12.

005 Lecture notes #1

12.005 Lecture Notes 1

Handouts - Course description & reserve list

Motivation for core course
Instructor background
Student background
Courses in physics and mathematics?
EAPS area of focus?
Textbooks

Mechanics: the study of the motion of matter and the forces that cause such motion.
Based on concepts of time, space, force, energy, matter.
Applications to point masses, solid bodies familiar from introductory physics.
Continuum mechanics - mechanics of parts of "bodies."

Continuum - define values of fields (e.g., density) as functions of position, i.e., at points.
Mn
Example : density (P) = lim
Vn 0 V n

Vi > Sequence of volumes converging on P

Mi > Mass enclosed by Vi

12.005 Lecture notes #1

Figure 1.1
Figure by MIT OCW.

Breakdown at small V (e.g., gas)

Figure 1.2
Figure by MIT OCW.

12.005 Lecture notes #1

Ignore fine structure. Assume:

Continuity: completely fills space (no pores or voids) and has properties
describable by continuous functions. [Scale dependent. E.g., sand will be treated as a
continuum]
Homogeneity - identical properties at all points [Scale dependent.]
Isotropy - properties same in all directions. [Often not true. e.g., schist]
Behavior describable in terms of partial differential equations subject to boundary
conditions. All functions will be "well behaved," except at a finite # of surfaces.
Logical contradiction between mathematical process of taking limits & physical
breakdown of continuum description at small scales ignored. (Assume - fields varying
slowly enough that math defined before description falls apart.)

At core: Newton's 2nd Law:

F = ma, or F -ma = 0.

for this course, usually a = 0 => the governing equation is F = 0!

Applies not just to particles, entire bodies, but to regions within bodies.
Free-body diagram - cut open body (thought experiment), examine forces of
interaction between surfaces.
Description of forces acting in the continuum - For taking limits, etc, forces/unit area
make sense. Consider tractions T: Force/unit area on faces (vector quantities).
Units: SI - Pascal (Pa); 1 Pa = 1 N/m2
cgs: dyne/cm2
"geophysics": bar ( 1 atmosphere): 1 bar = 106 dynes/cm2 = 0.1 MPa
A general description of the tractions acting on planes of arbitrary orientation requires:
stress - forces of interaction (2nd rank tensor)
Newton's 2nd > ij,j + fi = 0.
strain & rotation - kinematics of motion & deformation (2nd rank
tensors)
rheology (constitutive law) - relationship between stress & strain
solids - elastic, plastic, brittle
fluids - viscous, ideal
other - e.g., Silly putty, paint,

12.005 Lecture notes #1

We will quantify these quantities soon, but first a digression to a concrete example to
bring things into focus. When I moved to southern CA ~ 1980, I was impressed by press
coverage of a set of law suits pending against various governments (state & local). The
question to be settled was who was responsible for an expensive housing development
"sliding into the ocean." Briefly, the sequence of events was:
1) "Nature" deposited a terrace (mainly sand) beside the ocean.
2) The highway department made a road cut into the base of the sea cliff to put in a
coastal highway.
3) Expensive houses were built on top of the terrace.
4) Lawns were watered lavishly.
5) Landslides started wrecking the expensive houses.

Figure 1.3
Figure by MIT OCW.

12.005 Lecture notes #1

We will examine the "continuum mechanics" of this situation & come up with some
"expert opinion" about responsibility. We need to do two things: first, understand how
sand responds to "internal forces," then understand how to describe these forces.
Assertion: Sand pile behavior, as a first approximation, is governed by friction. So are
earthquake faults.
(reading: T & S, pp 351-353)
Consider an experiment to measure the frictional behavior of sand. One approach would
be to place some sand between a block & a table, subjecting the block to a downward
force F and a horizontal force V (in T&S notation). If the block has area A, there is a
normal traction
n = F/A
and a shear traction = V/A

Figure 1.4
Figure by MIT OCW.

Admonton's law (to summarize a lot of experiments) states that the shear force needed to
cause sliding is related to the normal force pressing the block down as
V = fs F
or, in terms of tractions
= fs n

12.005 Lecture notes #1

Digression: One very sensitive way of measuring fs (developed at MIT) is to use an

inclined plane & vary the inclination angle . Suppose this is done & loading is carried
out by varying the # of bricks squeezing the sand. How will depend on the # of bricks?

Figure 1.5
Figure by MIT OCW.

12.005 Lecture notes #1

Another digression: An amazing result from rock mechanics, "Byerlee's law," is that fs
is essentially independent of rock type!

Figure 1.6
Figure by MIT OCW.

Now, back to the sand pile. Admonton's law tells us that any time the shear traction on a
plane reaches a value of fs n, the sand pile will fail. The complication is that we no
longer have a brick & a bench to work with what we need is a mathematical
description of normal and shear tractions on arbitrary planes in a continuum. We start
out with a special description of the tractions on a surface oriented with the coordinate
system, then will go on to generalize.

12.005 Lecture notes #1

It is useful to break tractions into their component parts (normal & shear):
Define ij as force/unit area acting on face with normal ni in direction nj.
Consider now a small parallelepiped with faces in the co-ordinate planes and dimensions
dx1, dx2, and dx3. The tractions on the three faces can be resolved into their Cartesian
components, one normal and two tangential to the face on which the traction acts:
T(1)i = ( 11 , 12 , 13 )
T(2)i = ( 21 , 22 , 23 )
T(3)i = ( 31 , 32 , 33 )

where T(1)i denotes the traction on the face normal to the axis x1, and so on. The nine
components of the tractions form a 3 3 matrix
11 12
ij = 21 22

31 32

13
23

33

where for each component the first subscript denotes the co-ordinate axis to which the
surface is normal, and the second subscript denotes the direction in which the component
acts. For example, 23 denotes the component acting on the face normal to x2 in the
direction of x3 .

12.005 Lecture notes #1

Figure 1.7
Figure by MIT OCW.

Case 1: Water at rest: What are the relations among components?

Case 2: A cube at the corner of the table: What are the relations among components?

12.005 Lecture Notes 2

Back to the sand box

Figure 2.1
Figure by MIT OCW.

We believe in friction

= fs n

12 = fs ( 11 )

We understand shear and normal tractions on planes oriented along axes.

Figure 2.2
Figure by MIT OCW.

In this example, nature chooses the failure plane. Failure will occur if 1' 2 ' fs 1'1' .

How do we relate this to our prior coordinate system?

Tractions (or Stress Vector)

A surface force F that acts uniformly over planar surface of area A results in a traction
T of magnitude

F
T =

A
The traction is a vector that has the same direction as F and has units of pressure. In SI
units, tractions are measured in Pascals (Pa).

T=

F kg m
N
= 2 2 = 2 = Pa
A m s
m

Since tractions are vectors, they can be decomposed into normal and tangential vector
components. n and denote the magnitude of these components.
T = Ttan gential + Tnormal

n = Tnormal

= Ttan gential

Traction depends on the area of the surface over which it acts. Since area often changes
with the orientation of the surface, the values of n and do not behave like vector
components when the surface orientation changes.

The decomposition of tractions into normal and shear components is useful in writing
Admontons Law in terms of tractions. This new form of Admontons law is the link
between stress and failure across a plane.

Admontons law relates a normal force and a tangential force by a proportionality

constant called the coefficient of static friction fs:

Ftan gential = f s Fnormal

Dividing both sides of the equation by the area A over which the forces act leads to
Admontons law in terms of tractions:

= fs n

Consider a small planar element of area S passing through P. n is unit normal vector
(positive arbitrary, usually chosen outward for closed body).

F
~+

x3

n
P
S

x2

o
x1

Figure 2.3
Figure by MIT OCW.

Plane cuts material M. M+ is material on + side and M- is material on side.

In general, M+ exerts force F+ on M- and M- exerts force F- on M+ .

Tractions at a point

Tractions at a point are difficult to conceptualize because the area at a point over which a
force acts is infinitesimal. Despite this difficulty, the concept of point tractions is
extremely important because it allows one to find the traction on an arbitrary plane.

The Cauchy stress principle states that as the area around P shrinks to zero (see Figure
2.3), the following limit holds

F+
=T
S 0 S
lim

T is called the traction or stress vector at point P. The (n ) is a reminder that this stress
vector is defined only for a particular plane through P with normal vector (n ) .
Moments?
M+ could exert a moment on M-.
In general, moment 0 as S 0 .

Tractions are surface forces.

Surface moments dont exist.
Body moments can exist, e.g., magnetism.

12.005 Lecture Notes 3

Tensors
Most physical quantities that are important in continuum mechanics like temperature,
force, and stress can be represented by a tensor. Temperature can be specified by stating
a single numerical value called a scalar and is called a zeroth-order tensor. A force,
however, must be specified by stating both a magnitude and direction. It is an example of
a first-order tensor. Specifying a stress is even more complicated and requires stating a
magnitude and two directionsthe direction of a force vector and the direction of the
normal vector to the plane on which the force acts. Stresses are represented by secondorder tensors.
Stress Tensor
Representing a force in three dimensions requires three numbers, each referenced to a
coordinate axis. Representing the state of stress in three dimensions requires nine
numbers, each referenced to a coordinate axis and a plane perpendicular to the coordinate
axes.
Returning to determining traction vectors on arbitrary surfaces.
Consider two surfaces S1 and S2 at point Q.

Example

T(1)

S2

n1

n2

T(2)

n1

ramp

n2

page

S1
Figure 3.1
Figure by MIT OCW.

Tractions at a point depend on the orientation of the surface.

How to determine T , given n ?
For special cases n along axes.

23
21

22

x3

x2
x1
Figure 3.2
Figure by MIT OCW.

In vector notation, the tractions on the faces of the cube are written:
T(1) = 11 , 12 , 13
T(2) = 21 , 22 , 23
T(3) = 31 , 32 , 33

In matrix notation, the tractions are written:

T(1) 11 12 13

T(2) = 21 22 23

T
(3) 31 32 33
This matrix is generally referred to as the stress tensor. It is the complete representation
of stress at a point.

The Cauchy Tetrahedron and Traction on Arbitrary Planes

The traction vector at a point on an arbitrarily oriented plane can be found if

T(1) , T(2) , T(3) at that point are known.

Argument: Apply Newtons second law to a free body in the shape of a tetrahedron and
let the height of the tetrahedron shrink to zero.

Consider the tetrahedron below. The point O is the origin and the apices are labeled A, B,
and C.

x3
C

n
h

x1
Figure 3.3
Figure by MIT OCW.

The relevant quantities are defined as follows:

= density
Fi = body force per unit mass in the i direction

x2

ai = acceleration in the i direction

h = height of the tetrahedron, measured to ABC

S = area of the oblique surface ABC

Ti = the component of the traction vector on the oblique surface in the i direction
1
The mass of the tetrahedron is hS . The area of a face perpendicular to xi is ni S .
3

Force Balance on Tetrahedron

x3
-T(1) S1
-T(2) S2

T(n) S

x2

x1

-T(3) S3

bV
~

Body force/unit mass

Figure 3.4
Figure by MIT OCW.

Consider the force balance in the i=1 direction. Overbars denote values averaged over a
surface or volume.

F1 = ma1
1

F1 hS + T( n )1S 11 ( n1S ) 21 ( n2 S ) 31 ( n3S ) = hS a

3

Divide both sides by S.

1
1
F1 h + T( n )1 11 ( n1 ) 21 ( n2 ) 31 ( n3 ) = h a
3
3

Allow h to approach zero in such a way that the surfaces and volume of the tetrahedron
approach zero while the surfaces preserve their orientation. The body force and the mass
both approach zero.

T(n)1 = 11n1 + 21n2 + 31n3

Performing the same force balance in the other two coordinate directions leads to
expressions for the three traction components on an arbitrary plane.
T(n)1 = 11n1 + 21n2 + 31n3
T(n)2 = 12 n1 + 22 n2 + 32 n3
T(n)3 = 13n1 + 23n2 + 33n3

The set of these three equations is called Cauchys formula.

Different Notations

1. A general equation for the explicit expressions above is given by:

3

Ti = ji n j
j =1

2. Summation notation is a way of writing summations without the summation sign

. To use it, simply drop the and sum over repeated indices. The equation in
summation notation is given by:

Ti = ji n j

3. The equation in matrix form is given by

T1 11 21 31 n1


T2 = 12 22 32 n2
T3 13 23 33 n3

For example, consider sliding block experiment. = 30 o

x3

x2

n
30o

x1

Figure 3.5
Figure by MIT OCW.

On the sliding plane: What is the traction T in terms of ij ?

n = (0, cos 60 o , cos 30 o )

= (0, 1 / 2, 3 / 2)
11 21 31

T = (0, 1 / 2, 3 / 2) 12 22 32

13 23 33

Features of the Stress Tensor

The stress tensor is a symmetric tensor, meaning that ij = ji . As a result, the entire
tensor may be specified with only six numbers instead of nine.

21

x2
dx2

12

dx1
x1

Figure 3.6
Figure by MIT OCW.

Consider the torques t acting on an element with sides dx1 and dx2 .

t 3 = 2 12

dx1
dx
dx2 2 21 2 dx1 = 0
2
2

12 = 21

A similar argument shows 32 = 23 ; 13 = 31 .

Shears are always conjugate.

12.005 Lecture Notes 4

Stress Tensor (continued)
Before getting bogged down in rotation and notation, lets consider a special case:
rotation of the coordinate system by 45o about x3.

x2
x1'

x2'
45o

x1

x3, x3'
Figure 4.1
Figure by MIT OCW.

In parallel, consider the tractions on a plane with its normal in the x1 x2 plane at 45o to
x1 and x2.

2 / 2

n = ( 2 / 2, 2 / 2, 0)T = 2 / 2

11 + 12
2
12 + 22
T = n =

13 + 23

CASE 1
"Hydrostatic"

CASE 2

x2

"Uniaxial"

x2
1

P
x1

x1

x3

x3

CASE 3
"Plane Stress"

CASE 4
"Plane Stress"

+S

-S

Figure 4.2
Figure by MIT OCW.

Notes:
Case 1: T (on plane at 45o) is only n = p , no .
Case 2: T45o is still only n = 1 .
Case 3: T on faces 1 and 2 only n , no .
T45o is pure , no n .
Case 4: T on faces 1 and 2 all , no n .
T45o is all n , no .

Evidently, there are special directions, at least for these cases, where = 0 !
Is this true in general? Can we find the principal frame where

ij p

Require

11 p
0
0 1 0
0

p
= 0 22
0 = 0 2 0 ?
0
0
33 p 0 0 3

T is parallel to n

or

T =

n = n

where is tensor and is scalar (sometimes ),

or

I n = 0

The stress tensor can be represented in different ways to highlight particular features or
aid in solving geodynamic problems. This lecture explores how to represent the stress
tensor in terms of principle stresses and isotropic and deviatoric stresses. Representing
the stress tensor in terms of principle stresses makes visualizing the state of stress easier
because it reduces the stress tensor to only three numbers. It also makes some

calculations easier. Representing the stress tensor in terms of isotropic and deviatoric
stresses is helpful in determining the type of faulting produced by certain stresses.

Principle Stresses and Principle Axes

The stress tensor is a matrix that specifies the tractions on three mutually perpendicular
faces of an infinitesimal cube.

Figure 4.3
Figure by MIT OCW.

In general, these tractions are both parallel and perpendicular to the normal vectors of the
faces. At a certain orientation of the faces, however, the tractions are only parallel to the
normal vectors. The directions of these normal vectors are called principle directions and
the stresses are called principle stresses. Calculating the principle directions and stresses
requires using eigenvalues.

1. Begin with Cauchys formula:

Ti = ji n j

Require the traction Ti to be parallel to the normal vector by writing

Ti = ni

in which represents one of the three principle stresses.

2. Combine the above equations:

ji n j = ni
( ij ij )n j = 0
11
12
13 n1

22
23 n2 = 0
21
31
32
33 n3

In the middle equation, ij is the Kronecker delta. It equals 0 if ij and equals 1 if i=j.

3. Solve:

The matrix equation has a solution if

11
12
13
21
22
23 = 0
31
32
33

where the double bars indicate the determinant. This results in a third-degree equation in

3 + I 1 2 + I 2 + I 3 = 0

where
I1 = 11 + 22 + 33
I 2 =

11 12 11 13 22 23
+
+
21 22 31 33 32 33

11 12 13
I 3 = 21 22 23
31 32 33

Isotropic and Deviatoric Stress

The stress tensor can be divided into two parts: isotropic stress and deviatoric stress.
The isotropic stress ij 0 is defined as
1
3

ij 0 = kkij

It represents the mean normal stress or pressure. Subtracting the mean normal stress
from the stress tensor produces the deviatoric stress ij d

ij d = ij ij 0

The deviatoric stress represents the part of the stress that differs from a hydrostatic state.

12.005 Lecture Notes 5

Quantities in Different Coordinate Systems
How to express quantities in different coordinate systems?

x3'

x3
Direction Cosines

Axis

11

x1

x2'

23

x2

12

x1'

x1'

x1

x2

x3

11

12

13

x2'

21

22

23

x3'

31

32

33

Figure 5.1
Figure 5.1
Figure by MIT OCW.

Direction cosine ij is cosine of angle ij between primed axis i and unprimed axis j.
If xi ' and x j represent unit vectors that are the axes of two coordinate systems with the
same origin, they are related by the equation
3

j=1

j=1

x i ' = ij x j , x i ' = ij x j ij x j

where ij is the cosine of the angle between the primed axis xi ' and the unprimed axis x j .
For example, 12 is the cosine of the angle between x1 ' and x 2 . ij represents a 9component matrix called the transformation matrix. Unlike the stress tensor, it is not
symmetric (ij ji).

 11 12 13
ij = 21 22 23
31 32 33

In matix equations, the transformation law is written

x1 ' 11 12 13 x1

x ' =
2 21 22 23 x 2
x3 ' 31 32 33 x3

The inverse transformation law is written

x i = ji x j '

Consider the following transformation of coordinates:

x2
x1'

x2'
30o

x3, x3'

Figure 5.1a

Figure 5.1a
Figure by MIT OCW.

x1

The transformation matrix is

30 60 0
ij = 120 30 0
0
0 0

The explicit transformation equations are

x1 ' = cos 30 x1 + cos 60 x 2
x 2 ' = cos120 x1 + cos 30 x 2

Since xi ' and x j are both unit length, these equations are easy to verify from the picture.

First-order tensors

First-order tensors or vectors have two components in 2D coordinates and three

components in 3D coordinates. They transform according to the same laws as coordinate
axes because coordinate axes are themselves vectors.
If uj is a vector in the x j coordinate system and ui is a vector in the xi ' coordinate system,
then the following equations describe their transformation:

u i ' = ij u j
u i = ji u j '

Note that ij is positive if the angle is measured counterclockwise from xi ' to x j . It is

negative if the angle is measured clockwise.

Second-order tensors

The transformation law for second-order tensors like stress and strain is more
complicated than the transformation law for first-order tensors. It may be derived as
follows:

1. Begin with the vector transformation of traction Ti to Tk ' :

Tk ' = ki Ti

2. Rewrite Tk ' and Ti using Cauchys formulas:

Tk ' = ' kl nl ' and Ti = ij n j

Substitute Cauchys formulas into the original transformation equation:

' kl nl ' = ki ij n j

3. Transform the normal vector nj to nl and substitute into the previous equation:
n j = lj nl '

' kl nl ' = ki ij lj nl '

4. Cancel the nl term on each side and group the s:

' kl = ki jl ij

Note that changing the position of the last term changes the order of its
subscripts.

In vector notation, the equation is

' = T

where the double under ~ denote second-rank tensors and the superscript T
denotes the transpose of matrix .

Mohrs Circle

We explained before that an object resting on a slope will slide down when the shear
traction on the slope is greater than or equal to the product of the normal traction and the
coefficient of friction.

= f s n

On a shallow slope, n is large and the object will not slide. On a steep slow, is large
and the object will slide. For any plane with normal n , we can calculate if the plane will
fail if the stress tensor ij at the interface between the object and the slope is known.
Calculate n and , as follows:
Vector and Tensor Notation
T = n

Summation Notation
Ti = ijn j

n = T n

n = Ti n i

= T n n

= Ti Tk n k n i

This method is straightforward but cumbersome. A different approach involves rotating

the coordinate system such that x1 is along n . In this case n and are much easier to
derive:

n = '11
= '12
Mohrs circle may be derived in two or three dimensions. This lecture explains the
derivation in two dimensions because it is more straightforward and the results are easier
to graph and understand. The derivation assumes that x1, x2, and x3 are principle
directions.
Consider the following figure in which the xi coordinate system is rotated clockwise
about the x3 axis to xi:

x1

x1'

11

Figure 5.2
x2'

x2

22
21

Figure 5.2
Figure by MIT OCW.

The rotation matrix is:

11 12 13 cos
ij = 21 22 23 = sin
31 32 33 0

sin
cos
0

0
0
1

The stress tensor in the xi coordinate system is transformed to ' in the xi coordinate
system by the following equation:

' = T

cos sin

' = sin cos

0
0

0 11 0
0 cos

0 0 22 0 sin
1 0
0 33 0

11 cos 22 sin 0 cos

0 sin
= 11 sin 22 cos
0
0
33 0

sin
cos
0

sin
cos
0

0
1

0
1

11 cos2 + 22 sin 2 ( 11 22 )sin cos

0

2
2
= ( 11 22 )sin cos 11 sin + 22 cos 0

0
0
33

Use the double-angle identities for sine and cosine to simplify the expressions for the
normal stress 11 and the shear stress 12 become in the new coordinate system:
sin 2 = 2 sin cos
cos 2 = 2 cos 2 1 = 1 2 sin 2

11 ' = n =

11 + 22

11 22

2
2
22
12 ' = = 11
sin 2
2

cos 2

The expression for the normal stress and shear stress can be shown graphically in shear
space:

Allowable tractions

Figure 5.3
Figure by MIT OCW.

This figure is called Mohrs circle.

Mohrs circle plots in stress space. Admontons law may also be plotted in stress space
as a line with slope fs. When this line and Mohrs circle intersect, the criterion for failure
across a plane is met.

A note on signs.
As derived, assumed 1 > 2 . Simplest case: consider = 45 o

x2'

45 o

x1

x1

x2

x2
Figure 5.4
Figure 5.4
Figure by MIT OCW.

x1'

Consider several cases:

1
0

4
3

Figure 5.5
Figure by MIT OCW.

0) 2 = 60o
1) 2 = 90o
2) 2 = 180o
3) 2 = 270o
4) 2 = 300o

Back to the landslide:

Consider a common experiment in soil mechanics or rock mechanics in which scientists
apply a uniaxial stress 2 to a cylindrical sample confined by a uniform stress 1.

x2
x1
x3

Figure 5.6

1
1

Figure 5.6
Figure by MIT OCW.

Continue increasing 2 until failure. n of failure plane typically at ; 30o from 1.

Seal

Rubber jacket
Specimen

Figure 5.7
Confining pressure

Figure 5.7. Triaxial test deformation apparatus

Figure 5.7
Figure by MIT OCW.

55

3 = 5000 p.s.i.

50

35
30

25

25

Stress: k.p.s.i.

30

20
15
10
5
0

40

2
3 4
Millistrains

35
Stress: k.p.s.i.

35

Stress: k.p.s.i.

40

3 = 0

45

3 = 1000 p.s.i.

20

30
25
20

15

15

10

10

2
3 4
Millistrains

3 4
5
Millistrains

Figure 5.8. Stress-strain curves for Rand quartzite at various confining pressures.

Figure 5.8

1) Shear fracture.

2) Extension fracture.

3) Multiple shear fractures.

Figure 5.9

4) Extension fracture produced

by line loads.

5) Longitudinal splitting in
uniaxial compression.

Figure 5.9

Figures by MIT OCW.

9
8
1650

7
Strain (%)

685

500

4
3

3260

845

235

Figure 5.10

1
0

2490

0
0

1000

2000

3000

4000

5000

1- 3 (bars)

Figure 5.10. Stress-strain curves for Carrara marble at various confining

pressures. The numbers on the curves are confining pressures in bars.

Figure 5.10

15

Strain (%)

Figure 5.11

800 oC

10

500 oC
300 oC

25 oC

10

15

20

1- 3 (k.bars)

Figure 5.11. Stress-strain curves for granite at a confining pressure of 5 kilobars

and various temperatures.

Figure 5.11
Figures by MIT OCW.

Analyze this using a Mohr circle diagram:

| | = | n| + | 0|

| 0|
2

Figure 5.12
Figure 5.12
Figure by MIT OCW.

Mohr circle tangent to failure curve sample breaks.

= tan-1
2
2

n
Figure 5.13

Compressive stress negative

Compressive stress positive

Figure 5.13
Figure by MIT OCW.

 is the coefficient of friction.

= tan 1 ; 2 = 90 o

Figure 5.14
2 = 90o

Figure 5.14
Figure by MIT OCW.

0.0
0.6
1.0
lim

45
30
23
0

Since most rocks have a coefficient of friction of about 0.6, the normal vector to the
failure plane is typically 30from the direction of the least compressive stress. Another
way of saying this is that the failure plane is 30from the direction of the most
compressive stress.

Styles of Faulting

Faults are large-scale failure planes. Since the normals to failure planes are in the plane
containing the least compressive and the most compressive stresses and are typically at
30o from the direction of the least compressive stress, different styles of faulting can be

used to infer the directions of 1, 2, and 3.

The following diagram shows the deviatoric stresses associated with thrust faults, normal
faults, and strike-slip faults.

Total Stress

Figure 5.15
Deviatoric

Or

Deviatoric

Total

Or

Or
"Left Lateral"
Figure 5.15
Figure by MIT OCW.

"Right Lateral"

12.005 Lecture Notes 6

The coefficients of friction, , for a wide variety of rocks, are comparable, in the range
0.6 0.9. The only rocks with low coefficients of friction are clays, which are not stable
at the high pressures and temperatures characteristic of fault zones at depths of ~ 10 km
where earthquakes nucleate.

Figure 6.1
Note: the Navier-Coulomb failure law, || = || + 0, is also sometimes written

|| = || + S0.

For rocks, S0 ~ 0.1 - 4 kbars (10 - 400 MPa). How does this "breaking strength compare
with the "frictional stress?"
Plot stress predicted for the initiation of faulting as a function of depth, assuming that is
the lithostatic stress. (Recall that the "lithostatic stress gradient ~ 1 kbar/3 km.)

Figure 6.2

So, if Byerlee's law is correct, and if the stress in the lithosphere is not too far from
lithostatic, the shear stress on faults should be ~5 kbar at 15 km depth and about 250 kbar
at 700 km depth.

Figure 6.3

How can we test this prediction?

Model stresses associated with holding up mountains (e.g., Himalayas > 1.5 kbar)

Stress drops associated with earthquakes (usually ~ 3- 300 bars, very rarely > 1 kbar)

Work available from convection (dynamic "engine," <~1 kbar

Heat flow in fault zones (frictional heating rate ~ v => ~ 100 bars

Big discrepancies here - a frontier region of geodynamics.

Important factor - the role of pore fluid pressure in faults.

x3

x2
x1

Pressure in fluid

Figure 6.4
Figure by MIT OCW.

What is pressure in sample?

1 0
0

= 0 1 0
0
0 3

P=

( ) =

Tr
3

ii

i=1

Einstein summation:
If index repeated in a single term implicit sum.

ii

ii

i =1

T = n

Ti = ij n j

P=

[2 1 + 3 ]
3

Consider a common experiment in soil mechanics or rock mechanics: Apply uniaxial

stress 3 to a cylindrical sample confined by 1 . Continue increasing 3 until failure.

Figure 6.5

n of failure plane typically at ; 30 from 1 .

Empirical results: For a pore fluid pressure p, the Navier-Coulomb failure law becomes:
|| = |+p| + S0
(Note that because compressive pressure is positive, while compressive normal stress is
negative, the pressure decreases the amplitude of the "effective normal stress" |+p|.)
Graphically:

p
Figure 6.6

A qualitative explanation of the effect of pore fluid pressure is that the the fluid helps to
"support" some of the normal stress that is otherwise carried by solid grains. Consider a
simple model of a continuum made up of dry sand. Let fA represent the fraction of a
surface area that is made up of solid grain contacts. Then for a macroscopic normal
stress , the average (microscopic) normal stress at the solid grain contacts is /fA, since
the pore space in between can support no normal tractions. (Of course, in places the
actual value will be much larger than the average value.) If Admonton's law applies to
the contacts, then the microscopic shear traction needed to cause failure is /fA, and the
macropsopic shear traction is =.

If pore fluid pressure is now introduced, the pore fluid will support some of the
macroscopic normal stress, leading to a decrease to the normal stress at the grain
contacts.

Figure 6.7

Adding pore fluid pressure effectively "stretches" the n axis, giving, effectively, a lower
coefficient of friction (except that angles no longer work out!)

Figure 6.8

An alternative explanation for deep earthquakes is that the failure envelope bends over at
large n.

Figure 6.9

Finally, let's consider a medium that is anisotropic perhaps one which has a preexisting
fracture at an angle to the least principal stress. (For a preexisting fracture, the strength
S0 = 0.) Then if the double angle 2 is within the region shown, the rock will fail along
the preexisting fracture.

Figure 6.10

12.005 Lecture Notes 7

Newtons second law
For a point mass
F = ma

We can obtain F from a free-body diagram e.g., pendulum.

T
~

F
~
mg
~

Figure 7.1
Figure by MIT OCW.

Both F and a are vectors.

Fi = mai

i = 1, 2, 3

For a continuum, we can also construct a free-body diagram. Its easiest to do this
component-by- component.

Consider the following figure.

x1

33
x2
x3

13 +

33 +

33
x3

dx3

Figure 7.2
Figure by MIT OCW.

Consider forces on faces.

face

Traction (T3)

area

left

13

dx2 dx3

right

13 +

back

23

front

23 +

top

33

bottom

33 +

Force (F3) in x1 direction:

13
dx1
x1
23
dx2
x2
33
dx3
x3

13
dx1dx2 dx3
x1

dx2 dx3
dx1dx3
dx1dx3
dx1dx2
dx1dx2

13
x 1

dx1

 23
dx1dx2 dx3
x2
33
Force (F3) in x3 direction:
dx1dx2 dx3
x3

Force (F3) in x2 direction:

Body force: b3dx1dx2 dx3

Combining
13

dx1dx2 dx3 + 23 dx1dx2 dx3 + 33 dx1dx2 dx3 + b3dx1dx2 dx3 = a3dx1dx2 dx3
x1
x2
x3
Dividing through by V = dx1dx2 dx 3
13 23 33
+
+
+ b3 = a3
x1
x2
x3
Similar analysis gives
11 12 13
+
+
+ b1 = a1
x1
x2
x3
12 22 23
+
+
+ b2 = a2
x1
x2
x3
or
3

ij

x
j =1

or

ij
x j

+ bi = ai

i = 1, 2, 3

+ bi = ai

i = 1, 2, 3

(Einstein summation)

or
+ b = a

This three equations are known as the equilibrium equations, if ai = 0 .

To avoid accelerations, the stress tensor must satisfy equilibrium.
Aside By the continuum mechanics definition,
positive 11 extension
negative 11 compression
Geologist often use the opposite convention beware!

12.005 Lecture Notes 8

Assertion: most of the stress tensor in the Earth is close to "lithostatic,"
ij ~ -gd ij,
where is the average density of the overburden, g is gravitational acceleration, and d is
the depth of the point under consideration.
Consider the following table of the lithostatic pressure at various points of interest:

ocean ridge crest

abyssal plain
deep sea trench
base of crust
transition zone
Core-mantle bndry
Center of Earth

depth
(km)
2.5
4
12
30
600
2900
6400

Pa

bars

25 MPa
250
40 MPa
400
120 MPa 1.2 kbar
1 GPa
10 kbar
20 GPa 200 kbar
140 GPa 1.4 Mbar
360 GPa 3.6 Mbar

Assertion: Typical "tectonic" stresses have magnitudes in the range 0.3 - 300 MPa (33,000 bars).
Conclusion: In considering tectonic problems, it is usually useful to subtract some
isotropic reference state of stress. There are 2 common choices - the lithostatic stress,
which leaves the nonlithostatic stress tensor, and the pressure field, which leaves the
deviatoric stress field. But since we need to know the actual stress tensor in order to
calculate the pressure, which requires knowing the solution to the problem in advance, it
is most common to use the lithostatic stress field as the reference.
In order to place these comments in context, let's look at the situation described in T&S
problem 2-6, where a continent is in isostatic equilibrium with an ocean. To address the
possible state of (nonlithostatic) stress in the region, we will follow an iterative approach.
The "algorithm" can go as follows:
1) Assuming lithostatic stress, calculate the horizontal normal stress as a function of
depth beneath both the continental and the oceanic structures.
2) Evaluate the "reasonableness" of this assumption by making a free body diagram by
isolating a (thin) vertical "pill box" of material at the continent-ocean boundary and
calculating the total forces acting on its sides, given this assumption.

This can be done by integrating the normal traction as a function of depth to determine
the total normal force (per unit length into the picture) acting on each of the two vertical
faces of the pill box:
Moho

Fx length =

dy
n

surface

The normal traction exerted by the continent on the pill box is everywhere greater than
the normal traction exerted by the oceanic structure on the pill box. The total force on the
pill box is just the area shaded with horizontal hatchure in the figure below.

Figure 8.1
Figure by MIT OCW.

4) Since our assumptions have led to an unbalanced force on the pillbox that would lead
to an infinite acceleration as the thickness of the pill box shrinks, something must be
wrong with our assumption. So we need to modify our model to decrease the magnitude
of the normal traction exerted by the continent, increase the normal traction exerted by
the oceanic structure, or both.
5a) If we "fix" our problem in the continent, we require a nonlithostatic horizontal
extension.
5b) If we "fix" our problem in the oceanic region, we require a nonlithostatic horizontal
compression.
5c) Of course, another solution would be to have horizontal compression in the continent,
with even more horizontal compression in the oceanic area.
Note: Isostasy and lithostatic stress cannot exist simultaneously! Areas of thickened
crust tend to have a state of stress that is more extensional than nearby regions of thin
crust.
Examples: Tibet, Southern California Transverse Ranges.

Sources of tectonic stress are of great interest to geodynamics, paritcularly if variations in

stress can be interpreted in terms of interesting processes. Recall, from equilibrium,

ij , j + f i = 0
If we can ignore body forces, then, for example,

xx

= xz
x
z

Figure 8.2
Figure by MIT OCW.

Example:
Stress transition near back-arc basin

Figure 8.3
Figure by MIT OCW.

Measurements:

Flat jack null measurement

Figure 8.4
Figure by MIT OCW.

Given Tn across cut

Figure 8.5
Figure by MIT OCW.

Related techniques overcoring

Hydrofracture
Problems Near surface heterogeneity
Difficult to do remotely
Only gives present-day stress
Other techniques direction only
Cheap or free!

Borehole breakout relies on stress concentration around drill holes; spalling of

borehole walls

Figure 8.6
Figure by MIT OCW.

Changes in stress direction

Example subduction zone

Figure 8.7
Figure by MIT OCW.

Fore-arc: Compression xxdev < 0

xxdev > 0
Back-arc: Extension
mantle flow required
xx xz
Equilibrium:
+
=0
x
z
{Outer rise extension near surface due to bending (discussed later)}
Other examples of change in stress orientation:
US Mid continent E-W Compression
Basin and range E-W extension
California rotation of dev. Stress near faults
Earthquake focal mechanisms

Figure 8.8
Figure by MIT OCW.

Types of faults

Figure 8.9
Figure by MIT OCW.

Figure 8.10
Figure by MIT OCW.

Note: Angle of 1 to fault plane depends on fault properties.

12.005 Lecture Notes 9

Stress Rotation, after Zoback et al, 1987
The principal stress directions are observed to rotate in the vicinity of the San Andreas
fault (SAF). In the far-field (e.g., Nevada), the maximum compressive stress is oriented
at an angle to the fault trace ~ 55. But in the near field, this angle, now called , is
close to 85. (Note that this is the same as the angle between the least compressive stress
and the normal to the fault plane, the angle conventionally used in Mohr circle analysis.)
Assume that in the far-field stress has I = -68 MPa, II (assumed vertical and lithostatic)
= -136 MPa, and III = -204 MPa. (What style of faulting would this cause?)
Assume that the fault has strength C0, so that 12 becomes smaller approaching the fault.
(How could this happen, given Newton's second law?).
Also assume that the normal stress across the fault maintains its far-field value (22 in the
coordinate system fixed to the fault, as shown in the diagram). Assume that c remains
the same. (Does this agree with the style of faulting and the folding near the SAF?)
Then, from a Mohr's circle construction, it is straightforward to obtain a relation between
and .

Figure 9.1
Figure by MIT OCW.

Possible cause of weak faults

Preexisting fracture

Figure 9.2
Figure by MIT OCW.

Clay low

Figure 9.3
Figure by MIT OCW.

Pore fluid pressure model of fault weakening

Figure 9.4
Figure by MIT OCW.

Fault zone highly permeable

Darcy flow heat flow
Permeability conductivity
pT
Given source of water
High permeability high p

Figure 9.5
Figure by MIT OCW.

Question:
What are implication for stress direction in fault zone?
Is low in fault zone consistent with large outside?

Figure 9.6
Figure by MIT OCW.

12.005 Lecture Notes 10

A house built upon sand

Figure 10.1
Figure by MIT OCW.

Theory Sand follows Amontons Law: = n

(Mohr Coulomb)

Modify for pore fluid: = n

p f
Need to express and n on any arbitrary plane develop stress tensor (symmetric)

T = n
or

Mohr circle construction

Ti = ij n j

n = T n


Figure 10.2
Figure by MIT OCW.

Application to accretionary wedge

Figure 10.3
Figure by MIT OCW.

Assume

On verge of failure everywhere

, * b , b

Trigonometry relates , , , b ,
Pore Fluid

p f w gD

zz w gD

Often:

zz ; g(1 )z cos( )
Sandbox tectonics

Rheology Brittle failure friction dominates

[comp positive ]

Figure 10.4
Figure by MIT OCW.

Applications:
A)

Figure 10.5
Figure by MIT OCW.

Figure 10.6
Figure by MIT OCW.

B)

Figure 10.7
Figure by MIT OCW.

Figure 10.8
Figure by MIT OCW.

Consider stress only only variable for brittle failure

neglect strain, velocity, inertia,
Possible paradoxes:
Type A) int = repose strong steep
Type B) strong flat

Viscous fluid, type A&B weak falt

Complicating factors:
Pore pressure; based decollement

12.005 Lecture Notes 11

Sandbox tectonics Simple theory (after Dahlen 84)

Figure 11.1
Figure by MIT OCW.

Assumptions
Coulomb failure = (1 ) n

Based decollement may have different b (or b )

Material is on the verge of failure everywhere
Mohr circle construction
Inertia negligible F = ma = 0


4 parameters , , b ,b (?), 2 variables
Neglect complications of being under water (?)

(?) fluids in count?

Figure 11.2
Figure by MIT OCW.

Figure 11.3
Figure by MIT OCW.

surface slope
basal slope
0 angle between III and surface (or more fundamentally, I and z axis)
b angle between I and base

Equilibrium equations: ij, j + f i = 0

xx xz
+
g sin = 0
x
z
xz zz
+
+ g cos = 0
x
z
At z = 0
xz = 0

zz = 0
For now, ignore effects of pore fluid pressure will go back to this later.
Dry sand

Figure 11.4 Physical space

Figure by MIT OCW.

Figure 11.5 Mohr space

Figure by MIT OCW.

Argue:
No strength: s0 = 0 no scale length

= 0 & III functions of z only

x
0 is a constant

Plug in to equilibrium equations:

zz = gz cos

satisfies equilibrium equations and boundary conditions

xz = gz sin
How to relate 0 and ?
If stress is critical everywhere, can use Mohr circle.
+ III
III
Say c = I
r = I
2
2
Geometry of Mohr circle

csc =

c
r

tan 2 0 =
sec 2 0 =

xz

zz c
r

zz c

Equilibrium

xz
= tan
zz

Solving 4 equations tan =

tan 2 0

csc sec 2 0 1

The March of Science revisited (after S. V. Panasyuk)

To introduce the notation, let's draw a sketch:

Figure 11.6
Figure by MIT OCW.

We can use the well known Mohr's circle technique to describe the wedge behaviour.
Given coefficients of friction for the base b and the , and the principal stresses in the
wedge ( 1 , 3 ), we can construct the Mohr's circle diagram and the Coulomb failure
criterion as:

Figure 11.7
Figure by MIT OCW.

Using this diagram it is possible to determine the orientation of the principal stresses with
respect to the base ( b1 and b2 ) and the orientation of the fault planes, i.e. the slip lines
in the wedge () with respect to the principal directions ( 1 , 3 ). Since motion on the slip
planes participates in the creation of the actual slope of the wedge as well, this slope
might differ from that of the slip planes. To analyse the behaviour of the wedge in front
of a bulldozer, we derived equations to connect all angles ( , , 0 , b , , b ) with each
other:
i) from the picture of
relationship between angles:

the wedge (Fig. 1), we have the purely geometrical

+ = b 0

(1)

ii) from the equations of static equlibrium and boundary conditions (see previous
lecture notes):

tan 2 b
tan 2 b

(2)
= b , or b = arctan
tan( b ) =
csc sec2 b 1
csc sec 2 b 1
tan( ) =

tan 2 0
,
csc sec2 0 1

tan 2 0

or = arctan
csc sec 2 0 1

(3)

Note that in the second form of (2) and (3), angles (e.g. ) are given directly, which will be
useful below.
iii) from reasonable sense:
0 +
(4)

Having a system of 4 equations in terms of the 4 unknowns guarantees a solution, but

does not make apparent the "obvious" connection between angles. There are at least two
ways to solve this system and present results in a viewable form: the physical approach
(choosen by Nature and by the TA) and the mathematical approach (choosen by the
Princeton Professor).
I.
From a physical point of view the decollement dip (the base slope = ) is a more
fundamental quantity then the surface slope (the wedge slope = ). Together with the
material properties (coefficients of friction and b ) it determines the surface slope. For
example, suppose a sand wedge ( = tan ( ) = tan (300 )) is placed on sloping, slippery

ground ( b = tan ( b ) = tan (10 0 )) which is inclined at =35 to the horizon. Let's solve
the system of equations graphically to get all possible values for the other angles
, 0 , b :
Equation 1) relates and 0 : = ( b ) 0 . Its graph is:

Figure 11.8

Note: in these coordinate axes, the slope of this straight line is 45.

tan 2 b

Equation 2) determines b values from b = arctan

. Its graph is:
csc sec2 b 1
b 30
20

b = 100

10
0

b1 peak

b2

-10
-20
-30

-80

-60

-40

-20

20

Figure 11.9

40

60

80 b

Note: This function reaches peaks at

peak = ( 4 / 2) and b peak =

The value b intersects the curve at two points, b 1 and b 2 .

tan 2 0

. Its graph is:

Equation 3) defines in terms of 0 as = arctan
csc sec2 0 1
30
20
10

peak

peak

-10
-20
-30

-80

-60

-40

-20

20

40

80 0

60

Figure 11.10

Note: This function reaches peaks at peak = ( 4 / 2) and peak =

Since the first and third graphs can be plotted on the same axes, the values of and 0
which satisfy both equations are the coordinates of the intersection points of two curves.
Since the second and third graphs have the same dependence, we can use "combined"
axes. Now we are ready to find graphically all possible values for other angles , 0 , b
by plotting all three graphs in one:

Figure 11.11
Figure by MIT OCW.

From the graph (for = 30, b = 10, =35):

From the first intersection line b = 100 with the curve ( ), b 1 50 , and
( b1 ) 300 . Following down the 0 line (45 line in the equal unit axes) to its

intersection with the ( ) curves gives 01 110 and 1 190 . A similar approach
for the second intersection line b = 100 with the curve ( ), i.e. the second value
b 2 750 gives 02 150 and 2 250 .
II. From the mathematical point of view given in the approach is to choose a test
value for (from the stability diagram), determine 0 from the curve intersection for
tan( ) . The stability diagram is a function of the surface slope angle with respect to
the dip angle , and its graph can be obtained in the following way:
From equation 2) we can get two values of b , which are b1 and b2 for the given
friction coefficients in the base and wedge.
From equation 3) by varying the angle of 0 (say from /2 to /2) we can get the range
of and substitute all of these values into the equation 1) to get a formula for :

12 = b1,b2 0 ( 0 )
under the restriction 4).

12.005 Lecture Notes 12

Displacement Gradients
Quantitative description:
Suppose
D D'

P( xi ) P '( xi + ui )
Q( xi + dxi ) Q '( xi + dxi + ui + dui )

x3

D'

du i Q'
+
ui

Q
P

ui

P'

x2
x1
Figure 12.1
Figure by MIT OCW.

Suppose:
The deformation is continuous.
u
The first derivative i are continuous and very small.
x j
Then chain rule
u
u
u
u
dui = i dx1 + i dx2 + i dx3 = i dx j
x1
x2
x3
x j

Note:

ui
relates two vectors dui and dx j and is therefore a second rank tensor.
x j
u
du1 1

x1

u
du2 = 2

x1

du3 u3
x1

u1
x2
u2
x2
u3
x2

u1
dx1
x3
u2
dx2
x3
u3
dx
x3 3

For the Ventura Basin results shown, for the 2-D solution, for the stations HOPP-HAPYSNP,
ui 0.2 0.45
=
106 each year

x j 0.08 0.48

Note: We have no sensitivity to rigid body translations i 0 .

x

What about rotations?

x2

B'

A
A'

x1
Figure 12.2
Figure by MIT OCW.

u2
= tan
x1
u1
=
x2
1 u1 u2

12
2 x2 x1

So part of this displacement gradient tensor just gives rigid body rotation.

Rewriting
ui 1 ui u j 1 ui u j
=
+

x j 2 x j xi 2 x j xi
where

ui u j
+
= ij This is strain. It is symmetric.
x j xi

six independent components

ui u j

= ij This is rotation. It is antisymmetric.

x j xi

three independent components

For Ventura Basin, each year

0.20

0.26

0.19
0

106
ij =

0.26 0.48

ij =
0.19

Interpretation of ij

x3
dx2
Before
After
dx2 + du2

Figure 12.3
Figure by MIT OCW.

For example,

1 u2 u2 u2
+
=
2 x2 x2 x2

22 =

u
dx2 + du2 = 1 + 2 dx2 = (1 + 22 ) dx2
x2

Ventura Basin shows

N-S direction shortening
E-W direction lengthening

x2

Change in shape

x3

x3

Q'

''
P'
P

x2

Figure 12.4
Figure by MIT OCW.

'  tan ' =

u3
x2

''  tan '' =

u2
x3

'+ '' =

u3 u2
+
= 2 23
x2 x3

where 23 is distortion of x2 , x3 axes.

2 23 23 It is called engineering strain.

'

x2

12.005 Lecture Notes 13

Measurement of Displacement Gradient Tensor
Techniques of measurement of displacement gradient tensor:
Leveling
Triangulation
Trilateration
Very Long Baseline Interferometry (VLBI)
Global Positioning System (GPS)

Leveling
Telescope
(level)

Benchmark
Triangulation
Benchmark

Protractor

Telescope
Corner
Cubes

l2

Trilateration

l1

2li = cti

Laser
Quaser
VLBI
x= ct

GPS (Global Positioning System)

Figure 13.1
Figure by MIT OCW.

Sensitivity:

technique

angle

distance

height

orientation

Yes

leveling
triangulation

Yes

trilateration

Yes

Yes

VLBI

Yes

Yes

Yes

Yes

GPS

Yes

Yes

Yes

Yes

Consider plane strain 33 0 and a four-benchmark network

x2

x1
Consider the following strain tensors:

0 0

Case 1. 0 0
0 0 0

0 0

Case 2. 0 0
0 0 0

0 0

Case 3. 0 0
0 0 0

2 0 0

Case 4. 0 0 0
0 0 0

Can they be observed using triangulation?

Can they be observed using trilateration?
Which angles would change?
Which line lengths would change?
What would the change in line length be?

is a tensor, so ' = T , just as for stress

cos sin 0

= sin cos 0
0
0
1

x2
x2'

x1'

Figure 13.2
Figure by MIT OCW.

x1

Evaluating this for plane strain is straightforward:

11 ' =

11 + 22

11 22

cos 2 + 12 sin 2
2
2
+

22 ' = 11 22 11 22 cos 2 12 sin 2
2
2

12 ' = 11 22 sin 2 + 12 cos 2
2
+

Note: Shear strains are tough to measure.

These relationships are useful in relating longitudinal strains

l
l

l
l

to the total strain tensor.

can be easily measured using trilateration (10s km scale), strain gauge (10s mm

scale).

For example, delta rosetle

x2

x1'

x1''

30o

60o

Coils

30o
60o

60o

x1

Figure 13.3
Figure by MIT OCW.

Equilateral triangle of transducers records elongations in x1, x1, and x1directions.

These can be related to .

Summary of infinitesimal strain:

B
A

B'
A'

C'

Figure 13.4
Figure by MIT OCW.

The above figure shows translation and rotation.

l
Both length change and angle change depend on orientations.
l
l
>0
Solid lines
l
l
Dashed lines
<0
l
BAC decreases
ABC increases

Strain tensor:

11 12 13
ij = 12 22 23

13 23 33

diagonal

l
l

off diagonal

for lines along axes

for deformation of axes

The strain tensor depends on the orientation of the coordinate system.

Geologic Strain indicators pebbles elliptical

12.005 Lecture Notes 14

Finite Strain
The "paradox" of finite strain and preferred orientation pure vs. simple shear
Suppose:
1) Anisotropy develops as mineral grains grow such that they are
preferentially oriented with the "easy slip" direction facilitating deformation.
2) New grain growth is determined by the state of stress. (What else
would a mineral grain "know" about?)
3) The state of stress for pure shear and simple shear are identical pure
and simple shear stress differ only by a rotation of 45 in coordinate system.
4) In pure shear there is only one slip direction, whereas in pure shear, 2
systems could be active.
Paradox: If the stress state is the same, how does the rock "know" to have 1 preferred
orientation in simple shear, but not in pure shear?

Figure 14.1

Up to now, we have considered infinitesimal strain, where eij = (ui/xj + uj/xi)/2 <<1.
For the kinds of deformations common in structural geology, as well as for the deforming
telephone book example shown in class, this condition of "infinitesimal" deformation no
longer holds.
Life becomes complicated. In particular, it's important to know whether one is coming or
going:

Reference frame:
Consider a medium undergoing motion & deformation, with the displacement a (vector)
function of position.
B

x3
A

b
x2

x1

Figure 14.2

bj = f(ai)
or
bj = bj(ai)
Since the mapping ai <> bj is one-to-one, we could also say
aj= aj(bi).
It turns out (consider the telephone book example in class) that the description of strain
depends on whether the reference frame is:
Lagrangian fixed to particles (the "a" frame)
ui = bi(aj) - ai
Example: elasticity -- the particles return to where they started

Eulerian fixed to space (the "b" frame)

ui = bi -ai(bj)
Example: fluid mechanics, geology -- no idea where the particles started or where
they will end up; no stable frame fixed to the materials.
Finite strain depends on whether the reference frame is fixed to the particle or to "space:"
Lagrangian strain: Eij = (ui/aj + uj/ai + ul/ai ul/aj)/2
Eulerian strain: eij = (ui/bj + uj/bi - ul/bi ul/bj)/2
When strain is small, the 3rd term is small2, and both strains converge to the infinitesmal
value.
Physical interpretation:
diagonal elements - e.g., elongation, given by ds = (1 + E1) ds0.
x2
ds
dso

da1 db1

x1

Figure 14.3

For the Lagrangian system, E1 = (1 + 2E11)1/2 -1

off diagonal >change in angles of coordinate axes:
x2

x2

db'

da'

= /2

db

da

x1

Figure 14.4
sin a = 2E12/[(1 + E11) (1 + E22)]1/2
These quantities go to the infinitesimal values when Eij << 1.

x1

Because eij Eij, the principal values & principal directions of the Lagrangian and
Eulerian strain tensors differ. We will discuss these later.

Rotation
Consider two particles initially separated by dai (note: not just a function of strain tensor)
which undergo a displacement uj, and are then separated by dxi.
Lagrangian:
dx i = dai +

ui
u
daj = ij + i da j
aj
aj

This can be written, including the finite strain tensor,

1 u u u u 1 u u u u
dx i = ij + i + j + s s + i j s s daj
2 a j ai ai a j 2 a j ai ai a j

= ij + ij + ij da j

ij Lagrangian rotation tensor

if E ij = 0 rotation alone
Similarly, for the Eulerian case:
1 u u u u
ij = i j + s s
2 bj bi bi b j

dai = ij + eij + ij dx j

for small deformation: ij rotation needed to bring a fiber dai into position.
For large deformation: no easy way to separate into pure rotation and pure shearthe
two are intimately connected.
THIS IS OF UTMOST IMPORTANCE IN PROBLEMS LIKE THE DEVELOPMENT OF ROCK FABRIC.
ALSOREGIONAL TECTONICS.

Consider homogeneous strain

Figure 14.5
Figure by MIT OCW.

This holds if
Eij =

r r
1
( lm l m ij ) Eij ( Ri )
2
Ri R j

l
= constant R & r linearly related
 
j

Define Ri x, y ri x ', y '

In general
x ' = ax + by
y ' = cx + dy
a 2 + c 2 1 ab + cd
0

2
2
Eij = ab + cd
b + d 1 0
2

0
0
0

Examples
x ' = kx
y' = y

Figure 14.6
Figure by MIT OCW.

x ' = k1x
y ' = k2 y

Figure 14.7
Figure by MIT OCW.

x ' = kx
1
y' = y
k

Figure 14.8
Figure by MIT OCW.

x ' = x + 2sy
y' = y

Figure 14.9
Figure by MIT OCW.

Figure 14.10
Figure by MIT OCW.

Unstrained
1
x = 2 (dx ' by ')
h
1
y = 2 (cx '+ ay ')
h

Strained
x ' = ax + by
h2 = ad bc

y ' = cx + dy

Principal Axes A, B
Roots of
R 4 (a 2 + b2 + c 2 + d 2 )R 2 + h4 = 0
tan 2 =

2(ab + cd )
a + c 2 b2 d 2

tan 2 ' =

2(ac + bd )
a + b2 c 2 d 2
2

No Length Change
tan = [(1 B ) / ( A 1)]
2

B( A2 1)1/ 2
A(1 B 2 )1/ 2
relative to principal axis

1/ 2

tan ' =

relative to reciprocal axis

Maximum Shear
b +d a c
2(ab + cd )
relative to coord. axis
tan 2 =

tan ' = ( B / A)
relative to principal axis

Principal axis of infinitesimal

strain
+'
0 =
2

12.005 Lecture Notes 15

Elasticity
So far:
Stress angle of repose vs accretionary wedge
Strain reaction to stress but how?
Constitutive relations

ij = ij ( kl ) ; ij = ij ( kl )

For example,
Elasticity
Isotropic
Anisotropic
Viscous flow
Isotropic
Anisotropic
Power law creep
Viscoelasticity

Trade offs:
simplicity

realism

constant

variable

isotropic

anisotropic

elastic, viscous

viscoelastic

history

history dependent

independent

Tensors

Most physical quantities that are important in continuum mechanics like temperature,
force, and stress can be represented by a tensor. Temperature can be specified by stating
a single numerical value called a scalar and is called a zeroth-order tensor. A force,
however, must be specified by stating both a magnitude and direction. It is an example of
a first-order tensor. Specifying a stress is even more complicated and requires stating a
magnitude and two directionsthe direction of a force vector and the direction of the
normal vector to the plane on which the force acts. Stresses are represented by secondorder tensors.

Tensors are quantities independent of coordinate system.

Coordinate System

x3'

x3

Direction Cosines

x2'
x2
x1
x1'
Figure 15.1
Figure by MIT OCW.

x1

x2

x3

x1'

11

12

13

x2'

21

22

23

x3'

31

32

33

 ij = cos ij
where ij is the angle of primed to original.
xi ' = ij x j
xi = ji x j '

ij =

xi ' x j
=
x j xi '

Tensors:
a. 0th order (scalar) quantity dependent only on position
b. 1st order (31 components)

Ai ' = ij Aj

c. 2nd order (32 = 9 components) Aij ' = is jk Ask

d. 3rd order (33 = 27 components) Aijk ' = is jt kp Astp
e. 4th order (34 = 81 components) Aijkl ' = is jt kp lq Astpq
Linear Elastic Solid

ij = cijkl ekl
cijkl is elastic modulus tensor
cijkl = constants ( history, displacement, time)

34 = 81 components

ij = ji , eij = e ji cuts to 36
Strain Energy
1
U
U = ij eij ij =
2
eij

Write in terms of powers of eij

U = + ij eij + ijkl eij ekl

0 (to avoid spontaneous expansion, contraction)

U
= ( ijkl + klij ) ekl
eij

cijkl = cklij 21 individual components.

21 components data fitting ugh!

Lots of available information, but lots of hard work.
Reality 21 components (triclinic)

Simplicity derive in homework

Isotropic
cijkl = ij kl + ( ik jl + il jk )

and are Lame parameters.

ij = ekk ij + 2 eij
Can also express
eij = Sijkl kl

Compliance tensor
Aside
e11 = S1132 32

triclinic, trigonal

2
1

Figure 15.2
Figure by MIT OCW.

Sheer leads to lengthening.

Isotropic can relate cijkl and Sijkl directly

ii = ekk ii + 2 eii = ( 3 + 2 ) eii

2 eij = ij

ij
kk
2 + 3

Conventional moduli:

1. Hydrostatic comp.

Figure 15.3
Figure by MIT OCW.

ij = pij
ii = pii = 3 ekk + 2eii
= 3 p = (3 + 2 )eii

VP
p
=
K
V
eii

where K = + 2 / 3 is bulk modulus.

2. Uniaxial stress

Figure by MIT OCW.

11 = T
other ij = 0

2e1 = T

2 + 3

T
E (sometimes Y )
e1
where E =

(2 + 3 )
is Youngs modulus
+

Hooks law:
T = Ee
e22 e33
=

e11 e11

2e22 =

This is called Poissons ratio.

2 + 3

11 =

2( + )

= e11 + e22 + e33 = e11 (1 2 )

fluid: 0

1
2

most material: = 0.2 0.3

1
=
=
4

It is Poisson solid.

steel: ; 0.3 0.33

seismically measured
+ 2
vp =

vs =

compare v p , vs discriminate rock types

3. Simple shear

12 = 21 =
12 = 2e12 = 2Ge12
where G is shear modulus.
Note: Among , , K, , E, G only two are independent.

Useful forms:

11 = ( + 2 )e11 + e22 + e33

22 = e11 + ( + 2 )e22 + e33
33 = e11 + e22 + ( + 2 )e33
or
e11 =

11
E

e22 =
e33 =

22

11
E

11
E

33
E

22

22

33

33
E

12.005 Lecture Notes 16

Simple Example: Uniaxial Strain
Uniaxial strain is excellent approximation, but not exact.
Sediment loading erosion:
x1

x3
x2
Figure 16.1
Assume e22 = e33 = 0, e11 0

11 = ( + 2 )e11
22 = 33 = e11 =
11 =

+ 2

11 =

11

(1 )Ee11
(1+ )(1 2 )

= gd 0

0
1
3
0

0 = uniaxial

For = 0.25, 11 = gh ; 0.27 kbar/km (crustal rock)

22 = 33 = 11 ; 0.1 kbar/km deviatoric stress

3

0.27
0
0 0.15
0
0 0.12
0
0

0.09
0 = 0
0.15
0 + 0
0.06
0
0
0
0
0.09 0
0
0.15 0
0
0.06

Unloading:
Assume initially lithostatic

11 = 22 = 33 = gd
4

0
0
0 9

0 gd
0 = 0

2
0
0
gd 0

2
9

4
0

9
0 gd + 0

2
0

thrust faults

popups
give direction of compressional stress

Assume 2 < 3 tectonic stress

Type of faulting: (compression negative)
x1

x3 (E-W)
x2 (N-S)
Figure 16.2

2 < 3 < 1 thrusting on E-W plane

2 < 1 < 3 strike-slip
1 < 2 < 3 normal faulting on N-S plane
Stress increment from burial

4
9

0 gd

4

9

= 0

0
1
3
0

0
9

0 gd = 0

1
0

0
5
9
0

4
0

9
0 gd + 0

5
0

0
2
9
0

0 gd

2
9

Near surface in Ventura Basin, thrusting occurs. Stress tensor could look something like:
0 0

0
0 0

"tectonic" stress
0 1
1+
p=

Assume tectonic stress independent of depth.

For simplicity, assume

1
=
p=
2
2

Total deviatoric stress is

4

9
0

0
2
9
0

1
0

2
0 gd 0

2
0

Normal faulting occurs if

0
1
2
0

4
2
2
gd + < gd < gd
2 9
2 9
9
2
gd <
3

12.005 Lecture Notes 17

Special Cases
In general ij = eiiij + 2eij
Plane stress: (e.g., zz = 0, xz = 0 )
xx

= xy
0

p=

xy 0

yy 0
0

xx + yy
3

2 xx yy

dev = xy

zzdev =

xx + yy
3

Solving for strains:

1
( xx yy )
E
1
eyy = ( yy xx )
E
exx =

ezz =
exy =

xy
2

( xx + yy )

xy
2 yy xx
3
0

xx yy

3
0

Plane strain: (e.g., ezz = exz = 0 = zz dev xx dev = yy dev )

xx
= xy

0
xx + yy

xy
yy
0

p = zz
xx yy

dev = xy

xy
yy xx
2
0

1+
xx (1 ) yy
E
1+
eyy =
yy (1 ) xx
E
exx =

exy =

xy
2

Note:
Plate stress

2
+ 2

Plane strain
E
1 2

Plane stress: (e.g., 33 = 0)

11

22
Figure 17.1
Figure by MIT OCW.

1
( 11 22 )
E
1
e22 = ( 22 11 )
E
e11 =

e33 =

( 11 + 22 )

Tectonic Stress Often plane stress is useful (3D 2D).

Assume mostly in plate stress
Assume
zz = gz (lithostatic)

xx = gz + xx
yy = gz + yy

The non-lithostatic (as opposed to deviatoric) stress:

zz ' = 0

xx ' = xx
yy ' = yy

plane stress

Assume for simplicity xx = yy =

e11 = e22 =
e33 =

(1 )
compression for compression
E

extension for compression

dilation = (2 4 )

Look at the change in lithostatic pressure at a depth h h + h = h(1 2

gh ( + )g(h + h)

Plane stress, 1 = 2 =

Figure 17.2
Figure by MIT OCW.

1 = 2 =
3 =

(1 )

= 1 + 2 + 3 =

Now to conserve mass,

(2 4 )
2(1 2 )
=

E
E

V
=
=
V

New lithostat:

zz = ( + )g(h + h) + gh
2
= (1+ )g(1+ ) h gh
E

= ( +

2
) gh
E

zz 2(1 )
=
gh

Use

= 0.25
E = 100 GPa = 1 Mbar
h = 100 km

= 3 Mg/m 3 = 3 g/cm 3
zz 2(1 )
gh 5%
=
E

12.005 Lecture Notes 18

Dislocation in Elastic Halfspace Model of the Earthquake Cycle
Interseismic: Slip below depth D

xy

5 10 15
-15 -10 -5

X/D

y
-3

Displacement

X/D
Strain

Figure 18.1
Figure by MIT OCW.

Coseismic: Region above D catches up

xy

5
-15 -10 -5

y
-3

0
X/D

Figure 18.2
Figure by MIT OCW.

10 15

X/D

Displacements from Earthquakes, Fault Slip, etc

Consider a strike-slip fault with displacement S, independent of depth A screw
dislocation i.e., a slip discontinuity. Original ring (dashed) becomes helix (solid).

x3
ce
rfa
Su ault
f

Fault
displacement S

o
ati
c
o
sl
s
Di axi

x2

S/2
S/2

x1

Figure 18.3. Section across a mathematical model of a transcurrent fault.

Figure by MIT OCW.

Dislocations are used to describe defects in crystals, as well as fault motions. A crystal
disrupted by a screw dislocation is shown in the figure below.

9
10

b*
5

1
4

2
Figure by MIT OCW.

Figure 18.4. A screw dislocation in a cubic lattice constitutes a deformation that is

out of the plane of atoms illustrated. The two atoms denotes by solid circles are
essentially part of a second plane. The Burgers circuit indicated by the numbered
steps naturally moves into this second plane. Therefore in order to close the
circuit the Burgers vector b* must be perpendicular to the plane of atoms shown.

Types:

Edge

Screw

Figure 18.5
Figure by MIT OCW.

b*

10

5
8

10

5
8

Figure 18.6 Side view of an edge dislocation in a cubic lattice.

Figure by MIT OCW.

Dislocation motion helps crystals deform (dont all have to slip at one time). Also helps
Earth deform!

D
S

E
A

Figure 18.7 A dislocation.

Figure by MIT OCW.

Figure 18.8 Upper group: Slip by propagation of an edge dislocation EE. Lower
group: Slip by propagation of a screw dislocation SS.
Figure by MIT OCW.

Displacement (meters)

0
0

10

Distance from fault (km) NE

Figure 18.9 Displacement as a function of distance from a transcurrent fault.

Figure by MIT OCW.

12.005 Lecture Notes 19

Stress and strain from a screw dislocation
x3
ce
rfa t
u
S ul
fa

Fault
displacement S

o
ati
c
o
sl
s
Di axi

x2

S/2
S/2

x1

Figure 19.1
Figure by MIT OCW.

Need to get traction = 0 at surface. First consider medium.

Assume:
u1 = u3 = 0

u2 =

S
2

u = 0 no compression, only shear

Symmetry cylindrical coordinates, r, , z
with z parallel to x2 axis;

Solution is z =

ur = u = 0

uz S
=
r 2 r

We can get ij by coordinate transformation.

How to get traction on surface?

Trick image dislocation

x3

2D
Image dislocation
Surface

Actual dislocation

x1

Figure 19.2
Figure by MIT OCW.

Solution for matched image dislocation is whole space gives i3 = 0 on surface of

space!
Shear strain at surface:
x3
2D

Surface

Figure 19.3
Figure by MIT OCW.

r
x1

r 2 = x12 + x32
S

From each dislocation z =

2 r

Rotating strain tensor

12 = z cos = z

x3
r

At surface

2D x3
SD
S x3
=
=
+
2
2
2
2
2

2 x1 + x3
2D x3 + x12 (D + x1 )

D
12

where

x3

is actual dislocation

x + x32
2
1

2D x3

(2D x ) + x
2

2
1

is image dislocation

Displacement
u2 =

x1

D
12

dx1 =

x
S 2
1 tan 1 1
D
2

Aside slip discontinuity objectionable?

12 along x1 = 0 as x3 0
(stress singularity at tip of fault)

Alternative model
0
Apply uniform 12

Cut 0 x3 D, set 12 = 0

1/ 2

2
x1
S x1
u2 = 1+ 2
2 D
D

Displacement (meters)

6
1
3.5

0
0

10

Distance from fault (km) NE

Figure 19.4
Figure by MIT OCW.

Virtually indistinguishable!

St. Venants principle

Elastostatics if boundary tractions on a part S1 of the boundary of S are replaced by
statically equivalent traction dist, effects on stress dist are negligible at pts whose
distance from S1 is large compare to size of S.

Usual context long beam under end load (non-uniform)

See's average end load

Figure 19.5
Figure by MIT OCW.

Apply to loading 1/2 space

Figure 19.6
Pt source approximates in seismology.

12.005 Lecture Notes 20

Plates
Rock rheology function of T; T increases with depth.
Thin region near surface remains elastic on geologic times. Below this, mantle behaves as
a viscous fluid.
Plate theory:
Assume plate thickness h  L, with L a characteristic length scale
3-D equations simplify
Applications:

Seamount Loading

Subduction Zone
Fore-arc bulge
Trench

Folding

Laccolith
Magma

Figure 20.1
Figure by MIT OCW.

Main Feature

z
Fibers compressed

Neutral surface
}

Fibers extended

Figure 20.2
Figure by MIT OCW.

Fiber stresses large compare to tractions applied to surfaces.

Neutral surface smooth and centered

Kirchoffs assumption approximate, but useful.

yy = yx = yz = 0
h
2

xx , zz linearly through y

h
2

or
every straight line originally perpendicular to neutral
plane remains straight and perpendicular

More accurate formulations lead to nonlinear coupled equations, not solvable analytically.

To derive plate equations consider the following figures:

Moments

Surface Loads
q

Shear Forces

End Loads
V

Figure 20.3
Figure by MIT OCW.

x+dx
q(x)

w+dw

w
P
M

V+dV
V

Figure 20.4
Figure by MIT OCW.

V shear force / unit length

P horizontal force / unit length
M moment / unit length
q force / unit area

M+dM

Force balance vertical

qdx + dV = 0

dV
= q
dx

Torque balance (+ counterclockwise)

dM Pdw Vdx = 0
dM
dw
=V + P
dx
dx

Next relate M to w.

y
M
h/2

-h/2

Figure 20.5
Figure by MIT OCW.

h/2

M=

xx ydy

h / 2

For plane stress,

xx =

E
exx
(1 2 )

Bending the plate causes fiber strains

l y
exx = =
l
R

R
y
l = -y = -y l
R
l

Figure 20.6
Figure by MIT OCW.

Now let l get small

d 2w
xx = y 2
dx

d
R

+d

l
= - dw
dx

dx
Figure 20.7
Figure by MIT OCW.

dw

Then, substituting
h/2

E d 2w
M =
y2d y
2
2

(1 ) dx h / 2

M =

Eh3 d 2 w
12(1 2 ) dx 2

or
M = D

where D

d 2w D
=
dx 2 R

Eh3
12(1 2 )

(flexural rigidity)

Substituting

d 4w
d 2w
=
q
(
x
)

P
dx 4
dx 2

This equation is called plate equation.

12.005 Lecture Notes 21

Plates (continued)

d 4w
d 2w
Flexural equation: D 4 + P 2 + gw = q(x)
dx
dx
where D =

Eh 3
.
12(1 2 )

c
w
m

Figure 21.1
Figure by MIT OCW.

T = T0 cos kx = T0 cos

2 x

Harmonic load
t = t 0 cos kx , w = w0 cos kx , t 0 = T0 w0

d 4w
+ m gw = t 0 c g cos kx when P = 0
dx 4

k 4 Dw0 cos kx + m gw0 cos kx = t 0 c g cos kx

w0 =

t0

Dk 4
m
1+
c
c g
1/ 4

D
Call 2

c g

l flexural wavelength

For ? l , (l k)4 = 1 , w0 ;

t0

m
1
c

; isostacy

For = l , w0 ; 0 uncompensated

L
t
x

Figure 21.2
Figure by MIT OCW.

2 nx s
2 nx
+ t n sin
t(x) = t nc cos

L
L
n=0

Find t nc , t ns

Assume D, calculate wns , wnc

Synthesize w(x) , compare to observations.

Plate subject to an end load

y
L
x

Figure 21.3
Figure by MIT OCW.

Shear force:

dV
= q
dx

Since q = 0, V = const. = Va
Bending moment:
dM
dw
=V +P
dx
dx

Since P = 0,

dM
= V M = Va x + const 0 at x = L
dx

M = Va (x L)

Displacement:

d 4w
=0
dx 4

d 3w
= const
dx 3

d 2w
But M = D 2 = 0 ,
dx

dM
d 3w
= D 3 = Va
dx
dx

d 3w
V
= a
3
D
dx
d 2w
V
= a (x L)
2
D
dx
Subject to w,

w=

dw
= 0 at x = 0
dx

Va x 2
x
(L )
2D
3

cubic displacement

12.005 Lecture Notes 22

Plates (continued)

Va

y
L
h

Figure 22.1
Figure by MIT OCW.

w=

Va 2
x
x (L )
2D
3

Assumption xy = xx

xx =

E
xx
1 2

d 2w
xx = y 2
dx
M = D

xx =

d 2w
dx 2

y
M
D

xxmax =

E h 1
6V L 6V L
Va L = a2 = a
2
h h
h
1 2 D

 xy =

Va 1 h max
=
xx
h 6L

Response to a line load (e.g. volcanic chain under water)

Va

w
c

Figure 22.2
Figure by MIT OCW.

0 at x 0
d 4w
D 4 + ( m w )gw =
dx
V0 at x = 0

Solution to homogeneous equation

w = e x / (C1 cos

x
x
x
+ C2 sin ) + e x / (C3 cos + C4 sin )

1/ 4

4D
with =

( m w )g

is flexural parameter.

Invoke symmetry, boundedness. Determine solution only for x 0 .

dw
= 0 at x = 0
dx

C3 = C4

w0

w = C3e x / (cos

C1 = C2 = 0

x
+ sin )

Now to evaluate C 3 , go back to the end load problem (or original definition)
dM
dw
=V +P
dx
dx
C 3 depends on V0 .

1V
d 3w
=
3
2D
dx
1
d 3w
V0 = D 3
2
dx

=
x =0

4 DC 3

V0 is negative load, half supported by each side.

w=

V0 3 x /
x
x
e (cos + sin )
8D

w0 =

V0 3
8D

w = w0 e x / (cos

x
+ sin )

wb
0

x 0/

x b/

w0
w
Figure 22.3
Figure by MIT OCW.

x/

x0

= tan 1 (1) =

xb

= sin 1 (0) =

3
4

wb = w0 e = 0.0432w0
x ; 3o ; 300km

314

km ; 100km

Suppose plate is fractured.

V0/2

x=0

Figure 22.4
Figure by MIT OCW.

Same equation. Different boundary conditions different profile.

M = 0 at x = 0

d 2w
=0
dx 2

V0 3 x /
x
w=
e cos
4D

w0 =

V0 3
4D

wb
0

x 0/

x b/

x/

w0

w
Figure 22.5
Figure by MIT OCW.

x0 =

xb =

3
,
4

wb = 0.067w0

h ; 35km unbroken
h ; 50km

broken

Thickness of broken lithosphere is about 1.5 times thickness of the unbroken lithosphere.

Bending at a trench.

wb
x0

-V0

xb

-M0

Figure 22.6
Figure by MIT OCW.

w = e x / (C 3 cos

x
+ C 4 sin )

M 0 2
C4 =
2D
C3 = (V0 + M 0 )
w=

2
2D

(entire V0 supported)

2 e x /

x
x
M 0 sin + (V0 + M 0 )cos
2D

Observables -- x0 , xb , wb
tan

x0

= 1+ V0 / M 0

tan

xb

= 1 2M 0 / V0

xb x 0 =

wb =

2 e x

wb =

2D

2M0
2D

x
x
M 0 sin b + (V0 + M 0 )cos b

e[( xb x0 )/ ]e x0 /

sin ((xb x0 ) / )
cos (x0 / )

12.005 Lecture Notes 23

Plates Summary and Fluids
Plates Summary:
Successes moats and rises ubiquitous. Elastic flexure seems to apply.
Problems:

Geophysical measurements of w are difficult.

Reference level?
Other sources of topography.

Alternative approach use gravity

Consider same observed topography

Load

Crust
Stiff
"Airy" root
Floppy

Figure 23.1
Figure by MIT OCW.

Fourier Series t(x) t(k)

Apply load t c = t 0 cos kx
Get displacement D w = w0 cos kx

w0 =

t0

m
D 4
1+
k
c
c g

g = g0 cos kx
g0 : c (t c ekH w0 )
where H is crustal thickness and w0 depends on k and D.

g(k) g(x)
Compare g(x) with data.

Fluids
Fluids no memory of shape flow under applied tractions, body forces, stop flowing
(dont reverse flow) when driving forces removed.
Newtons concept of viscosity subject fluid to shearing

X2

u1

X1

Figure 23.2
Figure by MIT OCW.

12 =

u1
x2

where is shear viscosity.

Substance

(Pa sec) at 20 o C

Poise (10 Pa sec)

air

2 10 5

2 10 4

water

103

10 2

glycerine

10

ice ( 0 o C )

1013

1014

glass

1017

1018

Earth

1019 10 21

10 20 10 22

Physical cause gasses (vertical) motion of particles with different horizontal

velocities.

Fluids elastic resistance to distortion of atomic cages as atoms and molecules

slide by.

Figure 23.3
Figure by MIT OCW.

Solids diffusion of defects in the lattice (vacancies or interstitials); motion of

dislocations in lattice structure.

E
E

Figure 23.4
Figure by MIT OCW.

12.005 Lecture Notes 24

Fluids (continued)
For a general stress-strain, for a Newtonian fluid

ij = p ' ij + Dijkl kl
where Dijkl is viscosity tensor and kl is strain rate tensor.
For an isotropic fluid

ij = p ' ij + kk ij + 2 ij
For volumetric strain rate

kk = 3 p '+ (3 + 2 ) kk
kk

2
= p '+ ( + ) kk
3
3

Mean normal stress:

2
where + is bulk viscosity.
3
2
For many applications, + = 0 (Stokes fluid)
3
2
3

ij = p ' ij + 2 ij kk ij
For many applications, kk

ij = p ' ij + 2 ij
Often is used for viscosity

ij = p ' ij + 2 ij
Sometimes 0

(perfect fluid)

ij = p 'ij
1

Material Derivative
Laws of physics conservation of mass, conservation of energy, etc.
Express in reference frame of material, e.g. rod

T=0

T = T0
x
L

Figure 24.1
Figure by MIT OCW.

Steady state: T = T0 x / L;
Lagrangian frame: c p

T
=0
t

T
= k 2T + A
t

Eulerian frame material is moving. There would be a

T
for the above rod moving
t

through.
Marching band example.
Need to account for non physical change due to motion.
Above example:

T
T
T
where v
is advection term.
= v
x
x
t

Material derivative:
D
= + v
Dt t

Heat conduction
DT T
=
+ vT = k 2T + H
t
Dt

Conservation of Mass Continuity Equation

Consider motion in x2 direction:

X3
dx3

v2

] dx

v2 + (v2) dx2
x2

dx2

X2

X1
Figure 24.2
Figure by MIT OCW.

Sides: mass in - mass out = -

( v2 )dx2 dx1dx3 = ( v2 )dV

x 2
x 2

Front, back: mass in - mass out = -

( v1 )dV
x1

Top, bottom: mass in - mass out = -

( v3 )dV
x 3

For all 3 directions: -

( v1 )( v2 )( v3 ) =
x 2
x 3
t
x1

+
( vi ) = 0
t x i

v
+ vi
+ i =0
t
x i
x i
D
v
+ i = 0 (Law of conservation of mass)
Dt
x i

For an incompressible fluid with constant properties

p + 2 v + x =

Dv
Dt

or, with / (dynamic viscosity)

1 p
Dv
2 vi

+
+ xi = i
xi
x j x j
Dt

(Navier-Stokes equation)

Plane strain

V0
BOAT
X2

(water)
X1

Figure 24.3
Figure by MIT OCW.

t = 0, v = 0, v( x1 = 0) = (0, v0 , 0)
only have v2 0
in x2 direction

=0
x2

Subtract out hydrostatic

v2
2 v
= 22
t
x1
The solution becomes v = v0 (1 erf

where erf (y) =

x1
2 t

e d .

Velocity propagates downward a characteristic depth, x1 = 2 t .

Example: canoe 5 meters long,
v0 = 5 m/sec t : 1 sec

water : 10 2 cm2/sec x1 : 2 10 2 = 2 mm
A canoe will drag along about 2 mm water.

12.005 Lecture Notes 25

Navier-Stokes Equation dimensional form
2 vi
Dv
p

+
+ fi = i
Dt
xi
x j x j
Assume:
Characteristic velocity v0
Characteristic length L
Characteristic stress v0 / L
Characteristic time v0 / L

Choose non-dimensional variables

v ' = v / v0 x ' = x / L p ' =

or

v = v ' v0 x = x ' L p =
1
=
x L x '

f '=

L
f
v0

v
L
p t'= 0 t
L
v0

v0
L

p' t' =

L
t'
v0

v0
=
t L t '

Substitute into Navier-Stokes equation

2 vi '
v Dvi '
1 v0 p ' 1
1
+ 2 v0
+ 2 v0 f i ' = v0 0
x j ' x j ' L
L L xi ' L
L Dt '

or
2 vi '
v L Dvi '
p '

+
+ fi ' = 0
xi ' x j ' x j '
Dt '

where

v0 L
= Re

 Re gives importance of inertial terms relative to viscous terms.

Re

viscous forces balance

acceleration negligible

Re

inertia dominates

Note: in dimensionless form, Re is the only parameter in the Navier-Stokes equation.

for given geometry (boundary conditions) ALL equivalent

(non-dimensional) problems at same Re give same result!

Examples:
1. Low Reynolds number flow past a cylinder.

Re

Symmetry, like in the sphere problem.

Figure 25.1
Figure by MIT OCW.

2. Re = 10

vi
= 0 (steady)
t

vj

vi
0
x j

Asymmetry; eddies in wade.

Figure 25.2
Figure by MIT OCW.

The figure below is experimental.

102
CYLINDER
10
CD
1

10-1

10-1

10

102

103

104

105

106

107

Re
Figure 25.3. Variation of drag coefficient with Reynolds number for circular
cylinder.
Figure by MIT OCW.

Re < 1

CD 1 / Re

D V

102 < Re < 3 105

CD const.

D V2

Re 3 105

big drop in CD!

Recall
Earths mantle: Re 10-19
canoe: Re 2 105

Summary:
For low Re, inertia not important

Navier-Stokes equation linear

simple results (analytic theory)

For high Re, inertia important

Navier-Stokes equation nonlinear
time dependent
complicated experimental approach empirical

12.005 Lecture Notes 26

Growth and Decay of Boundary Undulations
Growth: Rayleigh - Taylor Instability
salt domes
diapirs
continental delamination
X3

u, u

= 0cos

l, l

2x1
= 0coskx

Figure 26.1
Figure by MIT OCW.

General problem: topography on an interface

= 0 cos kx1

k=

(1) If u < l topography decays as 0 et / .

(2) If u
> l topography grows.
Initially = 0 et /
.

Eventually many wavelengths interact, problem is no longer simple.

Characteristic time depends on , u , l , thickness of layers,

X1

Before glaciation

Subsidence caused by glaciation

Ice Sheet

Surface after melting of the ice sheet but prior to postglacial rebound

Full rebound

Figure 26.2. Subsidence due to glaciation and the subsequent postglacial

rebound.
Figure by MIT OCW.

Weight of ice causes viscous flow in the mantle.

After melting of ice, the surface rebounds postglacial rebound.

Different regions have different behaviors (e.g., Boston is now sinking).

X3

= 0coskx

Air

X1
m

Figure 26.3
Figure by MIT OCW.

Problem: how to reconcile physical boundary conditions with mathematical description?

Decay: Postglacial rebound (1/2 space, uniform )

k=

= 0 cos (kx1)
X3

Figure 26.4
Figure by MIT OCW.

Assume uniform

Subtract out lithostatic pressure P = p gx3

Assume g uniform

Use stream function

X1

v1 =

x3

v3 =

x1

4 = 0

Solution: = (A + Bkx3 )exp (kx3 ) + (C + Dkx3 )exp (kx3 ) sin kx1

Boundary conditions:
at x3 = 0 (mathematical, not physical)

33 = g
v1

13 = 0 =

x3

v3

x1

at x3 , must be bounded
C = D = 0

In order that 13 = 0 at x3 = 0 ,
2 2
+
=0
x32 x12
B=A

or = A(1+ kx3 )exp (kx3 ) sin kx1

Then

v1 = Ak 2 x3 exp ( kx3 ) sin kx1

v3 = Ak(1 + kx3 )exp ( kx3 ) cos kx1
at x3 = 0 v3 = = Ak cos(kx1 )

Now

33 = p + 2 33
33 = 0 at x3 = 0
p
2 vi
+
+ x1 = 0
To get p, use
xi
x jx j
for i = 1

2 v 2 v
p
+ ( 21 + 21 ) = 0
x1 x3
x1

Substitute for v1 and integrating p x

3 =0

= 2k 2 A cos kx1

But p = g A =
Or 0 =

g 0
2 k 2

g 0
g 0
=
2k
4

Or 0 = 0
where =

t =0

exp(

gt
t

) = 0 t =0 exp( )
2k

2k 4

=
g g

Solving for : =

g
4

For curves shown,

: 5000 yr
: 10 21 Pa
: 3000 km
Note: stream function exp(kx3 ) = exp(
Falls off to 1 / e at x3 :

2 x3

Senses to fairly great depth

postglacial rebound doesnt reveal the details of mantle viscosity structure,

but only the gross structure.

12.005 Lecture Notes 27

Flow in Porous Media
Problem of great economic importance (also scientific)
hydrology (ground water migration, toxic waste)
oil migration
soil stability, fault mechanics (pore pressure)
melt migration in mantle
geysers and hot springs

Porous medium voids porosity

volume fraction of voids

For example,
Sand: 40%
Pumice: 70%
Oil shales: 1020%

If pore connected permeable

Pressure gradient flow
k
Darcys law v = p

v volumetric flow rate

k permeability

We can use Poiseuille flow for simple geometries. For example, cubical matrix, circular
tubes or pipes.

Figure 27.1. An idealized model of a porous medium.

Circular tubes of diameter form a cubical matrix with
dimensions b.
Figure by MIT OCW.

1

12 b
3 2
4
2
=
=
b3
4 b2
dp
(one direction only)
dx
2 dp
[Poiseuille flow]
In each pipe (along x), u =
32 dx

Consider

1

4 u
2
4
2 = u = u
Darcy velocity: v =
b2
4b 2
3

v=

b 2 2 dp
72 dx

k=

1 2 2
b
72

Large b large v?

b2 =

Large large v?

k=

3 2
4

4
128 b 2

Compare to cubes separated along faces (channel flow)

Figure 27.2
Figure by MIT OCW.

1
6 b2

=3
= 23
b
b

Again,

dp
directed along one edge
dx

u=

1 dp 2
Z ( / 2) 2 )
(
2 dx
/2

1 dp Z 3 2 Z
5 2 dp
u=

2 dx 3
2 / 2
24 dx

Darcy velocity: v = 2

5 3 dp
5 b 2 3 dp
b
=

u
12 b dx
324 dx
b2

5b 2 3
k=
324
k is different depending on .
135 3
k=
324 b
Clearly, porosity distribution is important.

Figure 27.3
Figure by MIT OCW.

Also -- more easily measured than figured out theoretically more complicated
geometries numerical simulation.

Consider Lawn Sprinkler example flow in unconfined aquifer.

y
"Phreatic surface"

Land surface

h(x)

u(x)
x

Figure 27.4
Figure by MIT OCW.

h hydraulic head

u Darcy velocity

Dupuit approximation:

For

h
x

dp
h
= g
x
dx

1 flow is one-dimensional.

Darcys law: u =

k g h
x

Conservation of mass: Assume no input

Flux Q = u ( x)h( x) =

k g dh
h
= const.

dx

phreatic surface is a parabola

For h = h0 at x = 0
1/ 2

2Q x
h = h0 2

kg

Suppose we have a porous dam of width w. The relation between Q, h0 and h1 is:
Q=

kg 2
h0 h12 )
(
2 w

Q=

kg
( h0 h1 )( h0 + h1 )
2 w

or

y
w

Phreatic Surface

Dupuit Parabola

h0

B
Seepage face
C

Impermeable Layer

h1

Figure 27.5. Unconfined flow through a porous dam. The Dupuit parabola AC is the solution
if (h0-h1)/h0<<1. The actual phreatic surface AB lies above the Dupuit parabola resulting in a
seepage face BC.

Figure by MIT OCW.

12.005 Lecture Notes 28

Time Dependent Porous Flow
Assume 1-D flow through an unconfined aquifer.

Figure 28.1

Flux in

Qin = u(x,t)h(x,t)

Flux out Qout = u(x + x,t)h(x + x,t)

Qout Qin

h
(uh) x
x
t

But medium is porous, with porosity . For a given Q, the smaller , the larger h / t
Qout Qin = [h(t + t, x) h(t, x)] t
Mass conservation

h
+ (uh) = 0
t x

Darcys law (Dupuit approx) u =

Combining

h
x t
t

k g h
x

h k g h
=
(h ) (Boussinesq Equation)
t
x x

This is a nonlinear diffusion equation. We can simplify it by assuming

h = h0 + h'

h' = h0

(h ) = [(h0 + h') (h0 + h')]

x x
x
x
=

h'

[(h0 + h') ]
x
x

h0

2 h '
x 2

This gives the familiar diffusion equation

h' k gh0 2 h'
=
t
x 2
k gh0

equivalent to K in heat flow equation

equivalent to = / in fluid space

[note inverse relationship of fluid viscosity !]

Consider the sudden lowering of water in a river channel next to a saturated bank.
h(x + ,0) = h0
h(0,t) = h1
h(,t) = h0

Figure 28.2
Solution is like the solution to the heat flow equation for plate cooling, or the momentum
equation for crew shell:
Set f =

h
h0

f (0) =

h1
h0

=(

1/ 2 x
)
k gh0t
2

f ( ) = f (0)erfc( )
Finally, back to the tank demo

Figure 28.3
Flow between glass plates Darcy flow.
(recall 1 model of porosity)
Analyze a simple case-1 boundary

Figure 28.4
To conserve fluid

Figure 28.5

Qout Qin = dx

dh

dt

dQ
h
=
dx
t

but Q = u h , with u

dh
dx

k g h
= (uh) =
(h )
x x
t
x

nonlinear!

Linearize around h = h0 + h1
h' k gh0 2 h'
=
t
x '2

diffusion equation

exponential growth or decay

[Note: Dupuit approximation breaks down for large interface perturbation.]
Suppose h ' = cos kx , k = 2 / = wave number
2 h'
= k 2 cos kx
2
x

gh0 2

=
k where K is permeability.
t

Exponential growth or decay!

1
gh0 k 2

How does this compare to Rayleign-Taylor?