next up previous
Next: Taking stock Up: Cutting up Previous: How little area to

Laying out the tubes

Now we're in great shape. We have annexed our annulus, and although parts have been snipped away, no two snipped-away parts overlap. Within the sector defined by one of the snipped-away parts, the annexed area is on the order of the original area. (See Figure 20.)

  
Figure 20: Comparing areas within a snipped sector.

Thus we hope and expect that there is enough room for the tubes to grow. To verify this, let us specify precisely how the tubes are to run.

The region we are supposed to be cutting up is the portion of G between height h and height h/2 and lying above tex2html_wrap_inline1415 , the union of the base of the hemisphere and the annexed territory. Just as we divided the plane up into pieces, we will divide tex2html_wrap_inline1415 into pieces, and treat each such piece separately. We begin by cutting tex2html_wrap_inline1415 along the radii that extend to the tips and centers of the snips. (See Figure 21.)

  
Figure 21: Dividing up tex2html_wrap_inline1415 .

This yields pieces of three kinds: sectors, right triangles, and ``left triangles.'' In discussing how to lay out the tubes we will ignore left triangles, since they are just like right triangles.

For a given piece S, either a sector or a triangle, let tex2html_wrap_inline2571 be the portion of the plane at height z that lies above S, and let

Our task is to extend the tubes that have run into tex2html_wrap_inline2577 down through the region H until they run into tex2html_wrap_inline2581 . Since the tubes are to remain more or less vertical, this amounts to specifying a correspondence tex2html_wrap_inline2583 between tex2html_wrap_inline2577 and tex2html_wrap_inline2571 for each tex2html_wrap_inline2589 . Our strategy is to choose these correcpondences so that the available horizontal cross-section is shared equally among the tubes, that is, so that

for all tex2html_wrap_inline2591 .

This condition doesn't determine the tex2html_wrap_inline2583 's, so we require in addition that the tex2html_wrap_inline2583 's be nice and smooth, and take ``radii'' of tex2html_wrap_inline2577 to ``radii'' of S (z). To see what this entails, introduce polar coordinates tex2html_wrap_inline2601 in the (x,y)-plane so that the point r=0 is the center of the disk D and the ray tex2html_wrap_inline2609 lies just to the right of S. (See Figure 22.)

  
Figure 22: Polar coordinates.

Then H is determined by the inequalities

displaymath2534

displaymath2535

where

displaymath2536

when S is a sector, and

displaymath2537

displaymath2538

when S is a right triangle. Set

displaymath2539

and define

displaymath2540

displaymath2541

The function tex2html_wrap_inline2619 tells what fraction of the area of tex2html_wrap_inline2571 has tex2html_wrap_inline2623 -coordinate between 0 and tex2html_wrap_inline2623 . The function tex2html_wrap_inline2629 tells what fraction of the area of the infinitesimal strip between angles tex2html_wrap_inline2623 and tex2html_wrap_inline2633 has r-coordinate between tex2html_wrap_inline2637 and r. We make the tubes run along the curves tex2html_wrap_inline2641 , tex2html_wrap_inline2643 . The correspondences tex2html_wrap_inline2583 associate points in tex2html_wrap_inline2577 and tex2html_wrap_inline2571 that have equal values of u and v. These correspondences are illustrated in Figures 23 and 24. Figure 23 shows a sector; here the tubes are being deflected radially. Figure 24 shows a right and left triangle together; here in addition to being deflected radially the tubes are being squeezed from the center of the snip out towards the tips.

  
Figure 23: Tubes above a sector.

  
Figure 24: Tubes above a pair of triangles.

Have we succeeded in keeping the tubes more or less vertical? This comes down to checking whether there is a universal upper bound for

displaymath2542

where the subscript u,v indicates that u and v are to be held constant. For a sector, this fact is geometrically obvious. For a triangle, this fact is not quite so obvious, but nonetheless true. To see why, consider that the only free parameter is the ratio h/a, and the only way the expression above can fail to be bounded is if something bad happens as h/a goes to 0. So you just have to convince yourself that nothing bad happens as h/a goes to 0. As a last resort, this can be verified by computation.


next up previous
Next: Taking stock Up: Cutting up Previous: How little area to

Peter Doyle