Sunteți pe pagina 1din 3

Lecture 5: Limits of Finite Sums and the Definite Integral

Sep. 4, 2015
Lecturer: Chandra Vaidyanathan

6 Cauchys Mean Value Theorem


The version of Mean Value Theorem we studied last time was developed by Lagrange. Cauchy, generalised
the Lagranges form of Mean Value Theorem and we state and prove Cauchys version of Mean Value
Theorem, which can be thought of as the Generalised Mean Value Theorem.
Theorem 1. Cauchys Mean Value Theorem
Suppose functions f and g are continuous on [a, b] and differentiable through out (a, b) and also suppose
0
g (x) 6= 0 throughout (a, b). Then there exists a number c in (a, b) at which
0
f (c) f (b) f (a)
0 =
g (c) g(b) g(a)
Proof: Strategy: We shall appeal to Mean Value Theorem twice.
First we use it to show that g(b) g(a) 6= 0. For if g(b) g(a) = 0, then the mean value theorem would be
0
valid for g, giving that there exists a value c between a and b such that g (c) = 0, creating a contradiction to
0
the fact that g (x) 6= 0 in the interval (a, b). Next we will apply it to the function,
f (b) f (a)
F (x) = f (x) f (a) [g(x) g(a)].
g(b) g(a)
Clearly, this function is both continuous and differentiable where f and g are, and F (b) = 0 = F (a). By
0
MVT applied to F , we get, there is a c between a and b, for which F (c) = 0. This means,
0 0 f (b) f (a) 0
F (c) = f (c) [g (c)] = 0
g(b) g(a)
Remark 1. Note that Lagranges form of Mean Value Theorem occurs, when g(x) = x. Cauchys Mean
Value Theorem has a very interesting corollary, which is the LHopitals rule.
Theorem 2. LHopitals Rule
Suppose that f (a) = 0 = g(a), that f and g are differentiable on an open interval I containing a, and that
0
g (x) 6= 0 on I if x 6= a. Then,
0
f (x) f (x)
lim = lim 0
xa g(x) xa g (x)

assuming that the right side of this equation exists.


Proof: As with other limits, we establish the statement for the Right Hand Limit and the Left Hand Limit
and show them to be equal. We begin the proof for the RHL, but the arguement is identical for LHL, as well,
and hence, we will not bother to sketch it.
0
Suppose that x lies to the right of a. Then, g (x) 6= 0 and we can apply Cauchys Mean Value Theorem, to
the closed interval [a, x]. This step produces a number c between a and x such that
0
f (c) f (x) f (a)
=
g 0 (c) g(x) g(a)

1
But, f (a) = g(a) = 0, so
0
f (c) f (x)
0 =
g (c) g(x)

As x approaches a, c approaches a as well, since it is in between a and x. Therefore,


0
f (x) f (x) f (x)
lim = lim = lim 0
xa g(x)
+ ca g(x)
+ xa g (x)
+

EXAMPLE: 6.1. Find the following limits:


3x sin x
1. lim
x0 x

1+x1
2. lim
x0 x

1 + x 1 x/2
3. lim
x0 x2
x sin x
4. lim
x0 x3
Remark 2. LHopitals rule is equally applicable for the cases /, .0, . The indeterminate
cases of 1 , 00 and 0 can be handled by first taking logarithm of the function and then find the limit of
logarithm expression using LHopitals rule and then exponentiate back to find the result of the limit of the
original expression.

QUESTION: 6.1. Why is the procedure outlined above justified?

EXAMPLE: 6.2. Evaluate the following limits:


sec x
1. lim
x/2 1 + tan x

ln x
2. lim
x 2 x
ex
3. lim
x x2
 
1
4. lim x sin
x x

5. lim (1 + x)1/x
x0+

6. lim x1/x
x

2
7 Limits of Finite Sums
Suppose we have to find the area of the region that lies above the x-axis and below the graph y = 1 x2 ,
and between the vertical lines x = 0 and x = 1. Unfortunately, we donot have a readymade formula to find
this area and will have to come up with a way to estimate the area.
An approximation to the area can be found by using rectangles, say four of them, of heights, 1, 15/16,3/4
and 7/16 and of uniform width 1/4. Thus, we can get our first approximation to be 0.78125. But clearly,
this is an overestimate since, the region of interest is contained inside the region spanned by the rectangles.
This is the upper sum approximation to the area.
One way to correct for the over estimation is to take the same four rectangles of width 1/4, but of heights
15/16,3/4, 7/16 and 0, leading to an area of 0.53125, which is an underestimation. This called the lower
sum approximation to the area.
By using more and more rectangles to get a finer approximation, one could get arbitrarily close to the real
value of the area by using the methods above. If the interval [a, b] is subdivided into n subintervals of equal
widths, say x = (b a)/n, and if f (ck ) is the value of f at the chosen point ck in the k th subinterval, this
process then gives a finite sum of the form

(f (c1 ) + f (c2 ) + . . . + f (cn ))x

The choices for the ck could maximise or minimize the value of f in the k th subinterval or some value in
between. The true value lies somewhere between the approximations given by the upper and lower sums.

S-ar putea să vă placă și