
INTEGRABLE SPHERICAL AND PLANAR BALL BEARINGS 3
Main result 2.The spherical ball bearings problem (1.1) for B=Cis integrable for any εin
all configurations. Along with F1and F2, the system has a nonalgebraic first integral1
F±
3=±pD(A−C)F+DG −dCexp(±(1 −ε)pD(A−C)Φ).
For d= 0, the equations (1.1) coincide with the equations of motion of a Chaplygin ball with
the inertia tensor Ion a sphere (ε6= 1) and the plane (ε= 1) with slightly different definitions
of parameters εand D. Thus, the above results can be seen as natural extensions of well-known
integrable Chaplygin ball problems.
In [11] we also considered the associated planar problem. This is the case I in the notation of
the present paper, when the radius of the fixed sphere S0tends to infinity. We found an invariant
measure and proved the integrability by means of the Euler–Jacobi theorem [11]. Here, in Section
4, we perform an explicit integration of the reduced problem.
2. Rolling of a dynamically nonsymmetric sphere over nmoving homogeneous balls
and a fixed sphere
2.1. Kinematics. Let O~
e0
1,~
e0
2,~
e0
3,O~
e1,~
e2,~
e3,Oi~
ei
1,~
ei
2,~
ei
3be positively oriented reference frames
rigidly attached to the spheres S0,S, and the balls Bi,i= 1, . . . , n, respectively. By g,gi∈SO(3)
we denote the matrices that map the moving frames O~
e1,~
e2,~
e3and Oi~
ei
1,~
ei
2,~
ei
3to the fixed frame
O~
e0
1,~
e0
2,~
e0
3. Using the standard isomorphism between the Lie algebras (so(3),[·,·]) and (R3,×) given
by
(2.1) aij =−εijkak, i, j, k = 1,2,3,
the skew-symmetric matrices ω=˙
gg−1,ωi=˙
gig−1
icorrespond to the angular velocities ~ω,~ωiof
the sphere Sand the i-th ball Biin the fixed reference frame O~
e0
1,~
e0
2,~
e0
3attached to the sphere S0.
The matrices Ω = g−1˙
g=g−1ωg,Wi=g−1
i˙
gi=g−1
iωigicorrespond to the angular velocities
~
Ω, ~
Wiof Sand Biin the frames O~
e1,~
e2,~
e3and Oi~
ei
1,~
ei
2,~
ei
3attached to the sphere Sand the balls
Bi, respectively. We have ~ω =g~
Ω, ~ωi=gi~
Wi.
Then the configuration space of the problem is
Q=SO(3)n+1 ×(S2)n{g,g1,...,gn, ~γ1, . . . ,~γn},
where ~γiis the unit vector
~γi=−−→
OOi
|−−→
OOi|
determining the position of the centre of i-th ball Bi,i= 1, . . . , n. In the cases I and II, the velocity
of the centre of the i-th ball is ~vOi= (R±r)˙
~γi, while for the cases III and IV (n= 1) we have
~vO1=±(r−R)˙
~γ1. We will show in Proposition 2.1 that if the initial conditions are chosen such
that the distances between Oiand Ojare all greater than 2r, 1 ≤i < j ≤n, then the balls will
not have collisions along the course of motion. This is the reason why we do not assume additional
one-side constraints
(2.2) |~γi−~γj| ≥ 2r
R±r,1≤i<j≤n(cases I and II, n≥2).
Let A1, ..., Anand B1, B2, ..., Bnbe the contact points of the balls B1,...,Bnwith the spheres
S0and S, respectively. The condition that the rolling of the balls B1,...,Bnand the sphere Sare
without slipping leads to the nonholonomic constraints:
~vOi+~ωi×−−−→
OiAi= 0, ~vOi+~ωi×−−−→
OiBi=~ω ×−−→
OBi, i = 1, ..., n,
that is,
(2.3) ~vOi=±r~ωi×~γi, ~vOi= (R±2r)~ω ×~γi±r~γi×~ωi(cases I and II)
and
(2.4) ~vO1=±r~ω1×~γ1, ~vO1=±(2r−R)~ω ×~γ1±r~γ1×~ω1(cases III and IV).
The dimension of the configuration space Qis 5n+ 3. There are 4nindependent constraints in
(2.3), defining a nonintegrable distribution D ⊂ T Q of rank n+ 3. The phase space of the system,
Dconsidered as a submanifold of T Q, has the dimension 6n+ 6.
1Through the paper, the sign ±denotes + for cases I and III, and - for the cases II and IV.