You are on page 1of 7

Journal of Petroleum Science and Engineering 135 (2015) 261–267

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering


journal homepage: www.elsevier.com/locate/petrol

A new approaching method to estimate fracture gradient by correcting


Matthew–Kelly and Eaton's stress ratio
Bonar Tua Halomoan Marbun, Anisa Noor Corina n, Gde Valdy Arimbawa,
Ramadhana Aristya, Sigit Purwito, Aldiano Falah Hardama
Study Program of Petroleum Engineering, Faculty of Mining and Petroleum Engineering, Institut Teknologi Bandung, Jalan Ganesha 10, Bandung 40132,
Indonesia

art ic l e i nf o a b s t r a c t

Article history: Fracture pressure is one of the most important parameters in wellbore stability. Fracture pressure shows
Received 23 April 2014 the ability of rock to hold vertical stress before starts to fracture. The application of fracture gradient (FG)
Received in revised form affects the well design, such as mud weight profile, casing setting depth, and cementing operation.
24 July 2015
Fracture gradient could be determined directly by using leak-off test (LOT) and formation integrity test
Accepted 7 September 2015
Available online 14 September 2015
(FIT), and calculation from logging data. There have been a lot of studies in predicting fracture gradient
after the fundamental theory of fracture pressure was developed by Hubbert and Willis in 1957. However,
Keywords: most of those studies were performed without considering the geological characteristic, such as pre-
Fracture pressure cipitation environment, geology structure, and stratigraphy. A new methodology to predict fracture
Geomechanics
pressure from former calculations, Matthew–Kelly and Eaton, is proposed. The methodology character-
Poisson's ratio
ized the formation lithology from Poisson's ratio and stress ratio value which is corrected and analyzed
Stress ratio
Wellbore stability by two correcting constants, a and b. As the result, a new value of Poisson's and stress ratio of the
formation was generated and the accuracy of fracture gradient was improved. In addition, the flexibility
of this methodology indicates that this methodology could be applied in various drilling area.
& 2015 Elsevier B.V. All rights reserved.

1. Introduction 1.1. The tectonic stress field

Fracture gradient is a crucial parameter for developing a suc- Prior to drilling, rocks are in near-balanced state and the me-
cessful drilling program, especially in wellbore stability. An initial chanical stresses in the formation are less than the strength of rocks.
objective of this study is to develop a modification of previously Near-balanced state provides a naturally occurs stress in place which
fracture gradient prediction, Matthew–Kelly and Eaton, especially called the in-situ stress. There are three principal in-situ stress
by correcting the stress ratio constants. magnitudes, the vertical stress, Sv, equal to total weight of rocks and
This study shows improper fracture gradient estimation in fluid above that depth; the maximum principal horizontal stress, SH;
three exploration wells drilled in field X and drilling problem take and the minimum principal horizontal stress, Sh (Amoco, 1996).
place. It is necessary to address this concern by further in- The magnitudes of each stress influence the fault regime in
earth crust. Anderson (1951) classifies an area as being char-
vestigating the effect of vary lithology type in selected study area.
acterized by normal, strike-slip or reverse faulting depend on
However, several obstacles are encountered during this study,
whether (i) the crust is extending and steeply dipping normal
especially in experimental data availability including logging data,
faults accommodate movement of the hanging wall (the block of
leak-off test (LOT), or formation integrity test (FIT). Solving this
rock above the fault), (ii) block or crust are sliding horizontally
problem, this study proposed a new study methodology due to
past one another along nearly vertical strike-slip faults, or (iii) the
obtain the preferable results approaching the actual fracture crust is in compression and relatively shallow-dipping reverse
gradient. faults are associated with the hanging wall block moving upward
with respect to the footwall block (Zooback, 2007).
n
Corresponding author.
E-mail addresses: bonar.marbun@tm.ac.id (B.T.H. Marbun), 1.2. Fracture gradient theory
anisa.corina@gmail.com (A.N. Corina),
valdy.arimbawa@gmail.com (G.V. Arimbawa),
ramadhana.aristya@yahoo.com (R. Aristya), sgt.purwito@gmail.com (S. Purwito), 1.2.1. Fracture gradient estimation methods
aldiano.f.h@gmail.com (A.F. Hardama). Methods to determine formation fracture gradient consist of

http://dx.doi.org/10.1016/j.petrol.2015.09.006
0920-4105/& 2015 Elsevier B.V. All rights reserved.
262 B.T.H. Marbun et al. / Journal of Petroleum Science and Engineering 135 (2015) 261–267

Nomenclature Sv Vertical stress, psi


FG Fracture gradient, ppg
a, b Matthew Kelly and Eaton's fracture pressure correc- PR Poisson's Ratio
tion constants OBG Overburden gradient, ppg
D Depth, ft ECD Equivalent Circulating Density
Ki Matthew–Kelly stress ratio FIT Formation Integrity Test
pf Pore pressure, psi LOT Leak Off Test
pff Fracture pressure, psi LOP Leak Off Pressure
pn Normal pressure, psi
Sh Minimum horizontal stress, psi Unit conversion
SH Maximum horizontal stress, psi
ε1 Longitudinal contraction, ft 1 ft 0.3048 m
ε2 Lateral expansion, ft 1 psi 6994.75729 Pa
ν Poisson's ratio 1 g/cm3 8.33 ppg ¼0.0001 kg/m3
smin Effective minimum stress, psi 1 μs 1  10  6 s
sz Effective vertical stress, psi

predictive methods and verification methods. At the initial well 1.2.1.3. Eaton. Eaton (1969) proposed a calculation for fracture
planning, formation fracture gradient is calculated following the gradient by using another independent variable, Poisson's ratio
predictive methods. Afterwards, during drilling operation, fracture that shows a relationship between horizontal and vertical matrix
pressure is verified by pressure test after casing is cemented. The stress (Baker Huges INTEQ, 1996).
verified result is useful for well planning in the next drilling
ν = − ε2/ε1 (4)
operation.

1.2.1.1. Hubbert–Willis. Hubbert and Willis (1957) introduced fun-


damental principle of fracture gradient. In their statement, frac- pff /D = (ν/(1 − ν )). ( σmin/D) + pf /D (5)
ture gradient appears as the pressure needed to overcome the
where: v is Poisson's ratio; D is depth (ft); ε1 is longitudinal
minimum principle stress given by (Bourgoyne, 1991):
contraction (ft); ε2 is lateral expansion (ft); smin is effective mini-
pff /D = σmin/D + pf /D (1) mum stress (psi); pff is fracture pressure (psi); pf is formation
pressure (psi).
where: pff is fracture pressure (psi); D is depth (ft); smin is effective For any area outside Gulf Coast, Eaton's method can be applied
minimum stress (psi); and pf is formation pressure (psi). if the Poisson's ratio is available (Baker Huges INTEQ, 1996). The
Hubbert and Willis (1957) concluded that the minimum stress of Poisson's ratio can be obtain from overburden gradient data, actual
normal faulting regions, such as the U.S. Gulf Coast area, is equal to fracture pressure for several depth, and formation pressure data
horizontal matrix stress. Another assumption in this calculation is the (Eaton, 1969).
value of overburden stress gradient which equal to 1 psi/ft. Therefore,
the fracture pressure is approximately (Bourgoyne, 1991): 1.2.2. Fracture gradient verification
A pressure test called leak-off test is conducted by closing the
(
pff /D = 1 + 2pf /3D ) (2)
well at the surface and pumping drilling fluid into the closed well
where: D is depth (ft); pff is fracture pressure (psi); pf is formation at a constant rate until the pressure show a departure from the
pressure (psi). increasing pressure trend. Leak-off pressure (LOP) is shown by the
first point where the slope starts to decrease on the leak-off test
curve (Bourgoyne, 1991).
1.2.1.2. Matthew–Kelly. Matthews and Kelly (1967) published an-
Leak-off test (LOT) data is used as data verification for planning
other calculation of fracture pressure gradient. They stated that an
future field drilling and production operations because it measures
observed fracture pressure is exerted due to the force necessary to
the minimum horizontal stress (Sh). Sometimes the test is stopped
overcome the “matrix load”. As the result, Matthew and Kelly in-
until reach leak-off and the formation is only pressured up until a
troduced a new variable, “matrix stress coefficient”, which was
determined empirically from field data taken in normally pres- certain value. This test is called formation integrity test (Prassl, 1990).
sured formations, such as Gulf Coast sand reservoirs, as a function
of depth (Baker Huges INTEQ, 1996).
2. Methods
pff /D = ( Ki. σz )/D + pf /D (3)
This study proposes a new methodology to predict fracture
where: Ki is Matthew–Kelly stress ratio; sz is effective vertical
gradient by modifying the constant value in Eaton and Matthew–
stress (psi); D is depth (ft); pff is fracture pressure (psi); pf is for-
mation pressure (psi). Kelly equation with limited experimental data in field. The new
However, this method contains several weaknesses which may method to determine fracture gradient is shown in Fig. 1.
result in improper fracture gradient estimation. One of the weak-
nesses is that this method assume that the overburden stress is 2.1. Data collecting
equal to 1.0 psi/ft. Moreover, the matrix stress coefficient used in
their study is only limited in Gulf Coast area. Therefore, this The data was collected from mud logging and wireline logging,
method can be only used within a single field due to the variation including Gamma ray, Density, Porosity, Resistivity, and Sonic
of matrix coefficient (Baker Huges INTEQ, 1996). Logging, and also LOT and FIT as fracture pressure verification.
B.T.H. Marbun et al. / Journal of Petroleum Science and Engineering 135 (2015) 261–267 263

Data Collecting Data Filtering Data Classification Model Assumption

No Fit with LOT/FIT,


Correction by input
and Overburden Fracture Gradient
a- and b-constant
Pressure

Yes
Model Validation

Fit with LOT/FIT, Yes


and Overburden Result
Pressure

No

Fig. 1. Proposed study methodology on fracture gradient prediction. This is a new methodology that can be used to obtain the proper fracture gradient value.

2.2. Fracture gradient calculation (Well 1, 3, and 5) with total depth up to 4500 m MD. These fields
structure is a four-way dip horst-anticline closure structure with
At initial, the overburden pressure and pore pressure were trending NW-SE with the major E-W lateral strike-slip fault. These
calculated before modeling fracture gradient. Pore pressure was fields have the typical stratigraphy which contains 15 formations
calculated from mud logging data by using d-exponent based on formed in Late Permian until recent age which shown in Fig. 2.
modified Eaton's equation (Stephanie, 2013), and gamma ray, re- Field X is comprised by four dominant lithology; claystone,
sistivity, and sonic log data. Yet, the overburden pressure was limestone, sandstone, and shale. Shale is basically a sedimentary
calculated from neutron and density log. rock that composed of fine silt and clay. Regardless of its clay
Afterwards, fracture gradient was calculated by following Gulf component, shale cannot be considered as claystone. Claystone is
Coast (GC) constants because there is no available data to de- defined as sedimentary rock that is composed of clay only. Based
termine Poisson's ratio. on formation lithology and depth, each field X well is divided into
5 group which shown in Fig. 3. Claystone lithology is divided into
Ki − GC = − 0.00000000284D2 + 0.00011102903D 2 different groups, group 1 and 3 because there was a dis-
− 0.10621479633 (6) conformity found right after group 3 formed and it is possible to
have different type claystone in group 1 and 3.

υGC = 4.618272 × 10−14 D3– 2.03019285297 × 10−9


( ) ( ) 2.5. Fracture gradient correction
2 (7)
D + 0.0000327436946581803D + 0.281550687489
Fracture pressure value is influenced by lithology characteristic
where: Ki-GC is Matthew–Kelly stress ratio of Gulf Coast; v is because each lithology has different value of stress ratio and
Poisson's ratio of Gulf Coast; D is depth (ft). Poisson's ratio. Thus, a lithology grouping according to depth in-
crement was performed at initial stages. Fracture gradient of each
2.3. Study assumption group is calculated by using 2 models; Model 1 is calculated by
using Matthew–Kelly equation based on Gulf Coast model (Eq. (6)),
Model 2 is calculated by using Eaton's equation based on Gulf
1. The pore pressure and overburden pressure which calculated by Coast model (Eq. (7)). Afterwards, new constants of stress ratio and
modified Eaton's equation (1969) are valid. Poisson's ratio are calculated following fracture gradient and
2. There is no error in logging records. In other way, the tools had plotted as a function of depth. The flowchart to determine cor-
already calibrated well before the operation and the logging rected stress ratio and Poisson's ratio is shown in Fig. 4.
records had corrected to field condition during the operation. This study introduces a modified study of former equation by
3. LOT and FIT were executed based on general procedure and there introducing two constants, a and b, for correcting stress ratio and
was no error occurred during the test. Therefore, the LOT pressure Poisson's ratio value.
represents the real condition of minimum horizontal stress (Sh). For Matthew–Kelly's equation,
4. The minimum fracture gradient is equal with FIT pressure or
LOT pressure (based on data availability). And maximum frac- ( )
pff /D = (Ki a + b) Sv − pf /D + pf /D (8)
ture gradient is equal with overburden gradient.
For Eaton's equation,
5. The study was conditioned in a strike-slip fault regime in which
maximum horizontal stress (SH) has the greatest value than ( )
pff /D = (((ν/(1 − ν )) a) + b) Sv − pf /D + pf /D (9)
vertical stress (Sv) and minimum horizontal stress (Sh)
Those two constants have different influence in shifting frac-
2.4. Study area ture gradient curve; a-constant is effective on increasing or de-
creasing fracture curve gradient, meanwhile b-constant is effective
The study area focuses on field X which has 3 exploration wells on curve movement, either moving the fracture curve to the left or
264 B.T.H. Marbun et al. / Journal of Petroleum Science and Engineering 135 (2015) 261–267

Fig. 2. Field X Formation Marker. In this figure, a stratigraphic view of formation is described. The formation has 15 markers that described in Fig. 3.

EMW, ppg
Depth, ft

a=0.5 a=1
Fig. 3. Group Classification of field X. In this figure, markers from Fig. 2 are clas-
b=0 b=0.2
sified based on each lithology type.

Fig. 5. Curve movement by inputting various value of a-constant and b-constant. In


this figure, it is shown that there are curve movement because the addition of a-
and b-constant in the FG Eaton's equation. Each constant used in the equation may
cause different curve movement.
Lithology and Depth Grouping

Pressure Gradient, ppg

Fracture Gradient Calculation of Fracture Gradient Calculation of


Model 1: Matthew-Kelly Model 2: Eaton

Curve fitting based on minimum Curve fitting based on maximum


Depth (mMD)

constraint constraint

Averaging Data
(if only the depth contains FIT data)

Ki and Poisson’s ratio Correction

Fig. 4. A proposed fracture gradient prediction modeling. This figure describe the Fig. 6. Fracture gradient plot before correction. In this figure, it is shown that the
proposed methodology for predicting fracture gradient with calibration on LOT, FIT, FG by Matthew–Kelly and FG by Eaton do not give proper FG curve as the FIT has
and logging data. higher value than calculated FG by both methods.
B.T.H. Marbun et al. / Journal of Petroleum Science and Engineering 135 (2015) 261–267 265

right side. The example result is shown in Fig. 5. rather than FIT data as match point, because LOT data provide
The determination of fracture gradient was performed by in- valuable information to determine the minimum fracture pres-
putting a- and b- constants until the fracture curve reaches both sure. However, for any group which only contain FIT data, the
minimum and maximum constraint. By assuming that the field is a actual minimum value will lie along between FIT and overburden
strike-slip fault regime, the minimum constraint of fracture curve stress which resulting enormous minimum value possibilities and
are FIT or LOT value, and the maximum constraint is overburden lead to error. To minimize the error possibility given by FIT value,
the final value of fracture pressure is assumed equal with the
pressure. LOT data is preferred to be used as minimum constraint
averaging value from maximum and minimum curves. Afterwards,
Pressure Gradient, ppg
new constants of stress ratio and Poisson's ratio were calculated
following fracture gradient and plotted as a function of depth. A
verification of fracture gradient was performed in other wells
which have the similar geological characteristic to generate one
value for a- and b- constants.

2.6. Fracture gradient model validation


Depth (mMD)

To improve the accuracy, a validation is proposed in this study


by applying the corrected constants Ki and ν to other wells with
similar regime.

3. Study result

3.1. Before correction


Fig. 7. Minimum curve fitting to LOT of offset well value in Group 1.
The result of fracture gradient calculated by using the former
equation of Eaton and Matthew–Kelly (Eqs. (3)–(5)) is shown in
Pressure Gradient, ppg
Fig. 6. The result shows that fracture pressure curve of Matthew–
Kelly and Eaton do not meet LOT and FIT value.

3.2. After correction

An overview of curve fitting to minimum and maximum con-


straint in 2 groups of lithology, there are Group 1 and Group 3, is
Depth (mMD)

shown in Figs. 7–9.


In Fig. 7, minimum curve fitting constraint using LOT data as
match point rather than using FIT data, because LOT data more
valuable to give minimum fracture pressure. In Fig. 8, minimum
Matthew Kelly
curve fitting constraint using FIT data because in Group 3 there is
Eaton no available LOT data. The result of corrected fracture gradient in
one well was validated with other wells in similar field. After va-
lidation stage, the final result of corrected fracture gradient of
Fig. 8. Minimum curve fitting to FIT value in Group 3. 2 wells in the similar field is shown in Figs. 10 and 11.

Pressure Gradient, ppg


Depth (mMD)

Matthew-Kelly

Eaton

Fig. 9. Maximum curve fitting to overburden gradient value in Group 3.


266 B.T.H. Marbun et al. / Journal of Petroleum Science and Engineering 135 (2015) 261–267

Pressure Gradient, ppg


Depth (mTVD)

Fig. 13. Final corrected constants Poisson's ratio. This figure shows the corrected
Poisson's ratio which has already calibrated the FIT and LOT values.

Table 1
Value of C1, C2, and C3 constants for each model.

Fig. 10. Corrected fracture gradient in Well 3. Group Model's constant C1 C2 C3

1 Stress ratio (Ki)  5.68E  10 0.0000222 0.828757


Pressure Gradient, ppg Poisson's ratio  5.37E  10 9.25E  06 0.44462
2 Stress ratio (Ki)  2.84E  09 0.000111  0.0662148
Poisson's ratio  8.78E  10 2.33E 05 0.30724
3 Stress ratio (Ki)  2.84E  09 0.00011  0.0637148
Poisson's ratio  5.52E  10 1.71E  05 0.335109
4 Stress ratio (Ki)  2.84E  09 0.000111  0.22121
Poisson's ratio  3.80E  10 1.39E  05 0.305983
5 Stress ratio (Ki)  2.84E  09 0.000111  0.19121
Poisson's ratio  1.20E  10 6.91E  06 0.363987
Depth (mMD)

It is shown that claystone in Group 1 has greater stress ratio


than other group which caused by its plastic characteristic.
Meanwhile, lower stress ratio, such as sandstone in Group 4 and
claystone in Group 3 shows stiffer characteristic. Greater stress
ratio range which shown in Group 1 gives greater effective hor-
izontal stress. In addition, there are different stress and Poisson's
ratio of claystone in Group 1 and Group 3. However, both groups
Fig. 11. Corrected fracture gradient in Well 1. shown a similar stress and Poisson's ratio gradient which shown
that the internal properties of these groups is basically similar to
claystone properties. Regarding to the stress and Poisson's ration
difference, it may be caused by the difference of the depositional
time of Group 1 and Group 3. Therefore, Group 3 has stiffer
characteristic compared to Group 1. Gulf Coast model which
mostly comprised with sandstone lithology does not fulfill field X
characteristic and the proposed stress ratio tends to change within
the formation lithology.
The trend shows that the Poisson's ratio of sediment near the
surface has smaller value caused by younger sediments are more
elastic rather than lower sediments. In addition, the anisotropic
effect of shale is getting stronger with burial because the com-
paction make the minerals align at certain direction, thus causing
anisotropy. If the formation is in anisotropic condition (shale),
meaning that there is one direction that is stiffer than another so
Fig. 12. Final corrected constants stress ratio. This figure shows the corrected stress
the result of stress ratio is average of stress ratio from lateral di-
ratio which has already calibrated the FIT and LOT values.
rection and longitudinal direction. Both stress ratio and Poisson's
ratio function were generated as 3-polynomial function as follows:

3.3. Corrected stress ratio and Poisson's ratio Ki = C1D2 + C2 D + C3 (10)

The final result of stress ratio and Poisson's ratio as a function PR = C1 D2 + C2 D + C3 (11)
of depth is shown by Figs. 12 and 13. The result show big differ-
ences in the shallow depth and deeper depth of corrected Mat- The data is to be used to generate constant value (C1, C2, and
thew–Kelly stress ration and Poisson's ratio with Matthew–Kelly C3) is data from Figs. 12 and 13 of corrected Matthew–Kelly stress
stress ration and Poisson's ratio from Gulf Coast which is gener- ratio and Poison's ratio. The result of constant value (C1, C2, and C3)
ated using Eqs. (6) and (7). is shown in Table 1.
B.T.H. Marbun et al. / Journal of Petroleum Science and Engineering 135 (2015) 261–267 267

value, 5 data points are used which is shown in Table 2 for each
model. This analysis was performed by measuring and examining
the strength and direction of two variables. The correlation coef-
ficient value must be between  1 and þ1. If the correlation value
is approaching 0, those two variables are independent or there is
no association between those two variables. Positive correlation
value shows that as one variable value increases, another variable
value also increases, and vice versa (Wikipedia).
The result shows that correlation value between b-constant
value and C1, C2, and C3 in Matthew–Kelly's function obtain values
approach 71 which indicate that those constants tend to move
together. In this case, positive correlation is found in C1- and
Fig. 14. Plot between b-constant values with lithology group. C3-constant and negative correlation is found in C2-constant.
Meanwhile, the b-constant value in Eaton's model tend to move
Table 2 together with C3-constant rather with C1- and C2-constant
The final a- and b-constant for each model and lithology group. (Table 3).

Group Model Correction constants

a b 4. Conclusions
1 Matthew–Kelly 0.2 0.85
Eaton 0.4 0.65
An effective method to predict fracture gradient has been de-
2 Matthew–Kelly 1 0.04 veloped in this study. Specific conclusions are as follows:
Eaton 1 0
3 Matthew–Kelly 1 0.0425 a. This method is effective to be applied in wells without LOT
Eaton 1 0.005
data. Thus, empirical value of the constants could be de-
4 Matthew–Kelly 1  0.115
Eaton 1  0.115 termined from FIT data.
5 Matthew–Kelly 1  0.075 b. This method is effective to be applied in formation with var-
Eaton 1  0.07 ious lithology because of the constants could be adjusted fol-
lowing the lithology properties.
c. b-correction constant show a significant change due to geo-
Table 3
Correlation table between b-constant towards C1, C2, and C3 constants.
logical character variation, especially in field lithology. a-cor-
rection constant value not show significant change due to
Model Model's constant Correlation value of b-constant towards geological character variation.

C1 C2 C3

Matthew Kelly Matrix stress ratio 0.985  0.986 1.000 References


Eaton Poisson's ratio  0.195  0.319 0.899

Amoco, Drilling Handbook: Wellbore Stability, 1996.


Baker Huges INTEQ, 1996. Formation Pressure Evaluation Reference Guide. Hous-
3.4. Correlation between correction constants and corrected stress
ton, TX. Chapter 5, pp. 4–11.
ratio and Poisson's ratio Bourgoyne, A.T., et al., 1991. Applied Drilling Engineering. Society of Petroleum
Engineers, Richardson, TX, pp. 285–294.
Eaton, Ben A, 1969. Fracture Gradient Prediction and its Application in Oilfield
The behavior of b-correction constant result toward each li-
Operations, SPE-2163-PA.
thology group and the result is sketched in a plot shown in Fig. 14. Prassl, Wolfgang F., Drilling Engineering, 1990, Curtin University of Technology,
Fig. 14 shows that Group 1 have the highest b-constant value input Department of Petroleum Engineering, Master of Petroleum Engineering,
and Group 4 have the smallest b-constant input. However, the Drilling Engineering.
Stephanie, Y.F., 2013. A New Equation of Pore Pressure Prediction Strategy Based on
curves from Model 1 and Model 2 have the same trends. Eaton Model (Bachelor thesis). Petroleum Engineering Department, Institut
Based on result in Table 2, the a-constant value is slightly af- Teknologi Bandung, Indonesia.
fected by lithology rather than b-constant value. Another analysis Wikipedia. Pearson Product-Moment Correlation Coefficient. 〈http://en.wikipedia.
org/wiki/Pearson_product-moment_correlation_coefficient 〉 (accessed
was performed by analyzing the correlation between b-constant
01.05.13.).
values and stress ratio and Poisson's ratio constants function (C1, Zooback, M., 2007. Reservoir Geomechanic. Cambridge University Press, New York,
C2, and C3). To deliver constant value (C1, C2, C3) from b-constant pp. 8–146.

You might also like