restart; with( plots ): with( plottools ): Shrinking Sphere Problem Derivation of General Formula for Intersction of S and S_r, and Its Projection from the Top of S_r onto the z=0 Plane Douglas B. Meade 9 February 2007

S := a -> x^2+(y-a)^2+z^2=a^2; # fixed surface Sr := r -> x^2+y^2+z^2=r^2; # shrinking sphere P := r -> [ 0, 0, r ]; # top of shrinking sphere plotP := r -> plot3d( P(r), x=-1..1, y=-1..1, style=point, symbol=circle, symbolsize=10, color=blue ): plotS := a -> implicitplot3d( S(a), x= -a..a, y=0..2*a, z=-a..a, color=pink, style=patchnogrid, transparency=0.8 ): plotSr := r -> implicitplot3d( Sr(r), x=-r..r, y=-r..r, z=-r..r, color=cyan, style=patchnogrid, transparency=0.8 ): P1 := (r,a) -> display( [plotP(r),plotS(a),plotSr(r)], axes=normal, labels=["x","y","z"], orientation=[25,65] ): P1(1,2);

The intersection between these two spheres is a circle, parallel to the x=0 plane. Intersection := [allvalues( solve( {S(a),Sr(r)}, {x,y,z} ) )] ; The two parts to this solution are the top and bottom of the circle. y1,r1 :=eval( [y, sqrt(x^2+z^2)], Intersection[1] )[]: y2,r2 :=eval( [y, sqrt(x^2+z^2)], Intersection[2] )[]: y1=y2; simplify( r1=r2 ) assuming r>0, a>0; This shows that Q is the circle LywmKiQpSSJ4RzYiIiIjIiIiRikqJClJInpHRidGKEYpRiksJCoqI0YpIiIlRikpSSJyR0YnRihGKSwmKiRGMUYpISIiKiZGMEYpKUkiYUdGJ0YoRilGKUYpRjdGNUYp with L0kieUc2IiwkKigjIiIiIiIjRigpSSJyR0YkRilGKEkiYUdGJCEiIkYo . To construct the projection from the top of the shrinking sphere through Q onto the z=0 plane, parameterize the circle Q according to the angle made the positive x axis Q := unapply( [ r1*sin(theta), y1, r1*cos(theta) ], [theta,r,a] ): plotQ := (r,a) -> spacecurve( Q(theta,r,a), theta=0..2*Pi, color=gold, thickness=2 ): P2 := (r,a) -> display( [plotP(r),plotS(a),plotSr(r),plotQ(r,a)], axes=normal, labels=["x","y","z"], orientation=[45,60], scaling=constrained ): P2(1/2,2);

For each angle theta, the lines passing through P and the point Q(theta) can be parameterized in terms of the (scaled) distance measured along this line. LinePQ := unapply( expand( (1-alpha)*P(r) + alpha*Q(theta,r,a) ), [alpha,theta,r,a] ); The value of the parameter alpha when these lines hit the z=0 plane are given by alpha0 := unapply( [simplify( solve( LinePQ(alpha,theta,r,a)[3]=0, alpha ) ) assuming a>0, r>0][], [theta,r,a] ); Thus, the parametric representation of of the projected curve, R, in the z=0 plane is R := unapply( [simplify( LinePQ(alpha0(theta,r,a),theta,r,a) ) assuming a>0, r>0][], [theta,r,a] ); This completes the constructions needed to put all of this together in one animation. plotR := (r,a) -> spacecurve( R(theta,r,a), theta=0..2*Pi, numpoints=201, color=red, thickness=1 ): P3 := (r,a) -> display( [P2(r,a),plotR(r,a)] ): animQ := (r,a) -> animate( spacecurve, [LinePQ(alpha,theta,r,a), alpha=0..alpha0(theta,r,a)], theta=0..2*Pi, color=blue, thickness=2, orientation=[25,65], background=P3(r,a), scaling=constrained, frames=41 ): animQ(1,2); animQ(1/2,2);

These plots already illustrate the rapid convergence of every point on the curves R - except the one on the x-axis - to the origin (as r->0). Let's look at the parametric form of R. The three components are: X,Y,Z := R(theta,r,a)[]: x=X; y=Y; z=Z; Whenever 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, these expression are not indeterminate (as r->0) and so map( limit, [X,Y,Z], r=0, right ); But, Maple misses the special case when Ly1JJGNvc0c2JCUqcHJvdGVjdGVkR0koX3N5c2xpYkc2IjYjSSZ0aGV0YUdGKCIiIg==: eval( [X,Y,Z], theta=0 ); The remaining limit to be evaluated is the same one that was encountered in the Shrinking Circle Problem. map( limit, , r=0, right ) assuming a>0; These pointwise limits are nice, but they do no good in determining the limiting curve of the projected curves, R. The graphical evidence suggests that the limiting curve could be a circle. If so, then the pointwise limits tell us the only possible circle will be the circle passing through both [0,0,0] and NyUiIiEsJComIiIlIiIiSSJhRzYiRidGJ0Yj and lying in the z=0 plane. That is, LywmKiQpSSJ4RzYiIiIjIiIiRikqJCksJkkieUdGJ0YpKiZGKEYpSSJhR0YnRikhIiJGKEYpRiksJComIiIlRikpRi9GKEYpRik=, L0kiekc2IiIiIQ==. To confirm this, the first step is to verify for each positive value of r, the projected curve R is a circle: simplify( X^2 + (Y-2*a)^2 ); Now, as r shrinks to zero, the square of the radius clearly increases to limit( , r=0, right ); We close with a different animation that shows this convergence. to3d := transform( (x,y)->[x,y,0] ): plotR0 := a -> to3d( implicitplot( x^2 + (y-2*a)^2 = 4*a^2, x=-2*a..2*a, y=0..4*a, color=green ) ): animR := a -> animate( P3, [1-r,a], r=0..1, frames=30, numpoints=401, paraminfo=false, background=plotR0(a) ): animR(1);