584 Partial derivatives
lar region R of Figure 11.391 is a subset of the given rectangular region RI and,
moreover,
(14) G(x,yi) < 0 (xo - h < x < xo + h)
(15) G(x,y2) > 0 (xo - h < x < xo + h).
The negative and positive signs in Figure 11.391 serve to remind us that G(x,y) <
0 at points (x,y) on the lower edge of R and that G(x,y) > 0 at points (x,y) on
the upper edge of R. We are now pre-
(xo-h,y2) (X0' Y2) (xo+h,yz) pared to obtain results. Let xo - h <
x < xo + h. Since G0(x,y) > 0 when
y, S y 5 y2, we conclude that G(x,y) is
(x0 -h,Y1)
Figure 11.391
(xo, Yi) (xo+h,y,)
increasing over the interval y, S y S Y.
Since G(x,yi) < 0 and G(x,y2) > 0, we
conclude, with the aid of the intermedi-
ate-value theorem, that there is one
and only one number f (x) for which
yi < f(x) < y2 and G(x, f(x)) = 0.
Our proof is now about half done; we
have found our f but we need information about f'(x). To start getting this
information, suppose that xo - h < x < xo + h and xo - h < x + Ax < xo + h.
Then
(16) G(x, f(x)) = 0, G(x + Ax, f(x + Ax)) = 0
and consequently
(17) [G (x + Ax, f(x + Ax)) - G(x, f(x + Ax))]
+ [G(x, f(x + Ax)) - G(x, f(x))] = 0.
Applying the mean-value theorem (Theorem 5.52) to these differences shows that
there exists a number between x and x + Ax and a number 71 between f(x)
and f (x + Ax) such that
(18) G:(, f (x + Ax)) Ax +G.,(x,r7)[f(x + Ax) - Ax)] = 0
and hence
(19)
f(x+Ax) -f(x)__G,(,f(x+Ax))
Ax G,, (x,,?)
Since G. and G are, by hypothesis, continuous at the point (x, f(x)), the desired
result (11) will follow from this if we show that f is continuous at x, that is,
(20) lim f(x + Ax) = f(x).
To prove (20), let a be a positive number for which f (x) + e and f (x) - e lie
between yl and y2. Since G(x, f(x)) = 0 and G,,(x,y) is an increasing function
of y, we find that
G(x,f(x) - e) <0 <G(x,f(x)+e).
Since G is continuous, we can choose a positive number 8 such that
G(x + Ax, f (X) - e) < 0 < G(x -1- Ax, f (X) + e)