Journal of Chromatography A, 1271 (2013) 192–200

Contents lists available at SciVerse ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Development of a manual method for the determination of mineral oil in foods

and paperboard
Katell Fiselier a , Florian Grundböck a , Karsten Schön b , Oliver Kappenstein b , Karla Pfaff b ,
Christoph Hutzler b , Andreas Luch b , Koni Grob a,∗
a
Official Food Control Authority of the Canton of Zürich, P.O. Box, CH-8032 Zürich, Switzerland
b
German Federal Institute for Risk Assessment (BfR), Department of Product Safety, Max-Dohrn-Strasse 8-10, D-10589 Berlin, Germany

a r t i c l e i n f o a b s t r a c t

Article history: So far the majority of the measurements of mineral oil saturated hydrocarbons (MOSH) and
Received 25 September 2012 mineral oil aromatic hydrocarbons (MOAH) were obtained from on-line high performance liquid
Received in revised form chromatography–gas chromatography–flame ionization detection (on-line HPLC–GC–FID). Since this
12 November 2012
technique is not available in many laboratories, an alternative method with more easily available tools has
Accepted 14 November 2012
been developed. Preseparation on a small conventional liquid chromatographic column was optimized
Available online 21 November 2012
to achieve robust separation between the MOSH and the MOAH, but also to keep out the wax esters from
the MOAH fraction. This was achieved by mixing a small portion of silica gel with silver nitrate into highly
Keywords:
Mineral oil saturated hydrocarbons (MOSH)
activated silica gel and by adding toluene into the eluent for the MOAH. Toluene was also added to the
Mineral oil aromatic hydrocarbons (MOAH) MOSH fraction to facilitate reconcentration and to serve as a keeper preventing loss of volatiles during
Silver nitrate/silica gel solvent evaporation. A 50 ␮l volume was injected on-column into GC–FID to achieve a detection limit for
Separation from wax esters MOSH and MOAH below 1 mg/kg in most foods.
Toluene to support selective © 2012 Elsevier B.V. All rights reserved.
reconcentration

1. Introduction was extended to the MOAH in 2009 [6]. Since it requires special
equipment and expertise there has been the request to develop a
Mineral oil hydrocarbons (MOH) might be the quantitatively technically less demanding alternative.
most important food contaminant. Their presence is almost ubiq- The standard method applied in the far past, primarily in
uitous and concentrations may be high. Recently EFSA published environmental analysis, determined the sum of the mineral oil
their scientific opinion on mineral oil hydrocarbons and identi- hydrocarbons by infrared (IR) spectroscopy. Extracts in carbon
fied numerous sources [1]. Mineral oil products consist of complex tetrachloride were purified by retention of polar constituents on
mixtures. Here two fractions are analyzed: the mineral oil sat- Florisil or silica gel and analyzed by quantitative IR in the C H
urated hydrocarbons (MOSH), comprising paraffins (open chain stretching region. Detection limits were reported as 1 mg/kg for
alkanes) and naphthenes (hydrocarbons with at least one satu- feeds and 10 mg/kg for tissue [7]. However, this method does not
rated ring), and the mineral oil aromatic hydrocarbons (MOAH). distinguish the mineral hydrocarbons from hydrocarbons endoge-
The MOAH include polyaromatic compounds, but in contrast to nous in foods.
the polyaromatic hydrocarbons (PAH) widely analyzed, they almost Present methods are based on FID because of calibration prob-
exclusively exist as alkylated species and, therefore, exist in enor- lems encountered with other detection methods, particularly mass
mous numbers of isomers. spectrometry (MS) [2]. FID is the only method available for a quan-
Most data on mineral oil concentrations in food was generated titative determination of mixtures of hydrocarbons which are not
by on-line coupled high performance liquid chromatography–gas available as standards. As it provides virtually the same response
chromatography–flame ionization detection (HPLC–GC–FID), per mass of hydrocarbons, any standard can be used for determin-
recently reviewed in [2,3]. An analogous method with another ing any mineral oil. However, FID is of modest sensitivity, which is a
HPLC–GC interface was described in [4]. On-line HPLC–GC for particularly severe drawback for MOSH and MOAH analysis as they
the determination of mineral oil is from the early 1990s [5] and form broad humps. In fact, 50–100 ng MOSH or MOAH is required
to be measurable [2].
GC is the separation technique of choice because it enables the
∗ Corresponding author. Tel.: +41 43 244 71 31; fax: +41 43 244 71 01. distinction of the mineral oil hydrocarbons from hydrocarbons nat-
E-mail address: [email protected] (K. Grob). urally occurring in food. It also enables the characterization of the

0021-9673/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.chroma.2012.11.034
 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200 193

mineral oil hydrocarbons by the molecular mass distribution and 2. Experimental

the pattern of peaks on top of the hump of unresolved components.
However, MOSH and MOAH cannot be separated by GC, i.e. this 2.1. Materials
must be achieved by preseparation.
In mineral oil analysis by GC–FID, the challenges are (i) the Hexane from Brenntag (Schweizerhall AG, Basel, Switzerland)
preseparation of the MOSH and MOAH, (ii) the removal of inter- was redistilled. Alternatively “Baker Ultra Resi-Analyzed” hexane
fering material, in particular the wax esters, olefins and lipids, 9262 (J.T. Baker, Deventer, The Netherlands) or hexane Chro-
(iii) the reconcentration to achieve a limit of quantitation in the masolv (Sigma–Aldrich, Buchs, Switzerland) was directly used.
range between 0.1 and 1 mg/kg food as required from present Toluene Pestanal 34494 from Fluka (Buchs, Switzerland) was puri-
toxicological evaluation [8] and (iv) elevated fat contents in the fied by passing 10 ml through a column packed as for the samples.
samples. Dichloromethane was either Ultra Resi-Analyzed 9264 from Baker
Almost all previously described sample preparation methods or Chromasolv from Sigma–Aldrich. Ethanol 8462 from Baker
were limited to the analysis of the MOSH. Early methods started or Lichrosolve from Merck was used. 1,1,2-Trichloroethane was
with the saponification of lipids and were derived from those devel- from Fluka, silver nitrate and silica gel 60, 0.063–0.200 mm, from
oped for minor components in edible fats and oils, e.g. [9,10]. Merck, Darmstadt, Germany. The silica gel/10% silver nitrate from
This step can be avoided when the liquid chromatographic column Sigma–Aldrich/Fluka contained hydrocarbons (presumably olefins
has sufficient capacity to retain the lipids, first of all the triglyc- from the closure). It was purified by packing into a column and rins-
erides. ing with toluene, then dichloromethane and hexane, followed by
A method designed for isolating the MOSH from up to 100 mg drying in a rotary evaporator at 70 ◦ C.
lipids involved a small liquid chromatography column containing Standards for MOSH analysis were n-undecane (n-C11 ),
3.5 g activated aluminum oxide [11]. Of the reconcentrated elu- n-tridecane (n-C13 ), cyclohexyl cyclohexane (Cycy) and 5-alpha-
ate, 50 ␮l was injected into GC-FID by the on-column/retention cholestane (Cho); those for MOAH analysis were pentyl-benzene
gap technique to obtain a detection limit of around 1 mg/kg fat. (5B), 1- and 2-methyl-naphthalene (MN), 1,3,5-tri-tert-butyl-
If olefins from the edible oil, such as squalene isomerization prod- benzene (TBB) and perylene (Per; all from Fluka/Sigma–Aldrich).
ucts and sterenes, interfered, the edible oil was brominated prior to Verification standards used in the context of method develop-
preseparation to increase the retention of the olefins. Later it was ment were hexaethyl-benzene, 1,4-bis(2-ethylhexyl)-benzene,
found that activated silica gel not only retains olefins beyond the 1,4-bis(3,7-dimethyloctyl)-benzene, 1-phenyloctadecane and
MOSH, rendering bromination unnecessary, but also has increased 1,2,3,5-tetracyclohexyl-benzene (all from Fluka/Sigma–Aldrich).
capacity to retain lipids. 250 mg edible oil or fat could be loaded The wax ester stearyl heptadecanoate (WE35) was a gift from Carlo
onto 2 g silica gel activated at 400 ◦ C overnight [12]. Fiorini et al. Mariani (Milan). Gravex 913, a hydrotreated naphthenic mineral
[13] proposed using non-activated silica gel to obtain a narrower oil product used in printing inks from Shell was a gift from Shell
fraction and to inject in standard splitless mode. Hamburg (Germany).
After the finding of high mineral oil concentrations in Ukrainian Stock solutions of 100 mg C11 , C13 , Cycy, 5B, 1MN, 2MN and TBB
sunflower oils early in 2008 [14], many control laboratories started were prepared in 10 ml toluene or 1,1,1-trichloroethane. For stan-
analyzing MOSH, though at a rather high detection limit, as the dard solution 1, 12 mg each of Per and Cho was weighed into a 20 ml
European Commission had set a 50 mg/kg legal limit [15]. To measuring flask to which 600 ␮l of each stock solution (only 300 ␮l
support control, the EU Joint Research Centre (JRC)/Institute for of C13 ) was added. The flask was filled up with toluene or 1,1,1-
Reference Materials and Measurements (IRMM) organized a pro- trichloroethane, resulting at 0.3 mg/ml except for C13 (0.15 mg/ml)
ficiency test. Totally 55 laboratories provided results [16]. Of those as well as Cho and Per (0.6 mg/ml). For standard solution 2, solution
specifying the method, the majority worked with columns with 1 was diluted to 10 ␮g/ml (1:30).
10–30 g packing and a correspondingly high solvent consump-
tion. 2.2. Extraction of food and paperboard samples
Moret et al. [17] optimized the approach for MOSH analysis
toward small column size (1 g silica gel for 250 mg oil) with a low Extraction of the various types of samples was described in
solvent consumption. For the complete removal of the olefins with- detail in [2]. Dry food samples were extracted by immersion
out derivatization, the silica gel contained 10% silver nitrate. Of the in hexane, mostly overnight at room temperature. Homoge-
1.5 ml fraction of hexane containing the MOSH, 40 ␮l was injected nized paperboard samples were extracted by immersion in
by the on-column/retention gap technique, resulting in a limit of ethanol/hexane (1:1) at room temperature for 2 h. Completeness
quantitation in vegetable oil of 15 mg/kg. of the extraction was checked by re-extraction at accentuated con-
Moret et al. [18] extended this method to MOAH analysis: ditions, as extractability of solid samples may strongly vary. Wet
the MOAH were eluted from same column packed with silica samples were first blended into ethanol and allowed to stand for
gel/silver nitrate using hexane/dichloromethane 1:1. Fast program- 1 h to largely exchange the water in the pores against ethanol. Then
ming of the oven temperature (50◦ /min) accelerated the elution the ethanol was decanted and the residue extracted by immersion
and resulted in a gain in sensitivity. in hexane overnight. The ethanol and the hexane were combined
The method described in this paper aimed at separate MOSH and the hydrocarbons collected in the hexane phase by addition of
and MOAH analysis with a performance and sensitivity com- water.
parable to on-line HPLC–GC, but only requiring conventional
GC equipment. The target was a limit of quantitation suitable 2.3. Preseparation by column liquid chromatography
for the control of a 0.6 mg/kg limit for MOSH in the major-
ity of the foods. Since the aliquot of sample loaded onto the For the preseparation, 30 cm × 11 mm i.d. glass columns with a
column is limited by the fat content, the same compromise frit (Fisher Scientific, article B53275, Wohlen, Switzerland) were
was chosen as for on-line HPLC–GC: instead of using a large packed with silica gel containing 0.3% silver nitrate. Thirty-three
packed bed adjusted to the samples with highest fat content, the grams of silica gel coated with 1% silver nitrate were mixed with
method was configured for samples containing up to 20% fat, 67 g activated silica gel. Silica gel was coated with 1% silver nitrate
applying separate enrichment for those of a higher fat content by dissolving 0.5 g silver nitrate in 50 ml water and addition to
[2,3]. 49.5 g silica gel (round flask protected with aluminum foil against
194 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200

light). The mixture was homogenized and dried on the rotary evap- hexane, permanently keeping the oven at 80 ◦ C and inlet pressure
orator during 3 h at 70 ◦ C and 120 mbar, then heated out in a GC at 250 kPa. The duration of solvent evaporation was measured at
oven programmed from 70 to 125 ◦ C at 6◦ /min and left at 125 ◦ C for maximum attenuation by the width of the solvent peak at the top.
15 h. This duration minus 10 s was entered for the moment to reduce the
The glass column was placed onto the balance in a high beaker inlet pressure to 60 kPa.
and filled with 3 g silica gel/0.3% silver nitrate. The column was The oven temperature was programmed starting after the elu-
rinsed with 10 ml hexane, then loaded with 1 ml sample extract or tion of the C11 standard. As this retention time depended on the
solution of fat or oil (for recycled paperboard only 0.3 ml followed toluene concentration in the solution injected, it was established
by 0.7 ml hexane). After running the column almost to dryness, 2 ml with a mixture of 55% toluene in hexane, of which 50 ␮l was
hexane was added and the eluate discarded. Two round flasks were injected, with return to normal inlet pressure as determined above
prepared by rinsing with hexane, then 4 ml hexane added to the and leaving the oven isothermal at 80 ◦ C.
column and collected in the first flask. The eluent was switched
to a mixture prepared by adding 20 ml dichloromethane and 5 ml 2.6. Blanks
toluene into a 100 ml measuring flask and filling it up to 100 ml with
hexane. Of this eluent, 2 ml was added to the column and collected Blank tests ensured absence of sample contamination during
in flask 1 (MOSH fraction of totally 6 ml). Another 10 ml was added preparation. They were designed to take into consideration all
and collected in flask 2 (MOAH fraction). potential sources and were performed frequently to take into
The two fractions were reduced to a residue rich in toluene by account that such contamination may occur only occasionally, e.g.
solvent evaporation (the MOSH fraction after addition of 270 ␮l by somebody with cream on her/his hands or a vial which was
toluene) using a rotary evaporator with the water bath at 55 ◦ C, not properly cleaned (mineral oils may be difficult to wash off in a
first at 500 mbar for 90 s, then at 350 mbar during 3–4 min, until dishwasher). To this end samples virtually free of mineral oil were
200–300 ␮l of liquid was left. The system was aired through a tube reanalyzed to be as realistic as possible.
feeding the air close to the round flask in order to prevent recon- Analyses were not repeated within a series, but on different days
densed solvent to flow back into the sample. A restriction prevented in order to avoid that the same contamination was repeated in case
an excessive flow rate. there was one.

2.4. Optimization of selectivity in preseparation

2.7. Quantitative determination
Selectivity in preseparation to isolate MOSH and MOAH was
optimized using a mixture of marker substances: of n-triacontane The interpretation of chromatograms was described in detail in
(C30 ), 5-alpha-cholestane (Cho), tri-tert-butyl benzene (TBB), pery- [3,19]. Briefly, the baseline was transferred from blank runs and its
lene (Per) and heptadecyl stearate (wax ester 35, WE), a solution of level fitted to points outside the regions occupied by the humps of
1 mg/ml was prepared in toluene. Separation was investigated by MOSH or MOAH, which means usually at the beginning or toward
adding 20 ␮l of this solution onto the column and taking 2 ml frac- the end of the chromatogram. Then vertical cuts were introduced
tions. Of these, 10 ␮l was injected on-column onto a 7 m × 0.25 mm to delimit the fractions to be distinguished, for the MOSH at n-
i.d. GC column coated with a 0.13 ␮m film of PS-255 and equipped C16 (C10 –C16 fraction with a specific migration limit of 12 mg/kg in
with a 2 m × 0.53 mm i.d. uncoated, deactivated precolumn. food according to the BfR [20]) and at n-C35 , beyond which absorp-
tion into human metabolism was assumed to become negligible.
2.5. GC analysis For MOSH and MOAH extracted from packaging material, a cut was
introduced at a retention time corresponding to the end of the n-C24
GC involved a Trace chromatograph (Official Food Control peak for the estimation of the hydrocarbons which could be trans-
Zürich, KLZH; Thermo Scientific, Milan, Italy) equipped with a con- ferred to dry foods through evaporation/recondensation at room
ventional split/splitless as well as an on-column injector and a temperature.
TriPlus autosampler, or a 6890 chromatograph (German Federal
Institute for Risk Assessment, BfR; Agilent, Waldbronn, Germany) 3. Results concerning method development
equipped with an on-column injector and a xyz autosampler. The
separation columns were coated with a dimethyl polysiloxane sta- 3.1. Optimized selectivity of the preseparation
tionary phase (KLZH: 10 m × 0.25 mm i.d., 0.13 ␮m PS-255 from
Fluka; BfR: DB1-HT, 15 m × 0.32 mm i.d., 0.10 ␮m, Agilent, Wald- The selectivity of the liquid chromatographic preseparation
bronn, Germany). To enable 40–50 ␮l injections, these columns was optimized to achieve a robust separation between MOSH and
were equipped with a 7 m × 0.53 mm i.d. uncoated precolumn, MOAH, but also between the MOAH and other components of food
either consisting of raw fused silica (KLZH; BGB Analytik, Boek- or paperboard, such as oxygenated compounds of low polarity. It is
ten, Switzerland) or a 10 m × 0.53 mm i.d. (BfR; SGE, Griesheim, this latter aspect in which the method differs from that described
Germany), using a press-fit connector. in [18]. In particular it meant increasing the retention power for
On-column injection occurred slowly (KLZH: at 5 ␮l/s; BfR: MOAH only as much as necessary to separate them from the MOSH
“slow”) and the needle was withdrawn 10 s after depressing the and to keep the retention power for other compounds as high as
plunger. The detector base block for FID was thermostatted at possible.
370 ◦ C. The flame gases consisted of 35 ml/min hydrogen and For the optimization of the separation, marker compounds were
550 ml/min air (higher than usual to support complete combustion used starting from the experience gained in HPLC on silica gel [2]. In
of toluene). normal phase liquid chromatography on silica gel, the MOSH frac-
During injection, the carrier gas (hydrogen) inlet pressure was tion started with the paraffins being eluted with breakthrough, the
at 250 kPa in order to speed up the evaporation of the toluene. high molecular mass paraffins being eluted first owing to a weak
This pressure was reduced to 60 kPa shortly before chromatogra- size exclusion effect. The n-alkane C30 was used as a marker. The
phy started, i.e. shortly before the end of solvent evaporation. The naphthenes were somewhat retained. 5-Alpha cholestane (Cho)
moment for reducing the inlet pressure was determined by a 50 ␮l with four rings and a low degree of alkylation was chosen as a
injection of a mixture with relatively low toluene content (40%) in marker for the end of the MOSH fraction.
 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200 195

Fig. 1. Results from the optimization of preseparation. (A) Activated straight silica gel; (B) 1% silver nitrate on silica gel; (C) mixed packed bed of 2/3 activated silica gel and
1/3 silica gel with 1% silver nitrate eluted with 30% dichloromethane/hexane and (D) same mixed packing, but 30% toluene/hexane as eluent.

The retention of the MOAH increased with the number of aro- to 125 ◦ C, prepared as described in Section 2, 100% dichloromethane
matic rings and decreasing alkylation, i.e. highly alkylated benzenes was needed to get TBB eluted as a reasonably narrow band (10 ml),
were eluted first, polyaromatic hydrocarbons without alkylation but Per could not be recovered. The WE was partly coeluted with
last. The sterically well protected 1,3,5-tri-tert-butyl benzene (TBB) TBB (Fig. 1B). It was concluded that the retention power for the
was chosen as a marker for the beginning of the MOAH frac- MOAH had to be reduced whereas that for the wax esters needed
tion and, hence, Cho and TBB had to be separated well enough to being increased.
ensure robust separation of the MOSH and the MOAH. As marker The retention of the wax esters can be increased by activat-
for the end of the MOAH fraction, perylene (Per) was chosen, a ing the silica gel. However, silica gel with silver nitrate cannot be
non-alkylated 5-ring polyaromatic hydrocarbon with strong flu- heated above about 150 ◦ C without degradation. Hence, a packed
orescence, often even visible by eye in normal light. bed mixed of 1/3 silica gel with 1% silver nitrate and 2/3 silica gel
The wax esters (long chain alcohols esterified with long chain activated at 400 ◦ C was tested, combining a decreased retention
fatty acids) are food components eluted shortly after the MOAH of MOAH with increased retention of oxygenated compounds. The
fraction and may be present in concentrations severely overload- decrease of the amount of silver to 0.3% enabled the elution of the
ing GC when eluted in MOAH fractions. Stearyl heptadecanoate TBB by only 30% dichloromethane in hexane. As shown in Fig. 1C,
(WE35), one of the early eluted wax esters, was used as a marker. this increased the retention of the WE clearly beyond TBB, but Per
Since it had previously been observed that alkylated benzenes was still not be eluted.
were barely retained on non-activated silica gel, the first tests The elution of Per required a selective increase in eluent
involved silica gel activated at 400 ◦ C, initially with hexane, after strength. Toluene efficiently deactivates the retention power of sil-
fraction 3 with 50% dichloromethane/hexane as mobile phase. Plot ver nitrate for unsaturated hydrocarbons without being a strong
A in Fig. 1 shows the result in terms of peak areas of the marker eluent for activated silica gel. Using the same mixed packed bed
compounds in fractions of 2 ml each. There was some presepara- and 30% toluene/hexane, this approach was successful with regard
tion between C30 and Cho, but the saturated hydrocarbons tailed to the intended selectivity (Fig. 1D): the Cho/TBB separation was
into the fraction of the TBB, i.e. the MOSH/MOAH separation was excellent, Per was eluted closely after TBB and the WE was not
considered inadequate. Per was eluted soon afterwards, keeping eluted at all.
the MOAH fraction narrow, but it was partly coeluted with the wax However, fractions containing such a high proportion of toluene
esters (Fig. 1A). It was concluded that separation on silica gel was cannot be as strongly reconcentrated as required to achieve the
unsatisfactory, even when the silica gel was activated, because of detection limit: evaporation of toluene is accompanied by massive
the far lower separation efficiency compared to HPLC, where a gap losses of volatile hydrocarbons. Therefore the toluene concentra-
of 30 s is observed between Cho and TBB on non-activated silica gel. tion was reduced to 5% and substituted by 20% dichloromethane.
Silver nitrate strongly increases the retention of unsaturated In this way toluene even facilitated reconcentration as it acted
hydrocarbons [17,18], but easily to an extent that it becomes diffi- as a keeper and stopped the evaporation process. As shown in
cult to recover the MOAH. With 1% silver nitrate on silica gel heated Fig. 2A, Per was somewhat more retained, but the width of the
196 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200

Fig. 2. Preseparation as in Fig. 1D, but toluene partially substituted by dichloromethane: 5% toluene/20% dichloromethane/hexane for the elution of the MOAH. The cut
applied in the final method being marked. (A) Performance with standards and (B) with the addition of 270 mg sunflower oil to test the effect of an exaggerated amount of
lipids on the MOSH/MOAH separation.

MOAH fraction was still acceptable; the WE was absent. This was muesli, cereals or bakery ware) immersed in 20 ml hexane, which
the preseparation selected. means that such solutions cannot be reconcentrated. Samples con-
taining up to 4% fat, such as rice, noodles, couscous or semolina,
can be reconcentrated at least five times before the preseparation
3.2. Capacity of the column to retain fat and detection limit is performed, which enables a decrease of the detection limit by a
factor of 5.
The packing of the liquid chromatography column not only has
to achieve preseparation between MOSH, MOAH and other sample
components, but also has to retain the lipids, primarily the triglyc- 3.3. Reconcentration of MOSH and MOAH fraction
erides, as these would severely overload the GC column. With the
eluents used, triglycerides are well retained, but large amounts As a rule of thumb, 50–100 ng MOSH or MOAH must reach
deactivate the silica gel, with the effect that the following triglyc- the flame ionization detector to obtain a hump that can be inte-
erides enter deeper into the column. Such overloading is acceptable grated with adequate certainty [2]. If the limit for quantitation
as long as triglycerides leave enough packing for the preseparation in food should be around 0.5 mg/kg, an aliquot corresponding to
of the MOSH, MOAH and wax esters (triglyceride-coated silica gel 100–200 mg food must be injected into GC. If 20 g food is extracted
hardly retains MOAH and wax esters). A maximum load is of inter- with 20 ml hexane [2] without reconcentration before presepa-
est to maximize the sample aliquot that can be analyzed and the ration and conventional splitless injection of a 1 ␮l volume were
sensitivity achievable. It was assumed that up to half of the packing applied, the fractions obtained by the column described above had
could be exploited for the retention of the lipids without signifi- to be reconcentrated to 5–10 ␮l. This is not practical for a routine
cantly compromising the resolution of the target components [21]. method and would be linked with heavy losses of volatile MOSH
The capacity of the column was determined by loading 600 mg and MOAH. Therefore the fractions obtained must be injected in
vegetable oil and elution in the same way as for the MOSH large volumes. With a 50 ␮l injection, the sample must be recon-
and MOAH, i.e. initially with hexane, then with 5% toluene/20% centrated to 250–500 ␮l.
dichloromethane/hexane. The two fractions were collected into a If the method should enable the analysis of hydrocarbons as
tared round flask, evaporated to dryness and weighed. The residue from n-C10 , care is required to avoid losses during reconcentration
amounted to 200 mg for a commercial refined sunflower oil and of the sample to the volume derived above. Fractions must not be
185 mg for a crude olive pomace (sansa) oil. It was concluded that brought to dryness, but losses may also occur before. Solvent evapo-
400 mg lipids were retained by the packing, i.e. up to 200 mg could ration in a gas stream tends to result in high loss of hydrocarbons up
be loaded onto the column for our application. to about n-C16 [13,22]. Particularly solutions wetting glass surfaces
The effect of overloading the column by lipids on the are pulled upwards on the wall of the flask and when the solvent
MOSH/MOAH separation was checked by repeating the experiment is evaporated, the components are deposited onto a dry surface
shown in Fig. 2A, but adding 270 mg sunflower oil to the solution which is swept by the gas stream. In analogy to GC it depends on
(more than intended for the method). As shown in Fig. 2B, the bands the retention power of this surfaces up to which molecular mass
of C30 and Cho were broadened and the MOAH were eluted slightly the hydrocarbons are removed; uncoated surfaces in a precolumn
earlier (deactivation effect of the lipids), but the system was suf- exert a low retention power [23]. Losses are lower when the sample
ficiently robust to achieve the separations. This capacity of the 3 g contains a non-evaporating sample matrix like fat.
packing for 200 mg fat corresponds to ten times that of the HPLC Using rotary evaporators, the surface from which solvent evap-
column used for on-line HPLC–GC [2] and means that injection of orates is rinsed with every rotation and the solute material rinsed
a tenth of the eluted fractions into GC should result in the same back into the solution. This keeps losses during solvent evaporation
sensitivity. low until the residual solution withdraws rapidly and the rinsing
The capacity of the packing for lipids determines the aliquot of effect loses efficiency, usually when arriving at the last 5–10% of
food extract that can be loaded onto the column: the fat content the sample volume.
in the 1 ml extract or solution loaded must not exceed 200 mg. For A keeper reduces the losses of volatiles in the same way as a
edible oils and fats, the HPLC–GC method provided the dissolution stationary phase increases retention power in GC. Toluene cho-
of 300 mg sample in a 1.5 ml autosampler vial filled up with hexane. sen for the elution of the MOAH seemed to be a good choice.
Such 20% concentration also fits this method. When 10 g chocolate Firstly, it remains on the glass wall for some time after hexane
containing roughly 40% fat is dissolved/suspended in 20 ml hex- was evaporated and may retain the solutes until rotation brings
ane, the fat concentration in the solution is at the limit. The same them back to the bulk of the liquid. Secondly, its high boiling point
applies to a 20 g sample of dry food containing 20% fat (as some (110 ◦ C) renders the end point of the evaporation more robust. As
 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200 197

solvent evaporation strongly slows down with increasing toluene

concentration in the residual liquid, there is little risk of evapo-
rating the fraction to dryness. It was determined by GC–FID that
the MOAH fraction included approximately 3.5% toluene (includ-
ing some early eluent without toluene) and that the reconcentrate
of around 250 ␮l contained 40–50% toluene. Even evaporation pro-
longed by a minute neither strongly reduced the residual sample
volume nor the toluene content in the final solution.
The MOSH fraction does not include toluene to begin with. For
the above advantages, but also to achieve identical conditions for
injection of MOSH and MOAH, toluene was also added to the MOSH
fraction. The volume (270 ␮l) was selected such that the toluene
content in the resulting reconcentrate was similar to that in the
MOAH reconcentrate (which was a larger fraction and also con-
tained dichloromethane).

3.4. Solvent evaporation in GC

Large volume injection in capillary GC forms solvent vapor

in volumes which are large compared to the column flow rate,
i.e. risks to form broad solvent peaks. Programmed temperature
vaporizing (PTV) injection in solvent split mode avoids this by dis-
charging most vapors through the split outlet at a high flow rate
[24]. However, the retention of the volatile solutes in the injector
presupposes closure of the split outlet before solvent evaporation
is completed, which is tedious for mixtures containing variable
amounts of toluene. Solvent evaporation in the GC precolumn is
more robust since there is no exit other than the detector. This was Fig. 3. Evaporation of a test sample of n-alkanes using a rotary evaporator (n = 6).
(a, upper plot) peak areas of the six experiments bunched by the n-alkanes, pointing
the reason for preferring on-column injection or splitless injec- out variability of absolute areas resulting from differing final volumes. (b) Data for
tion with concurrent solvent evaporation (CSR). For the method a single experiment showing virtual absence of discrimination.
described here the on-column technique was utilized.
Efficient discharge of solvent through the column requires a high
flow rate of solvent vapors. This flow rate is determined by the Toluene tends to form black smoke in the flame ionization detec-
total flow rate (inlet pressure) and the concentration of the solvent tor, eventually resulting in sooting the detector and formation of
vapors in the gas phase (vapor pressure). The vapor concentration spikes in the chromatograms. The strong dilution in carrier gas
is a function of the oven temperature. At the solvent boiling point resulting from the low oven temperature chosen counteracts this.
at the given carrier gas pressure, the solvent vapors would com- Further, the air flow rate into the detector was doubled to ensure
pletely replace the carrier gas. However, since this temperature that there was an excess of oxygen.
would inhibit the solvent trapping process [2,25], a slightly lower
oven temperature must be selected, which may still be above the 3.5. Method validation
standard boiling point.
The injection of the reconcentrated fractions cannot really be To check the linearity of the response, 40 ␮l of a solution con-
optimized in this way. The boiling point is initially determined by taining n-alkanes from C11 to C40 in six concentrations ranging from
hexane (69 ◦ C at ambient pressure; dichloromethane in the MOAH 0.2 to 10 ␮g/ml was injected 5 times. It resulted in linear correla-
fraction was removed during reconcentration), then by toluene tion with R2 of at least 0.999 for all n-alkanes. This test also showed
(110 ◦ C). With a high inlet pressure, still an oven temperature of equal response for all n-alkanes for all 6 concentrations, i.e. absence
90 ◦ C would be possible. However, the analysis requires sufficient of discriminative effects.
separation of the verification standards C11 and 5B from the sol- On-column injections of 40 ␮l Gravex 913 (a mineral oil product
vent, primarily the toluene. As a short, thin film column is used in used for off-set printing inks; chromatograms shown in Fig. 5) in
order to keep the column bleed low, the oven temperature cannot hexane at 30, 75 and 150 ␮g/ml showed linearity corresponding to
be higher than 80 ◦ C, which keeps the vapor pressure of toluene R2 of 0.9975. The coefficients of variation for the integrated area of
unfavorably low. the broad humps were below 2% for each concentration (n = 4).
An acceptable width of the solvent peak was obtained by a Before GC analysis, MOSH and MOAH fractions were reconcen-
strong increase of the carrier gas flow rate/inlet pressure during sol- trated to a volume of 200–300 ␮l by means of a rotary evaporator,
vent evaporation. Since solvent trapping retains the solutes in the using toluene as keeper. Six samples of n-alkanes were reconcen-
condensed solvent, no chromatography occurs until solvent evap- trated as described in Section 2. The final volumes varied within
oration in the column inlet is completed. This high inlet pressure about 30%, but there was no discrimination according to volatility
should, therefore, be reduced to the level suitable for chromatogra- (Fig. 3). Due to the use of internal standards, the variation of the
phy shortly before the last solvent is vaporized. As the proportion volumes had no effect on quantitative analysis.
of the toluene in the injected sample varied, the time for reducing For the verification of the MOSH/MOAH separation optimized
the inlet pressure had to be adjusted to a sample with a low toluene by the markers Cho and TBB, additional compounds expected
content. The method for determining this moment was described to be eluted early in the MOAH fraction were tested: 1,4-bis(2-
in Section 2. Temperature programming was started after elution ethylhexyl) benzene, 1,4-bis(3,7-dimethyloctyl) benzene, 1-phenyl
of the verification standards C11 and 5B. This time was determined octadecane and 1,2,3,5-tetracyclohexyl benzene. They were all
by a mixture with a toluene concentration at the upper limit of that found quantitatively in the MOAH fraction (Fig. 4). None was
occurring in reality. detectable in the MOSH fraction.
198 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200

Table 1
Comparison of MOSH and MOAH concentrations measured in rice by on-line
HPLC–GC–FID and the manual method.

MOSH concentrations (mg/kg) MOAH (mg/kg)

C10–16 C16–25 C25–35

HPLC–GC–FID (n = 6)
Mean value 1.0 11.6 1.3 2.6
RSD (%) 5.3 1.3 6.4 3.2
Manual method
1 1.0 11 2.2 2.6
2 1.0 11.2 2.0 3.6
Mean value 1.0 11.1 2.1 3.1

initial weight. The mineral oil was composed of 72.5 ± 0.5% MOSH
and 27.5 ± 1.7% MOAH (Fig. 5), which was in agreement with the
result obtained by on-line HPLC–GC-analysis: 72.9% MOSH, 27.1%
MOAH.
A sample of rice contaminated from recycled paperboard (avail-
able in the BfR kit for testing method performance [26]) with
approximately 12 mg/kg MOSH and 3 mg/kg MOAH was analyzed
Fig. 4. Verification of the MOSH/MOAH separation by highly alkylated ben- by the manual method presented here as well as by on-line
zenes: hexaethyl benzene (1), 1,4-bis(2-ethylhexyl) benzene (2), 1,4-bis(3,7-
HPLC–GC. Fig. 6 shows good agreement of the chromatograms
dimethyloctyl) benzene (3), 1-phenyl octadecane (4), 1,2,3,5-tetracyclohexyl
benzene (5). 5B, MN, TBB and Per were marked with additional arrows.
obtained from the two methods, Table 1 that of the quantitative
data.
Fig. 7 shows the analogous comparison for a recycled paper-
The recovery of a mineral oil (Gravex 913) was determined board (also part of the BfR kit [26]). Table 2 lists the corresponding
working up 1 ml of a 1000 mg/l solution in hexane four times. The concentration data.
sum of the MOSH and the MOAH corresponded to 99.9 ± 2.3% of the Agreement between the proposed LV–GC–FID method with
amount determined by direct injection of the test solution and to manual sample preseparation and on-line-HPLC–GC–FID was also
96.3% when calculated by the internal standards according to the noted for a rapeseed oil spiked with 50 mg/kg Gravex 913 (Table 3).

Fig. 5. Chromatograms from a 1000 mg/l solution of Gravex 913 with direct injection (A; 963 mg/l) and after preseparation in MOSH (B; 694, mg/l) and MOAH (C; 264 mg/l).
 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200 199

Fig. 6. Chromatograms from contaminated rice obtained by on-line HPLC–GC (A and B) and the manual method (C and D).

Fig. 7. Chromatograms from recycled paperboard obtained by on-line HPLC–GC (A and B) and the manual method (C and D).
200 K. Fiselier et al. / J. Chromatogr. A 1271 (2013) 192–200

Table 2 in oils and fats the n-alkanes naturally present (usually C23 –C33 )
Comparison of data obtained from a paperboard sample analyzed with on-
may overload GC for MOSH analysis to such an extent that previ-
line HPLC–GC–FID or LV–GC–FID with manual preseparation; DIPN, diisopropyl
naphthalene. ous removal by chromatography on activated aluminum oxide is
required; (iii) olefins like squalene and its isomerization products
Method Concentrations (mg/kg)
present, e.g. in olive oil may seriously overload GC for MOAH analy-
MOSH MOAH DIPN sis and require epoxidation to enhance their polarity such that they
C10–16 C16–24 C24–35 <C35
are eluted after the MOAH. These tools were summarized in [2].
The lack of data on concentrations of MOSH and MOAH in dif-
HPLC–GC–FID (n = 4)
ferent food groups and from different countries mentioned in the
Mean value 21 407 332 54 19
RSD (%) 4.5 2.7 4.7 3.7 2.7 EFSA opinion [1] can be effectively tackled by these two methods.
Manual method 20 390 342 58 18
References
Table 3
Comparison of data for an edible oil spiked with 50 mg/kg Gravex 913 analyzed by [1] European Food Safety Authority (EFSA), Scientific opinion on
mineral oil hydrocarbons in food, EFSA J. 10 (6) (2012) 2704
on-line HPLC–GC–FID or the manual method.
http://www.efsa.europa.eu/en/efsajournal/pub/2704.htm
Laboratory Method Concentrations (mg/kg) [2] M. Biedermann, K. Grob, J. Chromatogr. A 1255 (2012) 56.
[3] M. Biedermann, K. Grob, J. Chromatogr. A 1255 (2012) 76.
MOSH MOAH Sum [4] P.Q. Tranchida, M. Zoccali, G. Purcaro, S. Moret, L. Conte, M. Beccaria, P. Dugo,
L. Mondello, J. Chromatogr. A 1218 (2011) 7476.
Manual method 37 13.5 50.5 [5] K. Grob, M. Lanfranchi, J. Egli, A. Artho, J. AOAC Int. 74 (1991) 506.
BfR HPLC–GC–FID 1 36.4 10.7 47.1 [6] M. Biedermann, K. Fiselier, K. Grob, J. Agric. Food Chem. 57 (2009) 8711.
HPLC–GC–FID 2 36.3 11.4 47.7 [7] D.G. Walters, K.V. Sherrington, N. Worrell, R.A. Riley, Formulation and analysis
of food-grade mineral hydrocarbons in toxicology studies, Food Chem. Toxicol.
KLZH (n = 8) HPLC–GC–FID 37.7 12 49.7
32 (1994) 549.
[8] JECFA, Joint FAO/WHO Expert Committee on Food Additives, 59th report,
2002, pp. 11–20; WHO Technical Report Series 913, http://whqlibdoc.who.
Originally the oil contained less than 1.5 mg/kg MOSH and less int/trs/WHO TRS 913.pdf
than 0.4 mg/kg MOAH. This sample is again part of the BfR kit and [9] L. Castle, M. Kelly, J. Gilbert, Food Addit. Contam. 10 (1993) 167.
[10] A. Guinda, A. Lanzón, T. Albi, J. Agric. Food Chem. 44 (1996) 1723.
additional information is available in [26]. [11] Ch. Wagner, H.-P. Neukom, V. Galetti, K. Grob, Mitt. Lebensm. Hyg. 92 (2001)
The data on reproducibility and the agreement with results 231.
obtained by on-line HPLC–GC–FID shows that measurement uncer- [12] K. Fiselier, K. Grob, Method shown during the workshop “Mineral Oil Material
in Foods: Analytical Methods, Occurrence, Evaluation” by EU-DG SANCO and
tainty resulting from the sample preseparation and LV–GC–FID
the Official Food Control Authority of Zurich, Zurich, Switzerland, September
analysis is low and, as for on-line-HPLC–GC–FID [2,3], usually small 17/18, 2008.
compared to that which may result from the interpretation of the [13] D. Fiorini, A. Paciaroni, F. Gigli, R. Ballini, Food Control 21 (2010) 1155.
[14] M. Biedermann, K. Grob, Eur. J. Lipid Sci. Technol. 111 (2009) 313.
chromatograms. Positioning of the baseline and the upper con-
[15] Summary minutes of the meeting of the Standing Committee on the Food
tour line of the hump against components other than MOSH and Chain and Animal Health, Brussels, 20 June 2008. http://ec.europa.eu/food/
MOAH creates uncertainty depending on the chromatogram and committees/regulatory/scfcah/toxic/summary20062008 en.pdf
the diligence of analyst. [16] Joint Research Centre (JRC), Scientific and Technical Report, EUR 23811EN,
Final report on proficiency test on the determination of mineral oil in sun-
flower oil, 2009. Available at: http://publications.jrc.ec.europa.eu/repository/
4. Conclusions bitstream/111111111/12584/1/eur%2023811%20en%20-%20minoil%202009
%20%20-%20%20lk tw.pdf
[17] S. Moret, L. Barp, K. Grob, L.S. Conte, Food Chem. 129 (2011) 1898.
An analytical method for the determination of MOSH and MOAH [18] S. Moret, L. Barp, G. Purcaro, L.S. Conte, J. Chromatogr. A 1243 (2012) 1.
in foods and paperboard with a detection limit usually below [19] Bundesinstitut für Risikobewertung (BfR) and Kantonales Labor Zurich,
Messung von Mineralöl – Kohlenwasserstoffen in Lebensmitteln und Verpack-
0.5 mg/kg was developed that avoids the need for special instru- ungsmaterialien, 2012. http://www.bfr.bund.de/cm/343/messung-von-
mentation, in particular an on-line HPLC–GC system. Preseparation mineraloel-kohlenwasserstoffen-in-lebensmitteln-und-
was optimized not only to separate MOSH and MOAH in robust verpackungsmaterialien.pdf
[20] Bundesinstitut für Risikobewertung (BfR), 7. Sitzung der BfR-Kommission
manner, but also to keep other compounds of low polarity out of
für Bedarfsgegenstände, Protokoll der Sitzung vom 14. April 2011,
the MOAH fraction, such as the wax esters. Verification with highly p. 9. http://www.bfr.bund.de/cm/343/7 sitzung der bfr kommission fuer
alkylated benzenes confirmed the selectivity of the separation bedarfsgegenstaende.pdf
[21] K. Grob, I. Kälin, A. Artho, J. High Resolut. Chromatogr. 14 (1991) 373.
between the MOSH and MOAH. The cross validation between the
[22] K. Grob, E. Müller, J. Chromatogr. 404 (1987) 297.
manual method presented here and on-line HPLC–GC–FID showed [23] K. Grob, H.P. Neukom, J. Chromatogr. 323 (1985) 237.
agreeing results for rice samples, edible oil and cardboard. [24] W. Engewald, J. Teske, J. Efer, J. Chromatogr. A 856 (1999) 259.
The method has the same limitations as on-line HPLC–GC–FID: [25] E. Boselli, K. Grob, G. Lercker, J. High Resolut. Chromatogr. 22 (1999) 327.
[26] Bundesinstitut für Risikobewertung (BfR), test kit accompanied by data
(i) it may be necessary to enrich samples with a high fat content and comments, http://www.bfr.bund.de/en/national reference laboratory
to reach a limit of quantitation below 0.5 mg/kg; (ii) particularly for materials in contact with food-10626.html