Adlard E.R. (ed.) Chromatography in the Petroleum Industry
Подождите немного. Документ загружается.
322
Chapter
I
I
Attempts to bring the same sort of control to mobile phase parameters in
capillary SFC have been less successful. The main objective
in
capillary SFC
is
to programme
a
decrease in the linear velocity of the mobile phase as the pres-
sure/density is increased in order to maintain high column efficiencies. The
cheaper simpler designs have a limited range over which they can alter the linear
velocity while restrictors with a wider and more practical dynamic range are
both expensive and complex.
Recently, a complete SFC system with capabilities for independent pro-
gramming of pressure (density), composition, flow and temperature was
launched. Significantly this system can be configured for either packed column
or capillary SFC modes and finally brings together all the operational parameters
for complete SFC optimization.
11.3
APPLICATIONS
At least
95%
of SFC applications in petroleum analysis occur in one of three
areas: simulated distillation, hydrocarbon group type separations and the analysis
of polycyclic aromatic hydrocarbons (PAH). Both the simulated distillation and
hydrocarbon-type analyses use predominantly C02 as the mobile phase because
they require the use of the FID for quantification. PAHs can be detected and
quantified using spectroscopic detectors, e.g. WD and FLD, and therefore a
wider range of fluids have been used to analyse these compounds. The analysis
of
chemical additives which are blended into fuels and lubricants as performance
enhancers has suprisingly received very little attention.
11.3.1
Simulated distillation
Distillation is the primary separation process in the petroleum industry.
Knowl-
edge of the boiling point distributions of crude oils and refined products is es-
sential for the operation and control
of
refinery plant and processes, for product
specification and for quality assurance. Distillation data are obtained using true
distillation methods (e.g. ASTM D86 and D1160) or by analytical techniques
which mimic the distillation process. By far the most widely used technique for
simulated distillation is GC-FID (e.g. ASTM D2887). Conventional distillation
yields true boiling point data and meaningful product fractions, which can be
fbrther characterized in terms of their physicochemical properties, but
is
expen-
sive and time consuming to perform. Simulated distillation methods are inex-
pensive, rapid and amenable to on-line monitoring but provide no additional data
on product quality.
Supercritical
fluid
chromatography
323
The upper boiling point limit for ASTM
D2887
is
538°C
which corresponds
to about
85%
of
a light crude (e.g. West Texas Intermediate or North Sea Brent),
70%
of a medium crude (e.g. Arabian Medium) and only
50%
of a heavy oil
(e.g. Boscan). Using high temperature
GC,
this upper limit can be raised above
800°C [47],
enabling approximately
95%
of
a light crude oil to be characterized.
To
achieve these high boiling points,
GC
column temperatures up to
430°C
must
be used and this has naturally raised doubts regarding the thermal integrity of
samples during the analysis
[48].
Schwartz et al.
[48]
presented evidence, using
thermo-gravimetric analysis, that an Arabian Heavy atmospheric residue started
to decompose at around
370°C.
In practice, laboratories performing routine high
temperature
GC
simulated distillation, including the author’s, have not reported
problems with Arabian Heavy or any other crudes
[49].
Even if current high temperature
GC
methods do not compromise the thermal
integrity
of
petroleum samples, it is questionable as to how much further
GC
can
go. The attraction of performing simulated distillation by
SFC
is that high tem-
peratures are not needed and the mechanism for extending the upper limit de-
pends on sample solubility not volatility. The concept of a simulated distillation
method based on the solubility of sample components in supercritical
C02
may
seem implausible at first. Polar high molecular weight components such as as-
phaltenes are not expected to be soluble in
C02
as a recent study comparing
Column:
51-17
X
50pm
SB--Phenyl-5
Program: Linear Density
Mobile Phase:
CO,
@
140%
Detector: FID
@
439°C
Time Split Injection
Injector
@
11
5%
t
L
1
0 16
32
40
60
TIME
IN
MINUTES
Fig.
11.5.
SFC
chromatogram
of
polyethelene standard
750.
Conditions:
C02
at
140°C;
linear
density program; FID at
430°C;
timed split injection with injector at
115°C.
References pp.
343-345
3
24
Chapter
I I
13
18
21
28
I1
18
43
48
53
RETENTION
TIME
(MINUTES1
Fig.
1
I
.6.
Calibration curve
of
retention time versus boiling point
for
polyethylene standard
750.
distillation fractions obtained by molecular distillation with supercritical fluid
extracts appears to confirm
[50].
A
tripartite relationship between volatility,
molecular weight and solubility is, however, generally recognized and has,
for
example been utilized by Boduszynski as part
of
an overall scheme to separate
and characterize crude oils
[5
11.
SFC and GC simulated distillation methods are basically the same. Normal
paraffins (e.g. a polyethylene standard like Polywax 750) are used to generate a
retention time versus boiling point calibration
for
the chromatographic column.
Fig.
11.5
illustrates the SFC chromatogram
of
Polywax
750
and its associated
SIMDIS
calibration plot using a capillary column. The oil sample, usually dis-
solved in carbon disulphide,
is
analysed under identical conditions to the stan-
dard with all data processing and reporting handled by specialized software
packages.
Either short capillary columns coated with non-polysiloxane phases
[52]
or
microbore columns (1 and
2.1
mm
ID)
packed with C18 bonded silica can be
used with equal success. Figure 11.6 shows the separation
of
Polywax
750
on a
short C18 silica column which compares favourably with the same analysis
achieved
on
a capillary column (Fig.
1
1
S).
Figure
1
1.6(b) illustrates the analysis
of
a
crude oil sample on the same microbore column. Crude oils are one of the
more difficult samples to analyse by either GC
or
SFC simulated distillation be-
cause
of
the wide boiling range. The analysis times for packed column SFC are
less than those for capillary SFC and can even be marginally shorter than
GC-
Supercritical jluid chromatography 325
TABLE
11.1
COMPARISON OF GC-
AND
SFC-SIMDIS DATA
FOR
PACS (TAKEN FROM REF. 37)
Compound True boiling point Deviation
from
TBP Deviation from TBP
(“C)
(“C)
for
GC-SIMIDS
(“C)
for
SFC-SIMDIS
Naphthalene 218
0
24
Fluorene 298
-1
I
20
Phenanthrene 338
-1
9
4
Pyrene 393
-1
1
0
Chrysene
447
-3 8
-14
Benzofluoranthene
480
-45
-1
1
Benzo[e]pyrene
493
-48
-10
Picene 519 -50
4
SIh4DIS methods
[60].
Retention time repeatability is,
of
course, critical for
simulated distillation. The capillary column holds a slight advantage in this re-
spect because
of
the long-term detrimental effect that pressure ramping is ex-
pected to have on the packing structure
of
the stationary phase in the microbore
column. Bartle and co-workers
[53]
have reported that the use
of
an
n-
octylpolysiloxane stationary phase in capillary columns leads to better corre-
spondence between the retention behaviour of
PAHs
and n-alkanes
of
similar
boiling point (Table
11.1).
Since the distillation behaviour
of
PAHs
does not
always relate directly to their boiling point, we should be careful not to conclude
that the SFC
SIMDIS
data approximates more closely to true distillation behav-
iour.
Many
of
the early problems with SFC simulated distillation concerned sample
injection. In particular injection
of
polywax standards was often more trouble-
some than the real samples because
of precipitation of the higher molecular
weight waxes in the injection valve and subsequent carryover effects. Dissolving
samples in high boiling solvents (decane, xylene, chlorobenzene) and the use of
heated injectors has largely removed this problem. Sample discrimination using
direct injection techniques (e.g. timed split) is certainly no worse than with high
temperature GC.
Newer SFC instrumentation is capable of programming beyond
400
atm (Fig.
1
1.6(a)) which, in conjunction with higher column temperatures, should extend
the upper boiling point limit even further. At present, the boiling range covered
by routine SFC is the same as that attainable by routine high temperature GC.
The addition
of carbon disulphide
(5
mol%) to COz eluted heavy hydrocarbon
wax material up to C150 (Fig.
11.7)
[54],
which
is
a little higher than that
re-
ported for high temperature GC
[57].
Note, however, that the SFC conditions
were severe and the sample had to be ‘injected’ onto the packed microbore col-
References
pp.
343-345
3
26
Chapter
II
1
-
*
0
10
20
30
Time
(min)
Fig. 11.7.
SFE-SFC-FID
characterization
of
a
heavy hydrocarbon
wax
using carbon disulphide-
modified carbon dioxide.
Peak
1
=
chloroform.
1
40°C,
100
x
1
mm
ID
experimental
C
18
modified
silica column; 180 bar
to
507
bar
at
15
bar/min.
umn by placing
it
in a small
SFE
vessel and extracting at
465
atm and
55OC
for
45
min.
Schwartz and co-workers were the first to demonstrate conclusively the valid-
ity of the
SFC
simulated distillation method
[48]. SFC-SIMDIS
data were com-
pared not only with true distillation but also with
GC-SIMDIS
and vacuum
thermogravimetric analysis (VTGA).
For
samples boiling in the range
260-
540°C
it was possible to demonstrate excellent correlation with actual distilla-
tion data.
Wax properties such as penetration and solidifying temperature depend on the
molecular weight range and content of low molecular weight material while
tensile strength is influenced by the amount and molecular weight distribution of
is0 and cyclo alkanes. Mellor et al.
[55],
using capillary
SFC,
analysed a number
of industrial paraffins and waxes and reported that
SFC
was suitable for finger-
printing high molecular weight paraffins and waxes. Thompson and Rynaski
1561
compared
SFC
and high temperature
GC
simulated distillation for the
analysis of wax samples isolated
from
underground crude
oil
storage caverns.
Average differences between the
two
types of analysis were less than
6°C. GC
sample injection required both the sample and the syringe to be heated to
60°C;
manual on-column injections had to be made as quickly as possible in order to
prevent a significant drop
off
in the
FID
response factors for the higher alkanes.
SFC
injection was performed by the timed split method using a heated
(1 15OC)
injection valve; lower injector temperatures resulted in poor recovery
of
the
Supercritical
fluid
chromatography
3
27
higher molecular weight alkanes. The close agreement between the GC and SFC
data implied little or no thermal decomposition in the former and no sample dis-
crimination problems in the latter.
The distillation characteristics of middle distillate feedstocks to an ethylene
cracker can be monitored using a process supercritical fluid chromatograph. En-
vironmental and economic benefits are claimed by using the
SIMDIS
data, in
conjunction with hydrocarbon group analysis by GC, to optimize the process
WI.
For certain products such as road bitumens and asphalts, where information
on the distillation characteristics is of little practical value, it more meaningful to
process the
SIMDIS
data in terms of carbon number distribution. Barbour and
Branthaver
[58]
used this approach to analyse non-polar asphalt subfractions as
part
of
the US Strategic Highway Research Program
(SHRP).
A
general model
for asphalt structure is that a non-polar predominantly hydrocarbon ‘solvent
phase’ acts as a dispersion medium for the more polar aromatic species which
are involved in weak associations (colloids/micelles). The dispersing power
of
this ‘solvent phase’ will be a function of its
compositiodconcentration
and will
influence some of the important physical properties of the bitumedasphalt.
Since petroleum heavy ends contain significant amounts of material (up to
50%)
which is not soluble in COZY these workers isolated the non-polar component of
the asphalt (the ‘solvent phase’) via three different routes: the neutral fraction
from ion exchange chromatography; the low molecular weight fraction from size
exclusion chromatography (SEC); and the maltenes fraction by deasphalting
with iso-octane. SFC-FID analysis was carried out on a short microbore column
(100
x
1
mm ID, polysiloxane stationary phase) using a combined pressure
(135-375 atm) and temperature (125-1 5OOC) gradient. Column degradation was
observed at higher temperatures and pressures. Under these conditions, a carbon
number range from 18-1 10 (relative to a polywax standard) could be studied.
For each asphalt studied, carbon number distributions were similar for all three
non-polar fractions
so
that any of these fractions might be considered represen-
tative of the ‘solvent phase’. For most of the samples it was possible to differen-
tiate between sol-type and gel-type asphalts.
For production purposes, crude oil is usually characterized in terms
of
pseudo-components defined in terms of carbon number. Carbon number distri-
butions of 26 stock tank oils were measured using GC and SFC simulated distil-
lation
[59].
Both methods were claimed to give the same carbon number distri-
butions within normal experimental scatter although it was reported that for most
crudes the SFC analysis eluted a significantly larger fraction of the oil than did
GC. For light crudes, where the SFC method eluted essentially
100%
of the hy-
drocarbons, it was possible to dispense with an internal standard without any
loss of accuracy in the data.
References
pp. 343-345
328
Chapter
I
I
11.3.2
Hydrocarbon group
type
analysis
Hydrocarbon group type refers to the separation and quantification of alkanes
(saturates), alkenes (olefins), aromatic hydrocarbons and polar compounds in
fossil fuels. Each of these four main group types may be further differentiated
into subgroups, e.g. normal, branched and cyclo-alkanes or mono-, di-, tri- and
poly-cyclic aromatic ring systems.
Until the advent of SFC, this type of analysis was performed using LC, GC or
mass spectrometry
(MS). MS
provides the most detailed information but is also
the most expensive option and is not often found in routine laboratories. GC is
restricted to the analysis of lighter distillates with FBPs around 220°C. Classical
open column
LC and HPLC methods are used extensively throughout the petro-
leum industry for routine analysis of middle distillate and higher boiling sam-
ples. Most
LC
methods rely on a time-consuming and laborious gravimetric
work-up to provide quantitative data because LC does not have a universal detec-
tor similar to the FID
in
GC. The main attraction of SFC, therefore, lies in its
ability to use the FID for simple and accurate quantification
of
separated hydro-
carbon groups.
The composition of petroleum changes with increasing boiling point
so
that a
separation method developed for a gasoline product may not be suitable for die-
sel fuels. Chromatographic methods therefore tend to be quite specific in their
scope with regards the sample types to which they can be applied. There are.
however, a number of common themes and problems which apply to all hydro-
carbon type separations. For example, all LC and SFC methods rely on adsorp-
tion chromatography to separate the different compound types and hence use the
same technology, i.e. silica or bonded silica polar stationary phases.
The SFC method operates with
two
fundamental differences: the mobile phase
and the FID. Since the mobile phase must be compatible with the FID, this limits
the choice to C02,
N20
or SF6. For obvious practical reasons, C02 is the most
popular fluid. Experience has shown that low fluid densities give GC-like sepa-
rations (i.e. retention increases with decreasing volatility) which are the antithe-
sis
of hydrocarbon type analysis.
In
order to avoid such band broadening effects
within each hydrocarbon group, SFC separations are usually performed with
moderate pressures at or near the critical temperature where fluid densities are
high. It should be noted that it
is
the density of the fluid which is important, and
not the temperature, implying a solubility and not a volatility phenomenon. An-
dersson has reported that the C02 fluid density should be
0.81
g/ml or higher for
hydrocarbon type separations
[61].
Lee and co-workers found that retention was
only slightly influenced by density in the range
0.88-0.93
g/ml
[62].
In practice
this means the solvent power of CO, can only be reduced by adding a weaker
FID-compatible fluid such as
SFs.
Supercritical jluid chromatography
329
For aromatic hydrocarbon type separations, the influence of alkylation on
the aromatic rings should be minimal. Retention behaviour under GC-like con-
ditions (retention times increase with increasing alkylation) is different to that
under LC-like conditions (retention times decrease with increasing alkylation).
For the majority of hydrocarbon type separations, LC-like conditions are pre-
ferred because they yield compact peaks with less overlap between the different
hydrocarbon types. However, for some analyses, where speciation of a hydro-
carbon group is required (e.g. measurement of naphthalenes in the di-aromatic
fraction of kerosenes), there may be an advantage in operating with GC-like
conditions.
On a solvent polarity scale for silica stationary phases, CO, at densities above
0.8
g/mI would be ranked slightly more polar than the liquid alkanes (pentane,
hexane, heptane) used as non-polar LC mobile phases. In general, hydrocarbon
type separations by SFC with COz are noticeably poorer than those obtained by
LC with hexane. Consequently, techniques used to improve resolution in LC
have also been tried with SFC. In particular, the most difficult separation be-
tween saturates and olefins requires either a more retentive stationary phase
(silver-modified silica retains olefins by complex formation) or a less polar
mobile phase (perfluorinated hydrocarbons in LC, SF6 in SFC). Both options
have been successfully applied to hydrocarbon type analysis by SFC.
SFC allows the use of mobile phase linear velocities above the optimum with-
out sacrificing too much in terms of resolution. The major limiting factor will be
the amount of gaseous C02 entering the FID. Packed capillaries of around
0.75
mm ID can be operated at optimum flow rates with all the column effluent
transferred to the FID. Smaller micropacked capillaries are therefore best for fast
analyses because flow rates above the optimum can be used without compromis-
ing the FID. Microbore columns
(1-2
mm
ID)
must be operated at non-optimum
flow rates if the whole column effluent is transferred to the FID. An option is to
operate at higher flow rates and only send a portion of the column to the FID, a
strategy which must be adopted for larger
(>2
mm ID) columns.
Finally, it is worth noting that the constant response of the FID to all hydro-
carbons is an approximation only. Non-uniform response may occur if the
FID
is
not operated under optimum conditions or the composition of the sample is un-
usual in the nature and amount of saturate and aromatic compounds present. The
presence of non-hydrocarbons, especially sulphur compounds, may exert undue
influence on the quantification which is almost always achieved by normaliza-
tion of the whole chromatogram. Some workers have reported variable FID re-
sponses for saturates and aromatics
[63]
and for the various aromatic types. As
with all forms of chromatography, problems with quantification can often be
traced to the injector and/or the detector which should be optimized for the type
of sample analysed.
References pp. 343-345
330
Chapter
I I
I
I.
3.2.1
Gasolines
The fluorescent indicator adsorption
(FIA)
method (ASTM
D
13 19/IP 156)
is
still
used routinely throughout the oil industry for the determination of saturates,
olefins and aromatics in gasolines and kerosenes. Norris and Rawdon
[65]
were
the first to suggest that this open column
LC
method could be replaced with an
SFC-FID
equivalent. Using a silica column connected in series with a silver ni-
trate
(20%)
coated silica column, they demonstrated the separation of a gasoline,
kerosene and diesel fuel into saturates, olefins and aromatics using
C02
(35"C,
210 atm) as the mobile phase. The
SFC
method was quicker (analysis time ca.
5
min), applicable to coloured samples and
full
range gasolines (cf.. ASTM
Saturates
Aromatics
r
50
Temperature
(
"C)
150
3
10
20
30
40
50
Time
(min)
Fig.
1
1.8.
25
x
1
mm
ID
column
packed
with
5
pm
silica
SF6
at
350
atm
and
50°C.
Supercritical fluid chromatography
33
1
D13
19
where gasolines must be depentanized and coloured contaminants re-
moved by distillation) and showed good repeatability. Results were reported as
%
mass by the SFC method compared with
%
vol.
by
FIA. The method was
submitted to ASTM for round robin testing but participating laboratories had
difficulty in reproducing the separation; this was thought to be due to instability
problems with the silver nitrate-modified silica column.
Schwartz and Brownlee
[66]
replaced the dual column system with a single
microbore column
(250
X
1
mm
ID)
packed with silica and switched to the less
polar
SF6
as the mobile phase. Although the separation between saturates and
Olefins
+
Saturates
Aromatics
1
Baddlush
w
10
20
lime
(rnin)
B
Y
io
20
Time
(min)
Fig.
11.9.
(a)
350 atm
10%
C02/SF6,50°C,
5
pm silica.
(b)
350
atm
Co2
at
40°C.
References pp. 343-345