Page images

so that from a geometrical point of view Yo, , Y2, Yz, ... are equidistant ordinates of the curve y = f(x).

Now these quantities Af(x), APF(X.), ... are called the successive differences of F (xo); and in all cases to which the preceding formulæ of approximation can be successfully applied, it is necessary that the values of these differences should rapidly decrease ; this circumstance may easily be verified, and it is found to be true in all cases of tabulated functions. Thus, if F(x) is a rational algebraical function of n dimensions, A*F(x)=0. On this hypothesis then we shall deduce from (213) successive approximate values to the definite integral.

Thus, if we omit all terms after a°F (wo), we have

[ocr errors]
[ocr errors]

which is a theorem frequently applied in mensuration.

Let n=4; /**(x) dx = mi (246–44; +593}. (217)

From (216) may be deduced a rule, originally discovered by Thomas Simpson, and frequently employed in mensuration.

Let the range xn— Xo be divided into n equal parts, where n is an even number, each of which =i; then as (216) gives the value of the integral for any two consecutive parts, taking the sum of all, we shall have,

[ocr errors][ocr errors][ocr errors][ocr errors]
[ocr errors][ocr errors][merged small]
[ocr errors]

Let n = 4; then

/**P(x) dx = 4 {24,–9z+2yz}. (221) If sof(x) = 0, as is the case when f(x) = x2, (220) and (221) give results strictly accurate.

115.] An approximation may also be made by the following process, which assigns certain limits within which the true value of the definite integral lies. Let F'(x) be the element-function of the definite integral ; and let f'(x) and $'(x) be two other functions of x, the values of which are respectively greater and less than F'(x), for every value of x within the range of integration; so that for every such value

f'(x) > F'(w) > $'(x); Moreover let us take f'(x) and $'(x), such that the definite integrals, of which they are the element-functions, may be determined: then evidently, by the definition of a definite integral,

[ocr errors]


[ocr errors]


XO Let us investigate by this method an approximate value of the

i dx definite integral si since for all values of x included

Jo (1 - 23) within the range of integration,

(1 – 22)}" u _m3 > 1;
pt dx pdx > ldx;
Jo (1 – x2) Jo (1 – x3y$ Jo

of p_dx _ 1

[ocr errors]
[ocr errors]
[ocr errors]
[ocr errors]
[ocr errors]
[ocr errors][merged small][ocr errors][ocr errors][ocr errors]

50 that limits are assigned very near together within which the value of the given definite integral lies. 116.] The definition given in (200) suggests other means of obtaining approximate values of definite integrals, and of ascertaining certain properties relating to those values.

Let f'(x) be the element-function: then if Xn is greater than Xo, and F"(x) does not change sign within the range of integration, | F'(x) dx has the same sign as F'(x). If however F'(x) changes sign, the integral will be positive or negative, according as the positive or negative part of the series is the greater.

Since, in the series on the right hand member of (200), X, X, 8,-X1, ... Xo— Xn--] are all quantities of the same sign, by preliminary Theorem III of Vol. I. the sum of the series is equal to the product of the sum of all these quantities and some mean value of the other factor: consequently

[ocr errors]
[ocr errors][merged small]

wherein 0 denotes an undetermined proper positive fraction. Then, the limits of the value of the definite integral, given by substituting 0 and 1 severally for 0, are (20,X.) F'(x) and (Xr—X) F(X), between which the value of the definite integral lies.

The geometrical interpretation of (221) deserves a passing notice. Let y = f'(x) be the equation to the curve P, PQP, in fig. 16; Then as the definite integral expresses the area P, P, M, Mo, and as M, M, = X,— Xo, the equation shews that the area is equal to that of a rectangle one of whose sides is m, Mn, and the other is an ordinate to the curve corresponding to an abscissa intermediate to x, and X

If the left-hand member of (200) is determined by means of its indefinite integral; then

F(X») — F(X) = (x,—X) '{xo+0 (Xn- %.)}, (225) which equation has been already found in Art. 111, Vol. I.

If the range x, — X, is infinitesimal,

[merged small][ocr errors][merged small]

which is the first term of the series, of which the definite integral is the sum.

If the element-function F'(x) is the product of two functions, f(x) and $'(x), of which o'(x) has the same sign for all values of x within the range of integration ; then

[ocr errors][ocr errors][ocr errors]

which = 0, when x = oc; and = 5, when x = 0, so that the mean value of this factor is mean between these quantities. Also


[blocks in formation]
[merged small][ocr errors]

po e-ar der
= 0, if a = 0.

(232) Jo em te- x's 117.] Of the various processes which have been devised for the purpose of approximating to the value of definite integrals, some consist in the development of the element-function into a series, and in the subsequent integration of each term separately; in others, series are formed, either by integration by parts, or by some other method, in general terms, and these are applied to the particular element-function.

The following series was devised by John Bernoulli.

Let/ *F'(x) dx be the definite integral whose value is required : then, integrating by parts, we have

[ocr errors]

This theorem may be proved, and also its general term may be found, by means of Taylor's series; and hereby the limits of the value will be determined.

By Taylor's series we have
F(x + h) = f(x) + F"(x)* +"(x) * 2 + ... + F"(x+6h)

Let h be replaced by -X, .: F(X) = F(0)+F'

10 + F"(x)

full 1 2 3 ......

...(-)17"(2–0) 1.2.0. so that replacing F(x) by F'(x) dx, and taking the definite integral of both members, we have



« PreviousContinue »