FOR PARTIAL FULFILLMENT OF
P452 COMPUTATIONAL PHYSICS, 2021-22
Solving a system of fluid-induced fault slip
May 2, 2022
Soumik Bhattacharyya
Roll No.- 1811155
Fourth year Integrated M.Sc
Guide:
Dr. Pathikrit Bhattacharya
Course Instructor:
Dr. Subhasis Basak
School of Physical Sciences
School of Earth and Physical Sciences
National Institute of Science Education and Research
Bhubaneswar
Solving a system of fluid-induced fault slip
Soumik Bhattachryya
SPS, NISER
1 Introduction
The action of injecting fluid into the subsurface level of Earth is often followed by a number of
earthquake events in the area. Figure 1 shows a 3D picture of injection of fluid through a borehole.
Now, the fluid goes to the subsurface level and starts to move in radially outward direction in a
region where there is already a damage zone in the rock. The host rock has a low permeability (k0 )
and high elastic modulus (G0 ) compared to the fault damage zone (k and G, respectively).
Figure 1: Fluid is injected through a borehole Figure 2: The pore pressure (p(r,t)) reduces the
with a rate Q(t). The fault damage zone (in yel- effective normal stress. The elliptical rupture
low) has a width of b and elastic modulus G. The grows radially outward the direction of r. [Fig-
host rock has an elastic modulus G0 >> G. [Fig- ure: P. Bhattacharya and R. C. Viesca (2019)]
ure: P. Bhattacharya and R. C. Viesca (2019)]
In figure 2, the picture is shown in a cylindrical geometry (r, θ , φ). The pore pressure is a purely
radial function, which decreases with distance from the borehole centre and increases with time.
This pore-pressure term effectively reduces the normal stress(σ̄).
σ̄ = σ − p(r̄, t) (1)
As a result this reduction in stress decreases the frictional resistance to sliding. This resistance
term given as a product of frictional coefficient (f) and σ̄. Apart from the frictional resistance, there
is a background shear traction, τ, which arises in the boundary between damage zone and host
rock. This helps the growth of an elliptical shaped rupture along the major axis (r), parallel to the
slip direction. The force balance equation of the problem is given by,
L(t)
∂
Z
Go 1 1
n r o n r o
f [σ − p(r, t)] − τ = − δ(s, t) E k + F k ds (2)
2π 0
∂s s−r s s+r s
In equation 2, the L(t) is the rupture length, which increases with time; δ (r,t) is the
pslip profile;
E and F are the complete elliptic integrals of respectively 2nd and 1st kinds; k(U) = 2 U/(1 + U);
1
s is another coordinate system along the same direction as r. We cannot take s and r as the same
systems as that will give an infinite coefficient term for ∂∂ δr at s= r.
Now, in this problem, the boundary value, i.e. L(t) is a function of t. However, the growth of
the rupture follows a diffusion equation and we can relate the the rupture length and time using a
dimensionless constant λ. p
L(t) = 4λ αt
We further substitute s and r to non-denationalise the equation.
s̄ = 2s/L − 1, r̄ = 2r/L − 1
The pore-pressure distribution p(r,t) is given by,
r2
p(r, t) = P∗ E1 (3)
4αt
E1 is the exponential integral function. Now, substituting all these in equation (2) leads us to,
τ σ
1− − E1 λ2 (r̄ + 1)2 =
f σ P∗
Z1 (4)
1 d δ̄ 1 r̄ + 1 1 r̄ + 1
§ ª § ª
− E k + F k ds̄
2π −1 ds̄ s̄ − r̄ s̄ + 1 s̄ + r̄ + 2 s̄ + 1
right hand side of the equation, then we will get a value of λ for a given
Now, if we can solve the
τ
value of T (T = 1 − f σ Pσ∗ ; in practical scenarios, if we can fit the diffusion equation to get the
value of λ, we can solve T. Please note that T is taken constant for this part, however, f can also be
a function of r and t as the slipping progresses.
Now, we proceed to find the dδ ds term to solve the equation. For an axi-symmetric non-singular
p
crack, we take solution to be in the form d δ̄/ds̄ = 1 − s̄2 φ(s)
We use the Gauss-Chebyshev quadrature of Type II to get the solution.
1
d δ̄ r̄ + 1 r̄ + 1
Z
1 1
§ ª § ª
I (r̄i ) = E k + F k ds̄
−1
ds̄ s̄ − r̄ s̄ + 1 s̄ + r̄ + 2 s̄ + 1
n
π ri + 1 ri + 1
X 1 1
= 2
1 − s̄ j E k + F k φj (5)
j=1
n + 1 s j − r i s j + 1 s j + r i + 2 s j + 1
n
X
= w i j Ki j φ j
j=1
2 Gauss Chebyshev Quadrature
In numerical analysis, a quadrature rule is an approximation of a definite integral of a function taken
as a weighted sum of function values at specified points within the domain of integration. In case
of rectangular quadrature, we take the quadrature points at equal intervals and take the weights
after solving a system of linear equations for known polynomials. However, Gaussian quadratures
give ourselves the freedom to choose not only the weighting coefficients, but also the location of
the abscissas at which the function is to be evaluated. Thus we have twice the degrees of freedom
compared to rectangular quadratures.
2
In Gaussian quadrature, we can arrange the choice of weights and abscissas to make the integral
exact for a class of integrands polynomials times some known function W(x). For a given weight
function, W(x) and a given order of quadrature (n), we can find a set of weights w j and abscissas
x j such that,
Z b n−1
X
W (x) f (x)d x ≈ wj f xj (6)
a j=0
The approximation gives exact solution if f(x) is in polynomial form.
Chebyshev calculated the values of w j and x j when the weight function is in the form
W (x) = (1 − x)α (1 + x)β
and α and β are ±1/2 each.
In our case, we have a non-singular crack which corresponds to the case when both α and β
are +1/2. The formula of weights and abscissas for the equation are given by the following set of
equations,
π 2i − 1
r̄i = cos
2 n+1
j
s̄ j = cos π (7)
n+1
π
wi j = 1 − s̄2j
n+1
Here ri are the points along which we calculate the slip gradients, i=1,2,3...n+1 for a chosen
degree n.
s j are the abscissas at each ri and j=1,2,3...n.
w j are the weight corresponding to each s j
3 Solving the equation using n-dimensional Newton-Raphson
Taking all terms of equation 4 at one side gives us the following force balance equation to solve,
τ∞ σ X 1
− E1 λ2 (r̄i + 1)2 +
Fi = 1 − Wi j Ki j φ j (8)
f σ P∗ j
2π
Here we have n+1 number of equations (i=1,2,3...n+1) and n+1 variables (λ, φ j ) to solve
(j=1,2,3..n). The set of equations are not linear, because the pore-pressure term (p(t)) is given in
exponential integral form. As a result, we cannot solve the equation using any matrix inversion
method, instead we go for n-dimensional Newton-Raphson method.
In traditional Newton-Raphson method for single variable x and a function f(x), we initiate the
process with a guess of x=x 0 , reasonably close to the actual solution, and continue the iteration as
follows,
f (x k )
x k+1 = x k − ′
f (x k )
Now, in case of n-dimensional Newton-Raphson, let us take the function F as,
3
Fi (x 1 , x 2 , ...x n ) = 0, i = 1, 2, 3..n
Let x denote the entire vector of values x i and F denote the entire vector of functions Fi . In the
neighborhood of x, each of the functions Fi can be expanded in Taylor series,
N −1
X ∂F i
Fi (x + δx) = Fi (x) + δx j + O δx2
j=0
∂ xj
The matrix of partial derivatives is called the Jacobian Matrix, J,
∂ Fi
Ji j ≡ .
∂ xj
F(x + δx) = F(x) + J · δx + O δx2
Now, neglecting the higher order terms and setting F(x + δx) as zero, we obtain
δx = −F · J−1
the J1 term is calculated by using numpy library for LU decomposition method. Finally we get the
next iteration vector x by using,
xnew = xold + δx
In our problem the Ji, j matrix is given by,
∂ Fi 2
= exp −λ2 (r̄i + 1)2
J1...(n+1),1 =
∂λ λ
∂ Fi 1 (9)
J1...(n+1),2...n = = Wi j Ki j
∂ φj 2π
τ
σ
We now solve the system of equations for [λ, φ j ] for different values of T ( 1 − f ∞
σ P∗ ).
4 Results
Continuity of the solution
Figure 3: Normalised summation of Fi with Figure 4: Normalised summation of Fi with
changing λ changing T
• Discussion:
The plot being continuous shows that there is no discontinuity in the taken range. However, we
cannot draw any further conclusion unless we can get a minima in the plot.
4
Relation between T, λ and slip gradient (δ r )
Figure 5: Change of λ with T Figure 6: Slip profile as a function of coordinate
• Discussion:
As expected, an increase in T essentially means a decrease in frictional coefficient (f), so the
rupture length (L(t)) increases more with time compared to higher friction. So, the constant of the
diffusion equation between L(t) and t, λ, also increases.
In the second plot (Figure 4), the coordinate is from -1 to 1, where -1 corresponds to borehole
centre and 1 corresponds to rupture length. As T increases, frictional coefficient decreases and
slipping increases as a result of that.
References
1. P. Bhattacharya, R. C. Viesca, Fluid-induced aseismic fault slip outpaces pore-fluid migration,
Science 364 (6439), 464-468
2. R. C. Viesca, D. I. Garagash, Numerical methods for coupled fracture problems, Journal of
Mechanics and Physics of Solids, 113, 2018, 13-34
3. W. H. Press et. al., Numerical Recipes The Art of Scientific Computing, Third Edition, Cam-
bridge University Press
