You are on page 1of 8

5456 Ind. Eng. Chem. Res.

2008, 47, 5456–5463

Correlation of Cement Performance Property Measurements with C3S/C2S Ratio


Determined by Solid State 29Si NMR Measurements
Christopher L. Edwards,†,‡ Rickey Morgan,‡ Lewis Norman,‡ Gary P. Funkhouser,‡ and
Andrew R. Barron*,†,§,|
Department of Chemistry, Rice UniVersity, Houston, Texas 77005; Halliburton Energy SerVices,
Duncan, Oklahoma 73536; Department of Mechanical Engineering and Materials Science, Rice UniVersity,
Houston, Texas 77005; and Energy and EnVironmental Systems Institute, Rice UniVersity, Houston, Texas 77005

The physicochemical and engineering performance properties of several API class G and H ordinary Portland
cements (OPCs) from various foreign and domestic sources have been investigated in comparison with the
tricalcium silicate/dicalcium silicate ratio (C3S/C2S) as determined by magic angle spinning (MAS) 29Si nuclear
magnetic resonance (NMR) experiments. XRF-derived oxide analysis appears to provide a lower C3S/C2S
ratio than determined by NMR analysis. Furthermore, oxide analysis suggests that all the cements have a
C3S/C2S ratio of 2-5, while our NMR method suggests the actual range is significantly broader. Determination
of C3S/C2S ratios by NMR provides an effective method of analysis for cements, owing to NMR’s direct
measurement of the minerals in question. NMR C3S/C2S ratios demonstrate predictive ability for the
determination of engineering performance properties. This is especially the case for prediction of strength
development; in keeping with generally accepted understanding of cement hydration behavior, the strength
development correlates with increasing C3S/C2S ratio, i.e., C3S content. The observed correlation between
NMR-derived silicate ratio and strength development holds for cements in the presence of either a retarder
(lignosulfonate) or a fluid loss additive (N,N-dimethylformamide/2-acrylamido-2-methylpropanesulfonic acid
copolymer). No significant correlation is observed between C3S/C2S ratio and the 72 h crush strength. The
lack of dependency of either thickening time or Young’s modulus to the C3S/C2S ratio as determined by
MAS 29Si NMR measurements suggests that these physical properties are independent of the relative silicate
composition. No correlations are observed between any physical property and the silicate ratio derived from
XRF data.

Introduction its correlations with SEM.3 Ghosh and Handoo reviewed the
progress in using infrared and Raman spectroscopies to monitor
Cement hydration is a very complex process, not solely
cement hydration behavior.4 Coveney, Fletcher, and co-workers
dictated by simple solvation and reprecipitation. The prediction
were able to build on the existing infrared knowledge base by
of the set kinetics, reaction to inhibitors, and the properties of
applying artificial neural net (ANN) analysis to diffuse reflec-
the set cement is presently not available. Instead of prediction,
tance (DRIFTS) spectra of cements.5,6
a series of experiments must be performed on each sample to
We have recently reported that for the MAS 29Si NMR of
obtain the property information required. It would be desirable
cements the longitudinal relaxation time T1 of C3S and C2S is
to have a simple analytical method (or short series of methods)
significantly altered because of the paramagnetic Fe3+ present
that would allow for the prediction of the chemistry of individual
in the C4AF matrix that makes intimate contact with each of
samples of cement from various sources. In this regard an
the silicate phases.7 The presence of the paramagnetic C4AF
attempt was made to correlate data from a wide range of
matrix thus allows for the rapid collection of spectra of cements.
spectroscopic and analytical techniques with experimental data
Furthermore, the relative concentration of tricalcium silicate
obtained by traditional methods. The ultimate goal of this project
(C3S) and dicalcium silicate (C2S) within a cement sample (i.e.,
is the development of a tool for the prediction of the hydration
C3S/C2S ratio) can be determined using the saturation recovery
chemistry of various cements based upon their chemistry.
method, offering the best combination of accuracy, speed, and
Various attempts in the past have been made to correlate
ease of processing. Our studies showed that the peak intensities
engineering performance properties with more conventional
are dependent on the relaxation times (τ). As the relaxation time
spectroscopic analyses. Barnes and co-workers were able to
is increased during the 29Si NMR saturation recovery experi-
identify several key factors to the role of gypsum in preventing
ment, the volume fraction of the silicate particle (crystallite)
“flash set” by means of time-resolved in situ X-ray diffraction.1
being “observed” increases until the entire particle is being
Parrott and co-workers applied a series of different tests,
sampled (Figure 1). Unfortunately, deconvolution analysis of
including quantitative X-ray diffraction and conduction calo-
the resulting spectra does not provide an absolute concentration
rimetry, to attempt correlations with degree of hydration.2
or weight percentage. However, the processed peak areas
Vlachou and Piau conducted a study of additive response and
attributed to the C2S and C3S content do allow for an accurate
* To whom correspondence should be addressed. E-mail: arb@ derivation of the C3S/C2S ratio.7 Using the data from the longest
rice.edu. relaxation times (i.e., after the intensity curve in Figure 1

Department of Chemistry, Rice University. plateaus) will provide the relative concentrations of C3S and

Halliburton Energy Services.
§
Department of Mechanical Engineering and Materials Science, Rice C2S. Thus, the C3S/C2S ratio can be calculated by a consider-
University. ation of the intensity ratio at τ values above which the slope in
|
Energy and Environmental Systems Institute, Rice University. the plot of intensity of the various saturation recovery parameters
10.1021/ie8000925 CCC: $40.75  2008 American Chemical Society
Published on Web 06/24/2008
Ind. Eng. Chem. Res., Vol. 47, No. 15, 2008 5457
Table 1. Mineral Composition of Portland Cements from XRF and
NMR Analysis
from XRF from NMR
cement C3S C2S C3A C4AF C3S/C2S C3S/C2S
BRZ 57.77 16.23 5.31 9.87 3.55 8.82
CAP 54.90 22.40 0.00 18.10 2.45 4.95
ELT 61.40 13.14 0.00 18.93 4.67 5.85
RUS 54.49 22.04 2.65 12.45 2.47 6.34
SDN 52.13 24.89 2.68 12.55 2.09 5.00
THA 57.46 21.78 1.37 13.62 2.63 5.37
TXI 55.99 22.13 5.72 9.74 2.53 3.07

Table 2. Consistometry Data for Cements without Additives, with


Retarder, and with Fluid Loss Additive
time to time to time to
fluid loss 50 Bc 70 Bc 100 Bc
Figure 1. Scaled signal intensity for C3S (9) and C2S (0) in a typical sample retardera additiveb (min)c (min)c (min)c
Portland cement (CAP) as a function of NMR relaxation (τ).
BRZ no no 53 56 63
plotted versus τ is essentially 0 (e.g., τ g 1 s in Figure 1). yes no 66 72 78
Determination of C3S/C2S ratios by NMR provides an effective no yes 51 61 70
CAP no no 78 95 108
method of analysis for cements, owing to NMR’s direct yes no 90 98 107
measurement of the minerals in question without the assump- no yes 73 74 75
tions of composition that are required with oxide analysis from ELT no no 90 120 134
X-ray fluorescence (XRF) and application of the Taylor- yes no 155 168 180
modified Bogue equations.8 no yes 94 98 102
RUS no no 69 71 79
In a typical series of experiments that must be performed on yes no 78 87 93
a given cement sample to match it to the specifications of a no yes 75 81 85
particular cementing job, there are four key measurements: SDN no no 80 96 110
thickening time determination in a consistometer, ultrasonic yes no 105 128 139
analysis of compressive strength development, Young’s modu- no yes 78 82 84
THA no no 54 65 73
lus, and crush strength (the latter two obtained from destructive yes no 78 83 92
mechanical testing). However, irrespective of the final analytical no yes 45 70 78
process, the Portland cement samples must be hydrated and set TXI no no 78 87 98
while measurements are performed. If there is to be an yes no 124 137 147
no yes 91 92 94
application of NMR as a tool for guiding downhole applications
a
of cement, then it is necessary to demonstrate whether there Lignosulfonate retarder at a loading of 0.5% bwoc. b Copolymer of
exist predictive correlations and for which parameters. N,N-dimethylformamide and 2-acrylamido-2-methylpropanesulfonic acid.
c
Bearden consistency units, proportional to the torque on the paddle in
Herein we have investigated the existence of correlations the rotating cup of the consistometer.
between the C3S/C2S ratio, as obtained by the MAS 29Si NMR
saturation recovery method, with a range of downhole simulation For the purpose of evaluating a manageable data set, three
performance parameters typically measured for cements. different cement formulations were prepared with each cement
sample. All cement formulations are based on a nominal 16.4
lb/gal slurry achieved by mixing cement with 39.4% water by
Experimental Section
weight of cement (bwoc). The first slurry considered was
In the present study seven different cements were studied. prepared without additives. The second formulation added a
Each sample was stored in a drybox under argon prior to NMR lignosulfonate retarder at a loading of 0.5% bwoc. The third
analysis. Three of the samples are API Class G cements from and final formulation added to the base cement a copolymer of
foreign countries: Brazil (BRZ), Russia (RUS), and Thailand N,N-dimethylformamide and 2-acrylamido-2-methylpropane-
(THA). The remaining four samples are API Class H cements sulfonic acid, a fluid loss additive used in oil well cement
from domestic cement manufacturers: Texas Industries (TXI), formulations, at a loading of 0.15% bwoc.
El Toro (ELT), Southdown (SDN), and Capitol (CAP). Class Thickening time, i.e., the time between initial mixing and an
G and H cements both use the same specifications for elemental apparent viscosity of 70 Bearden consistency units (Bc) at a
composition and restrict grinding aids to gypsum but differ with given temperature (<180 °F), is measured on an atmospheric
respect to fineness, with class G cements being the more finely consistometer according to API Schedule 5, which operates the
ground of the two. All cements were obtained from Halliburton run at 125 °F and a pressure ramped from 1000 to 5160 psi
Energy Services, Duncan, OK. All materials were used as over 28 min and maintained at 5160 psi until a consistency of
received. XRF mineral analysis for each cement sample is 100 Bc is achieved. Approximately 500 mL of slurry is stirred
summarized in Table 1. with a special paddle at 150 rpm in a thermostated canister.
Physical Measurements. Cement slurries are mixed by The test is designed to simulate pumping conditions to determine
independently weighing each component, dry blending the dry the time frame within which the slurry may be pumped before
additives into the cement powder, and mixing the water and setting. Consistometry data for cements are given in Table 2.
wet additives in a blender at low shear. The cement is then We note that it was found important to analyze samples of
added to the liquid under low to moderate shear within 10-15 cements that were freshly obtained and had not been exposed
s. The mixture is then blended under maximum shear conditions to humid air for any appreciable time.
for 35 s, with the operator scraping the sides of the blender to Compressive strength development is measured directly via
incorporate caked cement powder on the sides of the vessel. conventional destructive mechanical testing or indirectly and
5458 Ind. Eng. Chem. Res., Vol. 47, No. 15, 2008

Table 3. Ultrasonic Data for Cements without Additives of 6.00 kHz without high-power 1H decoupling. Spectra were
time to time to strength strength strength acquired using a 90° 29Si pulse of 5.75 µs and a spectral width
fluid loss 50 psi 500 psi at 24 h at 48 h at 72 h of 14 750 Hz unless otherwise specified. Data were acquired
sample retardera additiveb (min) (min) (psi) (psi) (psi) either with Bruker Biospin’s XWINNMR v2.6 or TopSpin v1.3.
BRZ no no 60 122 4897 5300 5428 The NMR data that served as the basis for correlations came
yes no 115 152 3774 4360 5280 from the saturation recovery experiments.7 T1 relaxation times
no yes 78 113 4308 4600 4801 measured via the saturation recovery method were collected
CAP no no 115 228 2330 3039 3450
yes no 177 281 2385 3138 3581 initially over a telescoping range of 16 delay periods τ primarily
no yes 180 247 2400 3160 3622 varying over a microsecond and millisecond time regime from
ELT no no 128 229 2877 3410 3775 1 µs (approximating a value of 0 s to accommodate pulse
yes no 220 317 2901 3486 3825 program timing considerations) terminating at 3.0 s for 128
no yes 181 247 3013 3577 3550
dummy scans and 40 000 acquisition scans at each value of τ,
RUS no no 102 156 3764 4314 4610
yes no 169 228 3839 4297 4544 with a 400-pulse saturation train with a 2 µs interstitial period
no yes 125 167 3873 4234 4400 and no relaxation delay. A second aliquot of sample was
SDN no no 122 205 3335 3920 4225 analyzed under time regimes of 1-70 s for 5000 scans and from
yes no 183 273 3469 4152 4385 70 µs to 10 ms for 200 000 scans. Otherwise, these additional
no yes 158 211 3646 4191 4294
THA no no 99 146 3557 4331 4725
saturation recovery experiments were performed under condi-
yes no 152 206 3780 4516 4970 tions identical to the first run. Data were processed using
no yes 129 164 4196 5046 5200 TopSpin version 1.3. Fourier transforms were applied without
TXI no no 155 245 2258 3210 3420 line broadening, phase corrections were made manually, and a
yes no 246 351 2324 3061 3259 fifth-order polynomial was fit to the baselines across the F2
no yes 180 249 2676 3092 3328
dimension. The data were then split into subspectra correspond-
a
Lignosulfonate retarder at a loading of 0.5% bwoc. b Copolymer of ing to individual values of τ for deconvolution analysis.
N,N-dimethylformamide and 2-acrylamido-2-methylpropanesulfonic acid.
Deconvolutions were carried out using a SIMPLEX routine
Table 4. Destructive Mechanical Testing Data for Cements without integrated into the solids line-shape analysis tool in TopSpin
Additives and with Retarder v1.3 with eight Gaussian peaks and at least two optimizations
72 h crush Young’s of 1000 iterations each. Line widths and intensities of all peaks
fluid loss strength modulus were free to be altered by the optimization independently, but
sample retarder additive (psi) (psi) no Lorentzian character was permitted. The SIMPLEX routine
BRZ no no 5690 799204 otherwise used default parameters. After the first line shape was
yes no 3980 526104 attempted by visual inspection and optimized, the parameters
no yes 5360 922284 were copied into subsequent data sets to be used as a starting
CAP no no 4872 843047
point for optimization after adjustments for peak intensities by
yes no 4028 738779
no yes 5623 1049038 visual inspection. Seven of the eight peaks were attributed to
ELT no no 4909 714051 C3S, while the remaining peak at -71.4 ppm was attributed to
yes no 5160 722237 C2S solely. Skibsted et al. used 13 peaks for their analysis of
no yes a a similar line shapes, with relative intensities of the 12 C3S
RUS no no 5730 876969
yes no 5740 932170
Gaussian and/or Lorentzian peaks fixed, but the work presented
no yes a a herein found seven peaks sufficient to fit the C3S portion of the
SDN no no 5850 1266380 line shape in the samples studied.10
yes no 5260 726191 Areas of fitted peaks are calculated by multiplying the half-
no yes 6540 981463
THA no no 5680 728749 widths (in hertz) of the fitted peaks by their respective heights
yes no 5170 811219 (in self-consistent arbitrary units). The C3S content is calculated
no yes 7860 1095458 from the sum of the seven peak areas attributed to C3S. In order
TXI no no 4080 649242 to compare data from runs with differing numbers of scans, the
yes no 4050 688539 areas from subspectra with matching values of τ are divided
no yes 5030 781451
and the average ratio is used as a scaling factor, which is then
a
Data unavailable. applied to the data set of more extreme τ values.
nondestructively in real time with the ultrasonic cement analyzer Table 5 tabulates an example of the processed peak areas
(UCA).9 Ultrasonic cement analysis for measuring compressive attributed to the C3S and C2S content in a sample as well as the
strength was obtained at 140 °F and ramped pressure from derived C3S/C2S ratio. Table 6 lists the characteristic times for
ambient to 3000 psi over 30 min and maintained at 3000 psi each sample as identified in the derivative analysis, and the
for the duration of the experiment (72 h total). The resulting calculated C3S/C2S ratios are shown in Table 1.
cement cylinder was crushed in a mechanical tester after 3 days Error Analysis. We propose that the errors associated with
to determine compressive strength and Young’s modulus. our NMR method are limited to measurement errors and
Ultrasonic data and destructive mechanical testing data are given variability of the sample as taken for NMR spectroscopy. The
in Tables 3 and 4, respectively. saturation recovery data were measured on at least two samples
NMR Measurements. Solid state MAS 29Si NMR spectra taken at random from the cement batch. The two samples were
were collected on a Bruker Avance 200 spectrometer at 39.76 chosen on different days (often 60-120 days apart). The overlap
MHz. Samples were measured using a 7 mm extended VT MAS of the low and central data and high and central data7 suggests
probe with 7 mm long barrel ZrO2 rotors and plugs and Kel-F that variability between samples is low. From the processed peak
fluoropolymer caps. Chemical shifts were referenced to hex- areas (e.g., Table 5) a variation due to sample choice of ca.
amethylcyclotrisiloxane (δ ) -9.66 ppm) by sample replace- 2-3% can be calculated. An estimate of the error inherent in
ment. All spectra were collected at a magic angle spinning speed the NMR measurement can be determined by a consideration
Ind. Eng. Chem. Res., Vol. 47, No. 15, 2008 5459
Table 5. Processed Deconvolution Data from BRZ Saturation
Recovery Experiments
τa aliquot C3S areab C2S areab C3S/C2S area ratio
300 µs 1 307233934.5 20085965.38 15.29595062
700 µs 1 827774190.5 51286225.81 16.14028284
700 µs 2 4421665493 979328685.2 4.514996405
1 ms 1 1269692417 103701393.8 12.2437353
1 ms 2 6986014080 568179270.9 12.2954399
3 ms 1 2981313319 237926952.9 12.53037238
3 ms 2 18766012084 2801917178 6.697561309
7 ms 1 5384913877 574563608.8 9.372180547
7 ms 2 37132204907 4930750757 7.530740598
10 ms 1 6806549942 653295701 10.41878882
10 ms 2 49529427385 5628982694 8.799001539
30 ms 1 12106608421 1195566591 10.12625186
70 ms 1 16252998231 1638427393 9.919877014
100 ms 1 18018586353 1824485898 9.875980065 Figure 2. Plot of “classical” C3S/C2S ratio (as determined by XRF) vs C3S/
300 ms 1 21805820315 2301266338 9.475574362 C2S ratio (as determined by MAS 29Si NMR) for the cements studied herein.
1s 1 24141665779 2662072342 9.068748959
1s 2 2317798481 286750669.8 8.082974952
3s 1 24774644566 2808135515 8.822453344 along the conventional hydration reaction scheme prior to the
3s 2 2334575361 316998473.4 7.364626512 or developed a poorly formed C-S-H layer, weakly percolating
7s 2 2354375187 306388876.6 7.684271089 the cement powder, resulting in an impediment to further
10 s 2 2343411843 312161240.5 7.507055774 hydration by conventional mechanisms. Additionally, the cement
30 s 2 2348358926 336692278.7 6.974792933
70 s 2 2363194752 306671128.1 7.705957736 setting is slowed without the additional heat provided by initial
a
rapid C3S, aluminate, and/or free lime hydration at the beginning
First aliquot with central τ values; second aliquot with extreme τ of a standard hydration run. Further hydration under standard
values. b Area calculated by multiplying peak intensity (arbitrary units)
by peak half-width (Hz) without rescaling to connect intermediate and conditions of a prematurely hydrated cement powder will take
extreme τ regimes. longer to reach the same strength by a given point in time
referenced from the beginning of “whole-hearted” intentional
Table 6. Characteristic Relaxation Timesa from Derivative Analysis hydration.
sample C3S τchar (ms) C2S τchar (ms)
BRZ 35.5 49.4 Results and Discussion
CAP 36.2 74.4
ELT 18.0 35.9
Figure 2 shows the plot of the calculated “classical” C3S/
RUS 29.0 94.7 C2S ratio based on oxide analysis from XRF versus the C3S/
SDN 15.6 38.2 C2S ratio determined by MAS 29Si NMR. Oxide analysis
THA 19.9 147.2 appears to provide a lower C3S/C2S ratio than determined by
TXI 33.7 139.4 NMR analysis. Furthermore, oxide analysis suggests that all the
a
τchar: x-intercept of second-derivative curve closest to global cements have a C3S/C2S ratio of 2-5, while our NMR method
maximum of first-derivative curve of the base intensity function. suggests the actual range is significantly broader. In looking at
relationships between the composition of the cement and
of the C3S/C2S ratio calculated for a particular sample at τ values physical properties, we used both C3S/C2S ratios.
above which the slope in the plot of intensity of the various Correlations for Cement. Initial studies were attempts to
saturation recovery parameters plotted versus τ is essentially 0. find a correlation between the C3S/C2S ratio and the physical
In most of the cases this involved τ above 1 s. Using the data measurements on cements in the absence of additives. Though
from the processed peak areas (e.g., Table 5), an error of 4% is cements are never used downhole without additives, it remains
calculated. Thus, the overall error in our NMR method for the instructive to consider the impact of varying degrees of
C3S/C2S ratio can be estimated. It is believed that this is far accessible mineral on unperturbed hydration reactions without
lower than the errors associated with the physical strength the added complication of the action of additives. Once a
development measurements. baseline understanding of the hydration correlations is estab-
It is difficult to determine the errors associated with the lished, one may evaluate the impact of additive response on
physical measurements; however, we can estimate some of the the correlations within the context of the simplified system.
errors based upon prior experience. Thickening times tend to A plot of thickening time measurements of slurries without
manifest a relative error of (10 min, while UCA data tend to additives versus the saturation recovery-derived silicate ratio
be less reproducible with a relative error of 10%. Reproducibility shows essentially no correlation. This suggests that there is no
of compressive strength measurements by direct destructive significant relationship between thickening time measurements
analysis often have an error of +5/-20%, and the Young’s and the saturation recovery-derived silicate ratio. In contrast,
modulus obtained from these destructive tests can manifest an as seen in Figure 3a, there is a good agreement between
error of approximately (15%. compressive strength development as a function of C3S/C2S ratio
Finally, the greatest failure of the traditional physical determined from NMR, i.e., the time to achieve a given
measurements is the susceptibility to premature hydration. Given compressive strength of slurries, and the saturation recovery-
that the initial C-S-H gel layer may form and age under derived silicate ratio. In agreement with conventional cement
irregular ion gradient conditions, partial hydration will have a chemistry wisdom, higher C3S content available to produce
significant effect on the strength development. If partial C-S-H on a short time scale results in faster strength
premature hydration occurs, then the cement will have either development. In contrast, there is no apparent relationship
“flash set”, owing to insufficient gypsum content to mitigate between compressive strength development and the C3S/C2S
the aluminate hydration, hydrated the aluminates and gypsum ratio determined from XRF (Figure 3b).
5460 Ind. Eng. Chem. Res., Vol. 47, No. 15, 2008

Figure 4. Plot of time for strength development (min) for Portland cement
Figure 3. Plot of time for strength development (min) for Portland cement as a function of (a) C3S and (b) C2S (R ) 0.603) peak intensity as determined
as a function of C3S/C2S ratio (a) as determined from MAS 29Si NMR for from MAS 29Si NMR for 50 psi.
50 psi (9, R ) 0.858) and 500 psi (0, R ) 0.673) and (b) analogous data
as determined from XRF.
saturation recovery-derived silicate ratio or a characteristic
relaxation time. No trend or dependence is observed for either
It is interesting to note that if the intensity of the individual parameter with the XRF-derived silicate ratio.
mineral peaks is plotted against the time for strength develop- Correlations for Cement and Retarder. Given that it is
ment (Figure 4), it is the C2S that appears to dominate the typical for cements to be used in concert with a retarder, there
correlation observed. This dominance is not completely coun- is interest in whether the correlations demonstrated for the
terintuitive, however, in the sense that higher C2S content means cements (above) are also maintained for experiments in the
potentially less C3S to readily percolate the cement slurry with presence of retarder. The retarder investigated is lignosulfonate.
C-S-H to solidify the material. Lignosulfonates are large, anionic polymers that have been
Figure 5a shows a plot of compressive strength development used as oil well cement retarders for many years. Although the
in terms of compressive strength attained by a given time of exact mechanism of lignosulfonates retarding the set of Portland
cements without additives versus the saturation recovery-derived cement is not well understood, it has been postulated that the
silicate ratio. This correlation is especially useful for field mechanism is a combination of adsorption and nucleation.11
engineers at the well site trying to schedule the time after cement Studies have shown sulfonate and hydroxyl groups adsorb onto
placement when the next phase of drilling may continue. It is the C-S-H gel layer of the hydrating cement. This fact has
interesting to note that as with strength development the led to the hypothesis that the sulfonate and hydroxyl groups
correlation with compressive strength is better for earlier times. present in lignosulfonates allow them to adsorb onto and,
Once again the XRF-based oxide analysis does not allow a consequently, incorporate into the C-S-H gel layer, causing
predictive correlation (Figure 5b) where NMR measurements a change in the morphology of the C-S-H gel and leading to
do. a more impermeable structure. This causes a type of waterproof-
A plot of the 72 h crush strength measured from destructive ing effect slowing further hydration. We have recently used 27Al
mechanical testing versus the C3S/C2S ratio shows a poor and 29Si solid state NMR spectroscopy to confirm this model.12
correlation (Figure 6), but a trend of increasing crush strength The results also show that while the effect of lignosulfonates
with increasing C3S/C2S ratio is apparent. However, the large on the C3A hydration should not be discounted, lignosulfonates
and asymmetric errors associated with measurements of crush predominantly affect the hydration kinetics of the C3S.
strength makes any predictive correlation difficult at this time. Again, as evidenced by the low correlation coefficients, there
No trend is observed for the XRF derived C3S/C2S ratio. is no observable correlation between thickening time measure-
In contrast to the data for strength development, the 72 h ments and the silicate ratio. This is unfortunate since a
crush strength appears to be controlled by the C3S rather than correlation of retarder response would have proven very valuable
C2S (see Figure 7). This is not surprising since C3S is the mineral to petroleum engineers wishing to better predict pumpability
that reacts on a time scale of minutes or hours, rather than days, windows.
months, and years, as is the case for C2S. The Young’s modulus Figure 8 shows a plot of compressive strength development
and thickening time for cements are independent of the in terms of time to achieve a given compressive strength of
Ind. Eng. Chem. Res., Vol. 47, No. 15, 2008 5461

Figure 7. Plot of time for 72 h crush strength (psi) for Portland cement as
a function of C3S peak intensity as determined from MAS 29Si NMR (R )
0.532).

Figure 5. Plot of compressive strength (psi) development for Portland


cement as a function of C3S/C2S ratio (a) as determined from MAS 29Si
NMR at 24 h (9, R ) 0.837), 48 h (0, R ) 0.734), and 72 h (b, R )
0.743) and (b) analogous data as determined from XRF.

Figure 8. Plot of time for strength development (min) as a function of


C3S/C2S ratio as determined from MAS 29Si NMR for 50 psi (9, R ) 0.697)
and 500 psi (0, R ) 0.743) for Portland cement in the presence of 0.5%
Figure 6. Plot of time for 72 h crush strength (psi) for Portland cement as lignosulfonate retarder bwoc.
a function of C3S/C2S ratio as determined from MAS 29Si NMR (R ) 0.445).

to retarder response is weaker than it is for the properties of


slurries with retarder versus the saturation recovery-derived the cement alone. It should also be noted that the data for BRZ
silicate ratio. As expected, the correlation indicates that shorter is lower than expected given the remaining data. Given the high
times to achieve target strengths are required when more of the C3S content of BRZ, the data seem to indicate that while
C-S-H-producing C3S is present. As observed for cement lignosulfonate is proposed to be an indiscriminant retarder, it
without additives there is no trend for XRF-derived silicate actually has a greater affinity for C3S than C2S. However, it
ratios. could also be that some aspect of the BRZ sample is unique
Figure 9a shows a plot of compressive strength development and is acting to enhance its strength development by means
for cements in the presence of 0.5% lignosulfonate retarder bwoc outside the scope of this analysis. Despite the lower correlation
in terms of compressive strength attained by a given time versus in comparison to cement in the absence of retarder, the potential
the saturation recovery-derived C3S/C2S ratio. It should be noted predictive nature of the NMR data is still far better than that of
that while a general trend remains, the correlation is significantly the XRF-derived silicate ratio (Figure 9b).
worse than the data without the retarder present. This is Whereas a plot of the 72 h crush strength measured from
unfortunate since it indicates that predictive power with respect destructive mechanical testing versus the C3S/C2S ratio showed
5462 Ind. Eng. Chem. Res., Vol. 47, No. 15, 2008

Figure 9. Plot of compressive strength development (psi) for Portland Figure 11. Plot of time for strength development (min) for Portland cement
cement in the presence of 0.5% lignosulfonate retarder bwoc as a function in the presence of 0.15% fluid loss additive bwoc as a function of C3S/C2S
of C3S/C2S ratio as determined from (a) MAS 29Si NMR at 24 h (9, R ) ratio as determined from MAS 29Si NMR for 50 psi (9, R ) 0.740) and
0.501), 48 h (0, R ) 0.436), and 72 h (b, R ) 0.667) and (b) analogous 500 psi (0, R ) 0.696) and (b) analogous data as determined from XRF.
data as determined from XRF.

Figure 10. Plot of time for 72 h crush strength (psi) for Portland cement in
Figure 12. Plot of time for compressive strength (psi) for Portland cement
the presence of 0.5% lignosulfonate retarder bwoc as a function of C3S/
in the presence of 0.15% fluid loss additive bwoc as a function of C3S/C2S
C2S ratio as determined from MAS 29Si NMR.
ratio as determined from MAS 29Si NMR for 24 h (9, R ) 0.505), 48 h
a trend without additives (see Figure 6), the presence of a (0, R ) 0.381), and 72 h (b, R ) 0.378).
lignosulfonate retarder appears to show deviation from a trend.
As shown by the data in Figure 10, the value for BRZ is thickening time derived from consistometry data appears
significantly lower than expected considering the remainder of independent of the silicate ratio derived from NMR or the
the data. Additional 72 h crush strength measurements would composition derived from oxide analysis.
have to be performed in order to ascertain whether any Correlations for Cement and Fluid Loss Additive. Fluid
significant trend was viable. It is likely that the BRZ sample loss additives prevent the loss of water from the cement slurry
crushed had an uneven surface making contact with the testing to porous rock formations downhole. These additives often have
bracket, which would have applied higher stress for a given the added effect of either retarding or viscosifying the cement
amount of force, resulting in an anomalously low value. slurry. The additive in question for this portion of the study is
A plot of Young’s modulus for cements set with retarder a copolymer of N,N-dimethylformamide and 2-acrylamido-2-
versus either the saturation recovery-derived silicate ratio or a methylpropanesulfonic acid, a viscosifying and nonretarding
characteristic relaxation time shows no correlation or trend. The variety of fluid loss additive. Polymers of this nature are
Ind. Eng. Chem. Res., Vol. 47, No. 15, 2008 5463
polyelectrolytes that act as flocculants that tie up free water in ing our investigations into the possible predictive protocols for
a slurry until the water can be consumed by the cement hydration cement performance with regard to downhole applications.
reactions. The ramifications of the addition of these polymers
to cements are not fully understood. Acknowledgment
There appears to be no distinctive trend for the 72 h crush Financial support for this work is provided by the Halliburton
strength measured from destructive mechanical testing versus Energy Services. The Bruker Avance 200 NMR spectrometer
either the NMR- or XRF-derived C3S/C2S ratio in the presence was purchased with funds from ONR Grant N00014-96-1-1146.
of a fluid loss additive. A plot of Young’s modulus for cements
set without additives versus the saturation recovery-derived Supporting Information Available: Processed deconvolu-
silicate ratio shows no correlation or trend. Variations in contact tion data from saturation recovery experiments for CAP, ELT,
surfaces for the samples could have accounted for the lack of RUS, SDN, THA, and TXI Portland cements; plots of 72 h crush
a correlation. strength (psi), strength development (min.), and compressive
As evidenced by the low correlation coefficients, there is no strength (psi) vs XRF-derived silicate ratio; plots of thickening
observable correlation between thickening time measurements time to reach an apparent viscosity of 50 Bc and Young’s
and the silicate ratio. In this particular case, however, a trend is modulus vs NMR-derived silicate ratio. This material is available
not necessarily expected, as it is impossible for the slurry to free of charge via the Internet at http://pubs.acs.org.
lose water in the HPHT consistometer, and this particular
additive is nonretarding. Like the previous series, a plot of Literature Cited
compressive strength development in terms of time to achieve
(1) Barnes, P.; Turrilas, X.; Jupe, A. C.; Colston, S. L.; O’Connor, D.;
a given compressive strength of slurries with a fluid loss additive Cernik, R. J.; Livesey, P.; Hall, C.; Bates, D.; Denis, R. Applied
versus the saturation recovery-derived silicate ratio shows a trend crystallography solutions to problems in industrial solid-state chemistry.
(Figure 11a), whereas the analogous plot using XRF-derived Case examples with ceramics, cements and zeolites. J. Chem. Soc., Faraday
silicate ratio shows no significant variation (Figure 11b). With Trans. 1996, 92, 2187–2196.
regard to the NMR-derived results, the same line of reasoning (2) Parrott, L.; Geiker, M.; Gutteridge, W. A.; Killoh, D. Monitoring
Portland cement hydration: Comparison of methods. Cem. Concr. Res. 1990,
for the cements without additives applies to the trend with 20, 919–926.
respect to variations in cements. (3) Vlachou, P.-V.; Piau, J.-M. Physicochemical study of the hydration
Figure 12 shows a plot of compressive strength development process of an oil well cement slurry before setting. Cem. Concr. Res. 1999,
for cements (in the presence of fluid loss additive) in terms of 29, 27–36.
(4) Ghosh, S. N.; Handoo, S. K. Infrared and Raman spectral studies in
compressive strength attained by a given time of cements with cement and concrete. Cem. Concr. Res. 1980, 10, 771–782.
fluid loss additive versus the saturation recovery-derived silicate (5) Fletcher, P.; Coveney, P. V.; Hughes, T. L.; Methven, C. M.
ratio. We note that, like the effect of retarder, the fluid loss Predicting quality and performance of oilfield cements with artificial neural
additive lowers the correlation observed. networks and FTIR spectroscopy. J. Pet. Technol. 1995, 47, 129–130.
(6) Coveney, P.; Fletcher, P.; Hughes, T. Using artificial neural networks
to predict the quality and performance of oil-field cements. AI Mag. 1996,
Conclusions 17, 41–49.
(7) Edwards, C. L.; Alemany, L. B.; Barron, A. R. Solid state 29Si NMR
Determination of C3S/C2S ratios by NMR provides an analysis of cements: comparing different methods of relaxation analysis
for determining spin-lattice relaxation times to enable determination of the
effective method of analysis for cements, owing to NMR’s direct C3S/C2S ratio. Ind. Eng. Chem. Res. 2007, 46, 5122–5130.
measurement of the minerals in question; determination of (8) ASTM C 856, “Standard Practice for Petrographic Examination of
characteristic relaxation times also proves useful in some Hardened Concrete”, Annual Book of ASTM Standards 4.02, Concrete and
situations. Unlike classic silicate analysis from XRF, the NMR- Aggregates.
(9) Keating, J.; Hannant, D.; Hibbert, A. Correlation between cube
derived silicate ratios demonstrate predictive ability for the
strength, ultrasonic pulse velocity and volume change for oil well cement
determination of engineering performance properties. This is slurries. Cem. Concr. Res. 1989, 19, 715–726.
especially the case for prediction of strength development; in (10) Skibsted, J.; Jakobsen, H.; Hall, C. Quantification of calcium silicate
keeping with generally accepted understanding of cement phases in Portland cements by 29Si MAS NMR spectroscopy. J. Chem.
hydration behavior, the strength development correlates with Soc., Faraday Trans. 1995, 91, 4423–4430.
(11) Chilingarian, G. V.; Nelson, E. B. DeVelopments in Petroleum
increasing C3S content, decreasing C2S content, C3S/C2S ratio, Science: Well Cementing; Elsevier Science Publishers B.V.: Amsterdam,
and/or the characteristic C3S relaxation time. In conclusion, we 1990; Vol. 28.
believe that MAS 29Si NMR offers superior analysis of silicate (12) Bishop, M.; Barron, A. R. Cement hydration inhibition with sucrose,
composition for Portland cements of the classic XRF-based tartaric acid, and lignosulfonate: analytical and spectroscopic study. Ind.
Eng. Chem. Res. 2006, 45, 7042–7049.
analysis.
The lack of dependency of either thickening time or Young’s ReceiVed for reView January 18, 2008
modulus to the C3S/C2S ratio as determined by MAS 29Si NMR ReVised manuscript receiVed May 16, 2008
Accepted May 19, 2008
measurements suggests that these physical properties are
independent of the relative silicate composition. We are continu- IE8000925

You might also like