University of North Georgia

Nighthawks Open Institutional Repository

Faculty Publications Department of Mathematics

9-9-2015

An example of Lagrangian for a non-holonomic

system
Piotr W. Hebda
University of North Georgia, [email protected]

Beata A. Hebda
University of North Georgia, [email protected]

Follow this and additional works at: http://digitalcommons.northgeorgia.edu/math_facpub

Part of the Mathematics Commons

Recommended Citation
Hebda, Piotr W. and Hebda, Beata A., "An example of Lagrangian for a non-holonomic system" (2015). Faculty Publications. Paper 3.
http://digitalcommons.northgeorgia.edu/math_facpub/3

This Article is brought to you for free and open access by the Department of Mathematics at Nighthawks Open Institutional Repository. It has been
accepted for inclusion in Faculty Publications by an authorized administrator of Nighthawks Open Institutional Repository.
An example of a Lagrangian for a non-holonomic system

Piotr W. Hebda,a) Beata A. Hebda

Department of Mathematics, University of North Georgia, Oakwood, Georgia, 30566, USA

An adjustable two-mass-point Chaplygin Sleigh is used as an example of a non-holonomic

system. Newtonian equations of motion based the assumption of zero virtual work done by
constraints are calculated. A Lagrangian that reproduces these equations as its unmodified Euler-
Lagrange equations is then explicitly given. The Lagrangian uses variables that are present in the
Chaplygin Sleigh equations of motion, as well as some additional variables. Some of the Euler-
Lagrange equations of that Lagrangian are non-differential. These non-differential equations
automatically and completely reduce out all of these additional variables, so that only the
variables that appear in the original equations of motion remain in the final dynamics of the
system.

I. INTRODUCTION

While the Newtonian equations of motion seem to be physically more fundamental than the

Lagrangian that produces these equations as its Euler-Lagrange equations, the Lagrangian is still

of great interest, since it provides a natural framework for further study of the system. For

example it is a starting point for calculating the Hamiltonian and the Poisson brackets structure.

The problem of constructing a Lagrangian for given equations of motion has been therefore

extensively studied, but it is still not completely resolved. 1)

Quite often we study a mechanical system for which a Lagrangian is already known, but

which is subsequently modified by imposing additional constraints. The constraints

_____________________________
a)
Author to whom correspondence should be addressed. Electronic mail: [email protected]
modify the original equations of motion, and the modifications then lead to the need of

modifications of the Lagrangian. Modifying the Lagrangian is quite simple if the constraints are

of holonomic type (constraints that could be expressed by restricting the allowable positions of

the system). In this case the new Lagrangian is obtained by adding the constraints, each

constraint multiplied by its own so called Lagrange multiplier, to the original Lagrangian.2) In

the case of non-holonomic constraints (these are constraints that involve velocities and cannot be

reduced to restricting the positions only) the situation is not so simple. Adding constraints

multiplied by the Lagrange multipliers to the original Lagrangian will produce equations of

motion that are acceptable from mathematical point of view, but they are different from actual

physical equations that result from such constraints. Specifically, in the case of non-holonomic

constraints, the constraints forces resulting from the use of Lagrange multipliers do not satisfy

the condition of zero virtual work, which is expected to be satisfied in real-world mechanics. 3)

Because of that fact the use of Lagrange multipliers is generally rejected in the case of non-

holonomic constraints. A commonly accepted approach for such a case is not to modify a

Lagrangian at all, but to obtain the Euler-Lagrange equations from the Lagrangian, and then

modify these equations to include forces resulting from the constraints.3) However, as the

resulting final equations of motion are not the usual Euler-Lagrange equations of a known

Lagrangian anymore, some advantages of using the Lagrangian are lost.

Another approach to the non-holonomic constraints can be done by adapting the Bateman-

Morse-Feshbach approach.4,5) In this approach a Lagrangian for an arbitrary system of equations

of motion is constructed by Lagrangian being equal to sum of all equations of motion, each

equation multiplied by a new variable. The method is somewhat analogous to Lagrange

multipliers method, extended not only to equations representing constraints, but to all equations

2
of the system. This approach results in getting correct equations of motion directly as the Euler-

Lagrange equations, but it also creates additional non-physical variables that were not existing in

the original equations of motion. The additional variables are then present in the Lagrange-

Hamilton formalism that follows, and it is not clear how to interpret them. So this approach,

while relatively simple, is not commonly accepted as a resolution of the Lagrangian construction

problem.

In this work we show a Lagrangian for one specific example of a non-holonomic system

using the explicit solutions of the equations of motion to construct the Lagrangian, rather than a

combination of the kinetic and potential energies used in a typical process of getting a

Lagrangian. On some level this is a quite satisfying approach, since one may claim that solutions

are more fundamental objects than kinetic and potential energies - solutions of equations are

directly observable and they always exists, while kinetic and potential energies are more abstract

constructs, and in some cases may not exists at all. On the other hand, generalization of our

approach to other examples may be problematic, because in many situations we do not have

explicit solutions of the equations of motion, and in these cases we will not be able to get the

Lagrangian explicitly, which may be a serious drawback. However, some preliminary results6)

suggest that even in such cases we can prove the existence of a Lagrangian, which by itself is an

interesting result.

Our approach produces the correct equations of motion directly from the Lagrangian, as the

Euler-Lagrange equations, with no further modifications necessary. Similarly to the Bateman-

Morse-Feshbach approach,4,5) we also use variables that do not appear in the original equations

of motion. We avoid the basic difficulty of the Bateman-Morse-Feshbach approach though,

because some of the Euler-Lagrange equations obtained from our Lagrangian are constrains

3
rather than differential equations. The constraints then automatically eliminate all variables that

do not appear in the original equations of motion, while leaving the original equations of motion

intact.

The organization of our presentation is as follows:

In section II, we define the adjustable two-mass-point Chaplygin Sleigh mechanical system, we

show that it is a non-holonomic system, and we derive its equations of motion.

In section III, we present the proposed Lagrangian and we derive and simplify its Euler-

Lagrange equations. We show that some of the equations are identical to the equations of motion

for the adjustable two-mass-point Chaplygin Sleigh presented in section II. We also show that

other Euler-Lagrange equations are constraints or time derivatives of the constraints, and that

these constraints eliminate all the additional variables used to create the Lagrangian, leaving in

the final dynamics only the variables that were present in the original equations of motion.

In section IV, we comment on a possibility to use Dirac’s Theory of Constraints7,8) to obtain

the Hamiltonian formalism for our Lagrangian.

II. THE ADJUSTABLE TWO-MASS-POINT CHAPLYGIN SLEIGH

Physically our mechanical system will be made of two particles, each moving freely in two

dimensions, and each having a unitary mass. Their position will be given by the usual variables

( x1 , x2 , x3 , x4 ), with the variables ( x1 , x2 ) describing the first particle, and variables ( x3 , x4 )

describing the second.

4
 Using the time derivatives (v1 , v2 , v3 , v4 ) of the variables ( x1 , x2 , x3 , x4 ) produces the following

equations of motion, still unmodified by the constraints, as:


v1  0 (1i)


v2  0 (1ii)


v3  0 (1iii)


v4  0 (1iv)


x1  v1 (1v)


x 2  v2 (1fvi)


x3  v3 (1vii)


x 4  v4 , (1viii)

where a dot above a variable means the time derivative.

To obtain an adjustable two-mass-point Chaplygin Sleigh from this free system, we impose

two constraints:

(v3  v1 )( x3  x1 )  (v4  v2 )( x4  x2 )  0 (2i)

5
(v1  v3 )( x4  x2 )  (v2  v4 )( x3  x1 )  0, (2ii)

that are supposed to be satisfied by all solutions of the equations of motion.

The constraint (2i) is holonomic, since it is obtained by taking a time derivative of

( x3  x1 )2  ( x4  x2 )2  r 2 .

r may vary with different initial positions of the particles, but the constraint (2i) assures that it

remains constant during the motion. The possibility of having different values of r is the reason

for calling our Chaplygin Sleigh “adjustable”.

v1  v3 v2  v4
The constraint (2ii) is non-holonomic. It can be interpreted as the velocity ( , )
2 2

x1  x3 x2  x4
of the center of mass ( , ) being parallel to the vector ( x3  x1 , x4  x2 ) which is
2 2

v1  v3 v2  v4
starting at the first particle, and ending at the second. The vector ( , ) is parallel to
2 2

( x3  x1 , x4  x2 ) , since the constraint (2ii) says it is perpendicular to ( x4  x2 ,( x3  x1 )).

This constraint (2ii) is non-holonomic since it allows a rotation of the particles around its

center of mass, and also it allows a translation along the vector starting at the first particle and

ending at the second. Combinations of these rotations and translations allow to reach all possible

positions of the particles, once the distance between the particles is established by the constraint

(2i). Therefore (2ii) is not imposing any restrictions on the possible positions of the system,

while imposing restrictions on possible velocities.

The constraints (2) may be written in standard form as

a jk ( x1 , x2 , x3 , x4 )  vk  0,
4

k 1
(3)

where j  1 represents the constraint (2i), and j  2 represents the constraint (2ii). Direct

comparison then gives:

6
a11  x1  x3 (4i)

a12  x2  x4 (4ii)

a13  x3  x1 (4iii)

a14  x4  x2 (4iv)

a21  x4  x2 (4v)

a22  x1  x3 (4vi)

a23  x4  x2 (4vii)

a24  x1  x3 (4viii)

We assume that the forces of the constraint are such that they do zero work during

instantaneous virtual displacements. It can be shown3) that from this assumption we get the

constraint’s forces to be

Fi   j a ji ,
2

j 1

These forces then modify the equations of motion (1), giving


v1  1 ( x1  x3 )  2 ( x4  x2 ) (5i)


v2  1 ( x2  x4 )  2 ( x1  x3 ) (5ii)


v3  1 ( x3  x1 )  2 ( x4  x2 ) (5iii)


v4  1 ( x4  x2 )  2 ( x1  x3 ) (5iv)


x1  v1 (5v)


x 2  v2 (5vi)

7

x3  v3 (5vii)


x 4  v4 , (5viii)

The constraints (2) must be preserved in time. So their time derivatives must be zero.

This fact and the equations of motion (5) give us, after somewhat tedious calculations, the

following expressions for 1 and 2 :

(v3  v1 ) 2  (v4  v2 ) 2
1  
2( x3  x1 ) 2  2( x4  x2 ) 2

v3v2  v4v1
2 
( x4  x2 ) 2  ( x3  x1 ) 2

Substituting that into the equations o motion (5), after some simplifications, gives:


(v3  v1 )2  (v4  v2 )2 v3v2  v4v1
v1   ( x3  x1 )   ( x4  x2 ) (6i)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 )2  ( x3  x1 )2


(v3  v1 )2  (v4  v2 )2 v3v2  v4v1
v2   ( x4  x2 )   ( x3  x1 ) (6ii)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 )2  ( x3  x1 )2


(v3  v1 )2  (v4  v2 )2 v3v2  v4v1
v3    ( x3  x1 )   ( x4  x2 ) (6iii)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 )2  ( x3  x1 )2


(v3  v1 ) 2  (v4  v2 )2 v3v2  v4v1
v4    ( x4  x2 )   ( x3  x1 ) (6iv)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 )2  ( x3  x1 ) 2


x1  v1 (6v)


x 2  v2 (6vi)


x3  v3 (6vii)


x 4  v4 , (6viii)

8
(v3  v1 )( x3  x1 )  (v4  v2 )( x4  x2 )  0 (6ix)

(v1  v3 )( x4  x2 )  (v2  v4 )( x3  x1 )  0 (6x)

The last two equations are constraints. The equations (6) are the final equations of motion of

the adjustable two-mass-point Chaplygin Sleigh.

III. THE LAGRANGIAN AND EULER-LAGRANGE EQUATIONS

Consider the Lagrangian

2 2 2
  
w1 w w
L  2  3  1 ( y1  w1 )  2 ( y 2  w2 )  3 ( y 3  w3 ) 
2 2 2
 1[ x1  w3 sin( y1 )  y2  w2  1  w2 cos  y1 ] 
y
w1

 w2 sin  y1 ] 
y1
  2 [ x2  w3 cos( y1 )  y3  w3 
w1

 w2 cos  y1 ] 
y1
 3[ x3  w3 sin( y1 )  y2  w2 
w1

 w2 sin  y1 ] 
y1
  4 [ x4  w3 cos( y1 )  y3  w3  (7)
w1
 1[v1  w1 w3 cos( y1 )  w1 w2 sin  y1 ] 
  2 [v2  w1 w3 sin( y1 )  w1 w2 cos  y1 ] 
 3[v3  w1 w3 cos( y1 )  w1 w2 sin  y1 ] 
  4 [v4  w1 w3 sin( y1 )  w1 w2 cos  y1 ] .

In the Lagrangian (7) all the 25 variables, namely

( x1 , x2 , x3 , x4 , v1 , v2 , v3 , v4 , y1 , y2 , y3 , w1 , w2 , w3 , 1 , 2 , 3 , 1 , 2 , 3 , 4 ,1 ,2 ,3 ,4 ) (8)

are treated on equal footing, as independent variables. (The specific formula for this Lagrangian

was obtained using the procedure described by authors in a separate work.6))

9
To obtain the equations of motion from the Lagrangian (7), we use the standard Euler –

d  L  L
Lagrange equations in the form  , where qi , i  1,...,25 represents all 25
dt   q  qi


 i

variables (8). We obtain 25 equations:


 w2 sin  y1 ] 
w2
 1  1[ w3 cos( y1 ) 
w1

 w2 cos  y1 ] 
w3
  2 [ w3 sin( y1 ) 
w1

 w2 sin  y1 ] 
w2
 3[ w3 cos( y1 ) 
w1
(9i)
 w2 cos  y1 ] 
w3
  4 [ w3 sin( y1 ) 
w1
 1[ w1 w3 sin( y1 )  w1 w2 cos  y1 ] 
  2 [ w1 w3 cos( y1 )  w1 w2 sin  y1 ] 
 3[ w1 w3 sin( y1 )  w1 w2 cos  y1 ] 
  4 [ w1 w3 cos( y1 )  w1 w2 sin  y1 ]


 2  1  3 (9ii)


 3   2  4 (9iii)

1  0 (9iv)

2  0 (9v)

3  0 (9vi)

4  0 (9vii)

1  0 (9viii)

2  0 (9ix)

3  0 (9x)

10
4  0 (9xi)

w2 y1 w3 y1 w2 y1 w3 y1
w1  1  1  2  2  2  3  2  4  2 
w1 w1 w1 w1
 1  [ w3 cos( y1 )  w2 sin( y1 )]  2  [ w3 sin( y1 )  w2 cos( y1 )]  (9xii)
 3  [ w3 cos( y1 )  w2 sin( y1 )]  4  [ w3 sin( y1 )  w2 cos( y1 )]

y1 y
w2  2  1  [  cos( y1 )]  2  sin( y1 )  3  [ 1  cos( y1 )]  4  sin( y1 ) 
w1 w1 (9xiii)
 1  [ w1 sin( y1 )]  2  [ w1 cos( y1 )]  3[ w1 sin( y1 )]  4 [ w1 cos( y1 )]

y1 y
w3  3  1  sin( y1 )  2  [  cos( y1 )]  3  sin( y1 )  4  [ 1  cos( y1 )] 
w1 w1 (9xiv)
 1  [ w1 cos( y1 )]  2  [ w1 sin( y1 )]  3[ w1 cos( y1 )]  4 [ w1 sin( y1 )]


y1  w1 (9xv)


y 2  w2 (9xvi)


y 3  w3 (9xvii)

 w2 cos y1 
y1
x1  w3 sin( y1 )  y2  w2  (9xviii)
w1

 w2 sin  y1 
y1
x2  w3 cos( y1 )  y3  w3  (9xix)
w1

 w2 cos y1 
y1
x3  w3 sin( y1 )  y2  w2  (9xx)
w1

 w2 sin  y1 
y1
x4   w3 cos( y1 )  y3  w3  (9xxi)
w1

v1  w1 w3 cos( y1 )  w1 w2 sin y1  (9xxii)

v2  w1 w3 sin( y1 )  w1 w2 cos y1  (9xxiii)

v3  w1 w3 cos( y1 )  w1 w2 sin y1  (9xxiv)

11
v4  w1 w3 sin( y1 )  w1 w2 cos y1  (9xxv)

Using equations (9iv) – (9xi) to simplify equations (9i) – (9iii) and (9xii) – (9xiv) and

rearranging the order of equations, we get:


1  0 (10i)


2  0 (10ii)


3  0 (10iii)


y1  w1 (10iv)


y 2  w2 (10v)


y 3  w3 (10vi)

w1  1 (10vii)

w2  2 (10viii)

w3  3 (10xix)

 w2 cos y1 
y1
x1  w3 sin( y1 )  y2  w2  (10x)
w1

 w2 sin  y1 
y1
x2  w3 cos( y1 )  y3  w3  (10xi)
w1

 w2 cos y1 
y1
x3  w3 sin( y1 )  y2  w2  (10xii)
w1

 w2 sin  y1 
y1
x4   w3 cos( y1 )  y3  w3  (10xiii)
w1

v1  w1 w3 cos( y1 )  w1 w2 sin y1  (10xiv)

v2  w1 w3 sin( y1 )  w1 w2 cos y1  (10xv)

12
v3  w1 w3 cos( y1 )  w1 w2 sin y1  (10xvi)

v4  w1 w3 sin( y1 )  w1 w2 cos y1  (10xvii)

1  0 (10xviii)

2  0 (10xix)

3  0 (10xx)

4  0 (10xxi)

1  0 (10xxii)

2  0 (10xxiii)

3  0 (10xxiv)

4  0 (10xxv)

The equations (10i) – (10vi) give us time derivatives of the variables (1 , 2 , 3 , y1 , y2 , y3 ) . The

equations (10vii) – (10xxv) are constraints. Constraints must hold as time progresses, so for

each constraint time derivatives of both sides must be equal. In general taking time derivatives

of existing constraints may create new constraints and/or give time derivatives of the variables

that were not included in the earlier equation. In case of equations (10) taking time derivatives of

constraints creates no new constraints. Instead it gives us time derivatives for all the variables

that had no time derivatives in equations (10), namely time derivatives of the variables

(w1 , w2 , w3 , x1 , x2 , x3 , x4 , v1 , v2 , v3 , v4 , 1 , 2 , 3 , 4 ,1 ,2 ,3 ,4 ) .

This somewhat tedious process produces the following formulas for the time derivatives of

all the 25 variables of the Lagrangian (7):

13

v1  w1 w3 sin( y1 )  w1 w2 cos y1 
2 2
(11i)


v 2  w1 w3 cos( y1 )  w1 w2 sin y1 
2 2
(11ii)


v3  w1 w3 sin( y1 )  w1 w2 cos y1 
2 2
(11iii)


v 4  w1 w3 cos( y1 )  w1 w2 sin  y1 
2 2
(11iv)


x1  v1 (11v)


x 2  v2 (11vi)


x3  v3 (11vii)


x 4  v4 (11viii)


y1  w1 (11ix)


y 2  w2 (11x)


y 3  w3 (11xi)


w1  0 (11xii)


w2  0 (11xiii)


w3  0 (11xiv)


1  0 (11xv)


2  0 (11xvi)


3  0 (11xvii)


1  0 (11xviii)

14

2  0 (11xix)


3  0 (11xx)


4  0 (11xxi)


1  0 (11xxii)


2  0 (11xxiii)


3  0 (11xxiv)


4  0 (11xxv)

and the constraints

 w2 cos y1 
y1
x1  w3 sin( y1 )  y2  w2  (11xvi)
w1

 w2 sin  y1 
y1
x2  w3 cos( y1 )  y3  w3  (11xvii)
w1

 w2 cos y1 
y1
x3  w3 sin( y1 )  y2  w2  (11xviii)
w1

 w2 sin  y1 
y1
x4   w3 cos( y1 )  y3  w3  (11xix)
w1

v1  w1 w3 cos( y1 )  w1 w2 sin y1  (11xx)

v2  w1 w3 sin( y1 )  w1 w2 cos y1  (11xxi)

v3  w1 w3 cos( y1 )  w1 w2 sin y1  (11xxii)

v4  w1 w3 sin( y1 )  w1 w2 cos y1  (11xxiii)

w1  1 (11xxiv)

15
w2  2 (11xxv)

w3  3 (11xxvi)

1  0 (11xxvii)

2  0 (11xxviii)

3  0 (11xxix)

4  0 (11xxx)

1  0 (11xxxi)

2  0 (11xxxii)

3  0 (11xxxiii)

4  0 (11xxxiv)

By direct calculations, using the constraints (11xvi) - (11xxiii) we get

(v3  v1 ) 2  (v4  v2 ) 2 v3v2  v4v1

  ( x1  x3 )   ( x4  x2 ) 
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2
(12i)
  w1 w3 sin( y1 )  w1 w2 cos y1 
2 2

(v3  v1 ) 2  (v4  v2 ) 2 v3v2  v4v1

  ( x2  x4 )   ( x1  x3 ) 
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2
(12ii)
 w1 w3 cos( y1 )  w1 w2 sin  y1 
2 2

(v3  v1 ) 2  (v4  v2 ) 2 v3v2  v4v1

  ( x3  x1 )   ( x4  x2 ) 
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2
(12iii)
  w1 w3 sin( y1 )  w1 w2 cos  y1 
2 2

16
 (v3  v1 ) 2  (v4  v2 ) 2 v3v2  v4v1
  ( x4  x2 )   ( x1  x3 ) 
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2
(12iv)
 w1 w3 cos( y1 )  w1 w2 sin  y1 
2 2

(v3  v1)( x3  x1)  (v4  v2 )( x4  x2 )  0 (12v)

(v1  v3 )( x4  x2 )  (v2  v4 )( x3  x1 )  0 (12v)

If we now use the equations (12i) – (12iv) to replace the right sides of the equations (11i) –

(11iv) and include equations (12iv) and (12v) with other equations, we obtain the following

system of equations of motion:


(v3  v1 ) 2  (v4  v2 ) 2 v3v2  v4v1
v1    ( x1  x3 )   ( x4  x2 ) (12i)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2


(v3  v1 ) 2  (v4  v2 ) 2 v3v2  v4v1
v2    ( x2  x4 )   ( x1  x3 ) (12ii)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2


(v3  v1 )2  (v4  v2 ) 2 v3v2  v4v1
v3    ( x3  x1 )   ( x4  x2 ) (12iii)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2


(v3  v1 ) 2  (v4  v2 )2 v3v2  v4v1
v4    ( x4  x2 )   ( x1  x3 ) (12iv)
2( x3  x1 )  2( x4  x2 )
2 2
( x4  x2 ) 2  ( x3  x1 ) 2


x1  v1 (12v)


x 2  v2 (12vi)


x3  v3 (12vii)


x 4  v4 (12viii)

17
with constraints

(v3  v1)( x3  x1)  (v4  v2 )( x4  x2 )  0 (12ix)

(v1  v3 )( x4  x2 )  (v2  v4 )( x3  x1 )  0 (12x)

v1  w1 w3 cos( y1 )  w1 w2 sin y1  (13i)

v2  w1 w3 sin( y1 )  w1 w2 cos y1  (13ii)

v3  w1 w3 cos( y1 )  w1 w2 sin y1  (13iii)

v4  w1 w3 sin( y1 )  w1 w2 cos y1  (13iv)

 w2 cos y1 
y1
x1  w3 sin( y1 )  y2  w2  (13v)
w1

 w2 sin  y1 
y1
x2  w3 cos( y1 )  y3  w3  (13vi)
w1

 w2 cos y1 
y1
x3  w3 sin( y1 )  y2  w2  (13vii)
w1

 w2 sin  y1 
y1
x4   w3 cos( y1 )  y3  w3  (13viii)
w1

1  w1 (13ix)

2  w3 (13x)

3  w3 (13xi)

1  0 (13xii)

2  0 (13xiii)

3  0 (13xiv)

4  0 (13xv)

18
1  0 (13xvi)

2  0 (13xvii)

3  0 (13xviii)

4  0 (13xix)


y1  w1 (14i)


y 2  w2 (14ii)


y 3  w3 (14iii)


w1  0 (14iv)


w1  0 (14v)


w3  0 (14vi)


1  0 (14vii)


2  0 (14viii)


3  0 (14ix)


1  0 (14x)


2  0 (14xi)


3  0 (14xii)


4  0 (14xiii)


1  0 (14xiv)

19
 
2  0 (14xv)


3  0 (14xvi)


4  0 (14xvii)

Let us stress that the entire system of equations (12), (13), and (14) is completely equivalent to

the system (9) of the Euler-Lagrange equations of the Lagrangian (7).

Let us now interpret the system of equations (12), (13), and (14). First, the constraints (13)

may be interpreted as implicit definition of all the variables

( y1 , y2 , y3 , w1 , w2 , w3 , 1 , 2 , 3 , 1 , 2 , 3 , 4 ,1 ,2 ,3 , 4 ) that do not appear in the adjustable two-

mass-point Chaplygin Sleigh (6), by the variables ( x1 , x2 , x3 , x4 , v1 , v2 , v3 , v4 ) that do appear there.

Moreover, it can be shown that all time derivatives (14) of the variables not appearing in the

Chaplygin Sleigh equations can be obtained directly by taking time derivatives of the constraints

(13), and then using equations (12).

This means that the variables ( y1 , y2 , y3 , w1 , w2 , w3 , 1 , 2 , 3 , 1 , 2 , 3 , 4 ,1 ,2 ,3 , 4 ) are

completely dependent of variables ( x1 , x2 , x3 , x4 , v1 , v2 , v3 , v4 ) . The former are defined by the

latter, and time derivatives of the former are the results of these definitions and the time

derivatives of the latter. Therefore the variables

( y1 , y2 , y3 , w1 , w2 , w3 , 1 , 2 , 3 , 1 , 2 , 3 , 4 ,1 ,2 ,3 , 4 ) are just redundant variables on the space

described by the variables ( x1 , x2 , x3 , x4 , v1 , v2 , v3 , v4 ) .

Please also notice that the equations (12) are identical to equations of motion of the

Chaplygin Sleigh (6) obtained in section II. Concluding, the Lagrangian (7) gives the correct

equations of motion for the Chaplygin Sleigh, while at the same time completely eliminating the

additional variables used for its construction from the final dynamics of the system.
20
V. A COMMENT ON A HAMILTONIAN

Since our Lagrangian is degenerate to the extreme, with no velocities expressible by the

canonical momenta, the Dirac’s Theory of Constraints7,8) is a natural choice for creating the

Hamiltonian formalism. Some preliminary results6) suggest that it will be possible to explicitly

calculate both the Hamiltonian and the Dirac’s Brackets for the adjustable two-mass-point

Chaplygin Sleigh shown in this work, and that Dirac’s Brackets of all variables appearing in the

Lagrangian (7), but not appearing in the equations of motion (6), as well as the canonical

momenta of these variables, will be equal to zero with every function using any variable of the

system.

REFERENCES
1
R. M. Santilli, Foundations of Theoretical Mechanics (Springer-Verlag, New York, 1978) vol. 1.
2
H. Goldstein, C. P. Poole and J. L. Safko, Classical Mechanics, 3rd Ed. (Addison-Wesley, Boston, 2002) p. 46
3
A. M. Bloch, Nonholonomic Mechanics and Control (Springer, New York, 2003) p.12
4
P. M. Morse and H. Feshbach, Methods of Theoretical Physics (McGraw-Hill, New York, 1953) p. 298.
5
H. Bateman, Phys. Rev. 38, 815 (1931).
6
P. W. Hebda and B. A. Hebda, Spontaneous Dimension Reduction and the Existence of a local Lagrange-

Hamilton Formalism for Given n-Dimensional Newtonian Equations of Motion (Faculty Publications, paper 2,

2015). http://digitalcommons.northgeorgia.edu/math_facpub/2
7
P. A. M. Dirac, Can. J. Math. 2, 129 (1950)
8
P. A. M. Dirac, Lectures on Quantum Mechanics (Belfer Graduate School of Science, New York, 1964)

21