SPE-199766-STU

Orientation of Hydraulic Fracture Initiation from Perforated Horizontal

Wellbores

Andreas Michael, Louisiana State University

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Calgary, Alberta, Canada, 30 Sep - 2 October 2019.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
The orientation of hydraulic fractures controls the productivity from hydrocarbon reservoirs. Productivity
from low permeability formations is greatly improved having multiple fractures oriented transversely rather
than longitudinally, relative to a horizontal wellbore. Analytical approximations from the literature for the
longitudinal and transverse fracturing stresses are modified to incorporate pore pressure effects and then
used to develop a criterion for the orientation of fractures initiating from perforated wells. The validity
of this criterion is assessed numerically and is found to overestimate transverse fracture initiation, which
occurs under a narrow range of conditions; when the formation tensile strength is below a critical value and
the breakdown pressure within a "window."
In horizontal wells, it is easier to achieve longitudinal fracture initiation, as transverse fracture initiation
only occurs over a narrow wellbore pressure-at-breakdown window, while longitudinal fracture initiation
occurs at comparatively higher wellbore pressures. The numerical study shows that in contradiction with
existing analytical approximations, the tangential stress which induces transverse fracture initiation, is a
stronger function of wellbore pressure just as the stress inducing longitudinal fracture initiation is. This
reduces the breakdown pressure window for transverse fracture initiation compared to what the derived
analytical approximations predict. Furthermore, this creates an additional constraint for transverse fracture
initiation; the critical tensile strength value, which determines the maximum tensile strength for which
transverse fracture initiation is possible for a given stress state.
The range of the in-situ stress states where transverse fracture initiation is promoted can be visualized
in dimensionless plots for perforated wells. This is useful for completion engineers; when targeting
low permeability formations, wells must be made to induce multiple transverse fractures. A numerical
simulation scheme performed on several stress states demonstrates frequent occurrence of longitudinal
fracture initiation, implying that the propagating fracture re-orients in the near-wellbore region to become
aligned perpendicular to the least compressive in-situ principal stress. This is the cause of near-wellbore
tortuosity, which in turn is a cause completions and production-related problems, such as early screenouts
and post-stimulation well underperformance.
2 SPE-199766-STU

Introduction
The industry has spent decades developing a reservoir stimulation technique based on the idea of creating
a network of cracks that enable hydrocarbons to flow out of the rock. This technique became known
as hydraulic fracturing; a process used today by virtually all wells for the production of commercial
hydrocarbon quantities from unconventional reservoirs. The industrial impact of this technology has been
massive and is responsible for the advances in the development of tight unconventional resources such as
shale gas, which would otherwise be non-producible (Michael, 2014; 2016a).
Fracture initiation often follows a plane different to the final fracture propagation plane (Valko and
Economides, 1995). Stress re-orientation in the near-wellbore region (Kirsch, 1898) may promote fracture
initiation of different orientation than the one dictated by the far-field stresses. The evaluation is performed
by considering the tangential stresses on two points (extremes) around the base of the perforation; one for
the initiation of transverse fractures and another for the initiation of longitudinal fractures, an idea first
presented by Hossain et al. (2000). In-situ stress state, wellbore pressure, perforation geometry and the
formation's mechanical properties are independent variables that are shown to determine the orientation of
the initiated hydraulic fractures; the dependent variable.
The hypothesis tested in this study is that the orientation of fracture initiation can be predicted a-priori,
if the stress state, rock formation properties and perforation geometry are known.
Hydraulic fracture orientation impacts the well productivity from hydrocarbon reservoirs. Productivity
from low permeability formations is significantly improved having multiple fractures oriented transversely
rather than longitudinally relative to a horizontal wellbore. This is useful for completion design; when
targeting low permeability formations, horizontal wells should be made to induce multiple transverse, with-
respect-to-the-wellbore, fractures, as opposed to longitudinal fractures which are more effective in higher
permeability zones and Frac-&-Pack operations (Economides and Martin, 2010).

Orientation of Fracture Initiation: Longitudinal vs. Transverse

For the limiting case of horizontal wells, initiated fractures can assume arbitrary orientations with the two
main extreme configurations with respect to the wellbore being longitudinal and transverse (Figure 1).
Longitudinal, or "axial" fractures, propagate in a plane parallel to the axis of the wellbore and transverse,
or "lateral" fractures, propagate in planes orthogonal to the wellbore axis. This study focuses on fracture
orientation in the early initiation stages, within a small region near the wellbore. This is more challenging to
predict than the fracture orientation in the later propagation stages outside this near-wellbore region, which
is dominated by the well trajectory compared to the in-situ stress state.

Figure 1—Fracture orientation configurations, (a) longitudinal and (b) transverse-to-

the-wellbore. The variable r represents a distance from the center of the wellbore and it
cannot be smaller than the wellbore radius, rw (modified from Michael and Gupta, 2019b)
SPE-199766-STU 3

Hubbert and Willis (1957) famously concluded that the final fracture propagation plane is perpendicular
to the least compressive stress, known as the preferred fracture plane (PFP) and solely depends on the well
trajectory compared to the in-situ principal stresses. This implies that drilling a well parallel to the least
compressive horizontal stress Shmin, which is a common industry practice, will yield transverse final fracture
propagation, regardless of the orientation of fracture initiation.
The motivation for this study came after a series of laboratory experiments (Figure 2) on highly-
concentrated gelatin samples failed to generate transverse fracture initiation as was expected, consistently
generating instead longitudinal fracture initiation (Michael, 2016b). The same observations were made from
true-triaxial experiments on hydrostone cement/gypsum plaster blocks (Alabbad, 2014).

Figure 2—Longitudinal fracture initiation generated from a perforated horizontal wellbore during scaled-
laboratory experiments in highly concentrated gelatin and hydrostone cement/gypsum plaster blocks

A number of experimental studies (Daneshy, 1973; Weijers, 1995; Wu et al., 2008) led to observation
of transverse fractures only through re-orientation of what was originally longitudinal fracture initiation.
Nonetheless, this is not a bar-none-fact about horizontal wells, since El Rabaa (1989) achieved transverse
fracture initiation in high stress block cement experiments, using motor oil as fracturing fluid and later
4 SPE-199766-STU

Feng and Sarmadivaleh (2019) with synthetic samples and honey. Therefore, while both longitudinal
and transverse fracture initiation have been observed in laboratory and field-scale studies (Barree and
Miskimins, 2015), the question we are asked to answer is, "what factors control the orientation of fracture
initiation and how"?

Analytical Approach
Stresses on the Wellbore Radius
The subsurface stress state can be characterized by three in-situ principal stresses, which are oriented
mutually-orthogonally. In most locations, those principal stresses consist of one in the vertical direction Sv
and two horizontal, SHmax and Shmin, with SHmax ≥ Shmin always true (Anderson, 1951).
When a well is drilled a pressurized cylindrical cavity is introduced to the subsurface medium, which
disturbs the stress field in the vicinity surrounding the wellbore. Thus, the three in-situ principal stresses
are no longer able to fully characterize the local stress state. Kirsch (1898) developed equations for the
magnitude of the stresses on a uniform plate with a circular hole under uniform tension, which were later
modified by Hubbert and Willis (1957) to include biaxial stresses and fluid pressure inside the hole. The
stress field in the near-well region is quantified by the radial stress σrr, the tangential (also known as "hoop"
or circumferential: normal stress in the tangential direction) stress σθθ, the out-of-plane stress parallel to
the wellbore σyy, and the shear stress σrθ, which is zero on the wellbore radius, rw. Figure 3 shows the
configuration of the three normal stresses (σyy, σθθ and σrr) at r = rw.

Figure 3—Spatial configuration of the near-wellbore stresses for a horizontal well drilled
parallel to Shmin. SHmax, Shmin and Sv are assumed to be along the x, y and z-axis, respectively
(modified from Michael and Gupta, 2019b). Shear stress is not shown (σrθ = 0 at r = rw)

In this study we assume a horizontal well drilled parallel to Shmin, the trajectory which promotes transverse
final fracture propagation (Hubbert and Willis, 1957) and hence yields the best chance of generating
transverse fracture initiation. The stress state at r = rw, can then be expressed in terms of the in-situ principal
SPE-199766-STU 5

stresses (Sv, SHmax and Shmin), wellbore pressure pw, formation pore pressure pp and formation Poisson's ratio
ν as,
(1)
(2)
(3)
(4)
where θ is the angle around the wellbore taken from Sv.

Approximations of the Fracturing Stresses from Perforated Wellbores

The presence of a perforation tunnel intersecting the cylindrical wellbore increases the geometrical
complexity of the problem, making derivation of a fully-exact analytical solution extremely challenging.
Nevertheless, following Hossain et al. (2000), application of three assumptions can reduce the problem to
a more simplified form, for which closed-form analytical solutions can be derived:
1. A micro-annulus existing between the cement and the rock formation, allowing the wellbore to be
modeled as an openhole.
2. The pressure in the perforation tunnel, pperf is equal to the wellbore pressure, pw.
3. Fractures initiate at the perforation base, which is cylindrical.
An expression (Eq. 5) for the tangential stress at the perforation base, can be derived by a second
"Kirsch analysis" using the solutions of the total axial and tangential stresses (Syy and Sθθ) at r = rw as
(5)
where θ represents the position which the perforation intersects with the wellbore radius, θp is the angle
around the perforation base taken from the wellbore axis, Shmin. Inserting the appropriate expressions of Syy
and Sθθ for a well parallel to Shmin (Eq. 1-2 without the subtracted pp term) gives

(6)

For a perforation on the side of the well, parallel to SHmax, θ = 90° and Eq. 6 takes the form

(7)

The value of at θp = 0° is the stress needed to overcome in tension to generate longitudinal fracture
initiation, hence it is called the longitudinal fracturing stress, σθL. Similarly, at θp = 90° is the stress
needed to overcome in tension to generate transverse fracture initiation, and is subsequently called the
transverse fracturing stress, σθT. The expressions of σθL and σθT can simplify to:
(8)

(9)
A similar designation can be made for a perforation at the top of the hole (at θ = 0°, parallel to Sv), by
interchanging the Sv and SHmax terms in Eq. 8-9.
6 SPE-199766-STU

Figure 4—Configuration of the longitudinal σθL and transverse σθT fracturing stresses on the
perforation base ( at θp = 0° and θp = 90°, respectively). The wellbore is aligned parallel
to the y-axis with perforations parallel to the x-axis (modified from Michael et al., 2018)

Make no mistake.
The Kirsch's (1898) theory is purely two-dimensional, assuming a state of plane strain. Hossain et al.'s
(2000) idea of applying this theory twice in an attempt to simulate a three-dimensional problem could not
yield analytically-exact solutions for σθL and σθT. Hence, the derived expressions (Eq. 8-9) are called the
"approximated" solutions of the two fracturing stresses. Similar approximations can be derived also for non-
linearly-elastic constitutive models (Michael and Gupta, 2019b; 2019c).

Conditions for Transverse Fracture Initiation

As hydraulic fractures are tensile failures, a longitudinal or a transverse fracture will be promoted if the
corresponding fracturing stress becomes more tensile than tensile strength of the rock formation, T. If this
is true for both σθL and σθT, then the orientation of fracture initiation is determined by which of the two
fracturing stresses is the most tensile (most negative according to the convention followed).
Using the aforementioned conditions, a set of closed-form analytical criteria (Eq. 10-11) can be derived
by setting σθT < −T and σθT < σθL, determining whether transverse fracture initiation will occur. If either of
those criteria is not satisfied, transverse fracture initiation cannot occur.
(10)
(11)
We examine the fracture initiation equations using numbers typical for the Barnett Shale. This is a shale
gas formation of Mississippian age located in the Fort Worth Basin of North Texas. The payzone is centered
at an average depth of 6,500 ft (Moos, 2012). Table 1 summarizes the stress parameters and mechanical
properties of the Barnett Shale, used as an application example throughout this study.

Table 1—Input parameters for the Barnett Shale in Texas

Parameter Value

Payzone depth 5,000-8,000 ft

Sv 7,215 psi
SHmax 4,550 psi
Shmin 3,900 psi
Pp 3,055 psi
E 2×105 psi
v 0.23-0.32
SPE-199766-STU 7

The two fracturing stresses σθL and σθT, are plotted against pw (equal to pperf) on Figure 5. For simplicity,
the quantities on both axes are normalized by Sv.

Figure 5—Variation of transverse and longitudinal fracturing stresses with wellbore pressure for Barnett
Shale's stress state. The perforation is assumed to be parallel to SHmax (modified from Michael et al., 2018)

Breakdown Pressure Window

As seen on Figure 5, σθL indicated by the red line, starts from a highly compressive value becoming more
tensile as pw increases (which is equal to pperf). Whereas, σθT indicated by the blue line starts from a small
tensile value and remains constant regardless independent of the pw value. This creates on the left of the σθL
and σθT lines intersection point, a region known as the "breakdown pressure window" for transverse fracture
initiation (Michael et al., 2018). The upper-bound of the pb window, pb/upper is given by setting Eq. 11 equal
to zero and solving for pw, such:
(12)
for perforations parallel to SHmax.
If the breakdown pressure, pb – the highest pressure reached during the hydraulic fracture treatment –
falls within this window range, transverse fracture initiation occurs. This is because in this pw range the two
conditions for transverse fracture initiation are satisfied. If pb is higher and falls outside this window, σθT >
σθL and thus longitudinal fracture initiation occur, because σθL is the most tensile fracturing stress.
This is an important condition as the practitioners to an extent can impact pb. Low pressurization rates
were shown to suppress pb (Zoback et al., 1977; Weijers, 1995) and can be used strategically to ensure pb
≤ pb/upper.
8 SPE-199766-STU

Allowable Stress State Diagrams

The range of stress states possible to occur in the Earth's subsurface can be visualized on a simple diagram
of SHmax/Sv against Shmin/Sv (Figure 6a, after Nelson et al., 2005 and Zoback, 2014). The allowable stress state
region is bound by the frictional limits of slip on a fault and the line of horizontal isotropy (SHmax = Shmin).
This critically-stressed fault theory constrains the ratio of the effective maximum to minimum stress, (S1 −
pp)/(S3 − pp) to be less than the critical minimum value required for slippage on a pre-existing fault as,

(13)

where μ is the friction coefficient of the fault, and S1 and S3 are the greatest and least compressive principal
stresses, respectively. According to Anderson (1951), S1 = Sv and S3 = Shmin for normal faulting, S1 = SHmax
and S3 = Sv for reverse faulting and S1 = SHmax with S3 = Shmin for strike-slip faulting. Typically the value of
friction factor for rocks is ~0.6 (Byerlee 1978), which puts the ratio (S1 − pp)/(S3 − pp) at 3.12; if the ratio
reaches this point fault slippage will occur. For a hydrostatic gradient for pp = 0.43 psi/ft, the allowable
stress state region becomes the region enclosed by the polygon shown in Figure 6a.

Figure 6—(a) Allowable region for in-situ stress states in the Earth's subsurface (after Michael et al., 2018).
The dotted lines represent horizontal isotropy (SHmax = Shmin) and the solid black lines represent the frictional
limits at μ = 0.6. The region bounded by the solid lines and the horizontal isotropy line represents the region
of possible subsurface stress states (NF: normal faulting, RF: reverse faulting and SSF: strike-slip faulting).
(b) Orientation criterion for fracture initiation for a horizontal well parallel to Shmin and perforated along SHmax
according to the modified Hossain et al. (2000) analytical approximations. The blue region indicates the stress
states where transverse fracture initiation occurs, while the red region indicates the stress states promoting
longitudinal fracture initiation. The magenta region represents those stress states where promoted fracture
initiation shifts from transverse to longitudinal after an increase in pw from 0.6 to 0.9 psi/ft (after Michael et al., 2018)

Following Nelson et al.'s (2005) use of the allowable region diagrams to demonstrate the stress states
promoting transverse drilling-induced tensile fracture (DITF) initiation and using Eq. 10-11, two lines are
added to the stress polygon; one for σθT = 0 and another for σθT = σθL assuming pp gradient 0.43 psi/ft (Figure
6b). The region behind the two lines (where σθT < 0 and σθT < σθL) represents the stress states where transverse
fracture initiation is promoted for a given pw gradient. Increasing pw from 0.6 psi/ft to 0.9 psi/ft, reduces
the range of stress states, which promote transverse fracture initiation. This is comparable with the effect
of pw within a specific stress state (Figure 5) where its increase causes a gradual shift from transverse to
longitudinal fracture initiation.
SPE-199766-STU 9

Fracturing Fluid Leakage Around the Wellbore

Evidence from laboratory testing (Alabbad, 2014; Michael, 2016b) and from the field (Barree and
Miskimins, 2015) showed that fracturing fluid can often leak all around the cross-section of the wellbore,
bypassing the perforation. This makes fracture initiation possible from all around the wellbore, not only
from the perforation tunnel.
The "upside" of such occurrence is that fluid leakage reduces the problem effectively to that of DITF
initiation; one with a fully-exact analytical solution (Nelson et al., 2005; Michael and Gupta, 2018; 2019a).
If the perforation is "ignored," the axial and tangential stresses (σyy and σθθ) will determine the orientation of
fracture initiation substituting σθT and σθL respectively, as the transverse and longitudinal-inducing fracturing
stresses. Figure 7 shows the overall effect of fluid leakage for parameters from the Barnett Shale; the pb
window shrinks significantly, suppressing the possibility of transverse fracture initiation.

Figure 7—The wellbore stresses: longitudinal fracture-inducing σθθ and transverse fracture-inducing σyy, at Sv (dashed-
dotted lines) where these stresses are the most tensile. These two wellbore stresses determine the orientation of fracture
initiation instead of σθL and σθT in the case of fracturing fluid leakage along the wellbore (modified from Michael et al., 2018)

The allowable stress state diagrams on Figures 8 shows the leakage effect on the stress state range
promoting either transverse or longitudinal fracture initiation for the two extreme θ values around the
wellbore; 0° and 90°. Due to the leakage, σyy and σθθ are used in the place of σθT and σθL respectively, yielding
visibly smaller blue regions of stress states promoting transverse fracture initiation than the corresponding
blue region for the no-leakage case shown in Figure 6b for the same pw gradient (0.6 psi/ft). This highlights
how good perforation practices aid transverse fracture initiation.
10 SPE-199766-STU

Figure 8—Allowable in-situ stress state diagram for fracture initiation for a well parallel to Shmin
experiencing leakage of fracturing fluid around the wellbore from, (a) the top of the hole (parallel
to Sv), and (b) the side of the hole (parallel to SHmax). In both cases, the blue region indicating
transverse fracture initiation is visibly smaller compared to the case of no leakage (Figure 6b)

Practical Problems Triggered by Longitudinal Fracture Initiation

The PFP of a horizontal well parallel to Shmin is transverse to the wellbore (Hubbert and Willis, 1957). Thus,
longitudinal fracture initiation, which is a non-PFP orientation, is expected to lead to fracture re-orientation
(Figure 9) at some point as the fracture propagates away from the wellbore, to align it transversely to the
well along PFP. This re-orientation is forced by the far-field, in-situ principal stresses (Sv, SHmax and Shmin),
which while moving away from the well, become more dominant with the near-wellbore stresses (Eq. 1-4)
diminishing.

Figure 9—Re-orientation of a fracture initiated longitudinally to the horizontal wellbore, to transverse, which is
the PFP (Hubbert and Willis, 1957) for a well drilled parallel to Shmin (modified from Valko and Economides, 1995)

This fracture re-orientation generates tortuosity in the near-wellbore region, which can cause serious
completion-related problems like early screenouts. The point of re-orientation restricts the flow of proppant
carried in the treatment fluid, something observed at the surface as a sudden rise in pump pressure. Another
SPE-199766-STU 11

problem caused by the near-well tortuosity is fracture fluid leakages across adjacent stages during multi-
stage fracture treatments. The results of a tracer test illustrated on Figure 10, show yellow fluid from the first
stage appearing in the second and third stages, despite all stages separated from each other using packers. A
similar observation is made for the red fluid of Stage 2 and the blue fluid of Stage 3. Barree and Miskimins
(2015) explain this by longitudinal fractures, which penetrate an as of yet untreated stage, providing possible
paths of low resistance for fluid entry.

Figure 10—Schematic showing the results of a tracer test during a multi-stage fracture treatment. Fracturing
fluid from each of the three stages was detected in adjacent stages, despite the stages being isolated from each
other by packers (modified from Barree and Miskimins, 2015). This is attributed to longitudinal fracture initiation
– prior to re-orientation to transverse – where the longitudinal fractures initiated extend beyond the stage length

The aforementioned completion-related issues suppress the efficiency of a hydraulic fracture treatment,
giving rise to production-related problems, such as poor, below expectations post-treatment well
performance. As a result, the overall net present value (NPV) of the fracture treatment is negatively impacted
(Balen et al., 1988; Michael, 2019).

Numerical Modeling
A numerical model was built to assess the validity of the orientation criterion developed using the analytical
approximations (Hossain et al., 2000) of the longitudinal and transverse fracturing stresses.
The model was built on FLAC3D (Fast-Lagrangian Analysis of Continua in 3 Dimensions) written by the
Itasca Consulting Group, which utilizes an explicit finite volume formulation that can simulate the complex
behavior of non-linear or unstable models exhibiting high degrees of strain. A rectangular block (Figure
11) is studied with a circular wellbore in the middle (a "1/8th model"). The block is loaded with the in situ
principal stresses (Sv, SHmax and Shmin). A perforation of radius significantly smaller than that of the wellbore
is placed perpendicular to the wellbore central axis at the side of the hole (θ = 90°).
12 SPE-199766-STU

Figure 11—The FLAC3D model mesh, with 174,496 grid points in total. A grid refinement
study was performed to optimize computational time without affecting the output accuracy

The block is modeled as a homogeneous, linear elastic continuum with constant Young's modulus, E and
ν. Constant stress or no displacement boundary conditions is used appropriately to set up the simulation
model. The pressure in the perforation, pperf is set equal to pw. As with the analytical approach, assuming
micro-annulus to exist between cement and the formation prior to fracture initiation, allows the model to
be simulated as an openhole completion intersected by a cylindrical perforation.

Perforation Geometry
A series of preliminary simulations was performed to investigate the impact of the length and radius of
the perforation on and to identify values for those parameters ideal for the simulation program of this
study. The length of the perforation, lperf was found to have only a minor impact on compared to the
perforation radius, rperf. Subsequently, these dimensions were fixed during this study at lperf = 16rw/9 and rperf
= rw/6 to ensure a fair assessment and accurate conclusions for geometry dimensions appropriate to those
used in the field.
A point far away from the perforation (> 10rperf) may be used as a verification point. At the verification
point, the perforation effects are negligible and the local stress field can be evaluated analytically for a
fully-exact solution. Shown in detail in Appendix A, the numerical model matches excellently the analytical
solution of the stress field at the verification point for (i) angle θ around the wellbore, and (ii) pw values
up to 10,000 psi.
SPE-199766-STU 13

Numerical (True-3D) Solution of the Fracturing Stresses

After verification, the FLAC3D numerical model was used to calculate the true three-dimensional, numerical
solutions of the longitudinal and fracturing stresses of a horizontal well parallel to Shmin with perforations
parallel to SHmax in the Barnett Shale. When the true-3D numerical solutions of σθL and σθT are compared to
the corresponding analytical approximations (Figure 12) it is clear that matching is extremely poor. Unlike
the prediction given by the modified Hossain et al. (2000) analytical approximation, σθT does not remain
constant as pw increases. FLAC3D outputs showed instead a positive relationship between the σθT and pw;
the former becoming more compressive as the latter increases. This impacts significantly the pb window,
which shrinks to 0-4,574 psi, about 60 percent smaller than the prediction of the analytical approximation
of 0-11,490 psi.

Figure 12—Comparison between the FLAC3D numerical model and the modified Hossain et al. (2000)
approximations for the two fracturing stresses (σθL and σθT) with Barnett Shale parameters. The well is
perforated along SHmax. The shaded region represents the "pb window" for transverse fracture initiation
(0–4,754 psi), evaluated numerically. The intercept of the σθT line at pw = 0 represents the critical tensile
strength, Tcrit; a second constraint on transverse fracture initiation (modified from Michael et al., 2018)

Disagreement between the numerical and analytical solutions is expected. The inability of the analytical
approximation in matching the true-3D numerical solution is attributed to the fact that it neglects the
interaction between the wellbore radial stress, σrr and the perforation radial stress. The way those two radial
stresses mutually interact, affects the magnitude of the axial and tangential stresses (σzz and σθθ) on the
wellbore radius. This in turn alters the magnitude of the tangential stress on the perforation base, from
which the values of σθL and σθT are obtained.

Critical Tensile Strength (Tcrit)

Due to the positive slope of the transverse fracturing stress, σθT with pw, an additional constraint is imposed
on transverse fracture initiation: the value of σθT at pw = 0 represents a critical tensile strength, Tcrit magnitude.
14 SPE-199766-STU

If the formation tensile strength T is larger than Tcrit (more tensile), σθT will never be more tensile than T
and hence transverse fracture initiation cannot occur regardless of the pb value. Contrary to the pb window
constraint, Tcrit depends solely on the rock properties and stress state and there is no way for the practitioner
to influence it.
Consequently, for transverse fracture initiation to be generated from a perforated well, three criteria must
be met in the following order:
a. T < Tcrit.
b. The value of pb falls within a "window" for transverse fracture initiation (i.e. σθT < −T).
c. The longitudinal fracturing stress is less tensile than the transverse fracturing stress at pw = pw (i.e.
σθT < σθL).
For the Barnett Shale example, the value of Tcrit given by the numerical model is 2,482 for perforations
along SHmax (Figure 12). Published data on the Barnett Shale by Rybacki et al. (2015) indicates a maximum
tensile strength of ~2,500 psi, so according to the modified Hossain et al. (2000) fracturing stresses
approximation for this case (Eq. 8-9), the fracture initiation observed in the Barnett Shale would most likely
be transverse if pb is kept under 4,574 psi.
Expanding the simulation over a range of stress states, we see that for a fixed pw value, the region of
stress states promoting transverse fracture initiation as predicted by the true-3D numerical model, is smaller
compared to what the analytically-approximated model predicts. Figure 13a indicates the stress states,
which promote transverse (versus longitudinal) fracture initiation according to the criterion developed using
the modified Hossain et al. (2000) analytical approximations for σθL and σθT, while Figure 13b shows the
same criterion developed using the true-3D numerical model constructed by plotting specific points. The
blue region of the stress states expected to generate transverse fracture initiation in Figure 13b is visibly
smaller and narrower than the corresponding blue region in Figure 13a, highlighting the extent to which the
analytical approximations for the fracturing stresses overestimate transverse fracture initiation.

Figure 13—Allowable stress state diagrams indicating the orientation of fracture initiation as predicted at pw = 0.6
psi/ft by, (a) the criterion developed using the modified Hossain et al.'s (2000) analytical approximations for the
longitudinal and transverse fracturing stresses (after Michael et al., 2018) and (b) the true-3D numerical model
SPE-199766-STU 15

Discussion
Validation with Laboratory Experiments
The orientation criterion developed using the true-3D numerical model (Figure 13b) is assessed considering
the results of laboratory-scale experiments, published in the literature. The model does an excellent job in
predicting the orientation of fracture initiation observed in various laboratory setups as shown on Figure
14. This enhances our confidence to the true-3D numerical model and highlights the rarity of transverse
fracture initiation occurrence, compared to longitudinal fracture initiation.

Figure 14—Orientation criterion developed by the true-3D numerical model (from Figure 13b) with stress state
points from four experimental studies; two yielding longitudinal (Wu et al., 2008; Michael, 2016b) and two yielding
transverse fracture initiation (El Rabaa, 1989; Feng and Sarmadivaleh, 2019) showing excellent agreement

El Rabaa (1989) obtained transverse fracture initiation in his true triaxial tests on cement blocks initiating
fractures using motor oil injection. A highly anisotropic stress state was used that would not be stable against
shear failure, thus not likely to occur in the Earth's subsurface (well outside the polygon). Nevertheless,
the stress state used by El Rabaa (1989) fulfills the criteria necessary for transverse fracture initiation if
the tensile strength of the cement block is neglected; σθT < 0 and σθT < σθL. Feng and Sarmadivaleh (2019)
performed true triaxial testing on synthetic samples of low tensile strength, using edible honey as fracturing
fluid achieving transverse fracture initiation also. A highly anisotropic stress state, close to the stress state
of El Rabaa (1989)'s experiments was used with a notably very low pressurization rate.
The same method predicts longitudinal fracture initiation for the stress state used in the highly-
concentrated gelatin block experiments of Wu et al. (2008) and Michael (2016b). Alabbad (2014) presented
several horizontal well design setups (Figure 2) attempted in many hydrostone cement/gypsum plaster block
16 SPE-199766-STU

samples. All those setups were deemed unsuccessful, due to what the author called an "observed tendency"
to generate longitudinal instead of requisite transverse fracture initiation. It is believed that the stress state
used could yield transverse fracture initiation however, leakage around the wellbore, which is clearly evident
from the photographs of the results provided (Alabbad, 2014), is the reason for the failure. This is attributed
to the weak interface between the aluminum wellbore tube used and the surrounding gypsum plaster; better
adhesion would have prevented leakage and generated transverse fracture initiation.
The successful validation of the numerical model enables the possibility of the model's use for the
development of correlations for predicting the orientation of fracture initiation from perforated wells, based
on the in-situ conditions. This can be achieved by using the true-3D, numerical model to derive correction
factors to the existing analytical approximations (Hossain et al., 2000; Michael et al., 2018) for specific
cases, at different pw values. The author would be happy to assist such attempts.

Preferred Perforation Alignment (PPA)

During conventional stimulation practices in the oilfield, the wellbore is perforated at several places around
its circumference meaning that perforations can exist at any angle θ on the wellbore radius. The analytical
approximations for σθL and σθT in Eq. 8-9 are valid for perforations parallel to SHmax (θ = 90°). Similar
expressions can be derived for arbitrary θ values, each yielding different pb/upper and Tcrit value and the same
is true for the numerical model. The perforation alignment for which pb/upper and Tcrit are maximized is known
as the preferred perforation alignment (PPA). Perforations parallel to PPA are the ones with the best chance
of generating transverse fracture initiation.
Continuing with the Barnett Shale example, which is under normal faulting stress regime (where, Sv >
SHmax > Shmin), for a horizontal well aligned with Shmin, the PPA is parallel to SHmax on the side of the borehole
(Figure 15), yielding a maximum possible value for pb/upper and Tcrit of 4,754 psi and 2,482 psi, respectively.
Subsequently, the values of pb/upper and Tcrit are minimized for perforations along "non-PPA," orthogonally
from PPA. Therefore, for wells parallel to Shmin in normal faulting stress states, non-PPA is parallel to Sv,
on top of the hole (Figure 15).

Figure 15—The preferred perforation alignment (PPA) configurations for horizontal wells drilled parallel
to Shmin under normal and strike-slip faulting stress regimes (modified from Michael and Gupta, 2019b)

Similar analysis (Michael and Gupta, 2019b) has shown that for strike-slip faulting stress states (where,
SHmax > Sv > Shmin) such as the Marcellus Shale, PPA is parallel to Sv. This is useful for completion engineers;
SPE-199766-STU 17

when targeting "sweet spots" strategically oriented perforations (Kurdi, 2018) along PPA can be used to
enhance the promotion of transverse fracture initiation by maximizing pb/upper and Tcrit.
The presence of casing and cement in the borehole, in the absence of a micro-annulus between the cement
and the rock formation, can shield the impact of pw on the surrounding rock. In the case of a good cement job,
the higher E of the casing is expected to suppress the stress changes caused by pw. The higher value of E of
the steel casing and the cement layers compared to the formation's suggests that the radial displacement the
wellbore experiences will be smaller compared to an openhole configuration at the same pressure. Appendix
B shows how cement can affect the analytical expressions for σθL and σθT in an extreme and rare case, which
can actually aid the promotion of transverse fracture initiation.

Impact of Natural Fractures

Interactions between induced hydraulic fractures with natural fractures were examined in experimental and
numerical studies (Lee et al., 2015). The presence of a natural fracture in the pathway of a propagating
hydraulic fracture may lead to either (i) crossing, (ii) divergence, or (iii) step-over, depending on various
parameters such as the angle at which the two fractures meet and the minerals present in the natural fracture.
Nonetheless, this falls into topics related to fracture propagation, which is beyond this study's scope.
In regards to fracture initiation per se, a natural fracture will likely impose an impact if it is located
on the perforation base, the assumed fracture initiation point. This is a rare, but possible situation and
numerical models, such as the one presented in this study can be employed to predict the outcome of such
unique occurrences. Moreover, heterogeneities induced by the presence of natural fractures, can diminish
the advantages of using PPA-oriented perforations rather than a conventional phase perforation technique.
An interesting spin-off research topic would be examining the role natural fractures can play in the re-
orientation of propagating hydraulic fractures, which did not initiate in the PFP orientation (Figure 9).

Conclusions
An orientation criterion was developed for fractures initiating from perforated wellbores using analytical
approximations for the transverse and longitudinal fracturing stresses. The validity of this criterion was
assessed using a true-3D numerical model, finding it to significantly overestimate transverse fracture
initiation. Based on the preceding analysis, the following conclusions are warranted:

• Transverse fracture initiation is a rarity, which occurs under a narrow range of conditions,
pertaining to low breakdown pressure, pb and formation tensile strength, T.
• A preferred perforation alignment (PPA) exists depending on the in-situ stress regime for which
the pb window and Tcrit for transverse fracture initiation are maximized.
• The presence of perforations per se, aids transverse fracture initiation and fluid leakage around the
wellbore during the treatment helps promote longitudinal fracture initiation.
• The frequent occurrence of longitudinal fracture initiation, suggests that fracture re-orientation as
it propagates away from the wellbore is likely field scenario.
• This fracture re-orientation can generate serious completion and production-related problems,
adversely impacting the overall NPV of the fracture treatment.
Considering the above set of conclusions, the following recommendations are made to practitioners for
enhanced promotion of transverse fracture initiation from perforated horizontal wells:
1. Slow pumping. Strategic use of low pressurization rates during the fracture treatment can keep the pb
within the window for transverse fracture initiation.
18 SPE-199766-STU

2. Strong casing/cement/rock interfaces. This will suppress leakage of the fracturing fluid around the
wellbore during the treatment, bypassing the perforations. This negates the impact of the perforations
in promoting transverse fracture initiation.
3. An oriented perforation strategy can, under ideal circumstances, promote transverse fracture initiation
by expanding the pb window. Nevertheless, heterogeneities in the medium, such as natural fractures
should be put into consideration, especially if an unexpected result is observed.
4. For laboratory studies, using a material of low T will improve the possibility of generating transverse
fracture initiation in the laboratory.

Acknowledgements
The author offers thanks to Ipsita Gupta and the Craft & Hawkins Department of Petroleum Engineering of
the Louisiana State University. Special greetings go to Jon E. Olson, Matthew T. Balhoff and the Hildebrand
Department of Petroleum & Geosystems Engineering of The University of Texas at Austin.

Nomenclature
E= elastic Young's modulus
lperf = perforation length
L, H, W = length, height and width of block used in the numerical simulations
pb = breakdown pressure
pb/upper = upper – bound of breakdown pressure window for transverse fracture initiation
pp = formation pore pressure
pperf = pressure in the perforation tunnel
pw = wellbore pressure
r= radius from the center of the wellbore
rperf = perforation radius
rw = wellbore radius
S1, S3 = greatest and least compressive in – situ principal stress, respectively
SHmax = maximum horizontal principal stress
Shmin = minimum horizontal principal stress
Sv = vertical principal stress (overburden)
Syy = wellbore axial stress (total)
Sθθ = wellbore tangential stress (total)
T= formation tensile strength
Tcrit = critical tensile strength

Greek Symbols
θ= angle around the wellbore taken from Sv
θp = angle around the perforation taken from the Syy
μ= friction factor
ν= Poisson's ratio
σrr = wellbore radial stress (effective)
σrθ = wellbore shear stress (effective)
σyy = wellbore axial stress (effective)
σθθ = wellbore tangential stress (effective)
= tangential stress at the perforation base (effective)
σθL = longitudinal fracturing stress (effective)
SPE-199766-STU 19

σθT = transverse fracturing stress (effective)

References
Alabbad, E. A. (2014). Experimental Investigation of Geomechanical Aspects of Hydraulic Fracturing Unconventional
Formations, Master's Thesis at the University of Texas at Austin.
Anderson, E. M. (1951). The Dynamics of Faulting. 2nd ed., Edinburgh: Oliver & Boyd.
Balen, R. M., Mens, H.-Z., & Economides, M. J. (1988). Applications of the Net Present Value (NPV) in the Optimization
of Hydraulic Fractures. Society of Petroleum Engineers. doi:10.2118/18541-MS.
Barree, R. D., and Miskimins, J. L. (2015). Calculation and Implications of Breakdown Pressures in Directional Wellbore
Stimulation. Society of Petroleum Engineers. February 3. doi:10.2118/173356-MS.
Byerlee, J. D. (1978). "Friction of Rocks." Pure and Applied Geophysics PAGEOPH, vol. 116, no. 4-5, 1978, pp. 615–626.,
doi:10.1007/bf00876528.
Daneshy, A. A. (1973). A Study of Inclined Hydraulic Fractures. Society of Petroleum Engineers. April 1.
doi:10.2118/4062-PA.
Deimbacher, F. X., Economides, M. J., and Jensen, O. K. (1993). Generalized Performance of Hydraulic Fractures With
Complex Geometry Intersecting Horizontal Wells. Society of Petroleum Engineers. doi:10.2118/25505-MS.
Economides, M., and Martin, A. N. (2010). How To Decide Between Horizontal Transverse, Horizontal Longitudinal,
and Vertical Fractured Completions. Society of Petroleum Engineers. doi:10.2118/134424-MS.
El Rabaa, W. (1989). Experimental Study of Hydraulic Fracture Geometry Initiated From Horizontal Wells. Society of
Petroleum Engineers. doi:10.2118/19720-MS.
Feng, R., & Sarmadivaleh, M. (2019). An Integrated Fracturing Pressure Diagnostic and CT Scan Approach for Hydraulic
Fracture Initiation and Propagation. In 81st EAGE Conference and Exhibition 2019.
Hossain, M. M., et al (2000), "Hydraulic Fracture Initiation and Propagation: Roles of Wellbore Trajectory, Perforation
and Stress Regimes." Journal of Petroleum Science and Engineering, vol. 27, no. 3-4, pp. 129–149., doi:10.1016/
s0920-4105(00)00056-5.
Hubbert, M. K., and Willis, D. G. (1957). Mechanics Of Hydraulic Fracturing. Society of Petroleum Engineers.
"Itasca Consulting Group." Software | Itasca Consulting Group, www.itascacg.com/software.
Kirsch, G. (1898). Die theorie der elasticitaet und die beduerfnisse der festigkeitskehre. VDI-Z 42, 707
Koskella, Dave, et al "Northside: Observation from an Underground Laboratory: An Integrated Approach to Unlocking
Performance in the Niobrara." SPE-GCS, Noble Energy, Inc., 21 Jan. 2015, www.spegcs.org/events/2713/.
Kurdi, M. (2018, September 24). A New Computational Model to Predict Breakdown Pressures in Cased and Perforated
Wells in Unconventional Reservoirs. Society of Petroleum Engineers. doi:10.2118/194038-STU.
Lee, H. P., J. E. Olson, J. Holder, J. F. W. Gale, and R. D. Myers (2015). The interaction of propagating
opening mode fractures with preexisting discontinuities in shale, J. Geophys. Res. Solid Earth, 120, 169–181,
doi:10.1002/2014JB011358.
Michael, A. (2014). Economic Implications Of The Current Geopolitical Forces Vis-à-vis Hydrocarbons On Global Energy
Markets. Society of Petroleum Engineers. May 19. doi:10.2118/169832-MS.
Michael, A. (2016). Financial Impact of Price Volatility on the Oilfield Services Sector of the Petroleum Industry. May
10. Society of Petroleum Engineers. doi:10.2118/179962-MS.
Michael, A. (2016). Hydraulic Fracturing Optimization: Experimental Investigation of Multiple Fracture Growth
Homogeneity via Perforation Cluster Distribution, Master's Report at the University of Texas at Austin.
Michael, A. (2019). The Net Present Value of a Hydraulic Fracture Treatment. Society of Petroleum Engineers. The Way
Ahead, www.spe.org/en/twa/twa-article-detail/?art=5557
Michael, A., and Gupta (2019). Analysis of Drilling-Induced Tensile Fracture Initiation in Porous, Permeable Media
Considering Fluid Infiltration. American Rock Mechanics Association. ARMA 19-2053.
Michael, A., and Gupta, I. (2018). Fluid-Driven Fracture Initiation During Loss of Control Situations. Abstract
(MR51B-0062) presented at 2018 AGU Fall Meeting, Washington, D.C., 10-14 December
Michael, A., and Gupta, I. (2019). Orientation of Fracture Initiation in Poroelastic Media: Application in Unconventional
Reservoirs. Society of Petroleum Engineers. doi:10.2118/195494-MS.
Michael, A., and Gupta, I. (2019). Orientation of Hydraulic Fracture Initiation in Poroelastic Media: An Analytical
Criterion for Perforated Wellbores. European Association of Geoscientists & Engineers. 81st EAGE Annual
Conference and Exhibition
Michael, A., Olson, J. E., and Balhoff, M. T. (2018). Analysis of Hydraulic Fracture Initiation From Perforated Horizontal
Wellbores. American Rock Mechanics Association. August 21
Moos, D. (2012). "Geosciences Technology Workshop (Hydraulic Fracturing)." AAPG, Search and Discovery Article
#80533, 14 Sept.
20 SPE-199766-STU

Nelson, E. J., et al (2005). "Transverse Drilling-Induced Tensile Fractures in the West Tuna Area, Gippsland Basin,
Australia: Implications for the in Situ Stress Regime." International Journal of Rock Mechanics and Mining Sciences,
vol. 42, no. 3, April, pp. 361–371., doi:10.1016/j.ijrmms.2004.12.001.
Rybacki, E., et al "What Controls the Mechanical Properties of Shale Rocks? â?" Part I: Strength and Young's Modulus."
Journal of Petroleum Science and Engineering, vol. 135, 2015, pp. 702–722., doi:10.1016/j.petrol.2015.10.028.
Vakó, P., and Economides, M. J. (1995). Hydraulic Fracture Mechanics. Chichester: Wiley, Print.
Weijers, L. (1995). The Near-wellbore Geometry of Hydraulic Fractures Initiated from Horizontal and Deviated Wells,
Ph.D. Thesis, Delft University of Technology, Delft, The Netherlands.
Wu, H., Golovin, E., Shulkin, Y., Chudnovsky, A., Dudley, J. W., & Wong, G. K. (2008, January 1). Observations of
Hydraulic Fracture Initiation and Propagation in a Brittle Polymer. American Rock Mechanics Association.
Zoback, M. D. (2014). Reservoir Geomechanics. Cambridge: Cambridge UP, Print
Zoback, M. D., Rummel, F., Jung, R., and Raleigh, C. B. (1977). Laboratory hydraulic fracturing experiments in intact
and pre-fractured rock. In International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts
(Vol. 14, No. 2, pp. 49-58). Pergamon. March
SPE-199766-STU 21

Appendix A
Numerical (True-3D) Model Verification
The numerical model, written on FLAC3D, was verified against the analytical solution of the elastic wellbore
stresses around a horizontal wellbore (Kirsch, 1898). A point on the wellbore far away from the perforation
(> 10rperf) was chosen for wellbore stress verification, where the effects of its presence are diminished. As
Figure A-1 illustrates, the FLAC3D numerical solutions for σyy, σθθ, σrr and σrθ indicate excellent agreement
with the corresponding analytical solutions for the same stresses at distance r = 1.02rw from the center of
the wellbore.

Figure A-1—Comparison between the wellbore stresses predicted by the FLAC3D numerical
model and the analytical solutions of Kirsch (1898) taken at a distance r = 1.02rw from the center
of the wellbore and 10rw from the perforation. The angle θ represents the angle from Sv. Barnett
Shale stress state parameters were used and pw was fixed at 5,000 psi (after Michael et al., 2018)

Moreover, the variation of the wellbore stresses with varying wellbore pressure was investigated for pw
ranging at 0-10,000 psi for Barnett Shale parameters Moos (2012). Figure A-2 illustrates the comparison for
σθθ and σyy. The FLAC3D model maintains an excellent agreement with the corresponding analytical solutions
of these stresses across the pw range.
22 SPE-199766-STU

Figure A-2—Variation of the stresses with pw as predicted by the FLAC3D numerical

model (points) and the corresponding analytical solutions (lines) based on Kirsch (1898)
taken at a distance r = 1.02rw from the center of the wellbore (after Michael et al., 2018)
SPE-199766-STU 23

Appendix B
Effect of Casing and Cement
The analytical results can be re-computed to mimic the effects of a cemented casing for the extreme case
where the tangential stresses around the wellbore are fixed at the point of cement curing. We assume this
to be a balanced condition, pw = pp. These wellbore stresses are then the "remote" loading at the base of
the perforation and hence we use the fracturing pressure as the perforation pressure pperf, which is equal
to pw and acts as a fracturing pressure. Using the Barnett Shale case (Figure B-1), it is apparent that the
shielding effect of the cemented casing makes both the transverse and longitudinal fracturing stresses at
the perforation base to become more tensile with increasing pw. This implies that good cement jobs without
annular leakage will be an crucial aid for transverse fracture initiation.

Figure B-1—Extreme case of longitudinal and transverse fracturing stresses for well-cemented boreholes for input
parameters from the Barnett Shale, showing all stresses becoming more tensile (at the same slope) as pw increases (modified
from Michael et al., 2018). For this situation, transverse fracture initiation is always promoted regardless of pw (pperf = pw)