68 MATHEMATICS AND THE LAWS OF NATURE
underestimate. Neither estimate is very accurate, of course, but the
idea of making simple over- and underestimates is almost enough
to solve the problem.
In symbols, our results, so far, look like this: PA ≥ F ≥ 0, where
we let the letter F represent the total force. Notice that we have
“trapped” F in the interval between 0 and PA. In mathematics,
however, what can be done once can usually be done twice, and
that is our next goal. Now imagine that we draw a line along the
wall exactly halfway between the surface of the water and the base
of the wall. We now repeat the procedure we just performed on
each half of the wall. The results are an upper and a lower esti-
mate for each half. If we add the halves together we get an upper
and lower estimate for F, the total force exerted on the wall by
the water. The procedure takes more work, but our new upper
and lower estimates are closer together than the previous ones. In
symbols this is the result of our second, more accurate estimate:
0.75 PA ≥ F ≥ 0.25 PA. Notice that this time we have trapped F,
the total force, in an interval that is half the size of that obtained in
our first calculation. As a consequence we now know more about
F than we did before.
We can continue to improve our results by dividing the wall
into thinner and thinner sections. If we divide the wall, as Stevin
did, into 1,000 equal horizontal strips, and compute an upper and
a lower estimate for each strip, and then add them together, the
result is 0.5005 PA ≥ F ≥ 0.4995 PA. Notice how close the upper
and lower estimates are now. The more finely he divides the wall,
the closer the upper estimate is to the lower estimate, or to put it
another way: The more finely he divides the wall, the smaller the
interval containing F becomes. The only number that belongs to
all such intervals is 0.5 PA. Stevin recognized this and concluded
that the force exerted by the water on the wall is 0.5 PA. He was
correct, but notice that no matter how finely the wall is divided,
neither the lower estimate nor the upper estimate equals 0.5 PA.
Instead Stevin’s method simply “squeezes” any other possible
answer out of consideration. This is a beautiful example of an
algorithm that utilizes infinite processes to solve a problem in
physics.