Estimation of Bitumen and Solids

Content in Fine Tailings Using

Low-Field NMR Technique
S.C. Motta Cabrera, J. Bryan, A. Kantzas, University of Calgary TIPM Laboratory
Abstract
The oil sands mining and extraction processes in Canada produce large volumes of tailings that are a mixture of mainly water,
clay, sand, chemicals and bitumen. This mixture is transported
to tailings ponds, where gravity segregation occurs. During this
process, a stable suspension called mature fine tailings (MFT) is
formed, which requires many years to fully consolidate. Therefore, land reclamation and water recirculation become significant
environmental issues. For this reason, it is important to understand the tailings content and their settling properties. This study
uses the low-field nuclear magnetic resonance (NMR) technique
to estimate the water, bitumen and solids composition of synthetic and real tailings samples through a bimodal compositional
detection method under ambient conditions. NMR measurements
were conducted in 15 minutes, which is a relatively fast measurement allowing for rapid monitoring of tailings compositions. The
results show that the NMR technique can be a potential on-site
fast measurement of composition and settling characteristics of
tailings.

Introduction
In recent years, a considerable fraction of bitumen production
in Canada comes from oil sands mining operations. The management of tailings is an important environmental concern, because
the production of one barrel of synthetic crude oil requires 2 m3 of
processed water (75% coming from recycled process water) and
produces 1.8 tonnes of solid tailings that are discharged into tailings ponds. Therefore, the tailings ponds require more volume than
the original oil sand, and as mining operations progress, the size of
these disposal ponds can become significant. In these ponds, sand
settles faster than fine clay solids, which form stable suspensions
that require thousands of years to reach full consolidation without
the aid of chemical additives. The accumulated volume of these
mature fine tailings (MFT) is expected to increase to over 1109
m3 by the year 2020(1).
Tailings are composed mainly of water (58% 62%), a small
amount of unrecovered bitumen (1% 5%) and solids (sand and
clays, 37%). The solids fraction is constituted of 95% clays and
5% sand. The three main clays present in the clay fraction are kaolinite, illite and sodium montmorillonite(2).
Although plant-scale improvements have been made to decrease the amount of remaining bitumen in tailings and to improve
their settling rates (i.e., the addition of flocculants and coagulants
to the water), there is a need for an on-site fast and real-time measurement of tailings composition to identify when upstream process modifications are needed in order to reduce the amount of
oil present in tailings. These modifications include activities such
as varying the flocculant concentration in response to the settling
response of the tailing, and can take place during the separation of
the aerated bitumen aggregates from the water-solids slurry.
8

This study proposes the use of the low-field NMR technique as

a non-destructive on-site application in the oil sand mining process
to perform rapid estimations of bitumen, water and solids content
in synthetic and real tailings samples.
The NMR technique is currently employed to establish compositions of oil/brine emulsions and the viscosity of heavy oil
and bitumen(3,4). Moreover, NMR logging tools are used in reservoir characterization and measuring properties, such as permeability(58), porosity(7,8), mobile and immobile fluids(912) and fluid
saturations(13).
Monitoring of tailings composition is usually accomplished by
Dean-Stark extraction. In comparison to this technique, NMR provides a faster composition measurement and requires a smaller
sample. Another advantage is that the NMR technique can provide
compositional results on-site or may even be implemented online
in a process, such as for oil sand processing.
Fundamentally, NMR measures the response of the nuclei of
certain atoms when they are immersed in a static magnetic field
and are then exposed to a second oscillating magnetic field. Some
nuclei experience this phenomenon, and others do not, depending
upon whether they possess a property called spin. The nuclei used
most often in NMR are hydrogen, phosphorus and Carbon 13. The
hydrogen atom is the most commonly studied nucleus for petroleum applications because it has the largest magnetic moment;
consequently it emits the strongest signal of any stable nucleus(14).
In addition, protons are present in hydrocarbons and water; thus,
NMR signals can be related to the volumes of both of these fluids
present in a given sample.
An NMR tool uses magnetic fields to polarize the protons in a
fluid (oil and water) and monitors the time it takes for these protons
to return to equilibrium. This time is known as the transverse relaxation time (T2). A low-viscosity fluid, such as water, presents a T2
relaxation time on the order of 2,000 milliseconds(9). In the case of
conventional oils, whose viscosity is higher than water, relaxation
times will vary from similar T2 values for water, down to a few
hundred milliseconds. Heavy oils and bitumen, which are more
viscous than either water or conventional oil, have T2 values under
10 milliseconds. The relaxation time is faster for oil and bitumen in
comparison to water, because protons movement in heavy oil and
bitumen is restricted by the viscous environment(3,8,11).
Each NMR measurement cycle, known as a Carr-Purcell-Meiboom-Gill pulse sequence (CPMG), follows a fixed pattern given
by four parameters known as time echo (TE) or echo spacing,
number of B pulses, wait-time between successive trains or repetitions and number of trains. With these parameters, the NMR tool
generates a spectrum showing the amplitude and relaxation time as
a function of the quality and quantity of fluids analyzed.
In the measurement of tailings, the appropriate determination
of the CPMG pulse parameters is fundamental because tailings
are characterized by having small amounts of highly viscous oil
and solids that are suspended in a continuous water phase. The
small fast-relaxing oil signal, coupled by the presence of a large
water signal (both bound and free water), makes it difficult to accurately quantify the oil content in the sample. This is particularly
Journal of Canadian Petroleum Technology

true if the NMR measurement is performed using the same CPMG

or pulse parameters to detect both water and oil signals simultaneously. Therefore, in order to improve the oil estimation in tailings
samples, a new method was developed during this study to detect
water and oil signals separately within a relatively short measurement time of 15 minutes. The method is denominated as the bimodal compositional detection method because it uses two sets
of CPMG pulse parameters for establishing water and oil signals
separately. The solids content in a sample is calculated by mass
balance.
The quantification of the fluids present in a sample through
NMR is determined based on the fact that the signal of a given
fluid is directly proportional to the measurable amount of hydrogen present in this fluid. This is known as the fluid hydrogen
index, which is defined as the moles of hydrogen per unit volume
of test liquid, referenced to the moles of hydrogen per unit volume
of water(11). Moles of hydrogen are not easy to measure directly, so
in practice the measured amplitude is simply related to the amount
(mass or volume) of fluid present. In experiments performed at the
Tomographic Imaging in Porous Media (TIPM) Laboratory(15), it
was proven that the relationship between the cumulative amplitude
of a given fluid (oil or water) and the mass of the fluid is linear.
This relation is presented in Equation (1):
m fluid =

At
AI ............................................................................................. (1)

where mfluid is the mass of a given fluid within the sample (g), At is
the cumulative total amplitude measured and AI is the amplitude
index of the single fluid. Therefore, if the amplitude index of oil
or water is previously known, then the mass of the fluid in an unknown sample can be calculated using Equation (1).
During this study, numerous fine tailings synthetic samples
were made with variable mixing configurations, quality and quantity of water, bitumen, sand and clay concentrations. These samples were tested in NMR tools under ambient conditions using the
bimodal compositional detection method. Real tailings synthetic
samples, provided by Shell Canada, were later analyzed using the
same method. The results of the experiments indicated the minimum water and oil detection limits that can be expected in an
NMR tool and the percentage errors of the mass predictions using
the new set of CPMG pulse parameters.

Experimental
Materials
All synthetic sample measurements were performed at ambient
temperature and pressure using two Ecotek NMR tools at a frequency of approximately 1.5 MHz. The denomination given to
these two machines were NMR1 and NMR2. Both tools were designed to be essentially twins of one another, but for sensitive measurements, such as tailings, it was necessary to identify the effect
of different electronics on the resulting fluid content estimations.
Cold Lake and Athabasca bitumen were the two types of viscous model oils used in this study. Their densities were measured
as 0.9960 g/cm3 and 1.0198 g/cm3, respectively. These values were
measured at 23C using an Anton Paar DMA 45 digital density
meter with a built-in thermostat. The viscosities for Cold Lake and
Athabasca bitumen at ambient temperature (23C) were found to be
66,670 mPas and 828,450 mPas, respectively. Viscosity measurements were made at four elevated temperatures and extrapolated to
ambient temperature (23C) using the Andrade-Guzman-Reynolds
equation(16). All viscosity measurements were performed using a
Wells-Brookfield cone and plate viscometer.
Although oil sands mining and the subsequent generation of
tailings usually involves Athabasca bitumen, the properties of the
oil (i.e., viscosity) may vary with depth or location in the reservoir(17). Certain processes also involve the addition of naphtha or
other solvents to the bitumen, which could also act to reduce its
July 2010, Volume 49, No. 7

viscosity. Therefore, there is value to analyzing the NMR response

to different ranges of high-viscosity bitumen.
Two types of water were used in the generation of synthetic
samples: distilled water and brine. The distilled water and the
brine had density values of 0.9970 g/cm3 and 1.0012 g/cm3, correspondingly, at 23C. The brine was supplied by Shell Canada, and
had the typical water composition used in industrial oil sands processing: salinity of 1,211 ppm, a sodium content of approximately
300 ppm and a calcium-magnesium content lower than 50 ppm.
The solids used were sand and three types of clays. The sand
was Lane Mountain 70, and the clays were illite (B-Clay, IRC),
kaolinite (Elmer Kaolin-KRC) and sodium montmorillonite (Wyoming SWy-2Montmorillonite, Na-MRC). Plainsman Clay, located
in Medicine Hat, Alberta, supplied the KRC and the IRC. These
two clays are mined in Troy, Idaho and Ravenscrag, Saskatchewan, respectively. Source Clays Repository in West Lafayette,
Indiana, supplied the montmorillonite clays. The Na-MRC was
mined in Apache Country, Arizona(18).

Procedure
Analysis of NMR spectra focuses on the monitoring of amplitudes and relaxation times as a function of the quality and quantity of the fluids present in a sample. Therefore, the use of two
qualities of water and bitumen in the tailings synthetic samples
requires previous knowledge about the behaviour of each one of
the single substances and their corresponding amplitude indexes.
These values were found by correlating the amplitudes to different
amounts of each fluid. Afterwards, mixtures of solids and water or
brine, and bitumen with water or brine were studied to observe the
behaviour of each system through time.
The variation of the water and brine amplitude indexes in the
presence of kaolinite and illite mixture was also analyzed through
time. After analysis of the behaviour of the previous mixtures, two
sets of synthetic tailings samples were prepared using a characteristic tailings composition: 1% 5 % bitumen, 58% 62% water
and 37% solids. This solids proportion was distributed as 5%
coarse solids or sand and 95% fine solids or clays. The clay proportion was prepared with 60% of illite, 35% of kaolinite and 5%
of sodium montmorillonite(19). These synthetic tailings samples
were also studied as a function of time and the mixing configuration. The substances used in each set of tailings synthetic samples
are shown as follows:
First set of samples: distilled water, Cold Lake bitumen,
sand and three clay types (kaolinite, illite and sodium
montmorillonite).
Second set of samples: brine, Athabasca bitumen, sand and
three clay types (kaolinite, illite and sodium
montmorillonite).
The purpose of the tests with different qualities of water and bitumen was to develop expected quantification limits for different
types of fluids. Subsequently, middlings and tailings samples were
prepared from real oil sands in a batch extraction unit (BEU) by the
Shell Canada Research Centre. These samples were also analyzed
through the NMR technique using the bimodal compositional detection method, and are denoted as real tailings samples.

Results and Discussion

Analysis of Single Substances: Water and
Bitumen
Different amounts of distilled water and brine (0.5g 12 g) were
measured in the two NMR tools to establish the amplitude indexes
of each type of water through mass balance. The CPMG pulse parameters used to detect water signals in all the samples are presented in Table 1.
Figure 1 shows the straight lines obtained from the correlation
between the amplitude data and different amounts of distilled water
and brine using the NMR2 tool. The slopes obtained represent the
9

Table 1: Parameters to detect water signal in a

sample.

CPMG Pulse Parameters

Time to echo (ms)
Number of B pulses
Wait time (ms)
Number of trains

Table 2: Parameters to detect bitumen in a sample.

Values
0.32
5,000
10,000
26

amplitude index values for each type of water. The values obtained
for distilled water and brine were 0.1424 and 0.1453, respectively.
Using the NMR1 tool, the distilled water and brine amplitude indexes were found to be 0.1656 and 0.1646, respectively. The difference in water amplitude index values between the two machines
is attributed to differences in the NMR electronics.
The similarity in the amplitude index values between brine and
distilled water on either machine shows that the water plant salinity
does not have a significant effect on the water amplitude index. It
has been demonstrated that in extremely saline brines this factor
may change(20), but in oil sands mining applications water or brine
amplitude index may be taken as a constant value.
Figure 2 shows the NMR spectra for 3 g of distilled water and
brine. Both qualities of water relaxed at approximately 2000 milliseconds, because for this low-salinity brine the viscosity of distilled water and brine were the same. As expected, there is no
difference in the NMR spectra of water and brine at low salinity.
The fact that water relaxes so slowly allowed for a definition of
1,000 milliseconds as the cutoff for bulk or free water. Faster relaxing water would be attributed to surface relaxation associated
with sand or clay.
The average percentage of errors found using the NMR2 tool for
varying distilled water and brine mass predictions was 1.9% and
1.2%, respectively. The results obtained using the NMR1 tool were
2.7% and 4.5% for distilled water and brine. In practice, therefore,
the error bars must be established for each machine use. The minimum detection limit for both qualities of water was established as
0.5 g in both machines.
In general, the NMR2 tool provided better and more consistent
results than NMR1 tool. However, water and brine used during
the measurements presented essentially the same response in either machine.

Analysis of Bitumen
Cold Lake and Athabasca bitumen are the two unconventional
oils analyzed in this study. As in the case of water, the NMR1 tool
was initially used to perform the measurements in order to determine the accuracy in the bitumen mass predictions.
The initial amounts of Cold Lake bitumen measured in the
NMR1 tool varied from 0.2 g 13 g. Measurements were first
made using the same CPMG pulse parameters as had been used
to detect the water signal. The correlation obtained between amplitudes and mass values was good; however, a significant

CPMG Pulse Parameters

Time to echo (ms)
Number of B pulses
Wait time (ms)
Number of trains

Values
0.16
200
100
1,000

deviation was observed when the amount of oil present in the

sample was less than 3.5 g. This deviation or inaccuracy in oil
mass predictions is caused mainly by the high viscosity of oil that
restricts the movement of its protons and generates oil peaks with
faster relaxation times.
The result is that for small amounts of oil, it becomes difficult for the NMR tool to accurately capture the fast-decaying oil
signal and separate this signal from possible noise in the first few
echoes. The amplitude values therefore do not correspond to the
real amount of oil present in a sample. In order to capture a better
oil signal, the CPMG parameters were required to be modified and
this was particularly true for tailings, where the oil content is usually small. For this reason, a new set of CPMG parameters was
established.
Table 2 shows the pulse sequence parameters used to detect the
bitumen content in the samples. In order to capture the oil signal,
fewer B pulses were used and significantly more trains were run.
Because of the short wait time, these measurements could still be
performed quickly. Thus, a more accurate measurement of viscous
oil could be made still within a short time (less than 5 minutes).
The same procedure was performed in the NMR2 tool to find
the amplitude index values using samples containing up to 6 g of
Cold Lake and Athabasca bitumen. Small oil masses were specifically chosen because the oil amount present in a tailings sample is
low. An amplitude index equal to 0.1152 for Cold Lake bitumen
and 0.0650 for the Athabasca bitumen were obtained from the
slopes in Figure 3. The fact that both oils have smaller amplitude
index values compared to water, and that the Athabasca bitumen
had a lower amplitude index compared to Cold Lake bitumen, is
related to the high oil viscosities(3).
The NMR spectra for 1.5 g of each type of bitumen are shown
in Figure 4. From these spectra it is possible to observe that both
viscous bitumens relax before 10 milliseconds, which is distinctly
different from water. This relaxation time allows for a definition
of the upper limit cutoff for bitumen any signal above 10 milliseconds will definitively not be the contribution of bitumen. It is
notable that Athabasca bitumen has less amplitude and a shorter
mean relaxation time than Cold Lake bitumen because of its higher
viscosity; consequently, NMR detection of this type of bitumen
is of greater difficulty. According to Figure 3, the minimal oil
detection limit for both types of bitumen was determined as 0.6 g.
This is higher than the detection limit of water, and is caused by the
fact that viscous bitumen is more difficult to measure than water.
The average percentage of errors found using the NMR2 tool for
Cold Lake and Athabasca bitumen mass predictions were 6% and
0.3

1.8
1.6

0.25

1.2

0.2

Amplitude

Amplitude

1.4

1
0.8
Distilled water
0.6

Brine

Distilled water
0.15

Brine

0.1

0.4
0.2

0.05

0
0

12

Mass, g
FIGURE 1: Amplitude vs. mass of different amounts of distilled water
and brine.

10

0
0.1

10

1000

T2, ms
FIGURE 2: NMR spectra for samples of distilled water and brine.

Journal of Canadian Petroleum Technology

0.45

0.7

Cold Lake

0.6

Athabasca

0.4

0 hours
24 hours

0.35
0.3

0.5

Amplitude

Amplitude

0.8

0.4
0.3
0.2

0.25
0.2
0.15
0.1

0.1

0.05

0
0

0.1

10

1000

T2, ms

Mass, g

FIGURE 5: NMR spectra of kaolinite and brine mixture.

FIGURE 3: Amplitude vs. mass for two types of bitumen.

10%, respectively. The errors obtained using the NMR1 tool were
10% for both types of bitumen. As in the case of water, the NMR2
tool provided better results than the NMR1 tool. Therefore, it can
be once again concluded that different tools will exhibit different
sensitivity to bitumen, but in general bitumen is more difficult to
quantify than water.

Fine tailings normally consist of 58 wt% 62 wt% water and 37

wt% solids. The solids generally contain a clay content of approximately 95 wt%, while the other 5 wt% is sand. Kaolinite, illite and
sodium montmorillonite are the three main types of clay present.
The high-clay fraction in tailings makes it important to study the
behaviour of each clay type mixed with water or brine through
time. Spectra analysis of each mixture allows for the determination
of typical relaxation times and the swelling capacity of each clay
type. The purpose of these measurements was to identify if signals
at early times could still be taken to be the contribution from oil, or
if the oil and clay-bound water signals would begin to overlap. In
order to determine the clay response, the three typical clays were
mixed separately with distilled water and brine in a proportion of
50/50 wt%. Based on the typical tailings composition, a synthetic
tailings sample of 20 g contained 7.4 g of water; thus, the amount
of clay added to each sample was also 7.4 g. To make these samples, water was added to dry clay at 0 hours and soaked into the
clay over a 24-hour period.
Figures 5 to 7 show the spectra obtained for kaolinite, illite and
sodium montmorillonite mixed with distilled water and brine at
0 and 24 hours using the parameters to detect water signals. The
spectra of kaolinite and illite mixed with both types of water
looked similar to one another. Both clays showed relaxation times
of approximately10 milliseconds 100 milliseconds. In contrast,
spectra of water in montmorillonite exhibited a faster relaxation for

0.4

0 hours
24 hours

0.35
0.3

Amplitude

Analysis of Water and Solids Mixture

0.45

0.25
0.2
0.15
0.1
0.05
0
0.1

10

1000

T2, ms
FIGURE 6: NMR spectra of illite and brine mixture.

both 0 and 24 hours. This phenomenon was especially significant

for the distilled water sample, and may be attributed to a higher
swelling capacity of this clay in the absence of salt.
Spectra of kaolinite and illite presented a peak after 1,000 milliseconds that corresponds to free water; the other peaks, between
10 milliseconds 100 milliseconds, represent the effect of surface
relaxation or available solid surface(20). After 24 hours, all spectra
exhibited faster relaxation, indicating that the movement of water
into small pores appeared to be a time-dependent process, as air
had to be displaced out of the clay, while water imbibed in.
The difference between illite/kaolinite and montmorillonite is
evident, both in the short and long T2 values. When the clay was
montmorillonite, there was no bulk water peak present even at

0.018

0.3

0.016
0.25

0 hours

0.014
0.2

Cold Lake

Amplitude

Amplitude

24 hours

Athabasca

0.012
0.01
0.008
0.006

0.15

0.1

0.004
0.05
0.002
0
0.1

10

T2, ms
FIGURE 4: NMR spectra for samples of bitumen.

July 2010, Volume 49, No. 7

100

0
0.1

10

T2, ms

1000

FIGURE 7: NMR spectra of sodium montmorillonite and brine mixture.

11

0.6

0.5

0 hours
24 hours

Amplitude

0.4

0.3

0.2

0.1

0
0.1

10

1000

T2, ms
FIGURE 8: NMR spectra of sand and brine mixture.

0 hours. This was significant because the initial mixture contained

50 wt% of water. This indicates the clay structure swelled, and all
the water was now associated with surface relaxation. In addition,
the water relaxation times were predominantly less than 10 milliseconds at 0 hours, but the water relaxed slowly after the fluids
had been allowed to re-distribute for 24 hours. This is opposite to
the behaviour of illite and kaolinite. These spectra also illustrated
that the relaxation times for water bound to clay were not always
greater than 10 milliseconds; consequently, no single cutoff T2
value could be assumed for separating the fraction of clay-bound
water from bitumen that could be present in the samples.
In the absence of bitumen in these synthetic samples, all amplitude data were related to water, so the total amplitudes were used
to predict water content in these kinds of mixtures. When small
amounts of bitumen are present, the oil and clay-bound water relaxation times are similar, which makes separation difficult(21). In
general, however, several conclusions could be made. If the clay
is predominantly illite or kaolinite, the relaxation times, even for
surface-bound water, do not begin until close to 10 milliseconds
and end at approximately 100 milliseconds. If the sample contains significant montmorillonite, the relaxation times for water
can overlap with that of the bitumen. Therefore, when significant
montmorillonite is present, NMR predictions of oil and water content will be less accurate.
Water mass prediction for the samples containing kaolinite, illite
and brine had an average error of 2.6% and 1.0%, respectively. The
water mass predictions for the same mixtures using distilled water
presented errors of 1.6% and 1.2%. This may have been caused by
differences in the clay-water interactions in the absence of salt.
In the case of the samples containing sodium montmorillonite,
the average percentage error found was 6.8%. This result is less accurate in comparison to the other type of solids. This could be attributed to higher iron content in this clay, which could affect the
2.6
0 hours
2.4

24 hours

After having studied the behaviour of each solid mixed separately with water or brine, the stability of the water amplitude
index through time was verified when brine or water was mixed
with both of the two typical clays found in the highest proportions
in tailings. The stability of the water amplitude index is important
because any variation in this value through time directly affects
the water mass prediction in a sample. Any systematic changes in
amplitude index will be caused by chemical interactions between
the clay and water or brine, while random fluctuations are a reflection of the machine stability.
The typical kaolinite and illite content in a tailings sample is 35
wt% and 60 wt% of the clay fraction, respectively(22). Thus, seven
samples of 20 g were prepared containing a solid phase consisting
of 60% illite, 40% kaolinite and distilled water ranging from 40
wt% 70 wt% of the total sample mass, in intervals of 5 wt%. Another seven samples were prepared with brine and the same clay
proportions. All samples were analyzed at 0, 24, 48, 96 hours and
1 week later in the NMR1 tool.
In Figure 9, the fluid mass is plotted against the NMR amplitude for the various samples at each of the times measured. The
slope of the line is the amplitude index of the water or brine. It is
evident that the brine amplitude index remains constant even after
1 week, with an average value of AI = 0.1670. The same stability
was obtained for the mixtures using distilled water and their amplitude index found was AI = 0.1688. The maximum average error
in the water mass predictions was found to be 5.0% for brine and
distilled water, respectively. The amplitude index values are essentially the same as those obtained with measurements of water alone
using the NMR1 machine; thus, these tests showed that the mixture
of kaolinite and illite with water or brine does not affect the water
and brine amplitude indexes over time. In this manner, NMR predictions of water content in tailings samples should be valid both at
early times to monitor conditions in the plant, or after the tailings
have already settled for extended times in ponds.

Analysis of Water and Bitumen Mixture

96 hours
2

Amplitude

Analysis of Water and Kaolinite-Illite Mixture

48 hours

2.2

One week

1.8
1.6
1.4
1.2
1
5

10

15

Mass, g
FIGURE 9: Amplitude vs. mass for kaolinite-illite and brine mixtures.

12

signal given by the NMR tool, or the measured water content may
have been incorrect because of additional water that already existed in the clay matrix(18). If the clay is predominantly montmorillonite, then the NMR water content predictions will therefore tend
to be less accurate. However, in general this clay type should not
have a significant effect on the tailings measurements because its
proportion in clay content is less than 5 wt%(22).
As mentioned previously, sand constitutes approximately 5 wt%
of the solid content in a tailings sample. Thus, distilled water and
brine were mixed separately with sand in the same proportion used
with the clay mixtures. Figure 8 shows the spectra for sand with
brine at 0 and 24 hours. The relaxation time found for sand was
in the range of several hundred milliseconds, only slightly faster
than the signal from bulk water. In the absence of confining pressure, the pore spaces in sand grains are much larger than in clays,
leading to slower relaxation. Previous studies have shown that the
presence of sand does not have a significant effect on the amplitude index of water(18), and this was also verified in this work. The
average error for the mixtures containing sand and brine was 0.9%,
and 1.5% for the mixtures prepared with distilled water.

Mixtures of water and bitumen were prepared with the two qualities of bitumen and water, and were measured using the established CPMG parameters. The initial mixtures were prepared with
different amounts of bitumen ranging from 0.6 g 6.0 g and approximately 10 g of water in each sample. These samples therefore had oil contents ranging from 5 wt% 37.5 wt%. The first
set of samples was prepared with Cold Lake bitumen and distilled
water, and in the second set of samples Athabasca bitumen and
plant brine were used. In both cases the oil was added first, and
then water was placed on top of the oil. Because of the difference
in density, the oil could then float to the top of the water.
Each sample was measured using the parameters to detect the
oil and water signals separately at 0 and 48 hours in the NMR1
Journal of Canadian Petroleum Technology

Table 3: Mixing configurations used in synthetic

samples.
0 hours

Amplitude

0.015

48 hours

0.01

0.005

0
0.1

10

1000

T2, ms
FIGURE 10: NMR spectra of Athabasca bitumen and brine mixture
using parameters to detect oil signal.

tool. The oil and water measurement took approximately 4 and 10

minutes, respectively; consequently, a complete measurement of
each sample in the NMR required approximately 15 minutes.
Figure 10 shows the spectrum generated using the parameters
to detect oil signals for the mixture of 0.6 g of Athabasca bitumen
with 10 g of brine at 0 and 48 hours. These spectra show well-defined peaks below 10 milliseconds for the small amount of bitumen
used. Because of small amounts of noise below 10 milliseconds, a
fixed cutoff could not be applied; rather, oil was taken as the amplitude up to the end of the first peak. The 10-millisecond cutoff
was still used as the upper limit, however; any signal above 10
milliseconds was known not to be oil. An additional observation
made from these samples was that the oil signal would begin close
to or before 1 millisecond. Therefore, if a given measured amplitude peak is to be attributed to bitumen, it should begin close to
or before 1 millisecond, and should end before 10 milliseconds.
The water signal in Figure 10 is not accurate because of the short
waiting times; only oil amplitude can be quantified from these
spectra.
Figure 11 presents the spectra for the same mixture, but using
the parameters to detect water signals. The water peaks in these
spectra were also well-defined after 1,000 milliseconds; this value
was used as the cutoff for water mass calculation.
The average error found in oil and water mass prediction was
7% and 1.5%, respectively, for Cold Lake and Athabasca mixtures, respectively. As expected, when the oil content was low, the
resulting error in the oil mass prediction could be significant. The
results of the oil/water mixture measurements were simply obtained to serve as an illustration of the approach used to determine
the fluid content separately, through multiple measurements on the
same samples.

Analysis of Synthetic Tailings Samples

At this stage, the stability of the NMR tool had been tested for
pure samples and mixtures of bitumen/water and water/solids. The
next step was to determine if accurate predictions of fluid content
could be made using synthetic tailings samples, containing known
masses of oil, water and solids.
Synthetic samples of 20 g were prepared using typical fine tailings compositions. The mixing configuration was the main parameter varied in these samples. The methodology used was to create
two large groups, depending on the oil and water, and then to subdivide each group into three smaller groups titled Test One, Test
Two and Test Three. Each small sub-group had five samples, comprising 15 samples in total per main group. In the first group, Athabasca bitumen and brine were used, while Cold Lake bitumen and
distilled water were the fluids used in the second group.
The samples of both groups were measured in the NMR2 tool.
The five samples of each sub-group contained bitumen from 1 wt%
5 wt%, water from 58 wt% 62 wt% and 37 wt% of solids. This
means that the first sample contained 1 wt% of oil, 62 wt% of
water and 37 wt% of solids. The second sample contained the same
solid percentage, but 2 wt% of bitumen and 61 wt% of water, and
July 2010, Volume 49, No. 7

Position of
Each Substance
in a Sample
Test 1
Test 2
Test 3

Bottom
Clay and sand Clay and sand Bitumen

Centre
Bitumen
Water
Water

Top
Water
Bitumen Clay and sand

so on with the other samples. The fixed solids component in these

samples was a mixture of clay (<22 m) and sand (>22 m), which
was prepared before each test. The solids consisted of 95 wt% clay
and 5 wt% sand. Illite (60 wt%), kaolinite (35 wt%) and sodium
montmorillonite (5 wt%) were the three type of clays used to prepare the clay fraction because these are the three main clays typically found in the fine tailings. The weight fractions of the various
clays were also decided based on typical tailings compositions(2).
The difference between the three tests is principally the mixing
configuration or the position of each substance in a sample. This
methodology therefore studied the interaction amongst oil, water
and solids (clay and sand) when they are layered differently in
order to determine if the configuration of the components would
have any effect on the subsequent NMR interpretations. Table 3
shows the different mixing configurations used for each test.
Initially in each test, the two substances that were placed in the
bottom and the centre of a sample were measured at 0 and 24 hours
in order to verify the stability of the machine through the consistency in the results. After 24 hours, the third substance was added
to the top of the sample, and the new mixture was once again measured at 0 and 24 hours. The final distribution of fluids was governed by gravity because the samples were not shaken. Settling
characteristics for agitated samples are the topic of a separate stage
of tailings characterization(23). The pulse parameters used to analyze each mixture were selected according to the type of substances mixed. For example, in Test One the initial bitumen and
solids (clays and sand) mixture was measured only with the pulse
parameters to detect the oil signal at 0 and 24 hours. When water
was added to these samples, the pulse parameters to detect both
water and oil signals separately were then used to perform the measurements. The results obtained for each test are presented in the
following section. All the NMR spectra shown in the figures correspond to the mixture prepared with brine and Athabasca bitumen,
because this was the system with the most similar characteristics
to real tailings samples.
Test One Results
The NMR spectrum obtained at 0 and 24 hours for the initial mixtures of Athabasca bitumen (O) and solids (S) were made
using the parameters to detect the oil signal. These spectra showed
a well-defined bitumen peak before 1 millisecond for the samples
containing 3, 4 and 5 wt% of bitumen. Figure 12 shows the spectra
obtained for the sample with 4 wt% of bitumen at both times. The
1
0.9
0.8

Amplitude

0.02

0.7

0 hours

0.6

48 hours

0.5
0.4
0.3
0.2
0.1
0
0.1

10

1000

T2, ms
FIGURE 11: NMR spectra of Athabasca bitumen and brine mixture
using parameters to detect water signal.

13

0.4

0.003

0.002

S+O @ 0 hours

Amplitude

Amplitude

0.0025

S+O @ 24 hours
0.0015
0.001

0.35

S+O+W @ 0 hours

0.3

S+O+W @ 24 hours

0.25
0.2
0.15
0.1

0.0005

0.05
0

0
0.1

10

0.1

100

FIGURE 12: NMR spectra of Athabasca bitumen and solids mixture at

0 and 24 hours, oil signal.

change in the apparent amplitude and relaxation time is caused

by the uncertainty in measuring only a small amount of viscous
bitumen.
The NMR tool did not detect the oil signal from the samples
containing 0.2 g (1 wt%), and 0.4 g (2 wt%) of bitumen, which
was a reflection of the minimum bitumen mass that could be detected in the NMR tool. The error in the bitumen mass predictions
for the samples containing 4 to 5 wt% of bitumen was found to be
5%. Moreover, it is noticeable that when the samples only contained bitumen and solids, the presence of the solids did not lead
to any monotonic change in the peak location with time. This was
expected because the oil is so viscous that there is no effect of surface relaxation on the oil peak location.
In Figure 13 it can be observed that when brine (W) was added
to the solids and bitumen mixture, an additional signal was measured from water relaxing in the pores of the clay/sand mixture. The
spectra obtained for samples containing 4 and 5 wt% of bitumen,
using the parameters to detect oil signals, still showed well-defined
bitumen peaks after 24 hours; in contrast, the sample containing 3
wt% of bitumen no longer showed an oil peak. Therefore, in the
presence of a large signal of slower relaxing fluid, the small oil
signal may be lost. The error in oil mass predictions for the samples containing 4 and 5 wt% of bitumen was 9%. The cutoff used
to determine the bitumen content in all samples was based on the
end point of the first peak found before or at approximately 1 millisecond. It was important to use the floating cutoff (at approximately 1 millisecond) to avoid attributing the amplitude from the
water bound to clays into the bitumen amplitude. If no signal was
observed until after 1 millisecond, then it was determined that the
sample did not contain bitumen in sufficient measurable quantities,
and only clay-bound water had been measured.
The NMR spectra illustrated in Figure 14 corresponds to the
same mixture containing 4 wt% of Athabasca bitumen, but
was measured using parameters to detect water signals. In the

1000

FIGURE 14: NMR spectra of Athabasca bitumen, brine and solids

mixture at 0 and 24 hours, water signal.

presence of sand and clay, no fixed water cutoff was used because
some of the water is relaxing through surface relaxation with the
solids, and these did not relax at the same time for all samples.
The peaks observed before or at approximately 100 milliseconds
correspond to clay-bound water or the surface relaxation effect(9).
The peak found after 1,000 milliseconds was denominated bulk or
free water. It was assumed that by using parameters to detect the
water signal, the signal from the oil was negligible; therefore, all
the measured amplitude was converted to the water mass. The average absolute error in water mass predictions for all the samples
containing brine was 3%.
The samples prepared with Cold Lake bitumen and distilled
water yielded similar results, compared to the set of samples already described. However, for these tailings the oil signal was
still detected in the samples containing 3 wt% of oil. This could
be attributed to the Cold Lake bitumen being less viscous than
Athabasca bitumen; therefore, its protons possess a little more
movement, which makes detection easier for the NMR tool. The
errors for oil and water mass predictions in these samples were 9%
and 7%, respectively.
Test Two Results
The samples prepared in the Test Two group contained a water
layer between the bitumen and the solids, which Takamura(24)
proposed as the structure that allows for the separation of oil from
the sand, or the water-wetting nature of the sand.
In the first part of Test Two, solids and water were mixed and
measured using the parameters to detect water signals at 0 and 24
hours. The NMR spectra and the total amplitude values obtained
at both times were found to be similar. Figure 15 shows the results
obtained for the sample that subsequently would have 3 wt% (0.6
g) of Athabasca oil. The total amplitude obtained in this figure
was caused by the signal coming only from water. As in Test One,

0.1

0.3
S+O+W @ 0 hours

0.09

S+W @ 0 hours

S+O+W @ 24 hours

0.25

0.08
0.07

S+W @ 24 hours

0.2

0.06

Amplitude

Amplitude

10

T2, ms

T2, ms

0.05
0.04
0.03
0.02

0.15

0.1

0.05

0.01
0
0.1

10

100

1000

T2, ms
FIGURE 13: NMR spectra of Athabasca bitumen, brine and solids
mixture at 0 and 25 hours, oil signal.

14

0
0.1

10

T2, ms

1000

FIGURE 15: NMR spectra of brine and solids mixture at 0 and 24

hours, water signal.

Journal of Canadian Petroleum Technology

0.35

0.14

0.3

0.12

S+W+O @ 0 hours
S+W+O @ 24 hours

0.2
0.15

0.08
0.06

0.1

0.04

0.05

0.02

0
0.1

10

W+O+S @ 24 hours

0.1

Amplitude

Amplitude

0.25

W+O+S @ 0 hours

1000

0.1

10

1000

T2, ms

T2, ms
FIGURE 16: NMR spectra of Athabasca bitumen, brine and solids
mixture at 0 and 24 hours, water signal.

the peaks observed between 10 milliseconds 1,000 milliseconds

correspond to the clays bound with water and after 1,000 milliseconds, the peak represents free water. The peaks are measured
at similar T2 values at both 0 and 24 hours, indicating that water
immediately entered and reached a good distribution between the
grains of the solids. Although the amplitude at 24 hours appears to
be lower than at 0 hours, if the actual amplitude values are summed
the total water amplitude is similar at both times.
After 24 hours, bitumen was added and the samples were measured again with both pulse parameters at 0 and 24 hours. The behaviour of this mixture is illustrated in Figures 16 and 17 using
parameters to detect water and oil signals, respectively. As in Test
One, the samples with 4 and 5 wt% of bitumen showed a bitumen
peak close to 1 millisecond at 0 and 24 hours.
For the samples with 1 and 2 wt% of oil, the first peaks measured
occurred close to 10 milliseconds, which cannot be interpreted as
bitumen because its relaxation time is too slow. This means that,
as was also the case in Test One, bitumen below 3 wt% could not
always be measured. The error in bitumen mass prediction for the
samples containing 3 wt% 5 wt% of bitumen was 11%.
In Figure 17 it was also possible to observe the presence of
water signals at approximately 10 milliseconds, which indicates
that the parameters to detect oil signals could also be used to identify fast-relaxing clay-bound water. In this study, however, the
total water quantification was only calculated from the spectra obtained using the parameters to detect the water signal. With these
parameters, the small bitumen signal was not detectable therefore
all the amplitude was attributed to water. The error in water mass
predictions was 2.5% for the samples containing brine. In the case
of the samples containing Cold Lake bitumen and distilled water,
the percentage errors for oil and water mass predictions were 8%
and 7%, respectively.

FIGURE 18: NMR spectra of Athabasca bitumen, brine and solids

mixture at 0 and 24 hours, oil signal.

Test Three Results

For this test, the initial mixtures were composed of bitumen
and water and were measured using both pulse parameters to detect water and oil signals at 0 and 24 hours. Well-defined bitumen
peaks at approximately 1 millisecond were observed for the samples containing upward of 3 wt% (0.6 g) of bitumen. As was also
observed previously during the analysis of oil/water mixtures, the
oil relaxed close to 1 milliseconds and no oil peak was measured
using the parameters to detect the water signal. In addition, it is
observed that the relaxation times for the different substances or
mixtures were similar to the relaxation times obtained in Tests One
and Two.
After 24 hours, the mixture of clay and sand was added to the
samples and settled through the oil and water. The spectra obtained
at different times, using the parameters to detect the oil signal, are
shown in the Figure 18 for the sample containing 3 wt% of Athabasca bitumen.
An oil peak was observed only in the spectra obtained after 24
hours; the peak obtained close to 10 milliseconds at 0 hours cannot
be assumed to be oil because its relaxation time was too slow.
Therefore, once again the strong influence of the clay-bound water
signal at 0 hours served to mask the signal from the oil; after the
system had settled to its equilibrium state the oil signal could once
again be measured.
The spectra obtained using the parameters to detect the water
signal are illustrated in Figure 19, which clearly show the presence
of clays at approximately 100 milliseconds. In this mixing configuration, the clays were now settling through the aqueous phase
and the final settled water distribution is no longer instantaneous.
Rather, the water relaxed faster after 24 hours, at which time the
solids had settled and the final pore size distribution had been obtained. The errors found for bitumen and water mass predictions
were 12% and 3.2%, respectively.
0.7

0.12
S+W+O @ 0 hours

0.6

S+W+O @ 24 hours

0.1

Amplitude

Amplitude

W+O+S @ 24 hours

0.5

0.08
0.06
0.04

0.4
0.3
0.2

0.02
0

W+O+S @ 0 hours

0.1
0
0.1

10

1000

T2, ms
FIGURE 17: NMR spectra of Athabasca bitumen, brine and solids
mixture at 0 and 24 hours, oil signal.

July 2010, Volume 49, No. 7

0.1

10

1000

T2, ms
FIGURE 19: NMR spectra of Athabasca bitumen, brine and solids
mixture at 0 and 24 hours, water signal.

15

In the samples prepared with Cold Lake oil and distilled water,
oil peaks were detected only for the samples containing 4 wt% or
5 wt% of oil. The errors for oil and water mass predictions were
found to be 9% and 8%.
When the mixtures contained only bulk water and viscous bitumen, any fast-relaxing signal could be measured and related to
oil content. In contrast, when there was also a surface water signal
that was changing with time, it was more difficult for both the actual acquisition and later the inversion software to discriminate
between the oil and bound water peaks. Despite these additional
uncertainties, the oil amplitudes still vary within the same range
as the water amplitude values. As in the other sample configurations, however, the small amounts of oil coupled with the lower
oil amplitude index will lead to greater errors in the predicted oil
fractions.
In general, the NMR tool presented difficulties to detect 1% or
2% of bitumen in the samples. Moreover, the mass predictions in
the tailings synthetic samples were more reliable after 24 hours
when all substances had already mixed.
The results obtained from NMR measurements on these synthetic tailings samples allow for several main findings to be summarized. Accuracy of oil content predictions was strongly dependent
on proper acquisition of the oil signal (i.e., a clean measured
signal) and a proper oil amplitude index. The NMR tool was able
to detect bitumen in amounts higher than 0.6 g and water could be
detected in all samples. Finally, in general NMR measurements
provide the same level of accuracy for tailings samples in all three
configurations tested. Therefore, the methods of spectrum analysis
developed in this study can be applied in predictions of actual real
tailings samples, which will have been mixed in unknown configurations of fluids and solids.

Analysis of Real Tailings Samples

Shell Canada Research Centre, located in Calgary, Alberta, provided a total of 34 blind tailings samples that had been prepared
from tests performed in a batch extraction unit (BEU). These samples were divided in two main groups, made up of 22 and 12 samples, respectively. These samples were initially measured twice in
the NMR tool using the new method; subsequently, the oil, water
and solids content of the same samples were determined by Dean
Stark (DS) extraction. This methodology was designed to test the
NMR predictions in blind samples against those obtained by DS,
which is the accepted method used in the industry to characterize
fluid content in tailings samples.
Moreover, Shell Canada prepared sister samples of the first and
second group of blind samples and sent them also for DS analysis. The goal of the testing with sister samples was to evaluate if
they possessed the same compositional characteristics as the other
blind samples. If so, then wider testing could eventually be done at
the plant-scale, by sending sister samples for NMR analysis, while
process modifications were made based on the original tailings.

First Group of Real Tailings Samples

The first group of real tailings samples was sub-divided into
two smaller groups by the Shell team. No information about the
samples was initially provided in order to evaluate the predictions
coming from the NMR technique using the bimodal compositional
detection method on blind samples. The two sub-groups were composed of Vials 1 10 and 11 22, respectively.
The samples were initially measured using the pulse parameters
to detect the water signal; the average water content for Samples 1
10 was found to be 7.88 g (17.1 wt%), compared to 20.19 g (69
wt%) for Samples 11 22. Assuming low oil content is present in
the tailings, this yields a solids content on the order of 80 wt% in
the first 10 samples, and 30 wt% in the second group of 12 samples.
Based on these results, it was inferred that the first set of samples
was representative of the tailings (i.e., the bottom) stream from separation processes, which should contain a higher weight fraction of
solids. These solids will also contain a proportionally higher fraction of coarse sand, compared to middlings. Samples 11 22 are
more representative of middlings. These samples were inferred to
contain fewer solids, but these solids would likely consist mainly
of clay suspensions. These assumptions were later confirmed when
the NMR results were compared with the results obtained by Shell
using the DS extraction method, and through identification of the
water peak locations in the measured samples.
The average error in water content for Samples 1 10 (tailings) and 11 22 (middlings) were found to be 0.79% and 0.43%,
in that order. These results show that the NMR technique can be
considered as a potential tool in the prediction of the water content
in middlings and tailings in the oil sands extraction process. The
graphical comparison of the NMR and DS results acquired for tailings and middlings samples are also illustrated in Figures 20 and
21, correspondingly.
Samples 1 22 were also measured in the NMR tool using the
pulse parameters to detect the oil signal. In Samples 1 10, oil
was detected only in 4 samples; however, the amounts predicted
were under 0.60 g, which had been established as the minimum
detection limit for the NMR tool through measurements of synthetic samples. Based on this fact and taking into account that the
water signal indicated that Samples 1 10 were tailings that usually contain little oil, it was presumed that the measured signal at
early relaxation times would be strongly affected by small amounts
of noise present during the early echoes. As a result, it was not
possible to identify how much of the measured amplitude was actually related to bitumen, and how much was noise. Thus, the predicted oil masses were not expected to be accurate. Well-prepared
tailings, which contain low bitumen content, are outside the detection limits of the NMR.
In the case of Samples 11 22, oil was detected in 9 out of 12
samples, but as in the other sub-groups, the amounts predicted were
still under 0.60 g. Once again, therefore, these samples should have
been below the detection limit for the NMR tool. However, based
on the results obtained from the water signal, these samples were
assumed to be middlings, which typically have a higher amount of

12

25

NMR Water Mass

20

Water Mass, g

Water Mass, g

10

DS Water Mass
NMR Water Mass

DS Water Mass

15

10

10

Sample No.
FIGURE 20: Comparison of water content results of Samples 1 10
(tailings) using NMR and DS techniques.

16

11 12

13

14

15

16

17

18

19 20

21

22

Sample No.
FIGURE 21: Comparison of water content results of Samples 11 22
(middlings) using NMR and DS techniques.

Journal of Canadian Petroleum Technology

0.6

0.6

DS Oil Mass
0.5

Real Oil Mass

NMR Oil Mass

NMR Oil Mass

0.4

0.4

Oil Mass, g

Water Mass, g

0.5

0.3

0.2

0.3

0.2

0.1

0.1

0
1

10

11 12

13

14 15

oil in comparison to tailings. This led to two possibilities: either

the detection limit was actually lower than what was measured for
the synthetic samples, or the signal obtained from the measurements was once again a combination of the amplitude of the bitumen and noise. The NMR oil predictions were compared with
those obtained by the DS method and it was found that in most of
the cases bitumen content in the samples was over-predicted by
the NMR technique. Therefore, it was concluded that the measured
amplitude values at early relaxation times (at approximately 1 millisecond) were strongly affected by noise in the measurements, and
the NMR predictions of oil mass will not be accurate even with optimized acquisition parameters. Figures 22 and 23 illustrate the differences resulting from both measurements.
Second Group of Real Tailings Samples
After the first set of 22 real tailings samples had been analyzed,
Shell Canada was encouraged by the preliminary findings and decided to provide a second set of blind samples for further analysis.
This set of 12 real tailings samples was provided based on their internal identification of varying qualities of oil sands. The purpose
of this test was to determine if bitumen content could be identified
below the detection limit obtained from the synthetic samples. The
weight range of these samples was between 25 g 43 g.
All samples were again initially measured using the parameters to detect the water signal. From the NMR predictions, it was
determined that the average water content for Samples 1 6 was
4.76 g (11.7 wt%), and was 18.03 g (66.3 wt%) for Samples 7
12. Additionally, the water peak locations in Samples 1 6 were
also at longer relaxation times than for Samples 7 12, indicating
a higher fraction of coarse solids in these samples. From the NMR
estimations of water and solids content, it was inferred that Samples 1 6 were tailings and Samples 7 12 were middlings.
It was then expected that more oil would once again be detected
in the second set of samples, when measured using the oil parameters. However, for these samples the NMR tool was only able to
detect a signal in the range of T2 values associated with oil for two
samples out the 12, and the amount predicted was again under the
established detection limit. Therefore, it was expected that these
samples most likely contained less oil than the original set of samples, and that the bitumen amplitude in the two samples where a
signal was measured would be strongly affected by noise.
The results were compared with those obtained through the DS
technique and the average error in water mass predictions was
found to be 6%. NMR oil mass predictions were only made for
the two samples where some amplitude was detected; in the other
samples the NMR was not able to measure any response from
bitumen.
In general, oil content could not be accurately quantified in samples that contained less than 0.6 g of oil. However, the results of
the NMR oil measurements are still valuable because the identification of a measurable oil signal could still serve as a warning
that there is too much oil present in the tailings. In this manner,
an application of NMR would not be used as a quantifiable tool,
July 2010, Volume 49, No. 7

17 18

19 20

21

22

FIGURE 23: Comparison of bitumen content results of Samples 11

22 (middlings) using NMR ad DS techniques.

but as an indicator that the process is off-spec. Moreover, it may

be possible to obtain more accurate measurements of water and oil
content through a combination of NMR and other measurements.
Possible alternatives include the addition of toluene to mix with the
oil and aid separation from water. Alternatively, after water content is measured through the NMR, water could be vaporized by
microwaves and the resulting hot bitumen would be easier to measure. These process alternatives could be tested in the future.
Comparison of Dean Stark Results for Sister Samples
In most tests that are carried out in order to test a new technology, an effort is made to eliminate bias in the predictions. In
this spirit, comparative studies are usually performed as blind samples in different laboratories. In this case, NMR tests were first
completed on samples that were provided by Shell Canada, but
at the same time, sister (supposedly twin) tailings/middlings were
also extracted for DS analysis.
The other goal of this study was to determine if NMR measurements would be similar to the results obtained with sister samples.
If this were possible, then ideally a wider testing program could be
initiated in actual plant samples, whereby some samples would be
collected for NMR analysis, while process modifications would
be based on measurements performed on these sister samples.
NMR predictions were first compared against the sister samples
DS numbers, and there was considerable scatter even in the water
predictions. Therefore, the samples that had been analyzed in the
NMR were later sent for DS analysis as well. Figure 24 shows
a comparison between the DS water mass results for the original samples vs. their sister samples. These samples belong to the
first group of samples analyzed (1 10), which were found to be
tailings. The oil mass results are also compared in Figure 25. Because the samples themselves had different total masses, the numbers are normalized against the solids content in order to properly
identify compositional differences between sister samples.
0.3

Original
Sisters

0.25

Water/Solids Ratio

FIGURE 22: Comparison of bitumen content results of Samples 1 10

(tailings) using NMR ad DS techniques.

16

Sample No.

Sample No.

0.2
0.15
0.1
0.05
0
1

10

Sample No.
FIGURE 24: Dean Stark water results for the original vs. sister
samples.

17

0.007

Sisters

0.006

Oil/Solids Ratio

The middlings contain faster relaxing water peaks, compared

to the tailings. This corresponds to water bound to clay; although
the solids content is lower in middlings, the water is all physically
constrained by the suspended clays. The significance of this observation is that the nature of the solids (sand vs. clays) will lead
to distinct differences in the settling behaviour of the tailings(23).
Therefore, the spectra obtained using the parameters to detect the
water signal do not just provide an indication of the water content
and the type of sample being measured, but they also provide information regarding the settling properties of the suspended solids.

Original

0.005
0.004
0.003
0.002
0.001
0

10

Sample No.
FIGURE 25: Dean Stark oil results for the original vs. sister samples.

The difference between the original and sister samples is especially significant in the oil content, although differences are also
present for water/solids content as well. This leads to an important observation regarding measurements of tailings samples: for
such heterogeneous systems, considerable error can be introduced
by comparing NMR results to sister samples. In order to evaluate
NMR applications in tailings, it is imperative to compare NMR
and DS on the same samples. In plant-scale operations, the fact
that middlings and tailings are so heterogeneous also means that
in order to determine that a given middlings or tailings stream has
certain oil, water or solids content, multiple measurements should
be taken and averaged. A single NMR measurement may not be
representative of the actual middlings or tailings composition in
the plant at that time.
Interpretation of Solids Properties From NMR Water
Measurements
The identification of a sample as being from the middlings or
tailings stream was related to the relative fraction of water and
solids, and also to the nature of the solids themselves. Unless a significant oil signal is measured, the middlings or tailings will contain little oil. Therefore, because it is known that the NMR water
content predictions are accurate, subtracting the NMR water mass
from the total mass of the sample gives a first-pass estimate of
the solids, and this is denominated solids by difference. As expected, the calculation of solids mass by difference is similar to
the measured solids mass from the DS extraction method. The average errors for the first and second groups of samples were found
to be 3.1% and 4.4%, respectively. Because the solids are determined using the NMR estimates of the water mass, the errors in
the solids will be directly proportional to the errors in the water
mass predictions.
In addition, the spectra generated using the NMR water parameters can also provide an indication of the nature of the solids. Figures 26 and 27 show an example of NMR spectra of a tailings and
middlings sample, respectively. These spectra were acquired using
the parameters to detect the water signal.

Conclusions
It has been demonstrated that an NMR bimodal compositional
detection method may be used to estimate water and bitumen content in synthetic tailings samples. These measurements, which
can be obtained in approximately 15 minutes, were then validated
against real tailings. Two qualities of water and bitumen were
used to prepare the synthetic samples, and the samples were analyzed using different mixing configurations, as a function of time.
Through the measurement of these samples, it was determined that
the NMR tool is able to detect bitumen in amounts higher than
0.6 g.
A total of 34 real tailings samples were measured in the NMR
tool, and the bitumen and water mass predictions were compared
against the results obtained using the DS extraction method. NMR
water contents were accurate when compared against DS values,
and the spectra could also be used to infer the nature of the solids
in the samples (i.e., coarse sand or fine clays). With this information, the samples can be identified as being either tailings or middlings. For all of the real samples provided, the oil masses present
were under the detection limit that had been determined from the
synthetic tailings samples. For samples containing a higher proportion of bitumen, it was observed that some oil amplitude could be
detected, but for under 0.6 g of bitumen this oil content could not
be accurately quantified.
Based on the results of this study, NMR can be considered as a
potential application for online determination of tailings streams
composition.

Acknowledgements
The authors wish to acknowledge the contributions of the staff
at the TIPM Laboratory, especially Sergey Kryuchkov. The advice and aid from Brad Komishke, Eva Anfort, Jill Amstrong and
Lauren Asplund from Shell Canada are also gratefully acknowledged. The collaboration of Pedro Pereira and Harvey Yarranton
from the University of Calgary is also appreciated.
The authors also acknowledge the Canada Research Chair in
Energy and Imaging affiliates (Shell, Nexen, Devon, PetroCanada,
Canadian Natural, ET Energy, Suncor, Schlumberger, Laricina
and Paramount). Finally, many thanks are extended to NSERC for
funding this research.

0.2

1.2

0.18

0.16

0.8

0.12

Amplitude

Amplitude

0.14

0.1
0.08
0.06
0.04

0.6
0.4
0.2

0.02

0
0.1

10

T2, ms
FIGURE 26: NMR spectrum of a tailings sample.

18

1000

0.1

10

1000

T2, ms
FIGURE 27: NMR spectrum of a middlings sample.

Journal of Canadian Petroleum Technology

Nomenclature
At
AI
mfluid
O
S
T2
TE
W

= cumulative total amplitude of the spectra

= amplitude index
= mass of a given fluid within the sample, g
= bitumen
= solids
= transverse relaxation time
= time echo
= water

References
1. Chalaturnyk, R.J., Scott, J.D., and Ozum, B. 2002. Management of Bitumen Sand Tailings. Petroleum Science and Technology 20 (910):
10251046.
2. Mikula, R.J., Kasperski, K.L., Burns, R.D., and Mackinnon, M.D.
1996. Nature of fate of oil sands fine tailings. In Suspensions: Fundamentals and Applications in the Petroleum Industry, ed. L.L. Schramm,
No. 251. Washington, DC: Advances in Chemistry Series, ACS.
3. Bryan, J., Kantzas, A., and Bellehumeur, C. 2005. Oil-Viscosity Predictions From Low-Field NMR Measurements. SPE Res Eval & Eng 8
(1): 4452. SPE-89070-PA. doi: 10.2118/89070-PA.
4. Bryan, J., Kantzas, A., and Moon, D. 2005. In-situ Viscosity of Oil
Sands Using Low Field NMR. J Can Pet Technol 44 (9): 2330.
5. Curwen, D.W. and Molaro, C. 1995. Permeability from Magnetic Resonance Imaging Logs. Paper 1995-GG presented at the SPWLA 36th
Annual Logging Symposium, Paris, 2629 June.
6. Kenyon, W. 1997. Petrophysical Principles of Applications of NMR
Logging. The Log Analyst 38 (2): 2143.
7. Straley, C., Rossini, D., Vinegar, H., Tutunjian, P., and Morriss, C.
1997. Core Analysis by Low-Field NMR. The Log Analyst 38 (2).
8. Prammer, M.G. 1994. NMR Pore Size Distribution and Permeability
at the Well Site. Paper SPE 28368 presented at the SPE Annual Technical Conference and Exhibition, New Orleans, 2528 September. doi:
10.2118/28368-MS.
9. Coates, G.R., Xiao, L., and Prammer, M.G. 1999. NMR Logging Principles and Applications. Houston, Texas: Gulf Publishing Company.
10. Ohen, H.A., Ajufo, A., and Curby, F.M. 1995. A Hydraulic (Flow) Unit
Based Model for the Determination of Petrophysical Properties from
NMR Relaxation Measurements. Paper SPE 30626 presented at the
SPE Annual Technical Conference and Exhibition, Dallas, 2225 October. doi: 10.2118/30626-MS.
11. Kleinberg, R.L. and Vinegar, H.J. 1996. NMR Properties of Reservoir
Fluids. The Log Analyst 37 (6): 2032.
12. Freedman, R. 2006. Advances in NMR logging. J Pet Technol 58 (1):
6066. SPE-89177-MS. doi: 10.2118/89177-MS.
13. Bryan, J., Mai, A., Hum, F., and Kantzas, A. 2006. Oil and Water Content Measurements in Bitumen Ore and Froth Samples Using Low
Field NMR. SPE Res Eval & Eng 9 (6): 654663. SPE-97802-PA. doi:
10.2118/97802-PA.
14. Belliveau, S., Mirotchnik, K., and Kantzas, A. 1996. Principles of nuclear magnetic resonance and magnetic resonance imaging. Enpe 513
course notes, University of Calgary, Calgary, Canada (Fall 2000).
15. Mirotchnik, K.D., Allsopp, K., Kantzas, A., Curwen, D., and Badry, R.
2001. Low-Field NMR Method for Bitumen Sands Characterization:
A New Approach. SPE Res Eval & Eng 4 (2): 8896. SPE-71208-PA.
doi: 10.2118/71208-PA.
16. Gutmann, F. and Simmons, L.M. 1952. The Temperature Dependence
of the Viscosity of Liquids. J. Appl. Phys. 23: 977978.
17. Larter, S., Adams, J., Gates, I.D., Bennett, B., and Huang, H. 2008.
The Origin, Prediction and Impact of Oil Viscosity Heterogeneity on
the Production Characteristics of Tar Sand and Heavy Oil Reservoirs. J
Can Pet Technol 47 (1): 5261. Paper No. 08-01-52. doi: 10.2118/0801-52.
18. Chung, P.M.K. 2007. Water uptake assessment of brine contaminated
soils using low-field nuclear magnetic resonance. MSc thesis, University of Calgary, Calgary, Alberta (October 2007).
19. Schramm, L.L. ed. 1996. Suspensions: Fundamentals and Applications
in the Petroleum Industry, No. 251, 677690. Washington, DC: Advances in Chemistry Series, ACS.
20. Hum, F.M. and Kantzas, A. 2005. Using Low-Field NMR to Understand Fines Behaviour. Paper SCA 2005-46 presented at the 19th
International Symposium of the Society of Core Analysts, Toronto,
Ontario, Canada, 2125 August.

July 2010, Volume 49, No. 7

21. Bryan, J., Mai, A., Hum, F., and Kantzas, A. 2006. Oil- and WaterContent Measurements in Bitumen Ore and Froth Samples Using
Low-Field NMR. SPE Res Eval & Eng 9 (5): 654663. SPE-97802-PA.
doi: 10.2118/97802-PA.
22. Shaw, R.C., Schramm, L.L., and Czarnecki, J. 1996. Suspensions in
the hot water flotation process for Canadian oil sands. In Suspensions:
Fundamentals and Applications in the Petroleum Industry, ed. L.L.
Schramm, No. 251. Washington, DC: Advances in Chemistry Series,
ACS.
23. Motta Cabrera, S.C., Bryan, J., Komishke, B., and Kantzas, A. 2009.
Study of the Settling Characteristics of Tailings Using Nuclear Magnetic Resonance (NMR) Technique. International Journal of Mining,
Reclamation and Environment 23 (1): 118.
24. Takamura, K. 1982. Microscopic structure of Athabasca oil sand.
The Canadian Journal of Chemical Engineering 60 (4): 538545.
doi:10.1002/cjce.5450600416.

This paper (2007-074) was accepted for presentation at the 8th Canadian
International Petroleum Conference (58th Annual Technical Meeting),
Calgary, 12-14 June, 2007, and revised for publication. Original manuscript received for review 2 May 2009. Revised paper received for review 9
April 2010. Paper peer approved 14 April 2010 as SPE Paper 138973.

Authors Biographies
Sandra Motta Cabrera is currently employed by Suncor Energy as a process engineer in training, where she works in tailings
management. She previously worked as an
oil sands research engineer in TIPM Laboratory, and as an environmental and quality
process engineer for Tetra Pak Colombia.
Motta was also a consultant of quality and
environmental management systems (ISO
9001 and ISO 14000) for different companies in Colombia. Her interests are in oil
sands mining, particularly in the characterization of tailings. She
holds an M.Sc. degree in chemical engineering from the University
of Calgary and a B.Sc. degree in chemical engineering from the National University of Colombia. She is a member of APEGGA.
Jonathan Bryan is a senior research and development engineer at the TIPM Laboratory.
His current interests are in thermal and nonthermal heavy oil recovery, and in petrophysics and reservoir characterization. Bryan
holds a B.Sc. degree in chemical engineering
with a petroleum minor from the University
of Calgary. He obtained his M.Sc. and Ph.D.
degrees from the University of Calgary in
the area of heavy oil and bitumen characterization using low-field NMR and enhanced
heavy oil recovery. Bryan is a member of
APEGGA and SPE.
Apostolos Kantzas is currently a professor
at the University of Calgary, holder of the
Canada Research Chair position in energy
and imaging and director of the TIPM Laboratory. He leads a 40-person research group
that is involved in research related to problems of flow through porous media, enhanced
oil recovery, soil remediation, reactor design
and tomographic imaging. Kantzas has authored or co-authored over 300 technical papers and over 200 technical reports. Before
his appointment at the university, he held a senior research engineer
position in the pipeline and oil technologies department at NOVA
Research and Technology Corporation. He holds a Dipl. Eng. in
chemical engineering from the Aristotle University of Thessaloniki
in Greece and M.A.Sc. and Ph.D. degrees in chemical engineering
from the University of Waterloo. Kantzas is a member of APEGGA,
SPE, CSChE, AIChE, CWLS and SCA.
19