3-node Quintic Hermite

Euler-Bernoulli Beam Element

Derivation and Static Condensation to

Standard 2-node Beam Element Stiffness
and Load Vector

Dr. Lonny L. Thompson

Sept 28, 2025

Repost this RETWEET

1 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 h (w ) h (w ) h /L (3 ) h /L (3 )
1 1 3 2 2 1 4 2

2-Node Cubic Hermite Beam Shape Functions

1
Dimensionless H (s)
i

0.5

-0.25
-1 0 1
Natural coordinate s 2 [-1,1]

H (w ) H (w ) H (w ) H /L (3 ) H /L (3 ) H /L (3 )
1 1 3 2 5 3 2 1 4 2 6 3

3-Node Quintic Hermite Beam Shape Functions

1
Dimensionless H (s)
i

0.5

-0.25
-1 0 1
Natural coordinate s 2 [-1,1]

2 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Euler–Bernoulli Beam Basics
Kinematic Assumptions
Cross-sections remain plane and perpendic-
ular to the neutral axis during bending.
Transverse shear deformation is neglected;
valid for slender beams with high length-to-
depth ratios.
The displacement field u(x, z) = ux ı̂ + uz k̂
is approximated by

ux(x, z) = −z θ(x)

uy (x, z) = w(x)
where w(x) is the transverse deflection, and
θ(x) = dw
dx section rotation assumed equal to
the slope.

3 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 Small deformation and small strain assump-
tions hold; geometric linearity is assumed.
No axial deformation uO (x) is considered;
only bending is modeled.

4 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Derivation of the 2-node cubic
C 1 Hermite shape functions
Each node has a transverse displacement w
and rotation θ = dw
dx , with associated shear
force V and bending moment M .

w1 w2
θ2
x
θ1 L
V1 V2 M2
x
M1 L
Two-node Beam Element. (Top) Kinematic DOF (wi, θi),
(Bottom) Nodal Actions (Vi, Mi)

Total DOFs per element: 2 × 2 = 4.

5 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 Using normalized coordinate to simplify the
algebra, ξ = x/L within the interval ξ ∈ [0, 1],
the nodal DOF are (2 per node):
w(0) = w1, w0(0) = θ1,
w(1) = w2, w0(1) = θ2

Kinematics and interpolation ansatz

For an Euler-Bernoulli beam, the transverse dis-
placement w(x) is C 1 (continuous function with
continuous slope) and the cross-section rotation
equals the slope:
dw
θ(x) =
dx
On an element of length L with local coordinate
ξ = x/L ∈ [0, 1], approximate w by a cubic:
w(ξ) = a0 + a1ξ + a2ξ 2 + a3ξ 3.
Then
dw 1 dw 1 2
!

θ(ξ) = = = a1 + 2a2ξ + 3a3ξ

dx L dξ L
6 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Nodal degrees of freedom and interpolation
conditions
Let the nodal DOFs be
   




w1  



w(0) 
   
   

θ1 θ(0)
   
   
   

{d} = = .
   
   
   
   
w2 w(1)
   
   
   
   
   
   
   


θ2 



θ(1) 


Impose these as four conditions on a0, . . . , a3 :

1. At ξ = 0 : w(0) = a0 = w1,
2. At ξ = 0 : θ(0) = L1 a1 = θ1 ⇒ a1 = Lθ1.
3. At ξ = 1 : w(1) = a0 + a1 + a2 + a3 = w2.
1
4. At ξ = 1 : θ(1) = L (a1 + 2a2 + 3a3) = θ2.

7 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Substitute a0 = w1 and a1 = Lθ1 into the last two
equations to solve for a2, a3 :


w1 + Lθ1 + a2 + a3 = w2,








θ1 + L2 a2 + L3 a3 = θ2.









Equivalently,
    

 1 1   a2  
 w2 − w1 − Lθ1 
= .
 
 
     
     
2 3 a3 L(θ2 − θ1)
     
     

Solve
a3 = L (θ2 + θ1) − 2 (w2 − w1) ,
a2 = 3 (w2 − w1) − L (2θ1 + θ2) .

Extract the shape functions

Insert ai back into w(ξ) and collect by DOFs:
2 3 2 3
! !

w(ξ) = 1 − 3ξ + 2ξ w1 + L ξ − 2ξ + ξ θ1
| {z } | {z }
N1(ξ) N2(ξ)
+ 3ξ − 2ξ 3 w2 + L −ξ + ξ θ2.
2 2 3
! !

| {z } | {z }
N3(ξ) N4(ξ)

8 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
From this, we identify the cubic Hermite shape
functions on the normalized interval ξ ∈ [0, 1]:

N1(ξ) = 1 − 3ξ 2 + 2ξ 3
2 3
!

N2(ξ) = L ξ − 2ξ + ξ
N3(ξ) = 3ξ 2 − 2ξ 3
2 3
!

N4(ξ) = L −ξ + ξ .

Matrix form:
 




w1 
 
 

θ1
 
  
 

w(ξ) = N1 N2 N3 N4 .
 
  
  
 
w2
 
 
 
 
 
 
 


θ2 


9 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Checks of Properties
Interpolation:
w(0) = w1.
N1(0) = 1, N2(0) = 0, N3(0) = 0, N4(0) = 0;

θ(0) = θ1.
N10 (0)/L = 0, N20 (0)/L = 1, N30 (0)/L = 0, N40 (0)/L = 0;

Similarly, at ξ = 1 for node 2.

C 1 continuity across elements is automatic

because both w and dw/dx are interpolated
continuously at nodes.
Partition of unity for displacement: N1 + N3 =
1 and N2 + N4 ≡ 0 at the nodes (displace-
ment field can reproduce constants and linear
slopes appropriately).
Cubic completeness: the four DOFs match
the four coefficients of a cubic in ξ, giving the
classic cubic Hermite polynomial basis.
10 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Hermite C 1 cubic functions on s = [−1, 1]
In the standard natural coordinate s ∈ [−1, 1] with
the affine map (coordinate change) ξ = (s + 1)/2,
the Hermite functions become
L
H1(s) = 14 (2 − 3s + s3) , H2(s) = 8 (2 − s − s2 + s3)
L
H3(s) = 14 (2 + 3s − s3) , H4(s) = 8 (−2 − s + s2 + s3) , s

The Hermite shape functions can also be ex-

pressed in factored form:
1
H1(s) = (1 − s)2(2 + s),
4
L
H2(s) = (1 − s)2(1 + s),
8
1
H3(s) = (1 + s)2(2 − s),
4
L
H4(s) = − (1 + s)2(1 − s),
8

11 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 h (w ) h (w ) h /L (3 ) h /L (3 )
1 1 3 2 2 1 4 2

2-Node Cubic Hermite Beam Shape Functions

1
Dimensionless H (s)
i

0.5

-0.25
-1 0 1
Natural coordinate s 2 [-1,1]

h (w ) h (w )
1 1 3 2

1
H (displacements)

0.5
i

-1 -0.5 0 0.5 1
s

h /L (3 ) h /L (3 )
2 1 4 2

Rotation Shape Functions (normalized): h /L, h /L

2 4
0.2

0.1
H /L (rotations)

-0.1
i

-0.2
-1 -0.5 0 0.5 1
s

12 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 In the natural coordinates s, the transverse
displacement approximation is

w(s) = H1w1 + H2θ1 + H3w2 + H4θ2.

These are equivalent to the [0, 1] forms via the

coordinate change. The mapping to −1 ≤
s ≤ 1 is natural since Gaussian quadrature
and other numerical integration rules are of-
ten derived in this symmetric, dimensionless
interval.

13 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 In natural coordinates on the interval s ∈ [−1, 1],
the Hermite interpolation conditions with x =
d d
(L/2)(1 + s), so dx = L2 ds are:
H1(−1) = 1, H1(1) = 0, H10 (±1) = 0

H2(−1) = 0, L2 H20 (−1) = 1, H2(1) = 0, L2 H20 (1) = 0

H3(1) = 1, H3(−1) = 0, H30 (±1) = 0

H4(1) = 0, L2 H40 (1) = 1, H4(−1) = 0, L2 H40 (−1) = 0.

14 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 The factored form gives the insights:
1. Factors (1 ∓ ξ)2 force both the value and
the first derivative to vanish at the opposite
node, exactly the Hermite conditions for C 1
interpolation.
2. Swapping s 7→ −s maps (H1, H2) ↔ (H3, −H4).
This mirrors the node 1/node 2 symmetry
and the sign change of the rotation shape
functions.
3. Near s = −1, H2 shows a cubic rise driven
by the local slope DOF; similarly H4 is the
mirrored version near s = 1.
4. The squared factor gives a ”flat” (zero-slope)
contact at the non-associated node, ensur-
ing C 1 continuity across elements.

15 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 5. The property H1 + H3 ≡ 1 is immediately
apparent after expansion; the ( 1 ± s)2 struc-
ture helps see why the displacement basis
reproduces constants while the slope basis
integrates to zero over the element ends.
6. In symbolic/numerical work, factoring out
(1 ± s) can reduce cancellation near s = ±1
and makes boundary checks trivial when
deriving B-matrices and end-force expres-
sions.

16 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Derivatives for curvature, moment, shear
d2 w 1 d2 w
κ(x) = 2 = 2 2 ,
dx L dξ
M = EIκ,
dM d3 w
V = = EI 3
dx dx
Differentiate the shape functions w.r.t. ξ and
scale by 1/L or 1/L2,
dN1 dN2
= −6ξ + 6ξ 2, 2
!

= L 1 − 4ξ + 3ξ , . . .
dξ dξ
then
d 1 d d2 1 d2
= , 2
= 2 2
dx L dξ dx L dξ

17 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
First derivatives:
dH1 3
= (s − 1)(s + 1),
ds 4
dH2 L
= (s − 1)(3s + 1),
ds 8
dH3 3
= − (s − 1)(s + 1),
ds 4
dH4 L
= (s + 1)(3s − 1).
ds 8
Second derivatives:
d2H1 3 d2H2 L
2
= s, 2
= (3s − 1),
ds 2 ds 4
d2H3 3 d2H4 L
2
= − s, 2
= (3s + 1).
ds 2 ds 4
Converting to physical coordinate x with x =
L
2 (1 + s), we have

2 2 2 2
d 2 d d 2 d 4 d
 

= , 2
=  
2
= 2 2.
dx L ds dx L ds L ds
 

18 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
So the curvature field is
d2w 2
4 d Hi
κ(x) = 2 = di
X

dx i=1 dx 2
 




w1 
 
 

θ1
 
4  d2H1 2 2
d H2 d H3 d H4 2
 




= 2  ds2 .
 
  
  
L ds2 ds2 ds2  
w2
 
 
| {z }  
 
 
Bκ(s) 





θ2 


Substituting the above second derivatives gives

the explicit B-row:
 

Bκ(s) =  L6s2 3s−1

L − L6s2 3s+1
L

 .

which is ready to use in the bending stiffness

integral,
Z
1 > L
k = −1 EIBκ Bκ ds.
2

19 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
For shear, the third derivatives are constants,
d3H1 3 d3H2 3L d3H3 3 d3H4 3L
3
= , 3
= , 3
=− , 3
= ,
ds 2 ds 4 ds 2 ds 4
Thus d3Hi/dx3 = (2/L)3d3Hi/ds3, which makes
V (x) = EIw000(x) piecewise affine in s (constant
in each element for a uniform EI and no dis-
tributed load).

20 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
For uniform beams with EI = constant, the inte-
gration can be done in closed form resulting in
the stiffness matrix,

1
Z
> L
k = EI −1 EIBκ Bκ ds
 2 




12 6L −12 6L 



 
 

6L 4L2 −6L 2L2

 
 
 

= kf
 
 
 
 
−12 −6L 12 −6L
 
 
 
 
 
 

6L 2L2 −6L 4L2

 
 
 

where kf = EI
L3 is a measure of the flexural stiff-
ness of the element.

21 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
The stiffness matrix is symmetric and relates
the nodal displacements and rotations (wi, θi) to
shear force and moments (Vi, Mi):
    
   
6L −12 6L
   
V1 12 w1

 
   


 
   


 
   


 
   


 
   


 
   


EI 6L 4L2 −6L 2L2 θ1

   
M1

 
   

 
   


 
   

 
= 3
    
 
 
L
    
V −12 −6L −6L w
   
12

 
   
 
2 2

 
   

 
  
 

 
  
 

    
 

 
   

 

6L 2L2 −6L 4L2

    
M2 θ2

 
   


 
   


 
 
 

   

References (classic treatments):

T. J. R. Hughes, The Finite Element Method:
Linear Static and Dynamic Finite Element
Analysis, Dover.
O. C. Zienkiewicz & R. L. Taylor, The Finite
Element Method, Vol. 1.
R. D. Cook et al., Concepts and Applications
of Finite Element Analysis.

22 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Derivation of the 3-node quintic
C 1 Hermite shape functions
For a 3-node beam element (nodes at x = 0, L/2, L)
with two DOF per node: transverse displacement
wi and rotation θi, i = 1, 2, 3.
Total DOFs per element: 3 × 2 = 6.
The nodal DOF are (2 per node):

w(0.0) = w1, w0(0.0) = θ1,

w(0.5) = w2, w0(0.5) = θ2,

w(1.0) = w1, w0(1.0) = θ3,

23 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 w1 w2 w3
θ2 θ3
x
θ1 L
V1 V2 M2 V3 M3
x
M1 L
Three-node Euler–Bernoulli beam element:
(a) nodal DOFs (wi, θi),
(b) nodal actions (Vi, Mi).

For a 3-node beam with 2-DOF per node, there

are six DOF. To match the 6 DOF, we define a
quintic polynomial with 6 amplitudes ai.

w(ξ) = a0 +a1ξ+a2ξ 2 +a3ξ 3 +a4ξ 4 +a5ξ 5, ξ ∈ [0, 1]

dw
For Euler-Bernoulli, θ = dx and (with x = Lξ )

1 dw
θ(ξ) =
L dξ

24 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Choose the nodal DOFs
   




w1  



w(0) 



   
1 0
   
θ1 L wξ (0)
   
   
   
   
   
   ! 

w 12
   


 w2 








{q} = 




 = 




!  .
1 0 1
θ2 L wξ 2
   
   
   
   
   
   
   

w3 w(1)
   
   
   
   
   
   
1 0
   
θ3 L wξ (1)
   
   

Imposing these 6 Hermite cardinal conditions to

solve for a0, . . . , a5 we can then regroup w(ξ) by
DOFs:
6
w(ξ) = Ni(ξ)di
X

1
= N1w1 + N2θ1 + N3w2 + N4θ2 + N5w3 + N6θ3.

25 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
where
N1(ξ) = 24ξ 5 − 68ξ 4 + 66ξ 3 − 23ξ 2 + 1,
4 3 2
!

N2(ξ) = Lξ 4ξ − 12ξ + 13ξ − 6ξ + 1 ,

2 2
!

N3(ξ) = 16ξ ξ − 2ξ + 1 ,
2 3 2
!

N4(ξ) = 8Lξ 2ξ − 5ξ + 4ξ − 1 ,
2 3 2
!

N5(ξ) = ξ −24ξ + 52ξ − 34ξ + 7 ,

2 3 2
!

N6(ξ) = Lξ 4ξ − 8ξ + 5ξ − 1 .

or as factored polynomials revealing the nodal

properties.

N1(ξ) = (ξ − 1)2(2ξ − 1)2(6ξ + 1)

N2(ξ) = Lξ(ξ − 1)2(2ξ − 1)2
N3(ξ) = 16ξ 2(ξ − 1)2
N4(ξ) = 8Lξ 2(ξ − 1)2(2ξ − 1)
N5(ξ) = −ξ 2(2ξ − 1)2(6ξ − 7)
N6(ξ) = Lξ 2(ξ − 1)(2ξ − 1)2

These satisfy the Hermite cardinal (intepolation)

conditions at ξ = 0, 12 , 1 :
26 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 At ξ = 0 : N1 = 1, N2/L = ∂w/ ∂ξ|0 = 1
(for the θ-shape), and the others vanish ap-
propriately; similarly for the mid and right
nodes.
The squared factors ξ 2, (ξ − 1)2, and (2ξ − 1)2
ensure that at non-associated nodes both
the value and slope vanish, delivering C 1
nodal continuity. The extra linear factors
(6ξ ± 1) and (2ξ − 1) ”tilt” the basis so that
each shape hits its unit target at its own
node (value or slope) while remaining zero
elsewhere.

27 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
On the interval s ∈ [−1, 1] with the affine map
ξ = s+1
2 . Then x = L
2 (1 + s) and d
dx = 2 d
L ds . Using th
DOF ordering {w1, θ1, w2, θ2, w3, θ3} at s = −1, 0, 1,
the Hermite functions become
s2 3 2
!

H1(s) = 3s − 2s − 5s + 4
4
L 2 3 2
!

H2(s) = s s − s − s + 1
8
H3(s) = s4 − 2s2 + 1
L 4 2
!

H4(s) = s s − 2s + 1
2
s2 3 2
!

H5(s) = −3s − 2s + 5s + 4
4
L 2 3 2
!

H6(s) = s s + s − s − 1
8

28 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
or in factored form,

H1(s) = 14 s2(s − 1)2(3s + 4),

H2(s) = L8 s2(s − 1)2(s + 1)
H3(s) = (s − 1)2(s + 1)2,
H4(s) = L2 s(s − 1)2(s + 1)2
H5(s) = − 14 s2(s + 1)2(3s − 4),
H6(s) = L8 s2(s − 1)(s + 1)2

29 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 H (w ) H (w ) H (w ) H /L (3 ) H /L (3 ) H /L (3 )
1 1 3 2 5 3 2 1 4 2 6 3

3-Node Quintic Hermite Beam Shape Functions

1
Dimensionless H (s)
i

0.5

-0.25
-1 0 1
Natural coordinate s 2 [-1,1]

H (w ) H (w ) H (w )
1 1 3 2 5 3

Displacement Shape Functions: H , H , H

1 3 5

1
H (displacements)

0.5

0
i

-1 -0.5 0 0.5 1
s

H /L (3 ) H /L (3 ) H /L (3 )
2 1 4 2 6 3

Rotation Shape Functions (normalized): H /L, H /L, H /L

2 4 6
0.2

0.1
H /L (rotations)

-0.1
i

-0.2
-1 -0.5 0 0.5 1
s

30 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Checking the cardinal conditions (interpolation
properties): H1(−1) = 1, L2 H20 (−1) = 1, H3(0) =
1, L2 H40 (0) = 1, H5(1) = 1, L2 H60 (1) = 1; all other
value/slope tests vanish accordingly.
Using the normalized coordinate ξ ∈ [0, 1] with
x = Lξ, then

d 1 d d2 1 d2
= , 2
= 2 2.
dx L dξ dx L dξ

Let the DOFs be {d} = [w1, θ1, w2, θ2, w3, θ3]> at
ξ = 0, 12 , 1. From the quintic (5th-order) Hermite
cardinal conditions (value and slope at the three
nodes), the displacement field is
6
w(ξ) = Ni(ξ)di,
X

i=1

Taking two derivatives for curvature gives a cubic

31 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
curvature field,
 




w1 
 
 
θ1
 
 
 
 
 
 

d2w 
 

w2 





κ(ξ) = 2 =  B1 B2 B3 B4 B5 B6

, 





dx | {z } θ2








 
Bκ(ξ) 




w3
 
 
 
 
 
 
 
θ3
 
 

1 d2Ni
Bi(ξ) = 2 2
L dξ

32 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Carrying out the differentiations,
2 3 2
!

B1(ξ) = 2
240ξ − 408ξ + 198ξ − 23 ,
L
2 3 2
!

B2(ξ) = 40ξ − 72ξ + 39ξ − 6 ,

L
32 2
!

B3(ξ) = 2
6ξ − 6ξ + 1 ,
L
16 3 2
!

B4(ξ) = 20ξ − 30ξ + 12ξ − 1 ,

L
2 3 2
!

B5(ξ) = 2
−240ξ + 312ξ − 102ξ + 7 ,
L
2 3 2
!

B6(ξ) = 40ξ − 48ξ + 15ξ − 1 .

L
Using this B-row, for pure bending with constant
EI, gives the closed-form element stiffness
Z
1 >
k = EI 0 Bκ BκLdξ

The same result can also be obtained with s ∈

[−1, 1], ξ = s+1
2 , x = L
2 (1 + s), so

d2 2
2 2
d
 

2
= 
 

dx L ds2
33 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Then
2 3 2
B1(s) = L 2 (30s − 12s − 15s + 4) ,
1 3 2
B2(s) = L (10s − 6s − 3s + 1)
16 2
B3(s) = L 2 (3s − 1) ,
1 3
B4(s) = L (40s − 24s)
2 3 2
B5(s) = L2 (−30s − 12s + 15s + 4) ,
1 3 2
B6(s) = L (10s + 6s − 3s − 1)
Z
1 > L
k = EI −1 Bκ Bκ ds
2
Evaluating the integral gives the symmetric 6 × 6
matrix (DOF order [w1, θ1, w2, θ2, w3, θ3] ):
 



5092 1138L −3584 1920L −1508 242L 


 
1138L 332L2 −896L 320L2 −242L 38L2
 
 
 
 
 
 
kf −3584 −896L 7168 0 −3584 896L
 
 
 
k= 


 .
35 1920L 320L2



 0 1280L 2
−1920L 320L 2




 
 
−1508 −242L −3584 −1920L 5092 −1138L
 
 
 
 
 

38L2 896L 5 · 64L2 −1138L 332L2

 

242L 

34 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 where kf = EI/L3. Interchanging the rows and
columns with the alternative DOF order

[w1, w2, w3, θ1, θ2, θ3] ,

grouped by displacements first and rotations

second, we can express the stiffness matrix in
block form:
 

EI  A LB ? 
k= 3
 
35L  L (B ?)> L2C 
 


35 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
with 3 × 3 matrix partitions,
 




5092 −3584 −1508 
 
 
A = −3584 7168 −3584 ,








 
 
 
−1508 −3584 5092
 
 

 




1138 1920 242 



 

B ? = −896
 





0 896





,
 
 
−242 −1920 −1138
 
 

 




332 320 38 



 
 
C = 320 1280 320








 
 
 
38 320 332
 
 

36 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
In this form,
the inner blocks are dimensionless; the L
and L2 prefactors on B and C immediately
show kww ∼ EI/L3, kwθ ∼ EI/L2 and kθθ ∼
EI/L.
A and C are symmetric Toeplitz-like (end
nodes mirror about the mid-node).
B ? shows the expected odd symmetry w.r.t.
the center (sign flips reflect the slope DOF
orientation).
The 6 × 6 stiffness matrix has rank = 4 (two
rigid-body modes: constant and linear w (ro-
tation) as expected and can be shown to be
positive semidefinite with zero determinant.
The matrix cannot be inverted without first
applying essential kinematic constraints to
remove these rigid body modes.

37 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Observations on the interior node DOF
vs. end nodes
At end nodes (1 and 3 ):
The displacement shape functions N1, N5
interpolate end values and have global
influence. Their associated slopes, N2
and N6, are odd-like partners. Because
these are not symmetric/antisymmetric
about the element midpoint, the stiffness
cross-terms kw,θ at ends do not vanish.
Because these cross-coupling stiffnesses
at the end nodes are nonzero, it means
that an imposed slope at the mid-node
produces a resisting force at the same
node in the displacement direction, or
vice versa.
At the mid node (2):
DOF 3 = w2 (transverse displacement at

38 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 the mid node).
DOF 4 = θ2 (slope at the mid node).
The displacement shape N3(ξ) = 16ξ 2(ξ − 1)2
is a bubble function: it vanishes at ξ = 0
and ξ = 1, and is only nonzero in the inte-
rior. Its slope partner N4(ξ) is antisymmetric
about ξ = 12 . Together, they form a sym-
metric/antisymmetric pair that cancels the
energy cross term.
N3 (mid displacement) is even about the
midpoint.
N4 (mid slope) is odd about the midpoint.
Their curvature functions B3(ξ), B4(ξ) inherit
this symmetry.
In particular, B3 is quadratic in ξ and sym-
metric about the element midpoint, while B4
is cubic and antisymmetric. Their product

39 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 integrates to zero over [0, 1]. That is,
1 1
k34 ∝
Z Z

0 B3(ξ)B4(ξ)dξ = 0 B4(ξ)B3(ξ)dξ = k43

is an integral of an even × odd function,

which is odd about the midpoint, hence zero.
That is the reason why the coupling between
the transverse displacement w2, and rotation
θ2 at node 2 is zero, k34 = k43 = 0.
In other words, k34 = k43 = 0 says there is no
direct stiffness coupling between the mid-
node displacement and mid-node slope.

40 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Physical interpretation
At interior node 2, displacement w2 and slope
θ2 are energetically orthogonal: imposing a
pure interior displacement bubble does not
create any nodal moment reaction at the
same location, and imposing a pure inte-
rior slope bubble does not create any nodal
shear reaction at that node.
This is in contrast to end nodes, where slope
and displacement are not orthogonal, be-
cause end displacements and slopes are
directly coupled through bending energy.

41 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Practical implications
w2 (bubble displacement) enriches the trans-
verse field to give better bending represen-
tation inside the element.
θ2 (bubble slope) enriches the slope distribu-
tion antisymmetrically.
They do not couple at the same interior node
2, but they still couple strongly with the end-
node DOFs (w1, θ1, w3, θ3).
As a result, the 3-node beam element im-
proves accuracy over the 2-node element
yet does not change the number of global
rigid-body modes (still 2: rigid translation
and rigid rotation).

42 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
In summary
N3 is a bubble → interior enrichment.
N3 is even, N4 odd → curvature functions
are orthogonal.
This orthogonality makes k34 = 0.
Physically, at the interior node, displace-
ment and slope DOFs represent indepen-
dent deformation modes; unlike the end
nodes, they don’t directly ”talk” to each other
in strain energy.

43 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Nodal forces and moments with equiv-
alent work as distributed loads
The principle of virtual work naturally extends
to derive the consistent nodal force vector for a
beam element under a distributed load, such as
a uniform load q(x). Here is the process.

Step 1. Virtual Work (Weak Form)

Let q(x) be the distributed transverse load
per unit length acting on the beam over x ∈
[0, L].
The weak form of the boundary value prob-
lem (BVP) for a beam under a distributed
load is:
L d2w d2δw L
·
Z Z

0 EI 2 2
dx = 0 q(x)δw(x)dx
dx dx
After introducing approximations for w and
δw, the left-hand side leads to the stiffness
matrix [ke].
44 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 The right-hand side is the virtual work done
by the external load, leading to the consis-
tent nodal force vector with equivalent work
as the original distributed load q(x) acting
on the beam.

Step 2. Finite Element Approximation

For the 2-node cubic beam element, intro-
duce the Hermite shape functions for the
virtual displacement.
4
δw = Ni(ξ)δdi
X

i=1
= [δw1, δθ1, δw2, δθ2] [N1, N2, N3, N4]>

45 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 Then:
L 1
· δw(x)dx = · δw(ξ) Ldξ
Z Z

0 q(x) 0 q(x(ξ))
 
>   




δw1 







N1(ξ)  



     
     

δθ1 N2(ξ)
     
     
1
  Z   

= 0 q · L dξ
     
     
     
     
δw2 N3(ξ)
     
     
     
     
     
     
     


δθ2 



N4(x)







| {z }
fe

and we identify the consistent nodal force

vector producing equivalent work to the orig-
inal distributed load q:
   

 

(Vq )1 N1(ξ)

 
 

 
 

 
 

 
 

 
  

 
  
 
(Mq )1 N2(ξ)

 
  

 
  
1

 
  

{fe} =  ·
Z
= 0 q L dξ
   
 
 
  
N3(ξ)

(Vq )2
 
  

 
  

 
  

 
  

 
  

 
  
   
(Mq )2 N4(ξ)

 
  

 
  

 

 

46 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 For a uniform load q = constant, for the 2-
node cubic Hermite beam element on ξ ∈
[0, 1], the consistentelement load vector is
     
1 qL

 

(Vq )1

 
    

 
    
2 2

 
    

 
    

 
    
qL2

 
    
L
 
(Mq )1

 
    

 
    
     
 
12 12
{fe} =  = qL =
     
   
   
1 qL

    
(Vq )2
 
    

 
    
2 2

 
    

 
    

 
    
qL2
     
L
 

− 12 − 12
     
(Mq )2

 
    

 
    

 

 

The total resultant force qL is equally divided

between the two nodes, and due to the cou-
pling of the interpolation between displace-
ment and rotation DOF, symmetric end node
moments with concave bending are applied.

47 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 qL qL 2
V1 = 2 V2 = 2 M2 = − qL
12
x
1 2
qL2
M1 = 12 L
2-node Euler–Bernoulli beam element: consistent nodal
loads for uniform q.

48 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Work-Equivalent Nodal Loads for 3-
node Quintic Beam Element
For the 3-node quintic Hermite element on ξ ∈
[0, 1] (with x = Lξ ), with DOF order [w1, θ1, w2, θ2, w3, θ3]>
Using the shape functions given earlier, the 6 × 1
work-consistent nodal load vector is
Z
L > Z
1 >
f= 0 N qdx = qL 0 N (ξ)dξ
     
7 7




N1  


 30







 30 qL




     
L 1 2
     
N2 qL
     
     
60 60
     
     
     
     
8 8
     
Z
1


 N3 





15






15 qL



= qL 0





 dξ = f = qL 




 = 




 .
N4
     












0 







0 



     
     
7 7
N5 30 qL
     
     
30
     
     
     
     
L
− 601 qL2
     
N6 − 60
     
     

49 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 qL 2qL qL 2
V1 = 6 V2 = 3 V3 = 6 M3 = − qL
12
M22= 0 x
1 3
qL2
M1 = 12 L
Three-node Euler–Bernoulli beam element with consistent
nodal loads for a uniform load q over L.

50 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Static Condensation of Interior Node
in Terms of Exterior Nodes
It is instructive to perform a static condensa-
tion of the mid-node (w2, θ2) from the 6 × 6 stiff-
ness matrix to show the structure and observa-
tions of the reduced 4 × 4 stiffness matrix (Schur
complement) recovered for the master DOFs
(w1, θ1, w3, θ3).

Partition the 3-node ( 6 × 6 ) stiffness

Use the DOF order [w1, θ1, w2, θ2, w3, θ3].
Split retained exterior r = [w1, θ1, w3, θ3] and de-
pendent interior u = [w2, θ2] DOF into partitions:
 




w1 
        
 

Krr Kru   dr fr θ1 w2 

 
      
      

= , dr = , du = .
      
      
         
         
Kur Kuu du fu w3 θ2
         
         
 
 
 
 
 


θ3 


51 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
From the 3-node quintic Hermite element (con-
stant EI ), the subblocks are (extracting the en-
tries that matter for condensation):
 

kf  7168 0 

Kuu =  ,
35  0 1280L2 


 




−3584 1920L 



 
 

kf −896L 320L2
 
 
>
 

Kru = , Kur = Kru .

 
 
 
35 −3584 −1920L
 
 
 
 
 
 
 

320L2
 


896L 


Two key observations right away are:

The condensed block Kuu is diagonal →
w2 and θ2 are energetically decoupled in-
ternal modes (orthogonality of the midnode
shapes).
This diagonal Kuu exposes symmetry.

52 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Static condensation (Guyan reduction)
Assuming no applied interior forces ( fu = 0 ), the
slave DOFs satisfy
−1
du = −Kuu Kur dr .

Insert this into the master equilibrium to get the

condensed stiffness
−1
K (c) = Krr − KruKuu Kur

Carrying out this product yields the 4 × 4 matrix

(in master order [w1, θ1, w3, θ3] ):
 




12 6L −12 6L 



 
 

6L 4L2 −6L 2L2

 
 
 

K (c) = kf .
 
 
 
 
−12 −6L 12 −6L
 
 
 
 
 
 

6L 2L2 −6L 4L2

 
 
 

53 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
or explicitly
 
12EI 6EI



 L3 L2 − 12EI
L3
6EI
L2




 
 
6EI 4EI
− 6EI 2EI
 
 
L2 L L 2 L
 

K (c) = .
 
 
 

− 12EI − 6EI 12EI 6EI

 
−
 
 
L3 L2 L 3 L2
 
 
 
 
6EI 2EI
− 6EI 4EI
 
 
 
L2 L L 2 L

Observations on the reduced stiffness

matrix
1. Exact recovery of the 2-node (cubic Her-
mite) beam stiffness.
The Schur complement equals the stan-
dard 2-node Euler-Bernoulli element. Thus,
a 3-node quintic element condensed at
the interior node reproduces the classic
2-node behavior at the ends exactly (for a
constant EI). That is, for EI = constant,
[w1, θ1, w3, θ3] are exact at end nodes for
any load.
54 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 2. Interior modes are true bubbles.
Because Kuu is diagonal, the mid-node
displacement w2 and slope θ2 act as in-
dependent bubble modes that enrich the
interior bending field. They improve intra-
element accuracy but do not alter the
end-end stiffness once condensed.
3. Using the 3-node element without conden-
sation gives better within-element fields (mo-
ments/shear) and higher accuracy per ele-
ment.

55 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
Static Condensation of Consistent Load
Vector
In the following, we derive the condensed con-
sistent load vector for a uniform transverse load
q and show that it exactly matches the 2-node
(cubic Hermite) result.
Partition into retained end DOF r = [w1, θ1, w3, θ3],
and dependent interior node 2 DOF, u = [w2, θ2]:
 
7
30 qL
 
 
 
 
   
 
1 2 8
qL qL
 
   
   
60 15
fr = , fu = .
   
   
   
7
   

30 qL 0
   
   
 
 
 
 

− 601 qL2
 
 
 

56 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
From the stiffness partition
 




Krr Kru 
 
 
Kur Kuu
 
 

the condensed master load vector (Schur com-

plement with fu included) is
 
1
2 qL
 
 
 
 
 
 
1 2
qL
 
 
−1
 

f (c) = fr − KruKuu fu = 12
.
 
 
 
1
 

2 qL
 
 
 
 
 
 

− 121 qL2
 
 
 

This matches the well-known 2-node cubic Her-

mite consistent loads. We find the 3-node el-
ement’s interior DOFs ( w2, θ2 ) are true bubble
modes: once condensed, they do not alter the
end-end equilibrium loads; we recover exactly
the 2-node nodal equivalents for uniform q (and
similarly for K ).

57 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
References
[1] [Link]
ernoulli_beam_theory.
[2] [Link]
lement_method_in_structural_mechanics

58 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.
 3-Node Quintic
Hermite Elements
based on Euler-
Bernoulli Beam Theory

Repost this RETWEET

– Powered by LaTeX �
Visit My LinkedIn Profile

59 of 59
LinkedIn � - Dr. Lonny Thompson, Clemson University, Sept 28, 2025.