Talanta 72 (2007) 289295

An automated FTIR method for the routine quantitative determination

of moisture in lubricants: An alternative to Karl Fischer titration
Frederik R. van de Voort
a,
, Jacqueline Sedman
a
, Robert Cocciardi
b
, Steve Juneau
c
a
McGill IR Group, Department of Food Science and Agricultural Chemistry, Macdonald Campus of McGill University, 21,111 Lakeshore Road,
Sainte-Anne-de-Bellevue, Quebec, Canada H9X 3V9
b
Thermal-Lube Inc., 255 Avenue Labrosse, Pointe-Claire, Quebec, Canada H9R 1A3
c
Naval Engineering Test Establishment, 9401 Wanklyn Street, LaSalle, Quebec, Canada H8R 1Z2
Received 16 August 2006; received in revised form 16 October 2006; accepted 19 October 2006
Available online 29 November 2006
Abstract
An accurate primary Fourier transform infrared (FTIR) method for the determination of moisture in mineral and ester based lubricants has
been developed based on the extraction of moisture into dry acetonitrile. FTIR evaluation of acetonitrile extracts from new and used lubricants
as well as common lubricant additives and contaminants which might co-extract indicated that phenolic constituents interfered signicantly with
moisture measurements. By measuring moisture at 3676 cm
1
on the shoulder of the asymmetric OH stretching band, spectral interferences from
extracted phenolic constituents were minimized. The spectra of calibration standards (02100 ppm), prepared by gravimetric addition of water
to dry acetonitrile, were recorded in a 1000-m CaF
2
transmission ow cell and produced linear standard curves having an S.D. of 20 ppm.
Lubricant sample preparation involved the vigorous shaking (20 min) of a 1:1.5 (w/v) mixture of lubricant and dry acetonitrile, centrifugation to
separate the phases, acquisition of the FTIR spectrum of the upper acetonitrile layer, and subtraction of the spectrum of the dry acetonitrile used
for extraction. A Continuous Oil Analyzer and Treatment (COAT

) FTIR system was programmed to allow the automated analysis of acetonitrile

extracts, and the methodology was validated by analyzing 58 new and used oils, independently analyzed by the Karl Fischer (KF) method. Linear
regression of FTIR versus KF results for these oils produced a linear plot with a between-method S.D. of 80 ppm. As implemented on the COAT

system, this FTIR method is capable of analyzing 72 acetonitrile extracts/h and provides a high-speed alternative to the KF titrimetric procedures
for the determination of water in lubricants.
2006 Elsevier B.V. All rights reserved.
Keywords: Acetonitrile; FTIR spectroscopy; Lubricants; Moisture content; Karl Fischer
1. Introduction
The level of moisture in lubricants is an essential parameter,
as water is one of the most widespread and destructive lubri-
cant contaminants. Water contamination can lead to corrosion
and excessive wear of machinery, and the presence of emulsi-
ed water in a lubricant can impair its effectiveness. Among the
variety of American Society for Testing and Materials (ASTM)
methods available for the determination of moisture in lubricants
andrelatedproducts, the ASTMD6304coulometric Karl Fischer
(KF) method [1], performed with a water evaporator accessory
(WAP), is considered the gold standard. The WAP is used to

Corresponding author. Tel.: +1 514 398 8618; fax: +1 514 398 7977.
E-mail address: frederik.vandevoort@mcgill.ca (F.R. van de Voort).
rapidly heat the oil sample under vacuumto vaporize water in the
sample, which is ushed with dry nitrogen into the KF reagent
solution. This procedure minimizes matrix effects commonly
associated with lubricant additives, which can undergo aldol
condensation and redox side reactions that can affect KF results
if oil is directly added to the KF reagent. Thus, ASTM D6304
with a WAP is considered the most comprehensive and accurate
procedure available for the analysis of moisture in lubricants
and petroleumproducts, including lubricant additives, base oils,
fully formulated lubricants and automatic transmission uids,
amongst others. Most oil analysis laboratories invest in a Karl
Fischer apparatus for the quantitative determination of mois-
ture along with other equipment for conducting other pertinent
analyses such as particulate contamination, acid number, vis-
cosity, ferrography, elemental analysis, oxidation and more, all
at substantial cost in terms of equipment and consumables.
0039-9140/$ see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.talanta.2006.10.042
290 F.R. van de Voort et al. / Talanta 72 (2007) 289295
Among the instruments often found in commercial oil
analysis laboratories is a Fourier transforminfrared (FTIR) spec-
trometer. This instrumentation is used for lubricant condition
monitoring, which involves the trending of a wide range of
parameters, including oxidation, nitration, sulfate by-products,
soot, glycol contamination, diesel fuel dilution, antiwear deple-
tion and water contamination (see Ref. [2] and references cited
therein). FTIR condition monitoring by direct trending [3] is
a semiquantitative procedure, in that no calibration against a
primary reference method is performed. Instead, results are
reported in terms of peak areas or heights (or baseline tilt in
the case of soot) and interpreted in relation to empirically estab-
lished criteria. Whereas this methodology provides a means of
detecting water contamination in in-service lubricants over time,
it does not allow for quantitative determination of the amount
of water present owing to overlap of the IR absorption bands
used to monitor water contamination with those of other OH-
containing constituents (often present in lubricating oils). On
the other hand, in the differential version of this methodology,
the spectral contributions of these constituents are, in principle,
eliminated by subtracting out the spectrum of a new oil of the
same formulation as the samples being analyzed, and the water
content may be estimated from a calibration equation obtained
using gravimetrically prepared standards. However, the use of
this methodology as the basis for the development of a generally
applicable quantitative method is highly problematic not only
because of the limitations associated with the need to have the
appropriate reference oil at hand but also because the absorption
bands of water are subject to matrix effects such as hydrogen
bonding, whichdifferential spectroscopycannot compensate for.
These complications have stymied the development of a quan-
titative FTIR procedure that would be competitive with the KF
method.
An alternative approach developed in our laboratory to over-
come these limitations made use of the stoichiometric reaction
of water with dimethoxypropane (DMP) to produce acetone,
coupled with differential spectroscopy to eliminate the spectral
contributions of the oil [4]. The moisture content of the sam-
ple was determined by measuring the intensity of the (C O)
absorption of acetone, calibrations being derived by standard
addition of water to a dry oil. This method accurately detects
moisture levels as lowas 50 ppmand is workable for compressor,
hydraulic, and gear oils. However, it gives erroneously high val-
ues for gasoline and diesel engine oils that are over-based with
calcium carbonate, because the acid catalyst required for the
reaction is neutralized by the base, thereby producing CO
2
and
more H
2
O. This method is also problematic for mineral-based
lubricants containing substantial levels of ester-based additives
as well as ester-based lubricants, as the (C O) absorption of
the acetone produced by the reaction is masked by the ester car-
bonyl absorption. This problem also prevents the FTIR-DMP
approach from being used for the analysis of moisture in edible
oils, as they are predominantly triacylglycerols. This limitation
led to the development of an alternative FTIR method based on
the extraction of moisture from edible oils using dry acetoni-
trile followed by measurement of the water absorption bands
in the spectrum and quantication based on a calibration devel-
oped using acetonitrile-water standards [5]. This paper describes
the adaptation of this spectroscopic approach to the analysis of
newand used lubricants and its implementation as an automated
method as well as its validation against the KF method.
2. Experimental
2.1. Materials
Additive-free (unformulated) mineral and ester base oils and
commercial additive packages were obtained from Thermal-
Lube Inc. (Pointe-Claire, PQ, Canada). Ester and mineral base
oils were blended with additive packages and/or constituents
commonly found in newor used oils, including esters, polyalkyl-
glycols, fatty acid rust inhibitors, phosphate-based antiwear
additives, sulfurized olen extreme pressure additives, and phe-
nolic and amine antioxidants. Reagent-grade acetonitrile (Fisher
Scientic, Nepean, ON, Canada) was dried and kept dry over 48
mesh 3

A molecular sieves and dispensed using a re-pipette,
with its air inlet protected by desiccant to prevent moisture
ingress. Formulated commercial lubricants, including diesel
engine, compressor, hydraulic, gear and other oils (Table 1),
were obtained from a variety of industrial suppliers. Samples of
used oils were provided by the Naval Engineering Test Estab-
lishment (NETE; Montreal, PQ, Canada), Insight Services Inc.
(Cleveland, OH), and the U.S. Armed Forces Technical Sup-
port Center (TSC) of the tri-service Joint Oil Analysis Program
(JOAP). From among these new and used oils, a subset of sam-
ples was selected to represent a variety of oil types and different
moisture contents and submitted to NETE for KF analysis.
2.2. Instrumentation
Details of the COAT

FTIR Oil Analyzer (shown in Fig. 1)

and its software have been described elsewhere [6]. Briey,
an ABB Bomem (Quebec City, PQ, Canada) WorkIR FTIR
spectrometer equipped with a DTGS detector and controlled
by an IBM-compatible Pentium 150-MHz PC running pro-
prietary Windows-based UMPIRE PRO
TM
(Universal Method
Platform for InfraRed Evaluation) software (Thermal-Lube,
Pointe-Claire, PQ, Canada) is connected to a Gilson autosam-
pler (Gilson Inc., Middleton, WI). The software controls the
spectrometer, autosampler, and pump and handles spectral data
processing automatically.
For the moisture analysis procedure, acetonitrile extracts
were placed in septum-capped vials and loaded into autosam-
pler trays. They were then sequentially aspirated under computer
control into a 1000-m CaF
2
transmission ow cell (Interna-
tional Crystal Laboratories, Gareld, NJ), using a sharpened
stainless steel needle to penetrate the sealed vials. Pump time
was set at 30 s/sample. The spectra of the samples were collected
by co-addition of eight scans at a resolution of 8 cm
1
with a
gain of 1.0. Weak Beer-Norton apodization was employed.
2.3. Analytical protocol
For sample preparation, 10 g of sample (0.0001 g) was
weighed on an analytical balance into a tared 50-mLclinical cen-
F.R. van de Voort et al. / Talanta 72 (2007) 289295 291
Table 1
Partial listing of commercial lubricants and the functional categories assigned to them by the manufacturer
Lubricant Category Lubricant Category
Rimula X SAE 40 Diesel engine oil Mobil Vacuoline 133 Bearing oil
Mil-L-9000 Diesel engine oil Mobil Vacuoline 525 Bearing oil
OMD 113 Diesel engine oil Mobil Gear 632 Gear oil
Mobil Delvac 1240 Diesel engine oil Mobil Rarus 424J Compressor oil
Mobil DTE BB Bearing oil Tellus T15 Hydraulic oil
RP ThermalGlyde 220 Gear oil Tellus 68 Hydraulic oil
Mobil DTE 832 Compressor oil Omala 220(P) Gear oil
Hydrex AW 32 Bearing oil Turbo T46 Turbine oil
Mobil Rarus 1024 Compressor oil Turbo P100 Turbine oil
Terrestic 220 Anti-friction bearing oil Tonna T220 Slide-way oil
Terrestic 68 Turbine oil Rimula X 15w/40 Diesel engine oil
Mobil Gear SHC 460 Gear oil Morlina 10 Bearing oil
Mobil Gear SHC 460 Gear oil Rando HD-46(P) Hydraulic oil
Mobil Delo 400 Diesel engine oil Meropa 320P Gear oil
Mobil DTE 26 Hydraulic oil Cosmolubric HF-130 Polyol ester oil
Mobil DTE Heavy Medium Turbine oil Hydro FE-56 Polyol ester oil
trifuge tube, and 15 mLof dry acetonitrile was dispensed into the
tube using a re-pipette. The samples were mixed vigorously for
20 min using a multi-sample, reciprocal action shaker (Eberbach
Corporation, Ann Arbor, MI), and then the oil and acetoni-
trile phases were separatedbycentrifugation(Eppendorf Canada
Ltd., Mississauga, ON) for 10 min at 10,000 rcf (relative cen-
trifugal force). The upper acetonitrile layer was transferred into
a 20-mL autosampler vial, which was then capped with a sep-
tum cap having a Mylar liner. The vials were loaded into the
autosampler tray from position #3 onward, with the rst two
Fig. 1. COAT FTIR Oil Analyzer used for the analysis of moisture. The system
consists of a micro gear pump (A), cell holder (B), FTIR spectrometer (C),
autosampler tray (D) and robotic arm (E).
vials in the tray being reserved for the dry acetonitrile used
to extract the samples, the rst serving as a rinse solvent and
the second to collect the spectrum of the solvent. The single-
beam spectra of the solvent and samples were ratioed against
an open-beam background spectrum collected at the beginning
of each run to give their respective absorbance spectra, and the
absorbance spectrum of the dry acetonitrile solvent was sub-
tracted from the absorbance spectrum of each sample to give its
differential spectrum. To compensate for the partial miscibility
of ester-based oils with acetonitrile, the spectra of their acetoni-
trile extracts were multiplied(prior tospectral subtraction) bythe
ratio of the height of the acetonitrile (C N) band at 2068 cm
1
(measured relative to a two-point baseline between 2125 and
2033 cm
1
) in the spectrum of the dry acetonitrile solvent to the
height of this band in the spectrum of the acetonitrile sample
extract.
2.4. KF analysis
The KF apparatus used by the NETE laboratory for oil anal-
ysis was a Mettler Toledo DL32 (Columbus, OH) equipped
with a WAP and used Hydranal (SigmaAldrich Inc., St. Louis,
MO) KF reagents for moisture quantication. The system was
validated using KF moisture standards (SigmaAldrich Inc.,
Milwaukee, WI) by loading 0.1 g of each standard and heat-
ing it to 220

C, ushing with dry nitrogen for 60 s and titrating

until the end point was reached, corresponding to a drift of less
than 5 g/min. Lubricant analysis used 1.0 g of sample heated
to 160

C with a nitrogen ush time of 210 s and then titrated

to the same endpoint as the moisture standards. Data obtained
from the KF analyses were used to evaluate the efcacy of the
FTIR moisture method.
2.5. Calibration and validation
Seven water/acetonitrile calibration standards were prepared
by gravimetrically adding distilled water to dry acetonitrile to
cover a range of 3002100 ppm moisture. The FTIR spectrum
292 F.R. van de Voort et al. / Talanta 72 (2007) 289295
of dry acetonitrile was subtracted from the spectra of the stan-
dards in a 1:1 ratio to remove the spectral contributions of
acetonitrile. Alinear regression calibration equation relating the
amount of moisture added (in ppm) to the absorbance value at
3676 cm
1
(relative to a baseline point at 3738 cm
1
) in the
differential spectra of the calibration standards was used to pre-
dict the moisture content of oil samples in order to assess the
extraction recovery, repeatability, reproducibility, and accuracy
of the method. Percent recovery was evaluated by spiking 10
different types of oil samples with increasing amounts of water
(2001000 ppmin200 ppmincrements) andassessedbydivid-
ing the FTIR-predicted values by their respective theoretical
gravimetric values and multiplying by 100. Repeatability and
reproducibility were evaluated by running three replicates of
12 new and used oil samples having high and low water con-
tents consecutively and on 3 different days and were assessed
by calculating the average standard deviation of the replicates.
Accuracy was evaluated using 58 validation samples, 23 of
which came fromJOAPwith the balance being commercial sam-
ples and samples prepared in our laboratory. All these validation
samples were analyzed by KF titration in the NETE labora-
tory. The accuracy of the FTIR method was assessed by linear
regression of the FTIR-predicted values against the reference
KF values and calculation of the standard error of prediction
(SEP).
3. Results and discussion
3.1. General considerations
Acetonitrile is a highly suitable solvent for the extraction of
water from lubricants and its subsequent quantitation by FTIR
spectroscopy for several reasons. First, it is sufciently polar
to be largely immiscible with both mineral- and ester-based
lubricants and is capable of efciently extracting water from
them. Moreover, because acetonitrile does not absorb strongly
in the portions of the mid-IR spectrum where water absorbs,
pathlengths of up to 1000 m may be employed for the analy-
sis of low levels of moisture, thereby providing high sensitivity.
Fig. 2 presents the OHstretching region in the spectra of a series
of calibration standards prepared by addition of water to acetoni-
trile fromwhich the spectrumof acetonitrile has been subtracted
out. Owing to the binding of water molecules exclusively to ace-
tonitrile rather than to each other at these high dilutions [7],
these spectra of water exhibit distinct symmetric and asym-
metric OH stretching bands (3630 and 3540 cm
1
), rather than
the single broad band normally observed. Both bands respond
linearly to added water and maintain consistent ratios relative
to each other over the concentration range examined (300
2100 ppm).
In addition to extracting water from oils, acetonitrile is also
capable of extracting some other polar constituents, such as alco-
hols, phenolic compounds, hydroperoxides, and organic acids,
but not ionic constituents. Since formulated lubricants contain a
wide variety of added constituents and can accumulate various
undened contaminants during the lubricants in-service life,
a number of species may be co-extracted with water into ace-
Fig. 2. Differential spectra of water in acetonitrile over a concentration range of
3002100 ppm, illustrating the two main water absorption bands, the symmetric
and asymmetric OH stretching bands at 3630 and 3540 cm
1
, respectively.
tonitrile. Such potential sources of spectral interferences were
examined by recording the spectra of acetonitrile extracts from
a wide variety of commercial new and used lubricants. Fig. 3
presents the 38503200 cm
1
region of the differential spectra
obtained after subtraction of the spectrum of dry acetonitrile
from the spectra of the standards and indicates that a variety of
constituents other than moisture were extracted into acetonitrile.
Although Fig. 3 shows that measurements at either of the
peak maxima of the water absorption bands would be subject to
substantial interference from various co-extracted components,
interference on the high-frequency side of the 3630-cm
1
band
was judged to be minimal. This is illustrated in Fig. 4, in which
the spectrum of the acetonitrile extract of a mineral-based oil
spiked with 286 ppm of water is superimposed on the spectra
obtained after extraction of the same oil spiked with 0.2 and
0.5% (w/w) 2,6-di-tert-butyl-4-methylphenol, corresponding to
the levels at which phenolic antioxidants are commonly present
in lubricant formulations. In the case of ester-based oils, the
bands of OH-containing constituents are broadened and shifted
toward lower wavenumbers as a result of hydrogen bonding
with the ester carbonyl group, thus generally being clear of the
Fig. 3. Differential spectra of acetonitrile extracts from a wide variety of com-
mercial newand used lubricants in relation to a differential spectrumof moisture
in acetonitrile, illustrating the spectral variability produced in the OH region.
F.R. van de Voort et al. / Talanta 72 (2007) 289295 293
Fig. 4. Differential spectrum of a mineral-based oil spiked with 0.2% (A) and
0.5% (B) 2,6-di-tert-butyl-4-methylphenol and extracted with acetonitrile and
a differential spectrum of 286 ppm water (C) spiked into the same base oil and
extracted into acetonitrile.
3630 cm
1
water band. Accordingly, the possibility of employ-
ing a location on the high-frequency tail of the 3630-cm
1
band
for the quantitation of water was investigated.
In principle, absorption measurements on the tail of a peak
are less reliable than measurements at the peak maximum since
they are more likely to be affected by changes in spectral band
shapes resulting fromintermolecular interactions such as hydro-
gen bonding. However, in the present case, such matrix effects
are expected to be minimal, as the samples are very dilute ace-
tonitrile solutions in which water has been shown to be well
dispersed without extensive hydrogen bonding (see Fig. 2). To
select the optimal measurement frequency, calibration equations
were derived by measuring the absorbance of water/acetonitrile
standards at 15 frequencies between 3700 and 3590 cm
1
at 8-
cm
1
intervals. These calibration equations were then used to
predict the amount of water in a mineral oil sample spiked with a
high level of phenol antioxidant (0.5%, w/w) and 286 ppmwater.
The results showed that the absorption of the phenol antioxidant
contributed to the predicted water content at wavenumbers lower
than 3676 cm
1
, with the magnitude of this contribution rang-
ing from40 ppm at 3660 cm
1
to 300 ppm at the peak of the
3630-cm
1
water absorption band. At 3676 cm
1
, the presence
of the phenol antioxidant contributed <20 ppm to the predicted
water content, and therefore 3676 cm
1
was chosen as the opti-
mal frequency for measuring water in acetonitrile extracts of
lubricating oils.
Asecond important consideration in developing this method-
ology was to compensate for the partial miscibility of certain
types of lubricating oils with acetonitrile, which would result
in underestimation of the moisture content of the oil owing to
dilution of the acetonitrile extract by oil. The most common
base oil types are mineral oils, synthetic hydrocarbon oils, vari-
ous synthetic ester-based oils, and polyalkylglycols. Mineral and
synthetic hydrocarbon base oils are effectively immiscible with
acetonitrile and separate completely with centrifugation, while
at the other extreme, polyalkylglycols are completely miscible.
Synthetic esters were found to have varying degrees of misci-
bility with acetonitrile, as evidenced by the observation of the
ester (C O) absorption at 1730 cm
1
in the spectra of their
acetonitrile extracts. To compensate for the partial miscibility
with acetonitrile of ester-based oils and hydrocarbon oils con-
taining substantial amounts of ester additives, the spectra were
multiplied by a dilution correction factor prior to the subtrac-
tion of the acetonitrile spectrum. The spectral multiplier was
calculated as the ratio of the height of the acetonitrile (CN)
band at 2068 cm
1
(measured relative to a two-point baseline
between 2125 and 2033 cm
1
) in the spectrum of the dry ace-
tonitrile solvent to the height of this band in the spectrum of
the acetonitrile sample extract. Based on the magnitude of this
dilution correction factor, the miscibility of the ester-based oils
with acetonitrile was estimated to fall within the range of 220%
depending on the type of ester base stock and hydrocarbon chain
length.
Given that various constituents and additive packages com-
monly found in mineral and hydrocarbon based oils could
affect the oils miscibility with acetonitrile, a variety of such
constituents were examined, including detergents, dispersants,
demulsiers, antioxidants, antiwear and extreme pressure addi-
tives, rust inhibitors, pour point depressants and viscosity
modiers, and these were not found to signicantly affect the
miscibility of the base oils with acetonitrile. Used mineral
based and synthetic hydrocarbon based engine oils, hydraulic,
compressor and gear uids also were not found to extract sig-
nicantly into acetonitrile (not more than 2%). As indicated
above, certain amounts of degradation by-products and addi-
tives were spectrally found to be extracted, but the resulting
dilution effect (assessed spectroscopically as described above
for the ester-based oils) was not signicant.
3.2. Calibration
Acalibration was developed using standards prepared simply
by adding water to acetonitrile at levels from 300 to 2100 ppm.
Owing to the low absorptivity of acetonitrile in the spectral
region of interest, a cell pathlength of 1000 m could be
employed. After subtraction of the spectrum of the dry acetoni-
trile from the spectra of the standards, the following calibration
equation was obtained using absorbance measurements made at
3676 cm
1
relative to a baseline at 3738 cm
1
:
H
2
O (ppm) = 5427.3A
3676/3738 cm
1 15.0,
R
2
= 0.999, S.D. = 10.9 (1)
As indicated by the regression S.D. of 11 ppm, the long path-
length allowed very high sensitivity to be attained. This equation
was programmed into the COAT

system to analyze the mois-

ture content of the acetonitrile extracts of lubricant samples
automatically, providing results directly in ppm moisture.
3.3. Validation and comparison to Karl Fischer titration
The analytical performance of the acetonitrile extrac-
tion/FTIR method was rst evaluated in terms of extraction
recovery, repeatability and reproducibility. Recovery was
assessed by spiking various amounts of water (2001000 ppm)
294 F.R. van de Voort et al. / Talanta 72 (2007) 289295
Table 2
Average percent recovery and S.D. for 10 different oil samples
Oil type Spiked additive Average recovery
a
(%) S.D. (%)
Mineral base oil 97.9 0.80
Mineral base oil Heavy-duty diesel engine oil additive (13.5%) 95.9 5.25
Mineral base oil Antiwear hydraulic oil additive (3%) 97.7 1.46
Mineral base oil Industrial gear oil additive (3%) 100.2 4.29
Ester base oil 101.1 0.52
Synthetic air compressor lubricant 101.9 5.65
Engine oil 92.9 2.78
Hydraulic uid 104.9 5.27
Turbine uid 96.5 4.54
Gear uid 94.4 6.41
a
Average recovery: average of % recovery values for same sample spiked with 200, 400, 600, 800 and 1000 ppm water.
into 10 samples of different types of oil, mixing vigorously on
a reciprocal shaker to ensure that the water was well dispersed
in the oil, and then analyzing the samples by the acetonitrile
extraction/FTIRmethod. For certain types of oil samples, recov-
ery increased as the time of mixing increased, as exemplied
by the following recovery data obtained for a fully formulated
hydraulic oil spiked with 600 ppm water: 2 min, 62%; 5 min,
80%; 10 min, 84%: 20 min, 92%. Average recoveries for the
10 different oil samples spiked with 2001000 ppm water in
200 ppm increments and mixed for 20 min were >90%, with
4 out of the 10 samples having 100% recovery (Table 2). Thus,
recovery was deemed satisfactory with a 20-min mixing time,
conrming the capability of acetonitrile to extract most of the
water present in lubricating oils.
For the assessment of repeatability and reproducibility, 12 oil
samples were analyzed in triplicate within the same run and on 3
different days, respectively. The data obtainedare summarizedin
Table 3. The average S.D. for repeatability was 8 ppm and the
average S.D. for reproducibility was 80 ppm. The much higher
S.D. of the data for reproducibility suggests that atmospheric
moisture level variations on different days may have affected
the results, indicating a certain level of moisture ingress during
sample preparation.
Generally, FTIR spectroscopy is considered a secondary
method requiring calibration against a recognized primary
method, such as KF titration; however, the FTIR method in this
case is in fact a primary method, traceable to gravimetrically
prepared water/acetonitrile standards, and, as such, the FTIR
moisture values predicted for lubricant samples should be accu-
rate. A performance comparison against the KF procedure was
undertaken to determine the validity of this assumption. Thus,
a set of used military vehicle oils (n =23) and a further 35 care-
fully selected samples were analyzed in the NETE laboratory by
the KF method and were also analyzed by the acetonitrile FTIR
method. Fig. 5 presents a plot of the FTIR moisture predictions
versus the KF results which yielded the following conventional
and zero-regression equations, respectively:
FTIRH
2
O = 36.6 + 0.987 KF H
2
O,
R= 0.950, S.D. = 80 ppm (2)
FTIRH
2
O = 1.063 KF H
2
O,
R= 0.950, S.D. = 83 ppm (3)
The intercept in Eq. (2) is well within the regression S.D., and the
change in the S.D. when the regression was forced through the
origin (Eq. (3)) was minimal, conrming that the FTIR results
are not biasedwithrespect tothe KFdata. Thus, the FTIRmethod
tracks the moisture content of the oils in the same manner as
the KF method, with an overall SEP of 80 ppm relative to
the KF reference values. The SEP is comparable to the S.D.
Table 3
Repeatability and reproducibility data for quantitation of water in 12 oil samples (mean and S.D. of three replicates)
Sample no. Type of oil Repeatability Reproducibility
Mean (ppm) S.D. (ppm) Mean (ppm) S.D. (ppm)
1 New motor oil 361.7 1.53 324.2 50.34
2 Used motor oil 755.7 10.12 666.6 108.31
3 Used motor oil 552.3 2.52 511.4 84.75
4 New bearing oil 5.0 8.54 39.3 50.95
5 New turbine oil 31.0 10.54 67.3 52.08
6 New engine oil 473.3 4.04 504.4 102.60
7 Polyol ester oil 342.7 10.69 467.6 207.43
8 New hydraulic oil 77.3 8.08 83.4 64.72
9 New gear oil 96.7 10.12 104.2 62.34
10 New turbine oil 86.3 8.39 92.1 53.24
11 Used motor oil 547.0 11.79 554.0 74.75
12 Used motor oil 590.7 9.29 584.2 65.24
F.R. van de Voort et al. / Talanta 72 (2007) 289295 295
Fig. 5. Regression of FTIR moisture predictions vs. the KF results for 58 used
and new oil samples.
for reproducibility of the FTIR method and indicates excellent
agreement between the FTIR and KF results.
Overall, the results of the validation study indicate that the
FTIR analysis of lubricant acetonitrile extracts is a viable and
accurate means of determining moisture content in a wide range
of lubricants. The rate of automated analysis on the COAT

system employed in this work is 72 acetonitrile extracts/h; in

addition, the sample preparation steps, apart from weighing out
the oil, can be performed in batch mode while the previous set
of samples is being processed. Accordingly, this FTIR method
provides a practical instrumental alternative to KF analyses,
allowingfor the rapid, automated determinationof moisture with
a high degree of accuracy. This methodology may be imple-
mented together with other FTIR methods developed in our
laboratory for the automated analysis of lubricants, including
acid number and base number [8] as well as lubricant condition
monitoring [2], thereby replacing several conventional, tedious
physical and chemical methods by FTIR methodology. As such,
FTIRanalysis offers substantial benets to commercial oil anal-
ysis laboratories in terms of both sample turnaround and cost
savings.
Acknowledgements
F.R. van de Voort would like to thank Thermal-Lube Inc. for
supporting his sabbatical sojourn at their research facilities in
Pointe-Claire, Quebec, Canada. Thanks also go to Dr. E. Urban-
ski of the Joint Oil Analysis Program run out of the U.S. Navy
Technical Support Center in Pensacola, Florida, for providing
lubricant samples. The very helpful collaboration of the Naval
Engineering Test Establishment, facilitated by its Director, Mr.
Mike Davies, is much appreciated. The authors acknowledge the
Natural Sciences and Engineering Research Council (NSERC)
of Canada for partial nancial support of this research.
References
[1] ASTM D6304: Standard Test Method for Determination of Water in
Petroleum Products, Lubricating Oils, and Additives by Coulometric Karl
Fischer Titration, ASTM International Standards.
[2] F.R. van de Voort, J. Sedman, R.A. Cocciardi, D. Pinchuk, Tribol. Trans. 49
(3) (2006) 410.
[3] ASTM E2412: Standard Practice for Condition Monitoring of Used
Lubricants by Trend Analysis Using Fourier Transform Infrared (FT-IR)
Spectrometry, ASTM International Standards.
[4] F.R. van de Voort, J. Sedman, V. Yaylayan, C. Saint Laurent, C. Mucciardi,
Appl. Spectrosc. 58 (2) (2004) 193.
[5] A. Al-Alawi, F.R. van de Voort, J. Sedman, Appl. Spectrosc. 59 (10) (2005)
1295.
[6] A. Al-Alawi, F.R. van de Voort, J. Sedman, A. Ghetler, JALA 11 (1) (2006)
23.
[7] A. Le Narvor, E. Gentric, P. Saumagne, Can. J. Chem. 49 (1971) 1933.
[8] F.R. van de Voort, J. Sedman, V. Yaylayan, C. Saint Laurent, Appl. Spectrosc.
57 (11) (2003) 1425.