Adlard E.R. (ed.) Chromatography in the Petroleum Industry
Подождите немного. Документ загружается.
392
Chapter
13
Concentration calculated from Eq.
(1
3.3)
is
only accurate when all solutes are
eluted and detected and the detector
is
linear over the range of concentrations
measured. It also assumes the areas are measured accurately.
13.6.3.2
Internal standard
The concentration of a solute,
ci,
is given by
(13.4)
where
As
is the peak area of an added internal standard,
Ris
is the relative re-
sponse factor of the solute to the standard (ratio
of
the
two
absolute response
factors),
Ws
is the weight of added standard and
W,,
is
the sample weight.
The internal standard method compensates for variations in sample volume as
the added standard peak area varies correspondingly. It also indicates when sol-
utes are retained on column
or
have not been detected because the concentrations
do not then add
up
to
100%.
Its shortcomings are that it
is
not always easy to
find
a
suitable standard
or
a convenient stretch of vacant baseline to place it, and
one solute has to be standard for a variety of other solutes of differing responses,
concentrations and perhaps peak shape.
I3.6.3.3
External standard
The external standard calculation compares peak areas under standard and
non-standard conditions:
4
c,
=-ws
AS
(13.5)
A
standard solution is prepared containing all the compounds of interest in the
unknown samples and analysed. Unknown samples are then analysed and the
solute concentrations obtained by comparison to the standard using Eq. (13.5).
Since solutes are compared against themselves, there is no need to use response
factors; they would be identical in Eq. (13.5) and would cancel.
This method relies on the injection volumes
of
standard and unknowns being
identical, which rules out manual injection techniques for the GC analysis
of
liquids as they are too erratic. External standard method is best used with
valve/sample loop injection of gases
(or
in
HPLC)
where the sample sizes are
relatively large.
13.6.4
Calibration
curves
Standard solutions are analysed over a range of concentrations to determine
response factors and to determine
or
compensate for detector non-linearity. Cali-
Modern data handling method
3
93
bration curves of area versus quantity are then calculated. If enough standard
solutions are analysed, the data are fitted to polynomials, which may turn out to
be linear.
Modern data processors either have all the standard calculation facilities built
into them, or they can be programmed to perform non-standard calcula-
tions. Some export the data into a computer which has the necessary software.
After entering details of the prepared standards and carrying out the analyses,
calculation of response factors and subsequent applications to unknowns is
automatic.
13.6.5 Empirical correction
of
analysis
errors
There are many causes of peak measurement inaccuracy, yet calibration often
produces accurate assays even for analyses where noise, overlap and asymmetry
are clearly present and creating the inaccuracies about which this chapter warns.
It is reasonable to question whether these inaccuracies are genuine and if so how
can any analysis give reasonable results. The answer lies in the empirical nature
of the response factor. When a response factor is determined from “bad” chro-
matography, it combines the lme response factor with another empirical factor
which converts the measured result into the correct answer. This empirical factor
is a function only of that instrument operating under the prescribed analytical
conditions; if the sample, method or chromatograph are changed, the system
must be recalibrated. However, if conditions do not change, then the empirical
factor remains constant and the analysis results are accurate. The measured re-
sponse factor compensates for bad chromatography. This fortunate compensation
has its limits. Problems arise when analyses slowly deteriorate, columns become
contaminated, analytes change peak size and shape, resolution changes; then the
compensation may no longer work. Unfortunately there is no noticeable bound-
ary where the compensation ceases to work, where calibrated analyses drift into
error. The analyst can only avoid problems by calibrating regularly and by
keeping the analytical conditions as stable and reproducible as possible between
calibrations.
13.7
VALIDATION AND STANDARD CHROMATOGRAMS
13.7.1 Meaning
of
validation
Validation in chromatography means the systematic proof of the reliability of
the analytical method in terms
of
precision and accuracy
[37].
The opportunities
for errors in peak measurement described earlier are just too many and varied to
References
pp.
398-399
394
Chapter
I3
accept results naively.
All
analyses should be validated by estimating the uncer-
tainty
of
each result. This specifically means calculating the accuracy error (i.e.
bias) in each result, and the coefficient of variation
or
relative standard deviation
of the experiment series, but calculating
CV
or
RSD
is
not always done in prac-
tice and quantifying bias may be impossible by its very nature.
The emergence and adoption of quality schemes in chemical manufacturing
has created the demand to routinely calculate
CVs
or
RSDs
but the pursuit of
accuracy is an unending task and success is never guaranteed.
13.7.2 System suitability
System suitability is the verification that
a
chromatographic system remains
within the operating limits established during validation throughout the series of
experiments
[37].
This means that once calibration has been made, the analyst is
confident that operating conditions do not drift between calibrations.
13.7.3
Validation and standard chromatograms
In
an ideal world, the integrator should be independently validated and cali-
brated by means of an independent standard chromatogram of accurately known
dimensions. It can then be used to validate and calibrate the chromatograph and
measure unknown samples. In practice, the chromatograph is used to validate an
integrator
or computer by injecting standard solutions and measuring them. The
“calibrated” integrator is then used to measure unknown samples on the chroma-
tograph. This interdependence means that systematic errors cannot be uncovered
and the measurement of unknowns can contain large inaccuracies.
With existing chromatographs, detector output during the analysis
of
a stan-
dard sample is not sufficiently repeatable or reproducible to be a good candidate
for standardization; in any case there are very few appropriate standards; a
“typical lab sample” remains the simplest and best.
Electronic signal generators and computer programs exist to generate chroma-
tograms of known dimensions
[29],
and although there are no agreed interna-
tional standards for them, their output is independent and fixed, and they offer
the preferred method for independent validation. They are also a very good
training aid for integrators.
Signal generators are independent of the chromatograph but have the disad-
vantage that they produce
a
single output “chromatogram” which bears
no
re-
semblance to any real samples that the lab measures. Computer programs are
better able to create a variety
of
independent standard chromatograms of known
Modern data handling metho&
395
dimensions, modelling real chromatograms. These can be measured and the re-
sults compared to the known answer. Accuracy as well as precision can be
checked on a synthetic chromatogram, relevant to the lab.
Historically, there have been several techniques for generating reproducible
signals including rotating drums with flashing peaks
[38],
tape recorders
[39],
simple signal generators
[
121
and, more recently, storing real chromatograms and
creating computer simulated chromatograms
[
13,3 11.
Electromechanical devices always suffer from wear: magnetic tapes stretch
and have a limited frequency spectrum, motor speeds wander and component
specifications vary with tolerance and ambient conditions.
Synthetic computer-generated chromatograms are an improvement because
they are fixed data files which cannot degrade (although files can be deleted).
Real chromatograms stored inside the data system which measured them have
relevance and variety. They are fixed and can be referenced at will. Such chro-
matograms are not
so
reliable ,when transferred to another integrator to compare
processing as the transferred signal will differ from the original detector signal
by having passed through the first integrator. When the signal is transferred to
the data processor of another manufacturer, the processing algorithms are almost
certainly different
[32,33].
Synthetic chromatograms based on realistic peak models such as the expo-
nentially modified Gaussian model, generated inside a computer and output
through a DIA converter open up the possibility of defining a standard chroma-
togram. It may prove impossible to agree upon a versatile international standard,
but analytical groups can agree among themselves on their
own
standards
which will be reproducible in different labs. The DIA converter is always subject
to component variations but these are offset by normalizing and ratioing peak
areas.
Comparisons of commercial integrators by the way they process a synthetic
chromatogram have shown that unless they are given relatively simple signals to
measure, they will not only process them incorrectly (as described by this chap-
ter) but also differently to each other
[32,33].
Manufacturers of compounds
which attract tariffs and duties (perfumes, wines and spirits for example), may be
tempted to buy the integrator that gives the lowest measure and attracts the low-
est tax invoice irrespective of whether it is the most accurate. Legal cases which
hinge on chromatographic analyses are another concern.
13.8
STRATEGIES FOR PE,QK MEASUREMENT
The strategy is to get rid
of
the problems systematically before the peaks are
measured.
References
pp.
398-399
396
Chapter
13
13.8.1 Noise
Noise
is
always present if the detector sensitivity
is
high enough, but sig-
nalhoise ratio can be improved by increasing the sample size and reducing the
detector sensitivity. If the column
is
regularly conditioned and kept in good
condition, there will be less unwanted bleed. Good quality carrier gas, hydrogen
and air will introduce fewer impurities into an
FID
flame and give lower back-
ground noise. The chromatograph will generate less noise if it
is
regularly serv-
iced.
13.8.2 Baseline drift
Baseline drift comes from poor experimental control. It is important therefore
to systematically root out the causes (which are just too numerous to list here).
Avoid temperature programming
if
it creates drift, minimize temperature ramps
if they cannot be avoided.
13.8.3 Peak
overlap
Peak overlap
is
a fact of life for most chromatographers. It can never be got
rid of entirely. Nevertheless, it
is
important to maximize resolution of peaks of
interest and keep overlapping groups of peaks as small as possible on a baseline
that is as flat as possible. Alternatively use a selective detector to pick out only
the required peaks.
Sometimes a shorter analysis
is
preferred to optimizing the flow rate for best
column performance. When speed and throughput are important, the analyst
implicitly accepts less accurate results.
13.8.4 Asymmetry
Poor injection port design has been the major cause of off-column asymmetry
but this
is
rapidly improving; dead volume has largely been eliminated.
Glass
inserts which restrict the expansive upstream flow of injected solutes remove
many of the problems. Off-column adsorption sites in the sample flow path, par-
ticularly septum shavings in the glass insert, must be removed.
On-column asymmetry is minimized by selecting
a
column of appropriate
po-
larity although modern column manufacturing techniques give rise to good peak
shape in most cases.
Modern data handling methods
397
An old chromatograph with
a
slow detector time constant will distort peak
shape. Please a salesman, throw it out and buy a new one.
13.8.5 Detectors
The linear dynamic range of detectors should be determined for all measured
analytes. It
is
important to note that detectors have different linear ranges for
different substances. It is possible to have a detector working linearly for one
component and non-linearly for another in the same sample.
When major peaks
of
interest go off scale, the analyst should occasionally in-
crease attenuation and check the tops of the peaks to confirm they have not been
clipped by the amplifier saturating.
13.8.6 Integrators
The analyst should know the limitations of integrators, read the manual and
learn how to program them. Give an integrator only those chromatograms which
it can measure accurately.
A
good chromatogram
is
one where solutes are well
resolved and reasonably sized for ease of measurement on a flat, noise-free
baseline.
13.8.7 Checking the analysis results
Results from an integrator or computer should never be accepted unquestion-
ingly.
13.8.7.1
Checking the chromatogram
A
chromatogram is good enough to be measured if:
-
-
-
-
it looks very similar to a “standard” or well known chromatogram
there are the correct number of peaks, i.e. the right sample was injected
the peaks are the right size, i.e. the injection volume was correct
the noise level
is
acceptable; the column does not need conditioning and
earlier solutes have completely eluted
the baseline is where it ought to be; the system had stabilized before in-
jection, and there are no dips afterwards
baseline drift is acceptable or absent
integration event marks are all present and correct
-
-
-
References
pp.
398-399
398
Chapter
13
The rule with event marks
is:
one event mark at every event (peak boundary
or
valley) and no event missed. It suggests the measurement parameters were set
correctly and the peaks were integrated correctly.
13.8.7.2
Checking the
analysis
report
If
the report is to be believed, all required peaks must be correctly identified.
If
they are not, the integrator time windows are probably not set correctly, in
which case the wrong calibration factors will be applied when calculating peak
quantities.
Peaks must have the correct measurement diagnostic indicating that they were
measured as single peaks
or
separated from others by a tangent
or
perpendicular.
If they
do
not, it is, once again, an indication that the data processor
is
not
pro-
grammed correctly.
13.8.7.3
Accepting the results
Finally, the actual peak areas and retention times must be repeatable within
previously accepted limits to show that experimental stability is good, and the
measured quantities must be within an accepted range to confirm that the prod-
uct is within specification.
13.9
REFERENCES
1
2
3
1
5
6
7
8
9
10
11
12
13
15
I6
17
18
19
N.
Dyson, in Chromatographic Integration Methods, Chromatography Monographs Series
(R.M. Smith, Series Editor), Royal Society
of
Chemistry
(1990)
ISBN
0-85186-587-9.
J.J.
Kirkland,
W.W.
Yau, 1-l.J. Stoklosa and C.H. Dilks Jr.,
J.
Chromatogr. Sci.,
15
(1977)
303.
D.C. Nelson,
V.
Dauciunas, R. Lysakowski and
M.
Duff, Int. Lab.,
32 (1992).
I.
Scott,
ICI
Chemicals and Polymers, Personal Communication
(1988).
D. MacDougall, Anal. Chem.,
52 (1980) 2242.
T.
Petticlerc and G. Guiochon,
J.
Chromatogr. Sci.,
14 (1976)
53
1.
P. Deng and
F.
Andrews, J. Chromatogr.,
349 (1985) 415.
A.
Savitsky and M.J.E. Golay, Anal. Chem.,
36 (1964) 1627.
J.
Steiner,
Y.
Termonia and J. Deltour, Anal. Chem.,
44 (1972) 1906.
D.L.
Massart, B.G.M. Vandeginste,
S.N.
Deming,
Y.
Michotte and L. Kaufman, Chernornet-
rics: A Textbook. Elsevier, Amsterdam
(1988)
ISBN
0-444-42660-4.
S.N.
Chesler and
S.P.
Cram, Anal. Chem.,
43 (1 97
I)
1922.
S.P. Cram. S.N. Chesler and A.C. Brown,
J.
Chromatogr.
126 (1976) 279.
M.R. Schure, J. Chromatogr.,
550
(1990)
5
I.
A.B.
Littlewood,
Z.
Anal. Chem.,
236 (1968) 39.
R.J.
Hunt. J. High Resolut. Chromatogr.
8 (1985) 347.
S.T.
Bake. in Quantitative Column Liquid Chromatography:
A
Survey
of
Chemometric
Methods’, Journal
of
Chromatography Library, Vol.
29,
Elsevier, Amsterdam
(1984).
A.N.
Papas, CRC Crit. Rev. Anal. Chem.,
20 (1989) 359.
J.
OlivC and J.O. Grimalt,
1.
Chromatogr. Sci.,
29 (1991) 70.
Modern data handling methods
3 99
20
J.G. Dorsey and J.P. Foley, J. Chromatogr. Sci.
22 (1984) 40.
21
M.S. Jeansonne and J.P. Foley,
J.
Chromatogr. Sci.,
29 (1991) 258.
22
C.P. Cai andN.S. Wu, Chromatographia,
31 (1991) 595.
22a
E.R. Adlard and R.E. Davies, in Petroanalysis
81.
Advances in Analytical Chemistry in the
Petroleum Industry (G.B. Crump, Ed.), Wiley, Chichester
(1982).
23
R. Amino, in Advances in Chromatography,
Vol.
15,
Signal and Resolution Enhancement
Techniques in Chromatography, Marcel Dekker, New York
(19??)
Chap.
2.
24
R. Bracewell, The Fourier Transform and its Applications, McGraw-Hill, New York
(1965).
25
Y. Hayashi, T. Shibazaki, R. Matsuda and
M.
Uchiyama, J. Chromatogr.,
407 (1987) 59.
26
G.H. Rayborn, G.M. Wood, B.T. Upchurch and
S.J.
Howard,
Am.
Lab,
18 (1986)
56.
27
L. Ghaoui and L.D. Rothman,
J.
High Resolut. Chromatogr.,
15 (1992) 36.
28
A. Westerberg, Anal. Chem.,
41 (1969) 1770.
29
B.G.M. Vandeginste and
L.
DeGalan, Anal. Chem.,
47 (1975) 2124.
30
B. Jover and J. Juhasz, J. Chromatogr.,
154 (1978) 226.
3 1
N. Dyson, Int. Lab.,
22(6) (1992) 38.
32
A.N. Papas and M.F. Delaney, Anal. Chem.,
59 (1987)
54A.
33
A.N. Papas and T.P. Tougas, Anal. Chem.,
62 (1990) 234.
34
G. Guiochon and C.L. Guillemin, in Quantitative Gas Chromatography for Laboratory Analy-
ses
and On-Line Process Control, Journal
of
Chromatography Library, Vol.
42,
Elsevier,
Am-
sterdam
(1988)
ISBN
0-444-42857-7.
35
P.W. Carr, Anal. Chem.,
52 (1980) 1746.
36
L.M.
McDowell, W.E. Barber and P.W. Carr, Anal. Chem.,
53 (1981) 1373.
37
T.
Crispin and R. Wirtz, Chromatographia,
22 (1986) 434.
38
R.D. Johnson,
J.
Gas Chromatogr.
6 (1968) 43.
39
J.T. Shank and H.E. Persinger, J. Gas Chromatogr.
5 (1967) 631.
This Page Intentionally Left Blank
E.R.
Adlard
(Ed.),
Chromatography
in
the
Petroleum
Industry
Journal
of
Chromatography
Library
Series, Vol.
56
0
1995
Elsevier Science
B.V.
AU
rights
reserved
40
1
CHAPTER
14
Capillary electrophoresis in the petroleum
industry
Tim
Jonesa and Gerard Bondouxb
a
Waters Chromatography Division, Mill@ore
(UK)
Ltd, Winster House, Heronsww, Chester Business Park,
Wrexham
Road,
Chester CH4
9QR,
UK
and bWaters Chromatography Division, Millipore
S.A.,
6
Rue Jean-
Pierre Timbaud, BP
307,
78054
St.
Quentin-en- yvelines,
France
14.1
INTRODUCTION
Capillary electrophoresis (CE) is a new, evolving analytical technique which
potentially offers considerable advances in the speed, efficiency and quality of a
number of application areas within the petroleum industry. CE has evolved fiom
classical electrophoretic techniques, the term “capillary” is used as the electro-
phoretic separation takes place in a fused silica capillary tube. Early electropho-
resis experiments in a glass tube were conducted by Lodge
(1886) [l],
who
monitored the migration of
H+
in
a
tube containing
a
phenophthalein jelly.
Compton and Brownlee
(1988)
[2]
have reviewed the last
100
years development
of CE. The use of small diameter glass tubes and silica capillaries has been re-
ported within the last
30
years by a number
of
authors
[3-91.
These efforts led to
the introduction of the first commercial instrument in
1989.
While CE is there-
fore considered a new technique, there is a significant history
of
work to support
more recent efforts.
The number of publications since the introduction of commercial CE instru-
ments reflects the exponential growth
of
the technique in the last
5
years. The
first published applications concerned the separation of proteins and bio-
molecules in general (Fig.
14.1).
The separation of other classes of compounds
such as pharmaceutical formulations, vitamins, inorganic ions, organic ions, car-
bohydrates and other natural molecules (as oppose to synthetic) have been ex-
plored. Even the separation
of
polymer particles by size has been demonstrated
[lo];
all the applications associated with classical electrophoresis can be
References
pp.
425-426