 for four data points was < 16, the sample was
which the reference samples were collected. Suspect samples taken deemed authentic. If S z2 was > 16 and < 36, the sample was
deemed suspect and if S z2 was > 36, the sample was deemed to be
Table 2 Maximum and minimum range data for major volatile congeners not authentic. The validity of the chosen limits was confirmed by
in five blended Scotch whisky brands A to E the complimentary results obtained by the two analytical tech-
niques (Fig. 6). The potential to refine the numerical analysis and
2- and judgement criterion will be investigated on an on-going basis as
3-Methyl more range data for a particular product is gathered.
Brand Methanol n-Propanol Isobutanol butanol
Various criteria were considered when selecting acceptance
Volatiles/g (100 l absolute alcohol)21 limits for these spectroscopic and chromatographic methods. As
already reported, the reference samples used to establish normal
Brand A 6.8–8.4 58–77 69–77 101–115 ranges were taken from retained genuine samples representing at
Brand B 6.0–9.2 58–89 63–76 80–96 least 10 production batches produced during the last two years.
Brand C 6.3–9.8 44–76 63–76 70–83 However, for brand B this represented only 5 to 10% of the total
Brand D 6.4–9.9 50–58 61–70 42–53
production batches produced for market. Suspect samples collected
Brand E 7.2–10.5 51–62 63–72 75–93
in the field could have been produced prior to the reference period

Fig. 5 Absorbance spectra of 20 suspect samples purporting to be brand B collected using the SAD Brand Authenticator®.

610 Analyst, 2004, 129, 607–612


Table 3 The sum of squared scores (S z2) results on 20 suspect samples (S1 to S20) and 15 other Scotch whisky brands (X1 to X15) each determined using
the SAD Brand Authenticator® and by gas chromatography

2- and
Sample Isobutyl 3-Methyl Authenticity S z2 by Authenticity
number Methanol n-Propanol alcohol butanol S z2 by GC by GC UV/VIS by UV/VIS

Volatiles/g (100 litres absolute alcohol)21

S1 4.3 31.9 33.0 39.2 401 Fail 629 Fail


S2 9.2 67.7 68.9 79.9 11 Pass 7 Pass
S3 1.4 2.2 3.4 8.5 1190 Fail 869 Fail
S4 7.2 62.1 66.1 87.3 6 Pass 25 Pass
S5 2.7 17.6 23.2 37.2 563 Fail 233 Fail
S6 3.3 35.3 14.2 18.8 804 Fail 619 Fail
S7 8.0 62.3 68.3 82.4 6 Pass 9 Pass
S8 7.0 36.1 39.9 44.0 283 Fail 387 Fail
S9 2.5 9.5 14.1 33.9 715 Fail 995 Fail
S10 5.2 32.6 33.5 39.6 381 Fail 626 Fail
S11 7.4 62.5 67.7 85.5 4 Pass 31 Pass
S12 6.4 65.6 70.9 83.2 7 Pass 7 Pass
S13 70.8 2.7 0.4 0.0 10672 Fail 475 Fail
S14 7.1 75.1 68.3 81.9 3 Pass 16 Pass
S15 3.1 18.8 24.1 28.0 628 Fail 820 Fail
S16 16.9 57.8 52.9 23.7 581 Fail 247 Fail
S17 0.0 0.1 4.6 11.4 1186 Fail 1215 Fail
S18 0.0 1.0 0.5 5.1 1329 Fail 1252 Fail
S19 0.0 1.0 0.5 5.0 1330 Fail 1256 Fail
S20 9.0 63.6 68.4 91.0 9 Pass 9 Pass
X1 6.7 79.1 51.4 31.8 298 Fail 469 Fail
X2 5.9 114.9 52.6 50.7 175 Fail 325 Fail
X3 6.4 69.9 60.1 53.2 110 Fail 531 Fail
X4 7.8 60.6 68.2 84.2 5 Pass 231 Fail
X5 6.3 66.5 64.1 63.9 54 Fail 236 Fail
X6 4.8 94.1 53.6 31.7 311 Fail 477 Fail
X7 11.2 57.2 55.8 61.6 112 Fail 377 Fail
X8 8.1 93.3 53.0 62.1 88 Fail 238 Fail
X9 11.1 55.7 60.5 57.5 119 Fail 272 Fail
X10 4.8 33.6 55.3 65.7 112 Fail 522 Fail
X11 5.0 83.7 53.6 50.1 160 Fail 721 Fail
X12 5.4 50.9 63.6 92.4 30 Suspect 326 Fail
X13 5.3 51.6 60.2 63.4 80 Fail 499 Fail
X14 5.7 51.7 96.3 221 1652 Fail 44 Suspect
X15 4.7 52.7 58.3 75.1 56 Fail 606 Fail

Fig. 6 The sum of squared scores (S z2) for 20 suspect samples and 15 other brands of Scotch whisky determined both spectrally, using the SAD Brand
Authenticator®, and by gas chromatography.

Analyst, 2004, 129, 607–612 611


or be derived from batches produced during this two-year period Conclusions
but out with the reference set. Older suspect samples may also have
been exposed to harsh environmental conditions such as bright This spectroscopic method has enabled the deployment of a novel
sunlight that can cause colour fading (with a resulting impact of method for Scotch whisky brand authentication in the field. It
lowering their UV/visible spectra). These effects could result in a clearly distinguishes counterfeits, most of which, are combinations
risk of higher values for spectral standard deviations for genuine of cheap local alcohol flavoured with a smaller proportion of
product collected in the field compared to those for genuine product whisky and colouring. It can also distinguish other Scotch whiskies
retained at production. Conversely, methanol and higher alcohol and, in particular, compliment gas chromatographic authentication
concentrations in such samples remain stable and therefore the gas where two distinct brands have similar malt whisky contents.
chromatographic standard deviations for retained and field samples Subsequent confirmation by gas chromatography not only confirms
of genuine product are similar. To allow for these affects, S z2 the spectroscopic result, it also gives compositional information on
values greater than the equivalent chi-squared limits were chosen, the non-authentic liquid.3
this being a factor easily reset for individual brands within the This method has three distinct benefits. First, the methodology
Brand Authenticator® software. An alternative approach would be enables laboratories not equipped with gas chromatography to
to take an inflated standard deviation value, which would result in conduct Scotch whisky brand authenticity analysis. Secondly, the
a correspondingly lower S z2 value closer to the chi-squared handheld spectrophotometer enables field-testing with the added
statistical limits. benefit of speeding up investigation processes (for example,
As greater experience is gained with the Brand Authenticator® enabling suspect goods to be seized pending full forensic
under field conditions and brand ranges extend over longer time examination). Thirdly, this fast screening technique can enable only
periods, it may be possible to tighten the S z2 acceptance criteria. those samples which failed the spectroscopic analysis to go forward
However, the width and consistency of the spectral ranges of for confirmatory analysis by gas chromatography, thus saving both
genuine product will vary from brand to brand depending on that time and analysis cost. The Brand Authenticator® takes less than
brand owner’s specific blend formulation. Therefore, during the one minute to analyse each sample compared to at least 20 min by
initial deployment of the Brand Authenticator®, it is recommended gas chromatography, thus offering advantages of speed, cost and
that appropriate acceptance limits are established brand by brand, mobility.
with the instrument used as a screening tool prior to gas This spectroscopic method has the potential to be used for brand
chromatographic confirmation. authenticating other categories of distilled spirit that have suitable
For consistency within this project, the gas chromatographic chi- UV/visible spectra such as other whiskies, brandy, rum and
squared statistical limits were inflated in the same way as those flavoured spirits. It also has the potential to find application in the
applied to the Brand Authenticator®. Review of the gas chromato- authentication of beers, wines and non-alcoholic beverages (al-
graphic result trends for brand B over a 20-year period indicated though sample filtration and degassing may be required prior to
that chi-squared statistical limits are appropriate for the GC method spectral analysis). There are also opportunities to develop more
if the correct production period segments are selected for the range sophisticated data handling post analysis using derivative spectros-
data. A similar approach for the Brand Authenticator® will be used, copy and chemometric techniques.
as spectral trends become available.
Finally, an operational process was then configured for the use of Acknowledgements
this mobile instrument under field test conditions. First, average The authors wish to thank John Ferguson, Bill Considine and Jon
and standard deviation spectral data for the required brands (up to Considine of Spectroscopic and Analytical Devices for designing
a memory capacity of 16 brands) were collected and held in the and manufacturing the Brand Authenticator® to our design brief.
Brand Authenticator® memory. The procedure for analysing a test Thanks are due to colleagues Jennifer Yeh (Diageo Taiwan),
sample was established as follows: Margarita Calderon (Diageo Colombia), Maria Camacho (Diageo
1. After the instrument has completed its initial warm-up Venezuela) and Marcial Cid Nieto (Diageo Spain) for their
procedure, flush and load the flow cell with 2 ml 40% ethanol in contributions to the field-testing of prototype instruments. Finally
order to reference the instrument. thanks to John Wilson (Diageo Scotland) for statistical advice and
2. The quality control procedure is then automatically activated. Brand Technical Centre colleagues who contributed to in-house
Check the instrument calibration against a quality control whisky of testing.
known spectral characteristics. Once the sample passes this test, the
spectrophotometer progresses to test samples. References
3. Select the brand whose authenticity is to be tested from the
1 R. I. Aylott, A. H. Clyne, A. P. Fox and D. A. Walker, Analyst, 1994, 119,
brand list held within instrument memory. Introduce liquid as
1741.
previously described. 2 R. I. Aylott, Chemistry Review, Philip Allan Updates, Deddington, March
4. The instrument records the spectra, calculates a S z2 score for 1999, p. 2.
the test sample, compares it against the acceptance criteria 3 R. I. Aylott, in Whisky. Technology, Production and Marketing, ed. I.
described above and displays its decision on sample authenticity. Russell, Academic Press, London, 2003, p. 277.
5. Samples whose authenticity is deemed “pass” (green light) are 4 S. A. Savchuk, V. N. Vlasov, S. A. Appolonova, V. N. Arbuzov, A. N.
Vedenin, A. B. Mezinov and B. R. Grigor’yan, J. Anal. Chem., 2001,
considered authentic. Samples whose authenticity is deemed “fail”
56(3), 246.
(red light) or “suspect” (yellow light) are retained and returned to 5 Protocol for the Food Analysis Performance Assessment Scheme
the central laboratory for confirmation by gas chromatography. (FAPAS), Organisation and Analysis of Data, Ministry of Agriculture,
6. Continue with the next suspect sample. Fisheries and Food, UK, 3rd edn., 1993.

612 Analyst, 2004, 129, 607–612

View publication stats

You might also like