Limits and Continuity, Continuous Function Attains its Maximum

Continuous Function Attains its Maximum

This theorem and the next one were developed by Bolzano (biography) and Weierstrass (biography).

Let f map a region R into the reals, where R is a closed bounded set in Rn. Assume f is continuous. We will show that f is bounded. We'll show this in 3 space, although the prove is valid in n dimensions.

Suppose f is unbounded and place the origin at the center of R. Draw the three planes perpendicular to the three coordinate axes. These planes divide R into 8 pieces. Each subregion includes the planes that separate it from the other subregions, hence each new region is closed. If f is bounded on all 8 regions it is bounded on all of R, hence f is unbounded on one of these regions. Call it R and repeat the above procedure.

with each step the region is smaller; its dimensions are cut in half. Points of R, drawn from these shrinking regions, cluster together. If you want to be formal about it, these points form a Cauchy sequence, and since Rn is a complete metric space, they approach a point that I will call p. Since R is closed it contains its limit points, hence p is in R.

Since f is continuous it is certainly continuous at p. Find a neighborhood δ about p that restricts f(x) to f(p)±1. Clearly f is bounded when restricted to this neighborhood, yet this neighborhood includes small regions that are allegedly unbounded. This is a contradiction, hence f is bounded on all of R.

This theorem fails when R is an open set, as p may lie on the boundary of R. The function 1/x is continuous on the open interval (0,1), yet it is unbounded.

since f is bounded on R, let u be the least upper bound. Let g(x) = 1/(u-f(x)). If f never attains the value of u, then the denominator is always nonzero, hence g is well defined and continuous. By the above, g has some upper bound b. This means f(x) never gets within 1/b of u, which is a contradiction, since u is the least upper bound. Therefore f(x) = u for some x in R. In other words, f attains its upper bound.

Cluster Point

Let R be a closed bounded region in Rn. Let S be an infinite set of points lying in R. These points cluster at a point p, which also lies in R. The proof is the same as that shown above. At each step, select the subregion that contains infinitely many points of S. The regions shrink to a point p, which lies in c by completeness and closure.