Initial Configuration: the S (paraboloid) and S_r (sphere) and the point P 

>	S  := (a,b,c) -> (x/a)^2-(y/b)+(z/c)^2=0;  # 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,b,c) -> implicitplot3d( S(a,b,c), x= -a..a, y=0..b, z=-c..c,                                     color=pink, style=patchnogrid, transparency=0.8, grid=[25,25,25] ):

 

>	plotSr := r       -> implicitplot3d( Sr(r), x=-r..r, y=-r..r, z=-r..r,                                     color=cyan, style=patchnogrid, transparency=0.8 ):

 

>	P1 := (r,a,b,c)   -> display( [plotP(r),plotS(a,b,c),plotSr(r)],                              axes=normal, labels=["x","y","z"], orientation=[25,65], args[5..-1] ):

 

>	P1(1,1,2,3, scaling=constrained);

 

>

 

Construction of Q: Intersection of S and S_r 

The intersection between these two solids is. 

>	Intersection := allvalues( solve( {S(a,b,c),Sr(r)}, {x,y} ) ):

 

>	simplify( [Intersection] ) assuming a>0, b>0, c>0;

 

>

 

There are two parts to this solution. 

>	xPOS := (p,q)->evalb( eval(x, eval(p,[r=1.,a=1.,b=1/2.,c=2.,z=1/2.]))>0 ) = q:

 

>	Q := [ unapply( eval( [x,y,z], select( xPOS, [Intersection], true  )[] ), [z,r,a,b,c] ),

 

>	unapply( eval( [x,y,z], select( xPOS, [Intersection], false )[] ), [z,r,a,b,c] ) ]:

 

>	piecewise( x>0, Q[1](y,r,a,b,c), x<0, Q[2](y,r,a,b,c) );

 

>

 

To construct the projection from the top of the shrinking sphere through Q onto the z=0 plane, the parameterization of Q can be done in terms of z. The minimum and maximum values of z occur when x=0. We now find the highest point, which we call Qstar. 

>	yzPOS := (p,q) -> evalb( eval([y,z], eval(p,[r=1.,a=1.,b=1.,c=2.]))::[positive,positive] ) = q:

 

>	Q0 := allvalues( solve( eval({S(a,b,c),Sr(r)},x=0), {y,z} ) ):

 

>	Qstar := unapply( eval( [0,y,z], select( yzPOS, [Q0], true )[] ), [r,a,b,c] ):

 

>	yM := unapply( Qstar(r,a,b,c)[2], [r,a,b,c] );

 

>	zM := unapply( Qstar(r,a,b,c)[3], [r,a,b,c] );

 

>

 

This curve is not easily identified. 

>	plotQ := (r,a,b,c) -> display( [seq(

 

>	spacecurve( Q[i](z,r,a,b,c), z=-zM(r,a,b,c)..zM(r,a,b,c), color=gold, thickness=2 ),

 

>	i=1..2 )] ):

 

>	P2    := (r,a,b,c) -> display( [plotP(r),plotS(a,b,c),plotSr(r),plotQ(r,a,b,c)],                               axes=normal, labels=["x","y","z"], orientation=[45,60], scaling=constrained ):

 

>	P2(1,3,2,1);

 

>

 

Construction of R: Projection of Q, from P, onto z=0 plane 

For each value of z, the lines passing through P and the corresponding point on Q can be parameterized in terms of the (scaled) distance measured along this line. 

>	LinePQ  := [seq( unapply( expand( (1-alpha)*P(r) + alpha*Q[i](z,r,a,b,c) ), [alpha,z,r,a,b,c] ), i=1..2 )]:

 

>	piecewise( x>0, LinePQ[1](alpha,z,r,a,b,c), x<0, LinePQ[2](alpha,z,r,a,b,c) );

 

>

 

The value of the parameter alpha when these lines hit the z=0 plane are given by 

>	alpha0   := unapply( [simplify( solve( LinePQ[1](alpha,z,r,a,b,c)[3]=0, alpha ) ) assuming a>0, r>0][],                     [z,r,a,b] );

 

>

 

Thus, the parametric representation of of the projected curve, R, in the z=0 plane is 

>	R := [seq( unapply( LinePQ[i](alpha0(z,r,a,b,c),z,r,a,b,c), [z,r,a,b,c] ), i=1..2 )]:

 

>	map( simplify, piecewise( x>0, R[1](z,r,a,b,c), x<0, R[2](z,r,a,b,c) ) );

 

>

 

This completes the constructions needed to put all of this together in one animation. 

>	plotR := (r,a,b,c) -> display( [seq( spacecurve( R[i](z,r,a,b,c), z=-zM(r,a,b,c)..  zM(r,a,b,c), numpoints=201, color=red, thickness=1 ), i=1..2 )] ):

 

>	P3    := (r,a,b,c) -> display( [P2(r,a,b,c),plotR(r,a,b,c)] ):

 

>

animQ := (r,a,b,c) -> display( [seq( animate( spacecurve, [LinePQ[i](alpha,z,r,a,b,c), alpha=0..alpha0(z,r,a,b,c)], z=-zM(r,a,b,c)..  zM(r,a,b,c),                                      color=blue, thickness=2, orientation=[25,65], background=P3(r,a,b,c),                                      scaling=constrained, frames=41 ), i=1..2 )] ):

 

>	P3(1,3,2,1);

 

>

 

>	animQ( 1 ,3,2,1);

 

>	animQ(1/2,3,2,1);

 

>	animQ(1/4,3,2,1);

 

>

 

Limit as r -> 0 

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 (for x>0): 

>	X,Y,Z := R[1](zeta,r,a,b,c)[]: x=X; y=Y; z=Z;

 

>

 

It is not surprising that Maple reports that every point converges to the origin. 

>	map( limit, [X,Y,Z], r=0, right );

 

>

 

We know there might be something interesting happening at the highest points on Q (along the positive y-axis). These points are given by 

>	Qstar := eval( [X,Y,Z], zeta=zM(r,a,b,c) ):

 

>

 

These expressions do not simplify much, but we do see that the z-component is 0. 

>	simplify( Qstar ) assuming a>b,b>c,c>0;

 

>

 

In the limit, we expect that x approaches zero and y approaches twice the radius of the limiting circle. The general limit is still the origin 

>	map( limit, Qstar, r=0, right );

 

>

 

but, at the highest point on Q: 

>	map( limit, Qstar, r=0, right ) assuming a>0,b>0,c>0;

 

>

 

This suggests that the limiting curve approached by R is the circle centered at [ 0, , 0 ] with radius :   .  Notice that the curvature of the restriction of S to the y=0 plane is .This result is consistent with the general conclusion (with = ). 

>

 

We close with an animation that shows this convergence. 

>	to3d := transform( (x,y)->[x,y,0] ):

 

>	plotR0 := p -> to3d( implicitplot( x^2 + (y-2*p)^2 = (2*p)^2, x=-2*p..2*p, y=0..4*p, color=green ) ):

 

>	animR := (a,b,c) -> animate( P3, [1-r,a,b,c], r=0..1, frames=40, numpoints=401, paraminfo=false, background=plotR0(c^2/2/b) ):

 

>	animR(2,1.5,1);  # Be patient! This animation could take a while to create.

 

>