750 J. Sep. Sci.

2012, 35, 750–757

Abdul Muttaleb Yousef Research Article

Jaber1
Nemr Ahmed Mehanna2
Abdalla Mahmoud
Abulkibash1
1 Chemistry Department, King
Fahd University of Petroleum &
Minerals , Dhahran, Saudi
Arabia
2 Industrial Support Services
Laboratories, Al-Bassam Group
of Companies , Dammam, Saudi
Arabia
Received June 09, 2011
Revised November 15, 2011
Accepted December 08, 2011
Simultaneous liquid–liquid extraction of
dibenzyl disulfide, 2,6-di-tert-butyl-p-cresol,
and 1,2,3-benzotriazole from power
transformer oil prior to GC and HPLC
determination
2,6-Di-tert-butyl-p-cresol (DBPC), dibenzyl disulfide (DBDS), and 1,2,3-benzotriazole (BTA)
are additives that may be found concomitantly in the oil matrix of power transformer.
DBPC and DBDS act as antioxidants while, BTA is a corrosion inhibitor that protects
copper conductors inside the transformer unit from corrosion. A powerful analytical method
is, therefore, required to determine these additives at trace levels in the transformer oil.
This work describes a unique single liquid–liquid extraction pretreatment step prior to the
determination of the components by gas chromatography (GC) and high-performance liquid
chromatography (HPLC) techniques. The optimum volume ratio used in the pretreatment
step was determined as 5:2:5 for mineral oil/n-hexane/acetonitrile, respectively. Relatively,
the method is simple and quick with a minimal use of solvents. Analytical results indicate
that the method is relatively sensitive, accurate, and precise for each of the three components
in fresh and used mineral oil. The calibration curves for the three components demonstrate
a significant increase in sensitivities. Detection limits found were, 100 mg L−1 (0.01% w/v),
0.80 mg L−1 , and 2.04 mg L−1 for DBPC, DBDS, and BTA, respectively. The Student’s
t values determined at 95% confidence level indicate that there is no significant difference
between the experimental means obtained by this method and the standard method for each
component.

Keywords: BTA / Corrosive sulfur / DBDS / DBPC / Liquid–liquid extraction

DOI 10.1002/jssc.201100505

1 Introduction 2,6-Di-tert-butyl-p-cresol (DBPC) is a common and useful

The chemical properties of the mineral insulating oil may
change as a result of aging or other factors where the dete-
rioration products could destroy the insulating property of
the oil. Moreover, a reaction may occur between corrosive
sulfur compounds, when available in the oil, and the cop-
per conductors inside the transformer unit to form copper
sulfide that eventually deposits on the surface of the insulat-
ing paper causing shorting faults between the turns of the
electric transformer. Mineral insulating oil quality can be im-
proved by using specific additives that retard the deterioration
processes.
Correspondence: Professor Abdul Muttaleb Yousef Jaber,
KFUPM, Chemistry Department, Dhahran 31261, Saudi Arabia
E-mail: ajaber@philadelphia.edu.jo
Abbreviations: BTA, 1,2,3-benzotriazole; DBDS, dibenzyl
disulfide; DBPC, 2,6-di-tert-butyl-p-cresol; PCB, polychlori-
nated biphenyles
primary antioxidant additive that is widely used as a radical
trapping agent, thus preventing the degradation of the oil
[1]. 1,2,3-benzotriazole (BTA) is an additive used as a passi-
vator that inhibits the corrosion effects of the organic sulfur
on copper conductors in the transformer unit. Also, many
mineral oils have been found to contain dibenzyl disulfide
(DBDS) [2] which is responsible for the formation of copper
sulfide [3]. DBDS has been used as an antioxidant and as a
stabilizer for petroleum distillates and polymers [4] and it has
been found in many samples of the commercially available
insulating oils as an undeclared artificial antioxidant [5]. It
has been found that BTA can compete with DBDS to form
a layer of stable copper complex [6, 7] on the copper surfaces
inside the transformer unit, thus, the corrosivity associated
with DBDS would be minimized. The resultant effect of these
additives depends primarily on their interrelated actions that
in turn depend on their concentrations and the operation
conditions of the power transformer unit. Thus, it is of prime
importance to determine the concentration of each of these
additives in insulating oil to check its quality.
Most standard methods and methods reported in the
literature for the determination of oil additives required a


C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com
J. Sep. Sci. 2012, 35, 750–757
pretreatment separation step for each additive from the oil
matrix prior to its determination by gas chromatography
(GC), or high-performance liquid chromatography (HPLC)
techniques. According to ASTM D 4768 [8], DBPC is sep-
arated by solid-phase extraction, SPE, using an alumina
column prior to its determination by the GC/FID (flame
ionization detector) system. To avoid the lengthy sample pre-
treatment, ASTM D 2668 method has been developed for the
direct determination of DBPC by using the Fourier transform
infrared (FTIR) technique [9]. Also, DBPC was determined
by HPLC [10], differential pulse voltammetry [11], and micel-
lar electrokinetic capillary chromatography [12] techniques
where the SPE or liquid–liquid extraction pretreatment step
was conducted prior to their determination.
The need for monitoring the passivator content in the
insulating oil matrix has also been investigated [13]. The pas-
sivator BTA has been determined using HPLC following an
SPE pretreatment step for new and used mineral insulating
oils according to the IEC-60666 [14] and ASTM WK24216 [15]
standard methods.
ASTM D 1275 and IEC-62535 are standard methods,
mainly applied for the determination of organic sulfur com-
pounds such as DBDS in mineral insulating oils [16–19]. The
results obtained from these methods are interpreted qual-
itatively by visual evaluation of the copper corrosion using
ASTM D 130 [20]. In conclusion, neither ASTM D 1275 nor
IEC-62535 standard methods can detect any form of organic
sulfur compounds if a metal passivator is present in the oil
matrix. The major challenge to detect DBDS in mineral insu-
lating oil using GC is to separate DBDS signal from the oil
matrix which requires a special pretreatment step. DBDS in
insulating oil was assessed quantitatively using SPE through
alumina cartridge followed by GC–MS (mass spectrometry)
determination [21]. This method is reported to be highly se-
lective for DBDS and can provide a quantitative conclusion.
However, using benzene as a desorbent renders the method
unpopular due to health hazard issues. Moreover, the SPE
is a tedious procedure and adds an extra cost to the method
of analysis. Recently, International Electrotechnical Commis-
sion (IEC) proposed a method for DBDS determination [22]
based on diluting the oil sample to about 20 times with a
solvent like isooctane prior to its determination by a GC–
ECD (electron capture detector) system. Although the dilution
helped the reduction of oil matrix from the chromatogram,
DBDS signal sensitivity was sacrificed due to the high dilu-
tion of the oil sample solution. IEC recently issued a standard
(IEC-60296) [23] that forces insulating oil manufacturers to
provide with each oil shipment a document to certify that the
product does not contain undeclared additives.
This article investigates the efficiency of a single liquid–
liquid extraction pretreatment step using a nonpolar dilu-
ent (n-hexane) for the oil sample and a polar solvent extrac-
tant (acetonitrile or methanol). Thus, DBDS, DBPC, and BTA
components are extracted from fresh and in service mineral
insulating oil samples prior to their determination by either
GC or HPLC techniques.

C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Sample Preparation 751
2 Experimental
2.1 Equipment and materials
Shimadzu GC, Model 17A, equipped with an FID an ECD,
AOC-201 auto injector, and DB-5 column of 30 m, 0.25 mm,
1.00 ␮m was used for DBDS and DBPC detection. Agilent
HPLC, Model 1200, equipped with ultraviolet (UV) detec-
tor, auto-injector, and ZORBAX Eclipse XDB-C18 column of
4.6 × 150 mm, 5 ␮m was used for BTA detection. Screw cap
glass tubes of 20 mL were used for the liquid–liquid extrac-
tion. The cartridges used for solid phase extraction include:
silica cartridge (500 mg SampliQ) and alumina cartridge
(AccuBond II Alumina N sorbents, with 50–200 ␮m parti-
cle size). A vacuum manifold for solid phase extraction from
Millipore was used.
The chemicals used include: DBPC, internal standards
diphenyl disulfide and diphenoxy benzene, DBDS, BTA
(Sigma Aldrich, St. Louis, MO, USA); fresh mineral insu-
lating oil free of DBPC, DBDS, and BTA (a gift from the Gulf
Chemical and Industrial Oils, Dammam, Saudi Arabia); ace-
tonitrile, methanol, and n-hexane (Merck, Kirkland, Québec,
Canada). All chemicals used were of analytical reagent grade.
A stock solution of DBDS, DBPC, and BTA was prepared
by dissolving accurately weighed samples of these materials
in fresh mineral insulating oil then diluting each with oil as
needed to give solutions of concentrations in ranges simi-
lar to those in mineral oil, namely, 10 000 mg L−1 (1.0% w/v)
DBPC and 100 mg L−1 DBDS and BTA. Various working stan-
dards of appropriate DBDS, DBPC, and BTA concentrations
were prepared from their stock solutions according to exper-
imental needs. A stock solution of internal standard mixture
of 0.2% diphenoxy benzene and 5 mg L−1 diphenyl disul-
fide was prepared in acetonitrile or methanol. The internal
standards mixture was used for DBDS and DBPC detection
through GC–ECD–FID. No internal standard was used for
the determination of BTA using the HPLC method.
2.2 Liquid–liquid extraction procedure
A 5-mL sample of mineral insulating oil containing DBPC,
DBDS, and BTA was diluted with a certain volume of n-
hexane (between 2 and 5 mL) in a 20 mL screw cap glass
tube and 5 mL acetonitrile or methanol were added to func-
tion as an extractant for the three components simultane-
ously. The mixture was shaken vigorously for about 10–15
s then centrifuged at 2000 rpm (revolutions per minute) for
1 min.
For the DBDS/DBPC determination, 2 mL of the acetoni-
trile or the methanol layer was transferred into another tube
and mixed with 2 mL of the internal standard stock solution.
This mixture was stored for GC–ECD–FID injection.
For the BTA determination, 1 mL of the acetonitrile or
the methanol layer was transferred into another tube, 4 mL
of deionized water was added to match the mobile phase
www.jss-journal.com
752 A. M. Y. Jaber et al.
matrix (water/acetonitrile, 80:20, v/v). The mixture was fil-
tered through a 0.45-␮m syringe filter and stored for HPLC
injection.
2.3 Solid phase extraction
Solid phase extraction trials were tested in this work according
to literature/standard procedures [8, 14, 21] with some mod-
ifications (instead of their home made clean up columns,
readymade cartridges were used in this study) to examine
the efficiency of solid phase extraction and the GC or HPLC
signal sensitivity when each of the three components are in-
dividually injected following the solid phase extraction step.
In all solid phase extraction trials, a mineral insulating oil
sample was diluted with n-hexane or n-heptane and eluted
through an alumina or silica cartridge using an appropriate
eluent. The sought component (DBDS or DBPC or BTA) was
washed out from the cartridge using a desorbent that was
benzene in the case of DBDS and methanol in the case of
DBPC and BTA.
3 Results and discussion
3.1 Assessment of the liquid–liquid extraction
process
The mineral insulating oil matrix usually introduces a high
background effect that obscures the chromatographic signals
of DBDS, DBPC, and BTA when determined by the GC or
HPLC techniques. According to literature reports [8, 14, 21],
SPE as a pretreatment step requires extensive dilution of the
oil sample, thus, the results obtained may suffer from low
sensitivity and poor precision. Consequently, a liquid–liquid
extraction step was introduced in this work and the results
are compared with those obtained from the SPE step.
In the preliminary liquid–liquid extraction trials, the
extraction procedure was conducted using a liquid extrac-
tant (acetonitrile or methanol) to extract, simultaneously, the
Figure 1. GC–FID signal for mineral insulating oil containing 2000 mg L−1 (0.2%) DBPC after extraction with acetonitrile using (a) oil/n-
hexane/acetonitrile, 5:2:5 and (b) oil/acetonitrile, 5:5.

C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. Sep. Sci. 2012, 35, 750–757
DBPC, BTA, and DBDS from a mineral insulating oil sample
diluted with n-hexane. The layer comprising the three com-
ponents was separated, mixed with internal standard when
necessary, then injected into the GC or HPLC system. The
chromatograms were recorded and the peak of each compo-
nent was used to assess its extraction efficiency. For the sake
of comparison, the known standard methods, ASTM D 4768,
IEC-60666, and IEC-62697 standard methods were applied
for the determination of DBPC, BTA, and DBDS. According
to these methods, GC and HPLC techniques were used for
the determination of DBPC and BTA, respectively, after solid
phase extraction pretreatment step; while DBDS was deter-
mined without a pretreatment step but with significant dilu-
tion with isooctane to minimize the oil matrix background
interference on the GC chromatogram.
The chromatograms obtained after liquid–liquid extrac-
tion showed that the background signal associated with the
oil matrix was almost eliminated as in the case of literature
methods in which solid phase extraction was used [8, 14, 21].
However, the chromatographic peak of each of the three com-
ponents was drastically enhanced in comparison with that
observed when determined by the standard methods men-
tioned above. The GC signal for DBPC increased almost
three times (Figs. 1a and 2) when liquid–liquid pretreatment
extraction step was used. The use of n-hexane in this liquid–
liquid extraction pretreatment has decreased the possibility
for the oil matrix extraction in the extractant (acetonitrile or
methanol) phase. n-Hexane was used earlier as a diluent
for edible oil samples when the synthetic phenolic antiox-
idants were extracted by acetonitrile polar solvent prior to
their determination by micellar electrokinetic capillary chro-
matography [12]. The use of n-hexane as a cosolvent in the
extraction was attributed to its role in decreasing the viscosity
of the oil sample, thereby enhancing the contact between the
oil and acetonitrile phases. When the same liquid–liquid ex-
traction pretreatment step was carried out in the absence of
n-hexane, the chromatogram showed a high background con-
tribution (Fig. 1b) indicating the presence of oil traces in the
acetonitrile layer. However, when a sample of this acetonitrile
layer was shaken with n-hexane (5:2, v/v), a chromatogram
www.jss-journal.com
J. Sep. Sci. 2012, 35, 750–757
Figure 2. GC/ECD signal for mineral insulating oil containing 50
mg L−1 DBDS after extraction with acetonitrile using oil/n-hexane/
acetonitrile, 5:2:5.
similar to that of Fig. 1a was obtained where the background
has almost disappeared. The nonpolar n-hexane solvent may
compete with the acetonitrile extractant for the oil matrix,
thus, it minimizes the extraction of oil into the acetonitrile
layer. The same argument holds for the solid phase extrac-
tion or liquid–liquid extraction of BTA prior to its injection
into the HPLC system. Though DBDS separation from the oil
matrix was not required when it was determined by the GC–
ECD system, according to the standard method IEC-62697,
a significant dilution of the oil sample prior to GC–ECD de-
termination was essential. This extensive dilution should be
responsible for the low chromatographic peak sensitivity and
consequently the high limit of detection (LOD) reported there
(≤5 mg L−1 ) [22].
These preliminary trials indicated that solid-phase extrac-
tion requires various elution steps that may dilute the ana-
lyte to be determined in the oil sample, thus, decreasing the
sensitivity and detection limit of the chromatographic sig-
nal. Furthermore, appreciable quantities of various solvents
and time will be consumed compared to the liquid–liquid
extraction pretreatment described in this work. For example,
according to the IEC-60666 standard method, BTA is deter-
mined in mineral insulating oil by SPE pretreatment where
about 40 mL of pentane and 10 mL of methanol are used for
one oil sample, while for DBPC, about 6 mL of pentane with
5 mL of methanol are required for one sample. Furthermore,
according to the IEC-62697 standard method for the DBDS
determination, the oil sample is diluted 20 times with isooc-
tane. However, according to this work, only 2 mL hexane (as
diluent) and 5 mL acetonitrile (as extractant) are used for an
oil sample comprising the three components. Thus, all the
three species are separated simultaneously from the matrix
of an oil sample using a single extraction step with only two
solvents and a total of 7 mL of solvents. The efficiency of
liquid–liquid extraction pretreatment step was assessed for
each of the three components to optimize the volumes of oil
sample and solvents that give the highest chromatographic
peak.

C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Sample Preparation 753
3.2 Optimizing of volume ratios in the liquid–liquid
extraction process
The focus of this work was to introduce a single, easy, and
efficient liquid–liquid extraction step with minimal use of sol-
vents where DBDS, DBPC, and BTA are extracted simultane-
ously from the oil matrix prior to their determination by GC
and HPLC techniques. The extraction efficiency was assessed
by varying the volume ratios: oil/oil diluent/extractant solvent
until the optimum ratio that produced the highest GC/HPLC
signal sensitivity was obtained; this would indicate the high-
est extraction efficiency. Two sets of oil/oil diluent/extractant
solvent volume ratios were studied. In one set the extrac-
tant solvent was 5 mL acetonitrile while in the other it was
5 mL methanol. The components of each set were as follows:
Y:X:5, where Y (mineral insulating oil) = 2 or 5 mL, and X
(n-hexane) = 2 or 5 mL. Thus, four combinations were pre-
pared for each extractant solvent. A series of solutions (a total
of four solutions for each extractant solvent) comprising vari-
ous DBDS, DBPC, and BTA concentrations (1–100 mg L−1 for
DBDS and BTA and 100–10 000 mg L−1 , i.e., 0.01–1% w/v for
DBPC) dissolved in fresh mineral oil (Y mL) were prepared
from each combination. The liquid–liquid extraction was car-
ried out according to the procedure mentioned above, and a
portion of each extract (acetonitrile or methanol layer) was
separated, mixed with the internal standard when needed,
and injected into the GC to determine DBDS, DBPC simul-
taneously where one line is passed into an ECD to determine
DBDS while the other line is passed into an FID to deter-
mine DBPC. Another portion of the extractant was injected
into the HPLC to determine BTA. Furthermore, two more se-
ries of standard solutions were prepared comprising DBDS,
DBPC, and BTA with the same range of concentrations but
dissolved in pure acetonitrile in one time and in methanol
in the other. Again, a portion of each of these sample solu-
tions was injected into the GC system and another portion
was injected into the HPLC system. Calibration curves were
constructed for each of the three components (DBDS, DBPC,
and BTA) in each of the four series. The extraction efficiency
for each volume ratio was calculated as the ratio of the slope
of the calibration curve constructed for solutions of a certain
component prepared in this volume ratio to the slope of the
calibration curve constructed for the same component in the
extractant solvent (acetonitrile or methanol) only. The extrac-
tion efficiencies calculated as percent extraction are presented
in Fig. 3.
It is obvious from Fig. 3 that acetonitrile achieved bet-
ter extraction efficiency than methanol for each of the three
components in all volume ratios, and the difference in percent
extractability was more pronounced with DBDS. This differ-
ence in extractability might be ascribed to the differences
in solubility of each of the three components in methanol
and acetonitrile, thus, the significant difference in the case of
DBDS might be attributed to its higher solubility in acetoni-
trile compared to that in methanol. Increasing the volume
of the oil diluent from 2 to 5 mL seems to be of little effect
on the extraction efficiency except for DBPC. This increase
www.jss-journal.com
754 A. M. Y. Jaber et al.
caused a decrease in extraction efficiency of about 7–9% for
DBPC in acetonitrile or methanol, however, the decrease in
extraction efficiency for the other two components were about
2–3%. This decrease may be attributed to the competition of
acetonitrile and n-hexane for each of the components. The
dramatic increase in the extraction efficiency was observed
when the volume of the oil sample was increased from 2
to 5 mL, however, the extractant volume was kept the same
(5 mL). Figure 3 shows that the highest extraction efficien-
cies for the three components were achieved when the min-
eral oil/oil diluent/extractant solvent was 5:2:5. Consequently
this volume ratio was recommended as the optimum value
for the liquid–liquid extraction step prior to GC and HPLC
determination of the three components. The change in the
extraction efficiency may be ascribed mainly to the dilution
factor that takes place when the components transfer from
the oil sample into the extractant (acetonitrile or methanol)
phase and partially due to the solubility of components in the
extractant phase. When the volume ratio increased from 2:5 to
5:5, the extraction efficiency for BTA increased almost by the
same proportion; however, the increase in the extraction effi-
ciency for both DBDS and DBPC was less than that expected
from the change in the volume ratio (Fig. 3). The variation
in the extraction efficiency for each of the three components
may be ascribed to their relative solubilities in acetonitrile or
methanol. Since BTA has the relatively highest solubility, it
achieved the highest extraction efficiency in both acetonitrile
and methanol phases.
As Fig. 3 indicates, the extraction efficiency did not reach
100%, hence, in conducting such determinations, standards
and oil sample solutions must undergo the same pretreat-
ment before the samples are injected into the GC–ECD–FID
and HPLC systems in order to compensate for the difference
in the extraction efficiency.
3.3 Linearity, sensitivity, and limit of detection for
the GC and HPLC calibration curves
The linearity was established for concentration ranges of
100–10 000 mg L−1 (0.01–1.00% w/v) DBPC, 1–100 mg L−1

C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. Sep. Sci. 2012, 35, 750–757
Figure 3. Extraction efficiencies for sev-
eral volume ratios (oil/oil diluent/extractant
solvent) using acetonitrile or methanol as
extractants.
DBDS, and 1–100 mg L−1 BTA solutions prepared in fresh
mineral insulating oil and extracted by acetonitrile liquid–
liquid extraction using the optimum volume ratio followed
by injection into the GC–ECD/FID and HPLC/UV systems
under the optimum operational and instrumental conditions
[8, 14, 21]. Diphenoxy benzene and diphenyl disulfide inter-
nal standards were used with DBPC and DBDS, respectively,
whereas BTA calibration curve was established using the ex-
ternal standard calibration method. Although according to the
standard method, IEC-60666 [14], it is recommended that the
HPLC–UV detector be set at 260–270 nm to measure the ab-
sorbance of BTA, it was set at 250 nm in this study, especially
with used oil, to minimize the chromatographic interference
of certain oxidation byproducts (furanic compounds such as
2-furaldehyde) that form in the oil as a result of oil and in-
sulating paper degradation and overheating. The maximum
absorbance wavelength for these furanic compounds falls in
the range of 272–280 nm [24], however, the wavelength of
maximum absorbance for BTA is 260 nm. The measurement
at 250 nm gave significant signal sensitivity for BTA and a
minimal absorbance for the furanic compounds. Thus, there
was no need for the standard addition method as proposed by
the IEC-60666 standard method [14] when dealing with used
oil.
Table 1 gives the linearity, slope, and limit of detection
determined according to Miller and Miller [25] for GC and
HPLC calibration graphs for DBPC, DBDS, and BTA fol-
lowing the liquid–liquid extraction pretreatment described
above. It also provides these values for the literature stan-
dard methods in which the pretreatment was only a dilution
by isooctane as in the case of DBDS or solid phase extrac-
tion as in the case of DBPC and BTA [8, 14]. Table 1 shows
that higher sensitivities and lower detection limits than those
of the standard methods were achieved. The sensitivity en-
hancement may be ascribed to the type of column and elu-
ent used. According to the IEC-62697 standard method, a
significant interference of polychlorinated biphenyles, PCB,
with DBDS signal was observed, however, this interference
was completely absent when the DB-5 column was used
(Fig. 4) under the same conditions and using the same ECD
detector.
www.jss-journal.com
J. Sep. Sci. 2012, 35, 750–757 Sample Preparation 755

Table 1. Linearity, slope, and limit of detection of the calibration graphs for DBPC, DBDS, and BTA using the liquid–liquid extraction and
the standard methods sample pretreatment prior to the GC and HPLC determination

Regression parameters DBPCb) DBDSc) BTAd)

ASTM D Liquid–liquid IEC- Liquid–liquid IEC- Liquid–liquid

4768 extraction 62697 extraction 60666 extraction

R2a) 0.995 0.999 0.999 0.997 0.997 0.999

Slopea) 0.81 1.84 0.01 0.04 2.53 5.03
LOD 500 mg L−1 (0.05% w/v) 100 mg L−1 (0.01% w/v) 1.80 mg L−1 0.80 mg L−1 5.30 mg L−1 2.04 mg L−1

a) Calibration ranges: 100–10 000 mg L−1 (0.01–1.00% w/v) DBPC and 1–100 mg L−1 DBDS/BTA.
b) Determined by GC/FID.
c) Determined by GC/ECD.
d) Determined by HPLC.

3.4 Accuracy and precision of GC and HPLC

determinations

The accuracy and precision in the determinations of DBPC,

Figure 4. The GC/ECD signal of 1 mg L−1 polychlorinated
biphenyles, PCB, and 10 mg L−1 DBDS in acetonitrile using DB-5
column.
DBDS, and BTA in mineral insulating oil using GC and
HPLC techniques preceded by liquid–liquid extraction pre-
treatment on the one hand and those for the standard meth-
ods on the other hand have been estimated for fresh and aged
mineral insulating oil samples. The results are presented in
Tables 2 (DBDS), 3 (DBPC), and 4 (BTA). The percentage
recoveries (the ratio of the quantity found by experiment for
a certain component to the quantity added to the oil sample)
for DBDS were in the range of 98.9–102.3% with percent
relative standard deviations (RSDs) of 0.45–6.28 for various
DBDS concentration ranges (1–150 mg L−1 DBDS). The two-
sided t-test was used to investigate whether the means differ

Table 2. Accuracy and precision for DBDS determination in mineral insulating oils using the liquid–liquid extraction and the standard
pretreatment methods prior to the GC determination

Quantity added Liquid–liquid extraction IEC-62697 Student’s t-test ASTM D1275 B ASTM D 130a)
DBDS (mg L−1 )
Quantity found Recovery Quantity found Recovery two-sided Corrosive–
DBDS (mg L−1 ) (%) DBDS (mg L−1 ) (%) (n = 3) noncorrosive

1.00 1.02 102.3 Not detected - - None corrosive Degree of corrosion is 1a

Percentage RSD 6.28 -
10.00 10.07 100.7 10.33 103.3 1.24 None corrosive Degree of corrosion is 1b
Percentage RSD 1.84 2.99
20.00 20.33 101.7 19.66 98.3 2.90 None corrosive Degree of corrosion is 3b
Percentage RSD 0.85 1.85
50.00 49.93 99.9 50.44 100.9 1.74 Corrosive Degree of corrosion is 4a
Percentage RSD 0.61 0.79
100.00 99.27 99.3 99.99 100.0 1.98 Corrosive Degree of corrosion is 4b
Percentage RSD 0.63 0.11
150.00b) 148.32 98.9 149.58 99.7 2.56 None corrosive Degree of corrosion is 1b
Percentage RSD 0.45 0.36

a) Qualitative visual corrosion classification as per ASTM D 130.

b) The oil sample was passivated with 30 mg L−1 BTA.


C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com
756 A. M. Y. Jaber et al. J. Sep. Sci. 2012, 35, 750–757

Table 3. Accuracy and precision for DBPC determination in mineral insulating oils using the liquid–liquid extraction and the standard
pretreatment methods prior to GC determination

Quantity added Liquid–liquid extraction ASTM D 4768 Student t-test

DBPC (% w/v)
Quantity found Quantity found two-sided
DBPC (% w/v) Recovery (%) DBPC (% w/v) Recovery (%) (n = 3)

0.01 0.010 110.0 Not detected - -

Percentage RSD 5.41 -
0.05 0.049 98.0 0.052 104.0 0.46
Percentage RSD 2.97 3.85
0.10 0.101 101.0 0.102 102.0 0.36
Percentage RSD 4.48 6.51
0.20 0.204 102.0 0.209 104.7 0.57
Percentage RSD 4.16 5.26
0.30 0.296 98.8 0.304 101.3 2.14
Percentage RSD 0.51 1.97
0.30a) 0.301 100.3 0.308 102.8 2.63
Percentage RSD 0.99 1.23

a) 0.30% w/v DBPC spiked in aged oil matrix.

significantly from zero. Student’s t values for three replicates
of measurements made on each sample at 95% confidence
level indicate that no significant difference existed between
the experimental means obtained from this GC–ECD follow-
ing the liquid–liquid extraction pretreatment and the IEC-
62697 standard method (Table 2). Table 2 shows also that
the IEC-62697 standard method did not detect DBDS at 1
mg L−1 level. This may be ascribed to the need for an oil
sample dilution of about 20 times to facilitate its elution
through the GC column. The ASTM D1275 B did not de-
tect the presence of DBDS even at concentrations of about
20 mg L−1 DBDS. Furthermore, when an oil sample con-
taining DBDS as high as 150 mg L−1 was passivated with
30 mg L−1 BTA, the ASTM D 1275 method gave a result of
noncorrosivity of the oil, a conclusion that gives a false indi-
cation to the presence of DBDS. The percent recoveries for
DBPC were in the range of 98–110% for concentration ranges
of 100–3000 mg L−1 (0.01–0.30% w/v). Student’s t values in-
dicate that no significant difference existed between the ex-
perimental means obtained from this GC–FID following the
liquid–liquid extraction pretreatment and the ASTM D 4768
standard method (Table 3). It is obvious from Table 3 that
100 mg L−1 (0.01% w/v) DBPC was not detectable when the
standard method was used. The percent recoveries for BTA

Table 4. Accuracy and precision for the BTA determination in mineral insulating oils using the liquid–liquid extraction and the standard
pretreatment methods prior to GC determination

BTA Quantity Liquid–liquid extraction IEC-60666 Student t-test ASTM D1275 B ASTM D 130a)
added (mg L−1 )
Quantity found Quantity found two-sided Corrosive–
BTA (mg L−1 ) Recovery (%) BTA (mg L−1 ) Recovery (%) (n = 3) noncorrosive

2.00b) 2.08 104.2 Not detected - - Corrosive Degree of corrosion is 4a

Percentage RSD 1.69 -
10.00b) 9.99 99.9 10.04 100.4 0.77 None corrosive Degree of corrosion is 3b
Percentage RSD 0.80 0.89
30.00b) 30.03 100.1 29.91 99.7 0.51 None corrosive Degree of corrosion is 2b
Percentage RSD 0.36 1.32
50.00b) 50.25 100.5 49.95 99.9 1.38 None corrosive Degree of corrosion is 1a
Percentage RSD 0.57 0.50
50.00c),b) 50.02 100.0 49.38 98.8 3.06 None corrosive Degree of corrosion is 1b
Percentage RSD 0.29 0.67
100.00c) 100.01 100.0 98.76 98.8 4.16 - -
Percentage RSD 0.18 0.50

a) Qualitative visual corrosion classification as per ASTM D 130.

b) The oil sample was spiked with 50 mg L−1 DBDS.
c) Aged oil matrix contained 1000 ␮g L−1 2-furaldehyde.


C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com
J. Sep. Sci. 2012, 35, 750–757
were in the range of 99.9–104.2% for concentration ranges of
2–100 mg L−1 BTA. Student’s t values indicated no significant
difference existed between the experimental means obtained
from this HPLC/UV following the liquid–liquid extraction
pretreatment and the IEC-60666 standard method (Table 4).
The percent recovery was about 100% when the liquid–liquid
extraction pretreatment for an aged oil sample containing
100 mg L−1 BTA and spiked with 1000 ␮g L−1 2-furaldehyde
was tested (Table 4). The method is accurate, precise, and can
be used successfully for different mineral insulating oil ma-
trices used for power transformers. Consequently, the liquid–
liquid extraction pretreatment facilitates the removal of the
oil matrix, preconcentrates the analyte (DBDS, DBPC, and
BTA), and reduces the analysis time of the three components
that may need to be tested together to judge the quality and
characteristics of the mineral insulating oil when fresh or
used. Thus, DBDS, the antioxidant and corrosive component
at the same time can be detected and controlled even at con-
centrations as low as 1 mg L−1 and in the presence of the BTA
passivator. These findings would help studying the kinetics
of copper corrosion in the transformer where the concentra-
tions of DBDS, DBPC, and BTA should be smaller than the
levels detected by the standards; a study that is currently under
investigation.
4 Concluding remarks
This work describes an efficient single and fast pretreatment
step for fresh and used transformer oil samples with minimal
consumption of solvents for the simultaneous liquid–liquid
extraction of three important mineral insulating oil additives,
namely, DBDS, DPBC, and BTA to facilitate their determina-
tion by GC and HPLC techniques. DBDS and DBPC are de-
termined simultaneously by GC/ECD/FID technique while
BTA is determined by HPLC technique. Comparison of the
results of this method for each of the three components to
those obtained by the standard methods revealed that this
method provides better sensitivity and lower detection lim-
its as a result of the preconcentration and removal of the
mineral insulating oil matrix that may interfere with the de-
termination of the components. Determining relatively low
concentrations of these three components would help study-
ing their interaction effects in the mineral insulating oil and
its corrosion and passivation actions under the transformer
operational conditions. Traditionally this interaction effect is
studied indirectly by investigating the copper sulfide forma-
tion. Currently, the corrosion process is being investigated
via the depletion of the DBDS concentration in the mineral
insulating oil.

C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Sample Preparation 757
King Fahd University of Petroleum & Minerals, the Industrial
Support Services Laboratories, and Gulf Chemical and Industrial
Oils are thanked for the support of this research project.
The authors have declared no conflict of interest.
5 References
[1] Ito, T., Morishima, Y., IEEJ Trans. 2009, 4, 422–424.
[2] Kawarai, H., Fujita, Y., Tanimura, J., Toyama, S., Yamada,
N., Nagao, E., IEEE 2009, 16, 1430–1433.
[3] Amimoto, T., Nagao, E., Tanimura, J., Toyama, S.,
Yamada, N., IEEE Trans. 2008, 16, 257–264.
[4] Scott, G., Atmospheric Oxidation and Antioxidants,
Elsevier, Amsterdam, London, New York 1965.
[5] Maina, R., Scatiggio, F., Kapila, S., Tumiatti, V., Tumiatti,
M., Pompili, M., Dibenzyl disulfide (DBDS) as corrosive
sulfur contaminant in used and unused mineral insulat-
ing oils, published on CIGRE SC2 web site, 2006.
[6] Wiklund, P., Levin, M., Pahlavanpour, B., IEEE 2007, 23,
1–14.
[7] Shin, J., Hong, J., Trans. Elect. Electron. Mater. 2010, 11,
288–291.
[8] J. ASTM D 4768-03, J. ASTM Int. 2008, 10.03.
[9] J. ASTM D 2668-07, J. ASTM Int. 2008, 10.03.
[10] Lamarre, A., Gendron, C., IEEE Trans. 1995, 2, 413–417.
[11] Foley, L., Kimmerle, F. M., Anal. Chem. 1979, 51, 818–822.
[12] Zamarreño, M. M., González-Maza, I., Sánchez-Pérez, A.,
Carabias Martı́nez, R., Food Chem. 2007, 100, 1722–1727.
[13] Augusta, M., Martins, G., Gomes, A., Pahlavanpour, B.,
IEEE 2009, 25, 23–27.
[14] IEC-60666, IEC, 2nd Ed., 2010, 22–30.
[15] J. ASTM WK 24216, J. ASTM Int. 2009 (a draft under
development).
[16] CIGRE WG A2-32, Copper sulfide in transformer insula-
tion, Final report—Boucher 378, 2009, 1–35.
[17] IEC-62535, IEC, 1st Ed., 2008, 6–13.
[18] ASTM D1275-06, J. ASTM Int. 2006, 10.03.
[19] Grant, D. H., IEEE Electrical Institution Conference,
Canada, June 2009, pp. 489–490.
[20] J. ASTM D 130-04, J. ASTM Int. 2006, 5.01.
[21] Toyama, S., Tanimura, J., Yamada, N., Nagao, E., IEEE
Trans. 2009, 16, 509–515.
[22] IEC-62697-1, IEC, 1st Ed., 2010, 10–27.
[23] IEC-60296, IEC, 3rd Ed., 2003, 6–7.
[24] J. ASTM D 5837-99, J. ASTM Int. 2005, 10.03.
[25] Miller, J. C., Miller, J. N., Statistics and Chemometrics
for Analytical Chemistry, Pearson, Harlow 2005, 5th Ed.,
pp. 121, 39.
www.jss-journal.com