You are on page 1of 101
THE GENERALIZED LAGRANGE INVARIANT 89 Lagrange invariant is the quantity n? dx dy dL dM, so that we have dx dy dp dq = dx’ dy’ dp' dq’, (6.33) where we have put p = nL, q = nM, p' = n'L', q! = n'M’. The proof of this result is most easily carried out by making use of some properties of matrices. By differentiating eqns (6.32) we have Op = —Vyg AX — Vey dy’ — Veg dx — Vey dy dq = ~Yyy dx! ~ Vy dy’ ~ Von dx — Vay dy (6.34) dp! = Veg dx! + Very dy! + Veg dX + Vey dy dq! = Vy_ dx! + Vyy dy! + Vyedx + Vyydy where V,. = @V/2x?, Viy = @V/éx Gy, ete. In matrix notation we can express eqns (6.34) as Vex Vey =. «\ [dx Veg —Vay .\ [ax Vow Vy VE dy | fa Vy. —1) [dy Vere Very —1 [| dpt Vex —Vey . | dp Vye Vyy . 1} \dg —Vyx —Vyy . e) Nag (6.35) Denoting the two matrices by A and B respectively and the column vectors of differentials by M’ and M, this equation may be written AM’ = BM, (6.36) so that M' = ABM. (6.37) This matrix equation can be expanded to ax — ax a pe ee = dx + — => — ey wo te aq 4 and three similar equations, so that ax Pg) on (6.38) Ax, Ys Ps) But this is simply the Jacobian which transforms the differential hypervolume element dx dy dp dg, i.e. we must have ax", y's Pg’) (x, Ys Ps) det (A“B) = dx’ dy’ dp! dq’ = dx dy dp da, (6.39) 90 OPTICAL INVARIANTS. so that eqn (6.33) will be proved if we can show that det (4-"B) = 1. (6.40) But det(A) | ww Volo 6.41 let(A) = , say . Voe Voy a and det (B) has the same value, remembering that V,, = Vs, etc. Also the determinant of the product of two matrices is the product of their determin- ants, from which it follows that det (A~!) = Q~!. Thus eqn (6.40) is proved and so eqn (6.33) is true. This result has been given by several authorst and it has been pointed out that it is the geometrical optics analogue of Liouville’s theorem in statistical mechanics; in the present application the phase space is four-dimensional, the optical direction cosines p and q are analogues of generalized momenta and the spatial coordinates x and y are like generalized position coordinates. The z-axis corresponds to the evolution of time. The analogy can be developed from Fermat's principle; this would usually be expressed by saying that the optical path length W between two fixed points is stationary along a physically possible ray path and W is given by W =§nds. (6.42) However if W is written in the form dx)? (dy) w = |n(xy,2) [} + (2) + (2) ic (6.43) the integrand is of the form of the Lagrangian L of generalized mechanics, with za the time axis. This formulation is a convenient starting point for develop- ing the geometrical optics of media of continuously varying refractive index, e.g. graded index fibres. Also in this analogy the skew invariant of Section 6.4 corresponds to conservation of angular momentum with no frictional forces. Returning to geometrical optics, the significance of the generalized Lagrange invariant is that in the geometrical optics model the light flux emitted by an element dx dy of a Lambertian source in the x-y plane over a range of solid angle specified by dLdM is constant through the optical system, apart from reflection and absorption losses, assuming that all the rays of this infinitesimal pencil do get through the system. Thus the volume element dx dy dp dq in phase space which represents this pencil of rays changes in shape as it travels through the system but its volume is constant. The analogy + For example, G. Toraldo di Francia, J. Opt. Soc. Am. 40, 600 (1950); O. N. Stavroudis, J. Res. nat. Bur. Stand. 63B, 31-42 (1959); and D. Marcuse, “Light Transmission Optics”, Van Nostrand- Reinhold, New York (1972), THE GENERALIZED LAGRANGE INVARIANT 91 —dl ax aaeane (a) ’ * x ax ' “ L Cs > L aL at (o) fe) (a) Fic. 68 The generalized Lagrange invariant traced in phase space with the paraxial Lagrange invariant H of Section 3.3 is obvious but the generalized Lagrange invariant applies to any optical system, symmetrical or not and image-forming or not. The directions of the coordinate axes used in defining the invariant can be chosen arbitrarily in the different spaces but in any one space the direction cosine elements must be referred to the same axes as the length elements dx and dy. It is, of course, assumed that the differentiations carried out in eqns (6.32) and (6.34) are permissible; this is true for optical systems with smooth surfaces provided singularities such as rays being tangent to refracting surfaces or suffering total internal reflection are avoided. It is also to be understood that if the characteristic function V is multivalued between the points P and P’ in Fig. 6.7 the invariant applies separately along each ray between these points. The meaning of the generalized Lagrange invariant can be illustrated by considering a pencil of rays in one plane only, so that it becomes n dx dL, as in Fig, 6.8a; the corresponding “volume” in phase space is as in Fig. 6.8b and after propagating some distance in the same medium the rays will have spread and the volume appears as in Fig. 6.8c. Finally after refraction the volume might appear at the conjugate plane to the original element asin Fig. 6.8d; the area of the figure multiplied by the refractive index of the medium is constant. The ordinary Lagrange invariant H, the optical sine theorem (eqn (6.31)) and other results could be deduced as special cases from the generalized Lagrange invariant. However, it is a somewhat abstract concept for ordinary glass optics, although the particle optics version is of great practical use; it therefore seems more appropriate to use special derivations for these theorems.

You might also like