Classical Dynamics

Michaelmas 2012

Small Oscillations, Stability and Normal Modes:

Proof that matrix F has real eigenvalues In lecture we reduced the linearised equations of motion of a system with n degrees of freedom to the = F and claimed that matrix F has real eigenvalues. Here were are going to prove this statement. form Let us start with Lagrangian L= and from Euler-Lagrange equation obtain Tij (q ) qj =
0 + i , and Now, consider small deviations i , so qi = qi

1 Tij (q )q i q j V (q ), 2 V (q ) . qi

(1)

(2)

and expand Tij

V (q ) . i in linear order in , while V (q ) in quadratic order. Tij (q ) j = Tij (q ) = Tij (q 0 ) + Tij qk k + O( 2 ),

q0

(3)

(4)

were it is clear why we dont need quadratic terms - they clearly will not contribute as T will be multiplied by (in fact we dont need the linear order either). For V (q ), however it is essential to include quadratic the term. V (q ) = V (q 0 ) + V (q ) qk k +
q0

1 2 V (q ) 2 qk ql

k l + O( 3 ).
q0

(5)

Here the second term on RHS vanishes as q0 is the point of minimum of V . Thus 2 V (q ) V (q ) = k . i qi qk q0 Therefore (3) becomes Tij (q 0 ) j = Vij j , were Vij = In the matrix form = V . T Thus, = T 1 V , and F = T 1 V.
2

(6)

(7) (8)

2 V (q ) qi qj

.
q0

(9) (10)

Now everything is ready for the nal step. Consider an eigenvector of F with eigenvalue : F = 2 T T 1 V = 2 T V = 2 T V = 2 T , (11)

were V R, T R, because T and V are symmetric. Also detT == 0 since T is invertible. Thus, 2 R and this completes the proof. 1