This document discusses generalized coordinates in classical dynamics. It begins by explaining that generalized coordinates are system-dependent coordinates that are chosen to have a visualizable geometric significance and make the principal equations of Lagrangian dynamics unchanged by a change of coordinates. It then provides an example of using generalized coordinates to solve the equations of motion for a constrained one-particle system. Finally, it generalizes this approach to solve for the equations of motion of a multi-particle system with holonomic constraints, using generalized coordinates to make the constraint equations trivial.
This document discusses generalized coordinates in classical dynamics. It begins by explaining that generalized coordinates are system-dependent coordinates that are chosen to have a visualizable geometric significance and make the principal equations of Lagrangian dynamics unchanged by a change of coordinates. It then provides an example of using generalized coordinates to solve the equations of motion for a constrained one-particle system. Finally, it generalizes this approach to solve for the equations of motion of a multi-particle system with holonomic constraints, using generalized coordinates to make the constraint equations trivial.
This document discusses generalized coordinates in classical dynamics. It begins by explaining that generalized coordinates are system-dependent coordinates that are chosen to have a visualizable geometric significance and make the principal equations of Lagrangian dynamics unchanged by a change of coordinates. It then provides an example of using generalized coordinates to solve the equations of motion for a constrained one-particle system. Finally, it generalizes this approach to solve for the equations of motion of a multi-particle system with holonomic constraints, using generalized coordinates to make the constraint equations trivial.
This document discusses generalized coordinates in classical dynamics. It begins by explaining that generalized coordinates are system-dependent coordinates that are chosen to have a visualizable geometric significance and make the principal equations of Lagrangian dynamics unchanged by a change of coordinates. It then provides an example of using generalized coordinates to solve the equations of motion for a constrained one-particle system. Finally, it generalizes this approach to solve for the equations of motion of a multi-particle system with holonomic constraints, using generalized coordinates to make the constraint equations trivial.
Abstract. This primarily deals with the conception of phase
space and the uses of it in classical dynamics. Then, we look into the other fundamentals that are required in dynamics before we start with perturbation theory i.e. topics like generalised coordi- nates and genral analysis of dynamical systems.
Date: 12 October, 2006.
cm i. 1 2 RAVITEJ UPPU
The general analysis of dynamics using the conception of phase space
is a very deep physical aspect explained in mathematical structure. Let’s intially look what is a phase space and the requirement of a phase space. Well before that let’s see the use of generalized coordinates.
1. Generalized coordinates 1.1. Why generalized coordinates? I have not particularily used the vector symbol over x. It is implicity asssumed to be so and when I speak of components, then I use it with an index (either a subscript or supeerscript) We always try to select coordinates such that they have a visual- izable geometric significance. They are mostly chosen on the basis (being independent usually helps us). They are system dependent un- like the usual coordiantes. Such coordinates are helpful principally in Lagrangian Dynamics, where the forms of the principal equations describing the motion of the system are unchanged by a shift to gen- eralized coordinates from any other coordinate system.
1.2. An approach mentioned. Say we have a N particle system with
K holonomic constraint(i.e. constraints of the form fi (x1 , x2 , ...xN , t) = 0 and i = 1, 2, ...K < 3N ) on the system, where, xj are position vectors of N particles. Now, let’s consider a one-particle system. Our surface is f (x, t) = 0 and from Newton’s equations we have mẍ = F~ + C ~ where C ~ is the unknown force of constraint. So, we have 6 unknwn things but only 4 equations. Our, SOS call has been answered by the considering that the force of constraint is normal to th surface at every point. Because, the only effect by the parralel forces (if they have any) is that they increase the acceleration along the surface. Now, how can we choose this force. We know that ∇f (~x) is always perpendicular to the surface unless ∇f = 0. So, let’s leave out this case in our analysis. (This can be mathematically ∂fi thought as,if K¿1, N¿1, the requirement is that matrix of ∂x α elements
are atleast of rank K). Hence the constraint force can be written as ~ = λ(t)∇f (~x, t) C Hence, we have 4 equations and 4 unknowns as C ~ has reduced to be dependent on λ which is dependent only on time. P hysical reason why C ~ should be perpendicular to the surface is that forces of constriant do no wrok other than keep reduce the availability of states to the system i.e. keep the particle confined to the surface. (Here, we do not consider the forces lik friction which are dissipative i.e our system become non-conservative). Such surfaces are called ”smooth”. GENERALIZED COORDINATES 3
Hence we have, mẍ = F~ + λ(t)∇f (~x, t) Now, let’s try to eliminate λ(t) . As λ∇f is perpendicular to the surrface, so we can eliminate by taking the component tangential to plane. Let ~τ be an arbitrary vector tangential to the surface at ~x at, i.e. only restriction on τ.∇f = 0 Hence, we have (mẍ − F~ ).τ = 0 This implies mẍ − F~ is perpendicular to surface at ~x and at time t. We can find such τ at each x. As τ is arbitrary vector tangential to surface, there are two linearly independent vectors at each ~x and hence two linearly independent vector functions of ~x and t. Hence we get two equations for the above equation. So now we have 3 equations and similarily 3 unknowns. Now, let’s generalize this above procedure for N particles and K con- straints. (W e do NOT use the summation notation below). Equation for the ith particle is mi~x¨i = F~i + Ci ~ ~ = P k We take C j=1 λj ∇i fi where ∇i is the gradient with respect to ~xj and λj and are K in number. P ∂f Now, if ∂V ∂t = 0 . As we shall see later, dE dt = j ∂tj . Let ~τ be N arbitrary vectors tangential to the surface. and hence wa have that XN τ~i .∇i fj i=1 Compoents of N vectors τi ⇒ 3N − K components are independent. Then we have (mi ẍi − F~i .~ τi = 0 The above principle is called that D’Alembert’s Principle. By the same arguments as with the one particle case, we get that 3N − K componenets of τ~i lead to 3N − K relations and the K constraint equations, totally counting to 3N equations and our unknows are the 3N unknown components of xi . Hence,can be solved. By doing all the above things, our problem has finally reduced to finding a suitable ~τ such that the ”only” condition on it is satisfied. In particle system, it is any arbitrary vector tangential to the surface. Here, in N particles, it is not easy to visualise (we will later come to know that this is nothing but a vector on the Tangent manifold T ∗ Q). But, we know tha tthe constraint equations define a 3N − K dimensional hypersurface in the 3N dimensional E3N . We call this Configuration Hyperspace or Configuration Manifold. We know the mathematical condition on ~τ . This defines a vector in 3N E which is a kind of tangent vector to the configuration manifold. Hence, it defines a 3-vector tangential to f = 0 in one particle system. 4 RAVITEJ UPPU
Similarily we get a 3N -vector tangential to fj = 0 hypersurface (Q).
Hence, picking taui means picking tangent vector in the 3N dimensions. Let’s define q α in the 3N space for which Q is the configuration hypersurface. Consider a region of E3N that has xi of Q and let q α where α = 1, 2, ..., 3N , be the new coordiantes of the region such they are invertible with the xi . q α = q α (x1 , ...xN , t) And these ar etwice differentible because they sould represent the ac- celeration also. Our objective is to pick them such that equations of constraints becomes trivial (i.e some of the q α are constant). When we write the equation of motion in these new coordinates those coor- dinates drop out which are constant. This an be don by choosing q α such that K of them depend on~ xi through function appearing in the constraint equations. Let the last K be of this kind. Hence, we have q n+j (~x) = Rj (f1 (~x), ..., fK (~x)); j = 1, ...K where we represent ~x for x~i and n = 3N − K (dimension of Q. These are the last K equations of 3N equations that give q α in xi , and they too must be invertible, which means that it must be possible to solve them for fj = fj (q n+j , ..., q n+K ). When the constraint conditions are imposed, they force the last K of the new coordinates to be constants. q n+j = R(0, ...0) This is what we mean by telling that the constraints are trivial in these coordiantes. Now, it’s all about the dependance of the remaining q α on time. q α can be used as well as xi to define a point in E3N . The last equation requires the point to lie in Q. Hence, q α form the coordinate system on Q. 1.3. A small digression which I made in the talk that was confusing. We have considered F~ = −∇V (x, t) and taking hte dot product with our equation we get d 1 m.ẍ.ẋ ≡ { mẋ2 + V } = −∇V.~x + λ∇f.~x˙ dt 2 ~ be a solution of equation of motion. Any particle remains Let x(t) df on the surface for f (~x(t), t) = 0 ⇒ dt = 0. But we know df ∂f = ∇f.~x˙ + dt ∂t dV ∂V = ∇V.~x˙ + dt ∂t Hence we have, dE d 1 ∂V ∂f = { mẋ2 } = − lambda dt dt 2 ∂t ∂t GENERALIZED COORDINATES 5
Hence, the total energy of the particle changes only if V or f are explicit functions of time (i.e if the potential depends on time or the surface is moving). These are already removed from our cases and analysis.
