GG303 Lecture 19 11/5/02 1

CAUCHY'S FORMULA AND EIGENVAULES (PRINCIPAL STRESSES) IN 3-D

I Main Topics
A Cauchy’s formula
B Principal stresses (eigenvectors and eigenvalues)
II Cauchy's formula
A Relates traction vector components to stress tensor components
(see Figures 19.1, 19.2, 19.3 for derivation)
B τi = σ ji nj
1 Meaning of terms r r r
r
a τ i=traction vector component: τ = τ1i + τ 2 j + τ 3k
b σ i j = stress component
c n =unit normal. The components nj of the unit normal are
the direction cosines between n and the coordinate axes.
Fi Fi A1 Fi A2 Fi A3
d = + +
A A1 A A2 A A3 A
2 This represents the physics directly
3 The traction component that acts in the i-direction reflects
the contribution of the stresses that act in that direction.
4 Note that the j's "cancel out" in equation II.B
5 Note that the subscripts on the τ and the n differ
6 σ is symmetric (σ i j =σ j i ), so …
C τi = σi j n j Standard form of Cauchy’s formula
1 The subscript j's still "cancel out"
2 The subscripts on the τ and the n still differ
3 Easier(?) to remember than “B”
D Full expansion
τi = σji n j τ i = σ ij n j
τ1 = σ11 n 1 + σ21 n 2 + σ31 n 3 = σ 11 n 1 + σ 12 n 2 + σ 13 n 3
τ2 = σ12 n 1 + σ22 n 2 + σ32 n 3 = σ 21 n 1 + σ 22 n 2 + σ 23 n 3
τ3 = σ13 n 1 + σ23 n 2 + σ33 n 3 = σ 31 n 1 + σ 32 n 2 + σ 33 n 3
E Matrix form
 τ1  σ11 σ 21 σ 23  n1   τ1  σ11 σ12 σ13  n1 
τ 2  = σ12 σ 22 σ 32 n2  τ 2  = σ 21 σ 22 σ 23 n2 
τ  σ   τ  σ σ 32 σ 33 n 3 
 3   13 σ 23 σ 33 n 3   3   31

Stephen Martel 19-1 University of Hawaii

GG303 Lecture 19 11/5/02 2

Derivation of Cauchy's Equation (2-D) 19.0

x2
τ2
σ11 τ
φ1
θ2 A
σ12
A1 θ1
τ1
φ2
x1
A2 σ21
σ22

ΣF1 = 0, so (ΣF1)/A = 0.

τ1(A/A) = (σ11)(A1/A) + (σ21)(A2/A).

Similarly, (ΣF2)/A = 0, so

τ2(A/A) = (σ12)(A1/A) + (σ22)(A2/A).

x2

φ1 A1/A = cosφ1 = cosθ1 = n1

θ A
A1 2
θ1 φ2
A2/A = cosφ2 = cosθ2 = n2
x1
A2

The ratio of the areas is given by the direction cosines of the normal to plane A.
This holds in 2-D and in 3-D also.

Substituting for the area ratios:

τ1 = (σ11)(n1) + (σ21)(n2).

τ2 = (σ12)(n1) + (σ22)(n2).

τi = σji nj

Stephen Martel 19-2 University of Hawaii

GG303 Lecture 19 11/5/02 3

Derivation of Cauchy's Equation C 19.1

C
Area A3
x2
Area A1 C σ32

σ31
τ2 τ
σ12
σ33
σ11 O B

O
σ13 τ1
τ3 O B
x1
D σ22
O B
D Area A σ21
Area A2
σ23
x3 D
ΣF1 = 0, so (ΣF1)/A = 0.

τ1(A/A) = (σ11)(A1/A) + (σ21)(A2/A) + (σ31)(A3/A).

Similarly, ΣF2 = 0 and ΣF3 = 0, so

τ2(A/A) = (σ12)(A1/A) + (σ22)(A2/A) + (σ32)(A3/A).

τ3(A/A) = (σ13)(A1/A) + (σ23)(A2/A) + (σ33)(A3/A).

Stephen Martel 19-3 University of Hawaii

GG303 Lecture 19 11/5/02 4

Derivation of Cauchy's Equation C 19.2

C
x2 Area A3

C P'''
Area A1

P''' B
P' O
O ω2
OP is normal to BCD
P
P' ω1
O B
ω3 x1
D
P'' O B
D Area A
P'' Area A2

x3 D

Note that ∆ DCB of area A projects onto the x1-x2 plane as ∆ OCB,
onto the x2-x3 plane as ∆ OCD, and onto the x3-x1 plane as ∆ OBD.

BOP' is perpendicular to CD, and because CD is a line in BCD,
BOP' is perpendicular to BCD. Similarly, COP'' is perpendicular to BD, so
COP'' is perpendicular to BCD. The intersection of BOP' and COP' is
perpendicular to BCD, and that intersection is OP.

ω1, ω2, ω3, are angles between OP and x1, x2, and x3, respectively.

A1 = 1/2 (base OCD)(height OCD) = (CD)(OP') = OP’

A 1/2 (base DCB)(height CBD) (CD)(BP') BP’

Stephen Martel 19-4 University of Hawaii

GG303 Lecture 19 11/5/02 5

Derivation of Cauchy's Equation 19.3

P' P'

ω1
P P

θ x1 θ x1
ω1 ω1
O B O B
Triangles BOP and BP'O Therefore, angle BP'o = ω1.
are similar right triangles;
they both have angle ODP A1 = OP' = cos ω1 = n1
(i.e., θ) in common. A DP'

Similarly, A2 = OP'' = cos ω2 = n2 and A3 = OP''' = cos ω3 = n3 so

A DP'' A DP'''

τ1(A/A) = (σ11)(A1/A) + (σ21)(A2/A) + (σ31)(A3/A) becomes

τ1 = (σ11)(n1) + (σ21)( n2) + (σ31)(n3). Similarly,

τ2(A/A) = (σ12)(A1/A) + (σ22)(A2/A) + (σ32)(A3/A) becomes

τ2 = (σ12)(n1) + (σ22)( n2) + (σ32)(n3), and

τ3(A/A) = (σ13)(A1/A) + (σ23)(A2/A) + (σ33)(A3/A) becomes

τ3 = (σ13)(n1) + (σ23)( n2) + (σ33)(n3).

So τi = σji nj, but σij= σji, so τi = σij nj

.
.

Stephen Martel 19-5 University of Hawaii

GG303 Lecture 19 11/5/02 6

I I I Principal stresses from tensor and matrix perspectives

Consider a plane with a normal vector n defined by direction cosines
n 1 , n2 , and n3 . The components of traction τ on the plane, by Cauchy’s
formula, are τ i = σ ij nj. They also are simply the components of τ : τ 1 = τ n 1 ,
τ2 = τn 2 , and τ3 = τn 3 . The components can be equated:
σ 11 σ 12 σ 13   n1   n1 
σ     
(1)  21 σ 22 σ 23  n2  = T n2  .
σ 31 σ 32 σ 33   n3   n3 
The right side of (1) can be subtracted from the left side to yield:
σ 11 − T σ 12 σ 13   n1 
(2)  σ σ 22 − T σ 23  n2  = 0 .
 21
 σ 31 σ 32 σ 33 − T   n3 
Equation (2) can be rewritten
(3) [ σ - I T] [n] =0, where I is the identity matrix
1 0 0
(4) I = 0 1 0. For any square matrix [A], [A][I] = [A].
0 0 1
According to theorems of linear algebra, equation (2) can be solved only if
the determinant |σ - I T| equals zero:
σ 11 − T σ 12 σ 13
(5)
σ 21 σ 22 − T σ 23 = 0
σ 31 σ 32 σ 33 − T

In many cases the components of σ are known but T is must be solved for.

Problems of the form of equation (3) are common in many branches of

mathematics, engineering, and physics, and they have a special name:
eigenvalue problems. The values of T (i.e., |T|, the principal values) that
solve the equation are called eigenvalues, and the vectors n (the principal
directions) that give the directions of T are called eigenvectors. Because
these problems are so common, many mathematics packages, including
Matlab, have special routines to solve for eigenvalues and eigenvectors.

Stephen Martel 19-6 University of Hawaii

GG303 Lecture 19 11/5/02 7

Solving (5) by hand requires finding the roots of a cubic equation (not
easy), so we consider the easier 2-D case, which yields a quadratic
equation.

σ 11 − T σ 12
(6) =0 Note: a b = ad − bc
σ 21 σ 22 − T c d

(7) (σ 11 − T )(σ 22 − T ) − (σ 12 )(σ 21 ) = 0

(8) T 2 − T (σ 11 + σ 22 ) + (σ 11 )(σ 22 ) − (σ 12 )(σ 21 ) = 0

(9a) T 2 − T (σ 11 + σ 22 ) + [(σ 11 )(σ 22 ) − (σ 12 )2 ] = 0 or (9b) T 2 − T ( I1 ) + [T2 ] = 0

The term T in equation (9) is solved using the quadratic formula:

2
( 1 0 ) T = (σ 11 + σ 22 ) ± (σ 11 + σ 22 ) − 4[(σ 11 )(σ 22 ) − (σ 12 )] = I1 ± I1 − 4 I2
2 2

2 2

( 1 1 ) T = (σ 11 + σ 22 ) ± (σ 11 + 2σ 11σ 22 + σ 22 ) − 4[(σ 11 )(σ 22 ) − (σ 12 )]

2 2 2

( 1 2 ) T = (σ 11 + σ 22 ) ± (σ 11 − 2σ 11σ 22 + σ 22 ) + 4[σ 12 ]
2 2 2

2
( 1 3 ) T = (σ 11 + σ 22 ) ± (σ 11 − σ 22 ) + 4[σ 12 ] = I1 ± I1 − 4 I2
2 2

2 2

Stephen Martel 19-7 University of Hawaii

GG303 Lecture 19 11/5/02 8

 2   2 
σ11 + σ 22   σ11 − σ 22   I1   I1 
T = 
±  2 
 + σ12 = [c ] ± [ r] =   ±   − I2  = σ1,σ 2


(14)  2    2   2  2 
   

An inspection of the diagram below shows that the first term in brackets
in equation (14) is the mean normal stress (i.e., the center of the Mohr
circle) and the second term in brackets is the maximum possible shear
stress (i.e., the radius of the Mohr circle). So the principal stresses lie at
the end of a horizontal diameter through the Mohr circle. The terms
c, r, I , and I are called invariants and are independent of the frame of
i 2
reference.

τs
(σ11, σ12)

σ1 + σ2
σ2 2 σ1

τn

(σ11, -σ12)

Stephen Martel 19-8 University of Hawaii

GG303 Lecture 19 11/5/02 9

Example
Suppose the stress state at a point is given by
( 1 5 ) σ ij = 10 3 , where dimensions are in MPa.
 3 2
 
Solving for the principal values using eq. (14) yields
(10 + 2) 10 − 2  2
(16) λ= ±  + 32 = 6 ± 25 = 11 and 1
2  2 
Now we substitute these back into (3)
(17a) 10 − 11 3   n1  0 
= ⇒ −1 3   n1  0 
= for T = σ 1 = 11MPa.
 3
 2 − 11 n2  0  3
 −9 n2  0 
(17b) 10 − 1 3   n1  = 0  ⇒ 9 3  n1  0 
= for T = σ 2 = 1MPa.
 3
 2 − 1 n2  0  3
 1 n2  0 
These relations yield
(18a) -n 1 + 3n2 = 0 (σ1 = 11MPa) (18b) 3x 1 + x2 = 0 (σ2 = 1MPa).
From (18a), for an eigenvalue (principal value) of 11 MPa, n1 = 3n2 .
From (18b), for an eigenvalue (principal value) of 1 MPa, n2 = -3 n1 .

1 MPa
x2 2 MPa x2' x2 n1 = cos θ1n
3 MPa 11 MPa n2 = cos θ2n
3 MPa = sin θ1n
x1' θ2n

10 MPa So n2/n1 = tan θ1n

-1
x1 1 x1 θ1n x1 θ1n = tan (n2/n1)
1 3 = atan2(n2,n1)

-3

For σ 1 = 11MPa*
x n n  n   1
θ = tan−1 2  = tan−1 2  = tan−1 2  = tan−1 = 18.5o
x1, normal
 x1 n   n1   3n2   3 
For σ 2 = 1Mpa*
x n n   −3n 
θ = tan−1 2  = tan−1 2  = tan−1 1 = tan−1(−3) = −71.5o

x1, normal
 x1 n   n1   n1 
The two eigenvectors are perpendicular, as they are supposed to be.

* In the first expression for θ , the normal direction is the x1’ direction,
and n1 and n2 are the direction cosines for a unit vector along x1’ . In the
first expression for θ , the normal direction is the x2’ direction, and n1 and
n 2 are the direction cosines for a unit vector along x2’ .

Stephen Martel 19-9 University of Hawaii

GG303 Lecture 19 11/5/02 10

V Matrix treatments of stress transformation

In matrix form, σ i′j ′ = ai′k a j ′lσ kl becomes (Mal & Singh, 1991, p. 37)

( 1 9 ) σ ' = [a] [σ ] [aT ],

where
a ′  a ′ 
1 1 a1′ 2 a1′ 3 1 1 a2′1 a3′1
( 2 0 ) a =  a2′1 a2′2 a2′3  and ( 2 1 ) aT = a1′ 2 a2′2 a3′2 
   
a3′1 a3′2 a3′3   a1′ 3 a2′3 a3′3 
The proper order of matrix multiplication is essential in order to
reproduce the expansions of lecture 17: [a] [σ ] [aT ]≠ [aT ] [σ ][a]!
In MATLAB, equation (19) would be written:
sigmaprime = a * sigma * a'
The term a' signifies [aT ]. Matlab also has a function “eig” to find
eigenvectors (given in terms of the direction cosines) and eigenvalues.
[V,D] = eig (sigma)

Example
»sigmaxy = [10 3;3 2]

sigmaxy =
10 3
3 2

»a = [3/sqrt(10) 1/sqrt(10);-1/sqrt(10) 3/sqrt(10)]

a=
0.9487 0.3162 The first row of matrix “a” is the negative
-0.3162 0.9487 of the first column of matrix V below.
The second row of matrix “a” is the negative
»sigmaprime = a*sigmaxy*a' of the second column of matrix V below.

sigmaprime =

11.0000 -0.0000
-0.0000 1.0000

»[V,D] = eig(sigmaxy)

V=
-0.9487 0.3162 Column 1 in V relates to column 1 in D
-0.3162 -0.9487 Column 2 in V relates to column 2 in D

D=
11 0
0 1

Stephen Martel 19-10 University of Hawaii