Chapter IX
The Logarithmic and Exponential Functions of a Real Variable

196. THE number of essentially different types of functions with which we have been concerned in the foregoing chapters is not very large. Among those which have occurred the most important for ordinary purposes are polynomials, rational functions, algebraical functions, explicit or implicit, and trigonometrical functions, direct or inverse.

We are however far from having exhausted the list of functions which are important in mathematics. The gradual expansion of the range of mathematical knowledge has been accompanied by the introduction into analysis of one new class of function after another. These new functions have generally been introduced because it appeared that some problem which was occupying the attention of mathematicians was incapable of solution by means of the functions already known. The process may fairly be compared with that by which the irrational and complex numbers were first introduced, when it was found that certain algebraical equations could not be solved by means of the numbers already recognised. One of the most fruitful sources of new functions has been the problem of integration. Attempts have been made to integrate some function f(x) in terms of functions already known. These attempts have failed; and after a certain number of failures it has begun to appear probable that the problem is insoluble. Sometimes it has been proved that this is so; but as a rule such a strict proof has not been forthcoming until later on. Generally it has happened that mathematicians have taken the impossibility for granted as soon as they have become reasonably convinced of it, and have introduced a new function F(x) defined by its possessing the required property, viz. that F(x) = f(x). Starting from this definition, they have investigated the properties of F(x); and it has then appeared that F(x) has properties which no finite combination of the functions previously known could possibly have; and thus the correctness of the assumption that the original problem could not possibly be solved has been established. One such case occurred in the preceding pages, when in Ch. VI we defined the function  log x by means of the equation

log x =dx x .

Let us consider what grounds we have for supposing logx to be a really new function. We have seen already (Ex. XLII. 4) that it cannot be a rational function, since the derivative of a rational function is a rational function whose denominator contains only repeated factors. The question whether it can be an algebraical or trigonometrical function is more difficult. But it is very easy to become convinced by a few experiments that differentiation will never get rid of algebraical irrationalities. For example, the result of differentiating 1 + x any number of times is always the product of 1 + x by a rational function, and so generally. The reader should test the correctness of the statement by experimenting with a number of examples. Similarly, if we differentiate a function which involves sinx or cosx, one or other of these functions persists in the result.

We have, therefore, not indeed a strict proof that logx is a new function—that we do not profess to give97—but a reasonable presumption that it is. We shall therefore treat it as such, and we shall find on examination that its properties are quite unlike those of any function which we have as yet encountered.

197. Definition of log x. We define log x, the logarithm of x, by the equation

log x =1xdt t .
We must suppose that x is positive, since (Ex. LXXVI. 2) the integral has no meaning if the range of integration includes the point x = 0. We might have chosen a lower limit other than 1; but 1 proves to be the most convenient. With this definition log 1 = 0.

We shall now consider how log x behaves as x varies from 0 towards . It follows at once from the definition that log x is a continuous function of x which increases steadily with x and has a derivative

Dx log x = 1/x;
and it follows from § 175 that log x tends to  as x .

If x is positive but less than 1, then log x is negative. For

log x =1xdt t = x1dt t < 0.
Moreover, if we make the substitution t = 1/u in the integral, we obtain
log x =1xdt t = 11/xdu u = log(1/x).
Thus log x tends steadily to  as x decreases from 1 to 0.

The general form of the graph of the logarithmic function is shown in Fig. 52. Since the derivative of  log x is 1/x, the slope of


pict

Fig. 52.

the curve is very gentle when x is very large, and very steep when x is very small.

Examples LXXXII. 1. Prove from the definition that if u > 0 then

u/(1 + u) < log(1 + u) < u.

[For log(1 + u) =0u dt 1 + t, and the subject of integration lies between 1 and 1/(1 + u).]

2. Prove that log(1 + u) lies between u u2 2 and u u2 2(1 + u) when u is positive. [Use the fact that log(1 + u) = u 0u tdt 1 + t.]

3. If 0 < u < 1 then u < log(1 u) < u/(1 u).

4. Prove that

limx1 logx x 1 = limt0 log(1 + t) t = 1.

[Use Ex. 1.]

198. The functional equation satisfied by log x. The function log x satisfies the functional equation

f(xy) = f(x) + f(y). (1)

For, making the substitution t = yu, we see that

log xy =1xydt t =1/yxdu u =1xdu u 11/ydu u = log x log(1/y) = log x + log y,

which proves the theorem.

Examples LXXXIII. 1. It can be shown that there is no solution of the equation (1) which possesses a differential coefficient and is fundamentally distinct from logx. For when we differentiate the functional equation, first with respect to x and then with respect to y, we obtain the two equations

yf(xy) = f(x),xf(xy) = f(y);
and so, eliminating f(xy), xf(x) = yf(y). But if this is true for every pair of values of x and y, then we must have xf(x) = C, or f(x) = C/x, where C is a constant. Hence
f(x) =C x dx + C= Clogx + C,
and it is easy to see that C= 0. Thus there is no solution fundamentally distinct from logx, except the trivial solution f(x) = 0, obtained by taking C = 0.

2. Show in the same way that there is no solution of the equation

f(x) + f(y) = f x + y 1 xy
which possesses a differential coefficient and is fundamentally distinct from arctanx.

199. The manner in which log x tends to infinity with x. It will be remembered that in Ex. XXXVI. 6 we defined certain different ways in which a function of x may tend to infinity with x, distinguishing between functions which, when x is large, are of the first, second, third, … orders of greatness. A function f(x) was said to be of the kth order of greatness when f(x)/xk tends to a limit different from zero as x tends to infinity.

It is easy to define a whole series of functions which tend to infinity with x, but whose order of greatness is smaller than the first. Thus x, x3, x4, … are such functions. We may say generally that xα, where α is any positive rational number, is of the αth order of greatness when x is large. We may suppose α as small as we please, e.g. less than .0000001. And it might be thought that by giving α all possible values we should exhaust the possible ‘orders of infinity’ of f(x). At any rate it might be supposed that if f(x) tends to infinity with x, however slowly, we could always find a value of α so small that xα would tend to infinity more slowly still; and, conversely, that if f(x) tends to infinity with x, however rapidly, we could always find a value of α so great that xα would tend to infinity more rapidly still.

Perhaps the most interesting feature of the function log x is its behaviour as x tends to infinity. It shows that the presupposition stated above, which seems so natural, is unfounded. The logarithm of x tends to infinity with x, but more slowly than any positive power of x, integral or fractional. In other words log x but

log x xα 0
for all positive values of α. This fact is sometimes expressed loosely by saying that the ‘order of infinity of  log x is infinitely small’; but the reader will hardly require at this stage to be warned against such modes of expression.

200. Proof that (log x)/xα 0 as x . Let β be any positive number. Then 1/t < 1/t1β when t > 1, and so

log x =1xdt t <1x dt t1β,
or
log x < (xβ 1)/β < xβ/β,
when x > 1. Now if α is any positive number we can choose a smaller positive value of β. And then
0 < (log x)/xα < xβα/β(x > 1).
But, since α > β, xβα/β 0 as x , and therefore
(log x)/xα 0.

201. The behaviour of log x as x +0. Since

(log x)/xα = yα log y
if x = 1/y, it follows from the theorem proved above that
lim y+0yα log y = lim x+(log x)/xα = 0.
Thus log x tends to  and log(1/x) = log x to  as x tends to zero by positive values, but log(1/x) tends to  more slowly than any positive power of 1/x, integral or fractional.

202. Scales of infinity. The logarithmic scale. Let us consider once more the series of functions

x,x,x3,,xn,,
which possesses the property that, if f(x) and φ(x) are any two of the functions contained in it, then f(x) and φ(x) both tend to  as x , while f(x)/φ(x) tends to 0 or to according as f(x) occurs to the right or the left of φ(x) in the series. We can now continue this series by the insertion of new terms to the right of all those already written down. We can begin with logx, which tends to infinity more slowly than any of the old terms. Then logx tends to more slowly than logx, logx3 than logx, and so on. Thus we obtain a series
x,x,x3,,xn,logx,logx,logx3,logxn,
formed of two simply infinite series arranged one after the other. But this is not all. Consider the function loglogx, the logarithm of logx. Since (logx)/xα 0, for all positive values of α, it follows on putting x = logy that
(loglogy)/(logy)α = (logx)/xα 0.
Thus loglogy tends to  with y, but more slowly than any power of logy. Hence we may continue our series in the form

x,x,x3,logx,logx,logx3, loglogx,loglogx,loglogxn,;

and it will by now be obvious that by introducing the functions logloglogx, loglogloglogx, … we can prolong the series to any extent we like. By putting x = 1/y we obtain a similar scale of infinity for functions of y which tend to  as y tends to 0 by positive values.98

Examples LXXXIV. 1. Between any two terms f(x)F(x) of the series we can insert a new term φ(x) such that φ(x) tends to more slowly than f(x) and more rapidly than F(x). [Thus between x and x3 we could insert x5/12: between logx and logx3 we could insert (logx)5/12. And, generally, φ(x) = f(x)F(x) satisfies the conditions stated.]

2. Find a function which tends to  more slowly than x, but more rapidly than xα, where α is any rational number less than 1/2. [x/(logx) is such a function; or x/(logx)β, where β is any positive rational number.]

3. Find a function which tends to  more slowly than x, but more rapidly than x/(logx)α, where α is any rational number. [The function x/(loglogx) is such a function. It will be gathered from these examples that incompleteness is an inherent characteristic of the logarithmic scale of infinity.]

4. How does the function

f(x) = {xα(logx)α(loglogx)α}/{xβ(logx)β(loglogx)β}
behave as x tends to ? [If αβ then the behaviour of
f(x) = xαβ(logx)αβ(loglogx)αβ
is dominated by that of xαβ. If α = β then the power of x disappears and the behaviour of f(x) is dominated by that of (logx)αβ, unless α= β, when it is dominated by that of (loglogx)αβ. Thus f(x) if α > β, or α = β, α> β, or α = β, α= β, α> β, and f(x) 0 if α < β, or α = β, α< β, or α = β, α= β, α< β.]

5. Arrange the functions x/logx, xlogx/loglogx, xloglogx/logx, (xlogloglogx)/loglogx according to the rapidity with which they tend to infinity as x .

6. Arrange

loglogx/(xlogx),(logx)/x,xloglogx/x2 + 1,{x + 1}/x(logx)2
according to the rapidity with which they tend to zero as x .

7. Arrange

xloglog(1/x),x/{log(1/x)},xsinxlog(1/x),(1 cosx)log(1/x)
according to the rapidity with which they tend to zero as x +0.

8. Show that

Dx loglogx = 1/(xlogx),Dx logloglogx = 1/(xlogxloglogx),
and so on.

9. Show that

Dx(logx)α = α/{x(logx)1α},D x(loglogx)α = α/{xlogx(loglogx)1α},
and so on.

203. The number e. We shall now introduce a number, usually denoted by e, which is of immense importance in higher mathematics. It is, like π, one of the fundamental constants of analysis.

We define e as the number whose logarithm is 1. In other words e is defined by the equation

1 =1edt t .
Since log x is an increasing function of x, in the stricter sense of § 95, it can only pass once through the value 1. Hence our definition does in fact define one definite number.

Now log xy = log x + log y and so

log x2 = 2 log x, log x3 = 3 log x,, log xn = n log x,
where n is any positive integer. Hence
log en = n log e = n.
Again, if p and q are any positive integers, and ep/q denotes the positive qth root of ep, we have
p = log ep = log(ep/q)q = q log ep/q,
so that log ep/q = p/q. Thus, if y has any positive rational value, and ey denotes the positive yth power of e, we have

log ey = y, (1)

and log ey = log ey = y. Hence the equation (1) is true for all rational values of y, positive or negative. In other words the equations

y = log x,x = ey (2)

are consequences of one another so long as y is rational and ey has its positive value. At present we have not given any definition of a power such as ey in which the index is irrational, and the function ey is defined for rational values of y only.

Example. Prove that 2 < e < 3. [In the first place it is evident that

12dt t < 1,
and so 2 < e. Also
13dt t =12dt t +23dt t =01 du 2 u +01 du 2 + u = 401 du 4 u2 > 1,
so that e < 3.]

204. The exponential function. We now define the exponential function ey for all real values of y as the inverse of the logarithmic function. In other words we write

x = ey
if y = log x.

We saw that, as x varies from 0 towards , y increases steadily, in the stricter sense, from towards . Thus to one value of x corresponds one value of y, and conversely. Also y is a continuous function of x, and it follows from § 109 that x is likewise a continuous function of y.

It is easy to give a direct proof of the continuity of the exponential function. For if x = ey and x + ξ = ey+η then

η =xx+ξdt t .
Thus η is greater than ξ/(x + ξ) if ξ > 0, and than ξ/x if ξ < 0; and if η is very small ξ must also be very small.

Thus ey is a positive and continuous function of y which increases steadily from 0 towards  as y increases from towards . Moreover ey is the positive yth power of the number e, in accordance with the elementary definitions, whenever y is a rational number. In particular ey = 1 when y = 0. The general form of the graph of ey is as shown in Fig. 53.


pict

Fig. 53.

205. The principal properties of the exponential function. (1)  If x = ey, so that y = log x, then dy/dx = 1/x and

dx dy = x = ey.
Thus the derivative of the exponential function is equal to the function itself . More generally, if x = eay then dx/dy = aeay.

(2) The exponential function satisfies the functional equation

f(y + z) = f(y)f(z).

This follows, when y and z are rational, from the ordinary rules of indices. If y or z, or both, are irrational then we can choose two sequences y1, y2, …, yn, … and z1, z2, …, zn, … of rational numbers such that lim yn = y, lim zn = z. Then, since the exponential function is continuous, we have

ey × ez = lim eyn × lim ezn = lim eyn+zn = ey+z.
In particular ey × ey = e0 = 1, or ey = 1/ey.

We may also deduce the functional equation satisfied by ey from that satisfied by  log x. For if y1 = log x1, y2 = log x2, so that x1 = ey1, x2 = ey2, then y1 + y2 = log x1 + log x2 = log x1x2 and

ey1+y2 = elog x1x2 = x1x2 = ey1 × ey2 .

Examples LXXXV. 1. If dx/dy = ax then x = Keay, where K is a constant.

2. There is no solution of the equation f(y + z) = f(y)f(z) fundamentally distinct from the exponential function. [We assume that f(y) has a differential coefficient. Differentiating the equation with respect to y and z in turn, we obtain

f(y + z) = f(y)f(z),f(y + z) = f(y)f(z)
and so f(y)/f(y) = f(z)/f(z), and therefore each is constant. Thus if x = f(y) then dx/dy = ax, where a is a constant, so that x = Keay (Ex. 1).]

3. Prove that (eay 1)/y a as y 0. [Applying the Mean Value Theorem, we obtain eay 1 = ayeaη, where 0 < η< y.]

206. (3) The function ey tends to infinity with y more rapidly than any power of y, or

lim yα/ey = lim eyyα = 0
as y , for all values of α however great.

We saw that (log x)/xβ 0 as x , for any positive value of β however small. Writing α for 1/β, we see that (log x)α/x 0 for any value of α however large. The result follows on putting x = ey. It is clear also that eγy tends to  if γ > 0, and to 0 if γ < 0, and in each case more rapidly than any power of y.

From this result it follows that we can construct a ‘scale of infinity’ similar to that constructed in §202, but extending in the opposite direction; i.e. a scale of functions which tend to more and more rapidly as x .99 The scale is

x,x2,x3,ex,e2x,ex2 ,,ex3 ,,eex ,,
where of course ex2 , …, eex , … denote e(x2) , …, e(ex) , ….

The reader should try to apply the remarks about the logarithmic scale, made in §202 and Exs. LXXXIV, to this ‘exponential scale’ also. The two scales may of course (if the order of one is reversed) be combined into one scale

loglogx,logx,x,ex,eex ,.

207. The general power ax. The function ax has been defined only for rational values of x, except in the particular case when a = e. We shall now consider the case in which a is any positive number. Suppose that x is a positive rational number p/q. Then the positive value y of the power ap/q is given by yq = ap; from which it follows that

q log y = p log a, log y = (p/q) log a = x log a,
and so
y = ex log a.
We take this as our definition of ax when x is irrational. Thus 102 = e2 log 10. It is to be observed that ax, when x is irrational, is defined only for positive values of a, and is itself essentially positive; and that log ax = x log a. The most important properties of the function ax are as follows.

(1) Whatever value a may have, ax × ay = ax+y and (ax)y = axy. In other words the laws of indices hold for irrational no less than for rational indices. For, in the first place,

ax × ay = ex log a × ey log a = e(x+y) log a = ax+y;
and in the second
(ax)y = ey log ax = exy log a = axy.

(2) If a > 1 then ax = ex log a = eαx, where α is positive. The graph of ax is in this case similar to that of ex, and ax as x , more rapidly than any power of x.

If a < 1 then ax = ex log a = eβx, where β is positive. The graph of ax is then similar in shape to that of ex, but reversed as regards right and left, and ax 0 as x , more rapidly than any power of 1/x.

(3) ax is a continuous function of x, and

Dxax = D xex log a = ex log a log a = ax log a.

(4) ax is also a continuous function of a, and

Daax = D aex log a = ex log a(x/a) = xax1.

(5) (ax 1)/x log a as x 0. This of course is a mere corollary from the fact that Dxax = ax log a, but the particular form of the result is often useful; it is of course equivalent to the result (Ex. LXXXV. 3) that (eαx 1)/x α as x 0.

In the course of the preceding chapters a great many results involving the function ax have been stated with the limitation that x is rational. The definition and theorems given in this section enable us to remove this restriction.

208. The representation of ex as a limit. In Ch. IV, § 73, we proved that {1 + (1/n)}n tends, as n , to a limit which we denoted provisionally by e. We shall now identify this limit with the number e of the preceding sections. We can however establish a more general result, viz. that expressed by the equations

lim n1 + x nn = lim n1 x nn = ex. (1)

As the result is of very great importance, we shall indicate alternative lines of proof.

(1) Since

d dt log(1 + xt) = x 1 + xt,
it follows that
lim h0 log(1 + xh) h = x.
If we put h = 1/ξ, we see that
lim ξ log 1 + x ξ = x
as ξ or ξ . Since the exponential function is continuous it follows that
1 + x ξξ = eξ log{1+(x/ξ)} ex
as ξ or ξ : i.e. that

lim ξ1 + x ξξ = lim ξ1 + x ξξ = ex. (2)

If we suppose that ξ or ξ through integral values only, we obtain the result expressed by the equations (1).

(2) If n is any positive integer, however large, and x > 1, we have

1x dt t1+(1/n) <1xdt t <1x dt t1(1/n),
or

n(1 x1/n) < logx < n(x1/n 1). (3)

Writing y for logx, so that y is positive and x = ey, we obtain, after some simple transformations,

1 + y nn < x < 1 y nn. (4)

Now let

1 + y n = η1,1 y n = 1 η2.
Then 0 < η1 < η2, at any rate for sufficiently large values of n; and, by (9) of §74,
η2n η 1n < nη 2n1(η 2 η1) = y2η 2n/n,
which evidently tends to 0 as n . The result now follows from the inequalities (4). The more general result (2) may be proved in the same way, if we replace 1/n by a continuous variable h.

209. The representation of logx as a limit. We can also prove (cf. §75) that

limn(1 x1/n) = limn(x1/n 1) = logx.

For

n(x1/n 1) n(1 x1/n) = n(x1/n 1)(1 x1/n),
which tends to zero as n , since n(x1/n 1) tends to a limit (§75) and x1/n to 1 (Ex. XXVII. 10). The result now follows from the inequalities (3) of §208.

Examples LXXXVI. 1. Prove, by taking y = 1 and n = 6 in the inequalities (4) of §208, that 2.5 < e < 2.9.

2. Prove that if t > 1 then (t1/n t1/n)/(t t1) < 1/n, and so that if x > 1 then

1x dt t1(1/n) 1x dt t1+(1/n) < 1 n1x t 1 t dt t = 1 n x + 1 x 2.
Hence deduce the results of §209.

3. If ξn is a function of n such that nξn l as n , then (1 + ξn)n el. [Writing nlog(1 + ξn) in the form

l nξn l log(1 + ξn) ξn ,
and using Ex. LXXXII. 4, we see that nlog(1 + ξn) l.]

4. If nξn , then (1 + ξn)n ; and if 1 + ξn > 0 and nξn , then

(1 + ξn)n 0.

5. Deduce from (1) of §208 the theorem that ey tends to infinity more rapidly than any power of y.

210. Common logarithms. The reader is probably familiar with the idea of a logarithm and its use in numerical calculation. He will remember that in elementary algebra log ax, the logarithm of x to the base a, is defined by the equations

x = ay,y = log ax.
This definition is of course applicable only when y is rational, though this point is often passed over in silence.

Our logarithms are therefore logarithms to the base e. For numerical work logarithms to the base 10 are used. If

y = log x = log ex,z = log 10x,
then x = ey and also x = 10z = ez log 10, so that
log 10x = (log ex)/(log e10).
Thus it is easy to pass from one system to the other when once log e10 has been calculated.

It is no part of our purpose in this book to go into details concerning the practical uses of logarithms. If the reader is not familiar with them he should consult some text-book on Elementary Algebra or Trigonometry.100

Examples LXXXVII. 1. Show that

Dxeax cosbx = reax cos(bx + θ),D xeax sinbx = reax sin(bx + θ)
where r = a2 + b2, cosθ = a/r, sinθ = b/r. Hence determine the nth derivatives of the functions eax cosbx, eax sinbx, and show in particular that Dxneax = aneax.

2. Trace the curve y = eax sinbx, where a and b are positive. Show that y has an infinity of maxima whose values form a geometrical progression and which lie on the curve

y = b a2 + b2eax.
(Math. Trip. 1912.)

3. Integrals containing the exponential function. Prove that

eax cosbxdx = acosbx + bsinbx a2 + b2 eax, eax sinbxdx = asinbx bcosbx a2 + b2 eax.

[Denoting the two integrals by IJ, and integrating by parts, we obtain

aI = eax cosbx + bJ,aJ = eax sinbx bI.
Solve these equations for I and J.]

4. Prove that the successive areas bounded by the curve of Ex. 2 and the positive half of the axis of x form a geometrical progression, and that their sum is

b a2 + b21 + eaπ/b 1 eaπ/b.

5. Prove that if a > 0 then

0eax cosbxdx = a a2 + b2,0eax sinbxdx = b a2 + b2.

6. If In =eaxxndx then aIn = eaxxn nIn1. [Integrate by parts. It follows that In can be calculated for all positive integral values of n.]

7. Prove that, if n is a positive integer, then

0ξexxndx = n!eξ eξ 1 ξ ξ2 2! ξn n!
and
0exxndx = n!.

8. Show how to find the integral of any rational function of ex. [Put x = logu, when ex = u, dx/du = 1/u, and the integral is transformed into that of a rational function of u.]

9. Integrate

e2x (c2ex + a2ex)(c2ex + b2ex),
distinguishing the cases in which a is and is not equal to b.

10. Prove that we can integrate any function of the form P(x,eax,ebx,), where P denotes a polynomial. [This follows from the fact that P can be expressed as the sum of a number of terms of the type Axmekx, where m is a positive integer.]

11. Show how to integrate any function of the form

P(x,eax,ebx,,coslx,cosmx,,sinlx,sinmx,).

12. Prove that aeλxR(x)dx, where λ > 0 and a is greater than the greatest root of the denominator of R(x), is convergent. [This follows from the fact that eλx tends to infinity more rapidly than any power of x.]

13. Prove that eλx2+μxdx, where λ > 0, is convergent for all values of μ, and that the same is true of eλx2+μxxndx, where n is any positive integer.

14. Draw the graphs of ex2 , ex2 , xex, xex, xex2 , xex2 , and xlogx, determining any maxima and minima of the functions and any points of inflexion on their graphs.

15. Show that the equation eax = bx, where a and b are positive, has two real roots, one, or none, according as b > ae, b = ae, or b < ae. [The tangent to the curve y = eax at the point (ξ,eaξ) is

y eaξ = aeaξ(x ξ),
which passes through the origin if aξ = 1, so that the line y = aex touches the curve at the point (1/a,e). The result now becomes obvious when we draw the line y = bx. The reader should discuss the cases in which a or b or both are negative.]

16. Show that the equation ex = 1 + x has no real root except x = 0, and that ex = 1 + x + 1 2x2 has three real roots.

17. Draw the graphs of the functions

log(x + x2 + 1),log 1 + x 1 x,eax cos2bx, e(1/x)2 ,e(1/x)2 1/x,e cot x,e cot 2x.

18. Determine roughly the positions of the real roots of the equations

log(x + x2 + 1) = x 100,ex 2 + x 2 x = 1 10,000,ex sinx = 7,ex2 sinx = 10,000.

19. The hyperbolic functions. The hyperbolic functions coshx,101 sinhx, … are defined by the equations

coshx = 1 2(ex + ex),sinhx = 1 2(ex ex), tanhx = (sinhx)/(coshx),cothx = (coshx)/(sinhx), sechx = 1/(coshx),cosechx = 1/(sinhx).

Draw the graphs of these functions.

20. Establish the formulae

cosh(x) = coshx,sinh(x) = sinhx,tanh(x) = tanhx, cosh2x sinh2x = 1,sech2x + tanh2x = 1,coth2x cosech2x = 1, cosh2x = cosh2x + sinh2x,sinh2x = 2sinhxcoshx, 2cosh(x + y) = coshxcoshy + sinhxsinhy, sinh(x + y) = sinhxcoshy + coshxsinhy.

21. Verify that these formulae may be deduced from the corresponding formulae in cosx and sinx, by writing coshx for cosx and isinhx for sinx.

[It follows that the same is true of all the formulae involving cosnx and sinnx which are deduced from the corresponding elementary properties of cosx and sinx. The reason of this analogy will appear in Ch. X.]

22. Express coshx and sinhx in terms (a) of cosh2x (b) of sinh2x. Discuss any ambiguities of sign that may occur.

(Math. Trip. 1908.)

23. Prove that

Dx coshx = sinhx,Dx sinhx = coshx, Dx tanhx = sech2x,D x cothx = cosech2x, Dx sechx = sechxtanhx,Dx cosechx = cosechxcothx, Dx logcoshx = tanhx,Dx logsinhx= cothx, Dx arctanex = 1 2 sechx,Dx logtanh 1 2x= cosechx.

[All these formulae may of course be transformed into formulae in integration.]

24. Prove that coshx > 1 and 1 < tanhx < 1.

25. Prove that if y = coshx then x = log{y ±y2 1}, if y = sinhx then x = log{y + y2 + 1}, and if y = tanhx then x = 1 2 log{(1 + y)/(1 y)}. Account for the ambiguity of sign in the first case.

26. We shall denote the functions inverse to coshx, sinhx, tanhx by argcoshx, argsinhx, argtanhx. Show that argcoshx is defined only when x 1, and is in general two-valued, while argsinhx is defined for all real values of x, and argtanhx when 1 < x < 1, and both of the two latter functions are one-valued. Sketch the graphs of the functions.

27. Show that if 1 2π < x < 1 2π and y is positive, and cosxcoshy = 1, then

y = log(secx + tanx),Dxy = secx,Dyx = sechy.

28. Prove that if a > 0 then dx x2 + a2 = argsinh(x/a), and dx x2 a2 is equal to argcosh(x/a) or to argcosh(x/a), according as x > 0 or x < 0.

29. Prove that if a > 0 then dx x2 a2 is equal to (1/a)argtanh(x/a) or to (1/a)argcoth(x/a), according as x is less than or greater than a. [The results of Exs. 28 and 29 furnish us with an alternative method of writing a good many of the formulae of Ch. VI.]

30. Prove that

dx (xa)(xb) = 2log{x a + x b} (a < b < x), dx (ax)(bx) = 2log{a x + b x}(x < a < b), dx (xa)(bx) = 2arctanxa bx (a < x < b).

31. Prove that

01xlog(1 + 1 2x)dx = 3 4 3 2 log 3 2 < 1 201x2dx = 1 6.
(Math. Trip. 1913.)

32. Solve the equation acoshx + bsinhx = c, where c > 0, showing that it has no real roots if b2 + c2 a2 < 0, while if b2 + c2 a2 > 0 it has two, one, or no real roots according as a + b and a b are both positive, of opposite signs, or both negative. Discuss the case in which b2 + c2 a2 = 0.

33. Solve the simultaneous equations coshxcoshy = a, sinhxsinhy = b.

34. x1/x 1 as x . [For x1/x = e(log x)/x, and (logx)/x 0. Cf. Ex. XXVII. 11.] Show also that the function x1/x has a maximum when x = e, and draw the graph of the function for positive values of x.

35. xx 1 as x +0.

36. If {f(n + 1)}/{f(n)}l, where l > 0, as n , then f(n)n l. [For logf(n + 1) logf(n) logl, and so (1/n)logf(n) logl (Ch. IV, Misc. Ex. 27).]

37. n!n/n 1/e as n .

[If f(n) = nnn! then {f(n + 1)}/{f(n)}= {1 + (1/n)}n 1/e. Now use Ex. 36.]

38. (2n)!/(n!)2n 4 as n .

39. Discuss the approximate solution of the equation ex = x1,000,000.

[It is easy to see by general graphical considerations that the equation has two positive roots, one a little greater than 1 and one very large,102 and one negative root a little greater than  1. To determine roughly the size of the large positive root we may proceed as follows. If ex = x1,000,000 then

x = 106 logx,logx = 13.82 + loglogx,loglogx = 2.63 + log 1 + loglogx 13.82 ,
roughly, since 13.82 and 2.63 are approximate values of log106 and loglog106 respectively. It is easy to see from these equations that the ratios logx : 13.82 and loglogx : 2.63 do not differ greatly from unity, and that
x = 106(13.82 + loglogx) = 106(13.82 + 2.63) = 16,450,000
gives a tolerable approximation to the root, the error involved being roughly measured by 106(loglogx 2.63) or (106 loglogx)/13.82 or (106 ×2.63)/13.82, which is less than 200,000. The approximations are of course very rough, but suffice to give us a good idea of the scale of magnitude of the root.]

40. Discuss similarly the equations

ex = 1,000,000x1,000,000,ex2 = x1,000,000,000.

211. Logarithmic tests of convergence for series and integrals. We showed in Ch. VIII (§§ 175 et seq.) that

1 1 ns,adx xs (a > 0)
are convergent if s > 1 and divergent if s 1. Thus (1/n) is divergent, but n1α is convergent for all positive values of α.

We saw however in § 200 that with the aid of logarithms we can construct functions which tend to zero, as n , more rapidly than 1/n, yet less rapidly than n1α, however small α may be, provided of course that it is positive. For example 1/(n log n) is such a function, and the question as to whether the series

1 n log n
is convergent or divergent cannot be settled by comparison with any series of the type ns.

The same is true of such series as

1 n(log n)2, log log n n log n .
It is a question of some interest to find tests which shall enable us to decide whether series such as these are convergent or divergent; and such tests are easily deduced from the Integral Test of § 174.

For since

Dx(log x)1s = 1 s x(log x)s,Dx log log x = 1 x log x,
we have
aξ dx x(log x)s = (log ξ)1s (log a)1s 1 s ,aξ dx x log x = log log ξ log log a,
if a > 1. The first integral tends to the limit (log a)1s/(1 s) as ξ , if s > 1, and to  if s < 1. The second integral tends to . Hence the series and integral
n0 1 n(log n)s,a dx x(log x)s,
where n0 and a are greater than unity, are convergent if s > 1, divergent if s 1.

It follows, of course, that φ(n) is convergent if φ(n) is positive and less than K/{n(log n)s}, where s > 1, for all values of n greater than some definite value, and divergent if φ(n) is positive and greater than K/(n log n) for all values of n greater than some definite value. And there is a corresponding theorem for integrals which we may leave to the reader.

Examples LXXXVIII. 1. The series

1 n(logn)2, (logn)100 n101/100 , n2 1 n2 + 1 1 n(logn)7/6
are convergent. [The convergence of the first series is a direct consequence of the theorem of the preceding section. That of the second follows from the fact that (logn)100 is less than nβ for sufficiently large values of n, however small β may be, provided that it is positive. And so, taking β = 1/200, (logn)100n101/100 is less than n201/200 for sufficiently large values of n. The convergence of the third series follows from the comparison test at the end of the last section.]

2. The series

1 n(logn)6/7, 1 n100/101(logn)100, nlogn (nlogn)2 + 1
are divergent.

3. The series

(logn)p n1+s , (logn)p(loglogn)q n1+s , (loglogn)p n(logn)1+s,
where s > 0, are convergent for all values of p and q; similarly the series
1 n1s(logn)p, 1 n1s(logn)p(loglogn)q, 1 n(logn)1s(loglogn)p
are divergent.

4. The question of the convergence or divergence of such series as

1 nlognloglogn, logloglogn nlognloglogn
cannot be settled by the theorem of p. 1267, since in each case the function under the sign of summation tends to zero more rapidly than 1/(nlogn) yet less rapidly than n1(logn)1α, where α is any positive number however small. For such series we need a still more delicate test. The reader should be able, starting from the equations

Dx(logkx)1s = 1 s xlogxlog2xlogk1x(logkx)s, Dx logk+1x = 1 xlogxlog2xlogk1xlogkx,

where log2x = loglogx, log3x = logloglogx, …, to prove the following theorem: the series and integral

n0 1 nlognlog2nlogk1n(logkn)s,a dx xlogxlog2xlogk1x(logkx)s
are convergent if s > 1 and divergent if s 1, n0 and a being any numbers sufficiently great to ensure that logkn and logkx are positive when n n0 or x a. These values of n0 and a increase very rapidly as k increases: thus logx > 0 requires x > 1, log2x > 0 requires x > e, log3x > 0 requires x > ee, and so on; and it is easy to see that ee > 10, eee > e10 > 20,000, eeee > e20,000 > 108000.

The reader should observe the extreme rapidity with which the higher exponential functions, such as eex and eeex , increase with x. The same remark of course applies to such functions as aax and aaax , where a has any value greater than unity. It has been computed that 999  has 369,693,100 figures, while 101010 has of course 10,000,000,000. Conversely, the rate of increase of the higher logarithmic functions is extremely slow. Thus to make loglogloglogx > 1 we have to suppose x a number with over 8000 figures.103

5. Prove that the integral 0a1 x log 1 xsdx, where 0 < a < 1, is convergent if s < 1, divergent if s 1. [Consider the behaviour of

εa1 x log 1 xsdx
as ε +0. This result also may be refined upon by the introduction of higher logarithmic factors.]

6. Prove that 01 1 x log 1 xsdx has no meaning for any value of s. [The last example shows that s < 1 is a necessary condition for convergence at the lower limit: but {log(1/x)}s tends to  like (1 x)s, as x 1 0, if s is negative, and so the integral diverges at the upper limit when s < 1.]

7. The necessary and sufficient conditions for the convergence of 01xa1 log 1 xsdx are a > 0, s > 1.

Examples LXXXIX. 1. Euler’s limit. Show that

φ(n) = 1 + 1 2 + 1 3 + + 1 n 1 logn
tends to a limit γ as n , and that 0 < γ 1. [This follows at once from §174. The value of γ is in fact .577, and γ is usually called Euler’s constant.]

2. If a and b are positive then

1 a + 1 a + b + 1 a + 2b + + 1 a + (n 1)b 1 blog(a + nb)
tends to a limit as n .

3. If 0 < s < 1 then

φ(n) = 1 + 2s + 3s + + (n 1)s n1s 1 s
tends to a limit as n .

4. Show that the series

1 1 + 1 2(1 + 1 2) + 1 3(1 + 1 2 + 1 3) +
is divergent. [Compare the general term of the series with 1/(nlogn).] Show also that the series derived from ns, in the same way that the above series is derived from  (1/n), is convergent if s > 1 and otherwise divergent.

5. Prove generally that if un is a series of positive terms, and

sn = u1 + u2 + + un,
then (un/sn1) is convergent or divergent according as un is convergent or divergent. [If un is convergent then sn1 tends to a positive limit l, and so (un/sn1) is convergent. If un is divergent then sn1 , and
un/sn1 > log{1 + (un/sn1)}= log(sn/sn1)
(Ex. LXXXII. 1); and it is evident that
log(s2/s1) + log(s3/s2) + + log(sn/sn1) = log(sn/s1)
tends to  as n .]

6. Prove that the same result holds for the series (un/sn). [The proof is the same in the case of convergence. If un is divergent, and un < sn1 from a certain value of n onwards, then sn < 2sn1, and the divergence of (un/sn) follows from that of (un/sn1). If on the other hand un sn1 for an infinity of values of n, as might happen with a rapidly divergent series, then un/sn 1 2 for all these values of n.]

7. Sum the series 1 1 2 + 1 3 . [We have

1 + 1 2 + + 1 2n = log(2n + 1) + γ + εn,2 1 2 + 1 4 + + 1 2n = log(n + 1) + γ + ε,
by Ex. 1, γ denoting Euler’s constant, and εnε being numbers which tend to zero as n . Subtracting and making n we see that the sum of the given series is log2. See also §213.]

8. Prove that the series

0(1)n 1 + 1 2 + + 1 n + 1 logn C
oscillates finitely except when C = γ, when it converges.

212. Series connected with the exponential and logarithmic functions. Expansion of ex by Taylor’s Theorem. Since all the derivatives of the exponential function are equal to the function itself, we have

ex = 1 + x + x2 2! + + xn1 (n 1)! + xn n! eθx
where 0 < θ < 1. But xn/n! 0 as n , whatever be the value of x (Ex. XXVII. 12); and eθx < ex. Hence, making n tend to , we have

ex = 1 + x + x2 2! + + xn n! + . (1)

The series on the right-hand side of this equation is known as the exponential series. In particular we have

e = 1 + 1 + 1 2! + + 1 n! + ; (2)

and so

1 + 1 + 1 2! + + 1 n! + x = 1 + x + x2 2! + + xn n! + , (3)

a result known as the exponential theorem. Also

ax = ex log a = 1 + (x log a) + (x log a)2 2! + (4)

for all positive values of a.

The reader will observe that the exponential series has the property of reproducing itself when every term is differentiated, and that no other series of powers of x would possess this property: for some further remarks in this connection see Appendix II.

The power series for ex is so important that it is worth while to investigate it by an alternative method which does not depend upon Taylor’s Theorem. Let

En(x) = 1 + x + x2 2! + + xn n! ,
and suppose that x > 0. Then
1 + x nn = 1 + n x n + n(n 1) 1 2 x n2 + + n(n 1)1 1 2n x nn,
which is less than En(x). And, provided n > x, we have also, by the binomial theorem for a negative integral exponent,
1 x nn = 1 + n x n + n(n + 1) 1 2 x n2 + > E n(x).
Thus
1 + x nn < E n(x) < 1 x nn.
But (§208) the first and last functions tend to the limit ex as n , and therefore En(x) must do the same. From this the equation (1) follows when x is positive; its truth when x is negative follows from the fact that the exponential series, as was shown in Ex. LXXXI. 7, satisfies the functional equation f(x)f(y) = f(x + y), so that f(x)f(x) = f(0) = 1.

Examples XC. 1. Show that

coshx = 1 + x2 2! + x4 4! + ,sinhx = x + x3 3! + x5 5! + .

2. If x is positive then the greatest term in the exponential series is the ([x] + 1)-th, unless x is an integer, when the preceding term is equal to it.

3. Show that n! > (n/e)n. [For nn/n! is one term in the series for en.]

4. Prove that en = (nn/n!)(2 + S1 + S2), where

S1 = 1 1 + ν + 1 (1 + ν)(1 + 2ν) + ,S2 = (1 ν) + (1 ν)(1 2ν) + ,
and ν = 1/n; and deduce that n! lies between 2(n/e)n and 2(n + 1)(n/e)n.

5. Employ the exponential series to prove that ex tends to infinity more rapidly than any power of x. [Use the inequality ex > xn/n!.]

6. Show that e is not a rational number. [If e = p/q, where p and q are integers, we must have

p q = 1 + 1 + 1 2! + 1 3! + + 1 q! +
or, multiplying up by q!,
q! p q 1 1 1 2! 1 q! = 1 q + 1 + 1 (q + 1)(q + 2) +
and this is absurd, since the left-hand side is integral, and the right-hand side less than {1/(q + 1)}+ {1/(q + 1)}2 + = 1/q.]

7. Sum the series 0Pr(n)xn n! , where Pr(n) is a polynomial of degree r in n. [We can express Pr(n) in the form

A0 + A1n + A2n(n 1) + + Arn(n 1)(n r + 1),
and

0P r(n)xn n! = A0 0xn n! + A1 1 xn (n 1)! + + Ar r xn (n r)! = (A0 + A1x + A2x2 + + A rxr)ex.]

8. Show that

1n3 n! xn = (x + 3x2 + x3)ex, 1n4 n! xn = (x + 7x2 + 6x3 + x4)ex;
and that if Sn = 13 + 23 + + n3 then
1S nxn n! = 1 4(4x + 14x2 + 8x3 + x4)ex.
In particular the last series is equal to zero when x = 2.
(Math. Trip. 1904.)

9. Prove that (n/n!) = e, (n2/n!) = 2e, (n3/n!) = 5e, and that (nk/n!), where k is any positive integer, is a positive integral multiple of e.

10. Prove that 1(n 1)xn (n + 2)n! = (x2 3x + 3)ex + 1 2x2 3/x2.

[Multiply numerator and denominator by n + 1, and proceed as in Ex. 7.]

11. Determine abc so that {(x + a)ex + (bx + c)}/x3 tends to a limit as x 0, evaluate the limit, and draw the graph of the function ex + bx + c x + a .

12. Draw the graphs of 1 + x, 1 + x + 1 2x2, 1 + x + 1 2x2 + 1 6x3, and compare them with that of ex.

13. Prove that ex 1 + x xn 2! + (1)nxn n! is positive or negative according as n is odd or even. Deduce the exponential theorem.

14. If

X0 = ex,X 1 = ex 1,X 2 = ex 1 x,X 3 = ex 1 x (x2/2!),,
then dXν/dx = Xν1. Hence prove that if t > 0 then
X1(t) =0tX 0dx < tet,X 2(t) =0tX 1dx <0txexdx < et0txdx = t2 2!et,
and generally Xν(t) < tν ν!et. Deduce the exponential theorem.

15. Show that the expansion in powers of p of the positive root of x2+p = a2 begins with the terms

a{1 1 2ploga + 1 8p2 loga(2 + loga)}.
(Math. Trip. 1909.)

213. The logarithmic series. Another very important expansion in powers of x is that for  log(1 + x). Since

log(1 + x) =0x dt 1 + t,
and 1/(1 + t) = 1 t + t2 if t is numerically less than unity, it is natural to expect104 that log(1 + x) will be equal, when 1 < x < 1, to the series obtained by integrating each term of the series 1 t + t2 from t = 0 to t = x, i.e. to the series x 1 2x2 + 1 3x3 . And this is in fact the case. For
1/(1 + t) = 1 t + t2 + (1)m1tm1 + (1)mtm 1 + t ,
and so, if x > 1,
log(1 + x) =0x dt 1 + t = x x2 2 + + (1)m1xm m + (1)mR m,
where
Rm =0x tmdt 1 + t.

We require to show that the limit of Rm, when m tends to , is zero. This is almost obvious when 0 < x 1; for then Rm is positive and less than

0xtmdt = xm+1 m + 1,
and therefore less than 1/(m + 1). If on the other hand 1 < x < 0, we put t = u and x = ξ, so that
Rm = (1)m0ξumdu 1 u,
which shows that Rm has the sign of (1)m. Also, since the greatest value of 1/(1 u) in the range of integration is 1/(1 ξ), we have
0 < Rm < 1 1 ξ0ξumdu = ξm (m + 1)(1 ξ) < 1 (m + 1)(1 ξ) :
and so Rm 0.

Hence

log(1 + x) = x 1 2x2 + 1 3x3 ,
provided that 1 < x 1. If x lies outside these limits the series is not convergent. If x = 1 we obtain
log 2 = 1 1 2 + 1 3 ,
a result already proved otherwise (Ex. LXXXIX. 7).

214. The series for the inverse tangent. It is easy to prove in a similar manner that

arctan x =0x dt 1 + t2 =0x(1 t2 + t4 )dt = x 1 3x3 + 1 5x5 ,

provided that 1 x 1. The only difference is that the proof is a little simpler; for, since arctan x is an odd function of x, we need only consider positive values of x. And the series is convergent when x = 1 as well as when x = 1. We leave the discussion to the reader. The value of  arctan x which is represented by the series is of course that which lies between 1 4π and 1 4π when 1 x 1, and which we saw in Ch. VII (Ex. LXIII. 3) to be the value represented by the integral. If x = 1, we obtain the formula
1 4π = 1 1 3 + 1 5 .

Examples XCI. 1. log 1 1 x = x + 1 2x2 + 1 3x3 + if 1 x < 1.

2. argtanhx = 1 2 log 1 + x 1 x = x + 1 3x3 + 1 5x5 + if 1 < x < 1.

3. Prove that if x is positive then

log(1 + x) = x 1 + x + 1 2 x 1+x 2 + 1 3 x 1+x 3 + .
(Math. Trip. 1911.)

4. Obtain the series for log(1 + x) and arctanx by means of Taylor’s theorem.

[A difficulty presents itself in the discussion of the remainder in the first series when x is negative, if Lagrange’s form Rn = (1)n1xn/{n(1 + θx)n}is used; Cauchy’s form, viz.

Rn = (1)n1(1 θ)n1xn/(1 + θx)n,
should be used (cf. the corresponding discussion for the Binomial Series, Ex. LVI. 2 and §163).

In the case of the second series we have

Dxn arctanx = D xn1{1/(1 + x2)} = (1)n1(n 1)!(x2 + 1)n/2 sin{narctan(1/x)}

(Ex. XLV. 11), and there is no difficulty about the remainder, which is obviously not greater in absolute value than 1/n.105]

5. If y > 0 then

logy = 2 y 1 y + 1 + 1 3 y 1 y + 13 + 1 5 y 1 y + 15 + .

[Use the identity y =(1 + y 1 y + 1)(1 y 1 y + 1). This series may be used to calculate log2, a purpose for which the series 1 1 2 + 1 3 , owing to the slowness of its convergence, is practically useless. Put y = 2 and find log2 to 3 places of decimals.]

6. Find log10 to 3 places of decimals from the formula

log10 = 3log2 + log(1 + 1 4).

7. Prove that

log x + 1 x = 2 1 2x + 1 + 1 3(2x + 1)3 + 1 5(2x + 1)5 +
if x > 0, and that
log (x 1)2(x + 2) (x + 1)2(x 2) = 2 2 x3 3x + 1 3 2 x3 3x3 + 1 5 2 x3 3x5 +
if x > 2. Given that log2 = .6931471 and log3 = 1.0986123, show, by putting x = 10 in the second formula, that log11 = 2.397895.
(Math. Trip. 1912.)

8. Show that if log2, log5, and log11 are known, then the formula

log13 = 3log11 + log5 9log2
gives log13 with an error practically equal to .00015.
(Math. Trip. 1910.)

9. Show that

1 2 log2 = 7a + 5b + 3c,1 2 log3 = 11a + 8b + 5c,1 2 log5 = 16a + 12b + 7c,
where a = argtanh(1/31), b = argtanh(1/49), c = argtanh(1/161).

[These formulae enable us to find log2, log3, and log5 rapidly and with any degree of accuracy.]

10. Show that

1 4π = arctan(1/2) + arctan(1/3) = 4arctan(1/5) arctan(1/239), and calculate π to 6 places of decimals.

11. Show that the expansion of (1 + x)1+x in powers of x begins with the terms 1 + x + x2 + 1/2x3.

(Math. Trip. 1910.)

12. Show that

log10e x(x + 1)log10 1 + x x = log10e 24x2 ,
approximately, for large values of x. Apply the formula, when x = 10, to obtain an approximate value of log10e, and estimate the accuracy of the result.
(Math. Trip. 1910.)

13. Show that

1 1 xlog 1 1 x = x + 1 + 1 2 x2 + 1 + 1 2 + 1 3 x3 + ,
if 1 < x < 1. [Use Ex. LXXXI. 2.]

14. Using the logarithmic series and the facts that log102.3758 = .3758099 and log10e = .4343, show that an approximate solution of the equation x = 100log10x is 237.58121.

(Math. Trip. 1910.)

15. Expand logcosx and log(sinx/x) in powers of x as far as x4, and verify that, to this order,

logsinx = logx 1 45 logcosx + 64 45 logcos 1 2x.
(Math. Trip. 1908.)

16. Show that

0x dt 1 + t4 = x 1 5x5 + 1 9x9
if 1 x 1. Deduce that
1 1 5 + 1 9 = {π + 2log(2 + 1)}/42.
(Math. Trip. 1896.)

[Proceed as in §214 and use the result of Ex. XLVIII. 7.]

17. Prove similarly that

1 3 1 7 + 1 11 =01 t2dt 1+t4 = {π 2log(2 + 1)}/42.

18. Prove generally that if a and b are positive integers then

1 a 1 a + b + 1 a + 2b =01ta1dt 1 + tb ,
and so that the sum of the series can be found. Calculate in this way the sums of 1 1 4 + 1 7 and 1 2 1 5 + 1 8 .

215. The Binomial Series. We have already (§ 163) investigated the Binomial Theorem

(1 + x)m = 1 + m 1 x + m 2 x2 + ,
assuming that 1 < x < 1 and that m is rational. When m is irrational we have

(1 + x)m = em log(1+x), Dx(1 + x)m = {m/(1 + x)}em log(1+x) = m(1 + x)m1,

so that the rule for the differentiation of (1 + x)m remains the same, and the proof of the theorem given in § 163 retains its validity. We shall not discuss the question of the convergence of the series when x = 1 or x = 1.106

Examples XCII. 1. Prove that if 1 < x < 1 then

1 1 + x2 = 1 1 2x2 + 1 3 2 4x4 , 1 1 x2 = 1 + 1 2x2 + 1 3 2 4x4 + .

2. Approximation to quadratic and other surds. Let M be a quadratic surd whose numerical value is required. Let N2 be the square nearest to M; and let M = N2 + x or M = N2 x, x being positive. Since x cannot be greater than N, x/N2 is comparatively small and the surd M = N1 ±(x/N2) can be expressed in a series

= N 1 ±1 2 x N2 1 1 2 4 x N2 2 ±,
which is at any rate fairly rapidly convergent, and may be very rapidly so. Thus
67 = 64 + 3 = 8 1 + 1 2 3 64 1 1 2 4 3 642 + .

Let us consider the error committed in taking 8 3 16 (the value given by the first two terms) as an approximate value. After the second term the terms alternate in sign and decrease. Hence the error is one of excess, and is less than 32/642, which is less than .003.

3. If x is small compared with N2 then

N2 + x = N + x 4N + Nx 2(2N2 + x),
the error being of the order x4/N7. Apply the process to 907.

[Expanding by the binomial theorem, we have

N2 + x = N + x 2N x2 8N3 + x3 16N5,
the error being less than the numerical value of the next term, viz. 5x4/128N7. Also
Nx 2(2N2 + x) = x 4N 1 + x 2N2 1 = x 4N x2 8N3 + x3 16N5,
the error being less than x4/32N7. The result follows. The same method may be applied to surds other than quadratic surds, e.g. to 10313.]

4. If M differs from N3 by less than 1 per cent. of either then M3 differs from 2 3N + 1 3(M/N2) by less than N/90,000.

(Math. Trip. 1882.)

5. If M = N4 + x, and x is small compared with N, then a good approximation for M4 is

51 56N + 5 56 M N3 + 27Nx 14(7M + 5N4).
Show that when N = 10, x = 1, this approximation is accurate to 16 places of decimals.
(Math. Trip. 1886.)

6. Show how to sum the series

0P r(n)m n xn,
where Pr(n) is a polynomial of degree r in n.

[Express Pr(n) in the form A0 + A1n + A2n(n 1) + as in Ex. XC. 7.]

7. Sum the series 0nm n xn, 0n2m n xn and prove that

0n3m n xn = {m3x3 + m(3m 1)x2 + mx}(1 + x)m3.

216. An alternative method of development of the theory of the exponential and logarithmic functions. We shall now give an outline of a method of investigation of the properties of ex and logx entirely different in logical order from that followed in the preceding pages. This method starts from the exponential series 1 + x + x2 2! + . We know that this series is convergent for all values of x, and we may therefore define the function expx by the equation

expx = 1 + x + x2 2! + . (1)

We then prove, as in Ex. LXXXI. 7, that

expx ×expy = exp(x + y). (2)

Again

exph 1 h = 1 + h 2! + h2 3! + = 1 + ρ(h),
where ρ(h) is numerically less than
1 2h+ 1 2h2 + 1 2h3 + = 1 2h/(1 1 2h),
so that ρ(h) 0 as h 0. And so
exp(x + h) expx h = expx exph 1 h expx
as h 0, or

Dx expx = expx. (3)

Incidentally we have proved that expx is a continuous function.

We have now a choice of procedure. Writing y = expx and observing that exp0 = 1, we have

dy dx = y,x =1ydt t ,
and, if we define the logarithmic function as the function inverse to the exponential function, we are brought back to the point of view adopted earlier in this chapter.

But we may proceed differently. From (2) it follows that if n is a positive integer then

(expx)n = expnx,(exp1)n = expn.
If x is a positive rational fraction m/n, then
{exp(m/n)}n = expm = (exp1)m,
and so exp(m/n) is equal to the positive value of (exp1)m/n. This result may be extended to negative rational values of x by means of the equation
expxexp(x) = 1;
and so we have
expx = (exp1)x = ex,
say, where
e = exp1 = 1 + 1 + 1 2! + 1 3! + ,
for all rational values of x. Finally we define ex, when x is irrational, as being equal to expx. The logarithm is then defined as the function inverse to expx or ex.

Example. Develop the theory of the binomial series

1 + m 1 x + m 2 x2 + = f(m,x),
where 1 < x < 1, in a similar manner, starting from the equation
f(m,x)f(m,x) = f(m + m,x)
(Ex. LXXXI. 6).

Miscellaneous Examples on Chapter IX107

1. Given that log10e = .4343 and that 210 and 321 are nearly equal to powers of 10, calculate log102 and log103 to four places of decimals.

(Math. Trip. 1905.)

2. Determine which of (1 2e)3 and (2)1 2π is the greater. [Take logarithms and observe that 3/(3 + 1 4π) < 2 53 < .6929 < log2.]

3. Show that log10n cannot be a rational number if n is any positive integer not a power of 10. [If n is not divisible by 10, and log10n = p/q, we have 10p = nq, which is impossible, since 10p ends with 0 and nq does not. If n = 10aN, where N is not divisible by 10, then log10N and therefore

log10n = a + log10N
cannot be rational.]

4. For what values of x are the functions logx, loglogx, logloglogx, … (a) equal to 0 (b) equal to 1 (c) not defined? Consider also the same question for the functions lx, llx, lllx, …, where lx = logx.

5. Show that

logx n 1log(x + 1) + n 2log(x + 2) + (1)n log(x + n)
is negative and increases steadily towards 0 as x increases from 0 towards .

[The derivative of the function is

0n(1)rn r 1 x + r = n! x(x + 1)(x + n),
as is easily seen by splitting up the right-hand side into partial fractions. This expression is positive, and the function itself tends to zero as x , since
log(x + r) = logx + εx,
where εx 0, and 1 n 1 + n 2 = 0.]

6. Prove that

d dxn logx x = (1)nn! xn+1 logx 1 1 2 1 n.
(Math. Trip. 1909.)

7. If x > 1 then x2 > (1 + x){log(1 + x)}2.

(Math. Trip. 1906.)

[Put 1 + x = eξ, and use the fact that sinhξ > ξ when ξ > 0.]

8. Show that {log(1 + x)}/x and x/{(1 + x)log(1 + x)}both decrease steadily as x increases from 0 towards .

9. Show that, as x increases from 1 towards , the function (1 + x)1/x assumes once and only once every value between 0 and 1.

(Math. Trip. 1910.)

10. Show that 1 log(1 + x) 1 x 1 2 as x 0.

11. Show that 1 log(1 + x) 1 x decreases steadily from 1 to 0 as x increases from 1 towards . [The function is undefined when x = 0, but if we attribute to it the value 1 2 when x = 0 it becomes continuous for x = 0. Use Ex. 7 to show that the derivative is negative.]

12. Show that the function (logξ logx)/(ξ x), where ξ is positive, decreases steadily as x increases from 0 to ξ, and find its limit as x ξ.

13. Show that ex > MxN, where M and N are large positive numbers, if x is greater than the greater of 2logM and 16N2.

[It is easy to prove that logx < 2x; and so the inequality given is certainly satisfied if

x > logM + 2Nx,
and therefore certainly satisfied if 1 2x > logM, 1 2x > 2Nx.]

14. If f(x) and φ(x) tend to infinity as x , and f(x)/φ(x) , then f(x)/φ(x) . [Use the result of Ch. VI, Misc. Ex. 33.] By taking f(x) = xα, φ(x) = logx, prove that (logx)/xα 0 for all positive values of α.

15. If p and q are positive integers then

1 pn + 1 + 1 pn + 2 + + 1 qn log q p
as n . [Cf. Ex. LXXVIII. 6.]

16. Prove that if x is positive then nlog{1 2(1 + x1/n)}1 2 logx as n . [We have

nlog{1 2(1 + x1/n)}= nlog{1 1 2(1 x1/n)}= 1 2n(1 x1/n) log(1u) u
where u = 1 2(1 x1/n). Now use §209 and Ex. LXXXII. 4.]

17. Prove that if a and b are positive then

{1 2(a1/n + b1/n)}n ab.

[Take logarithms and use Ex. 16.]

18. Show that

1 + 1 3 + 1 5 + + 1 2n 1 = 1 2 logn + log2 + 1 2γ + εn,
where γ is Euler’s constant (Ex. LXXXIX. 1) and εn 0 as n .

19. Show that

1 + 1 3 1 2 + 1 5 + 1 7 1 4 + 1 9 + = 3 2 log2,
the series being formed from the series 1 1 2 + 1 3 by taking alternately two positive terms and then one negative. [The sum of the first 3n terms is

1 + 1 3 + 1 5 + + 1 4n 1 1 2 1 + 1 2 + + 1 n = 1 2 log2n + log2 + 1 2γ + εn 1 2(logn + γ + ε),

where εn and εn tend to 0 as n . (Cf. Ex. LXXVIII. 6).]

20. Show that 1 1 2 1 4 + 1 3 1 6 1 8 + 1 5 1 10 = 1 2 log2.

21. Prove that

1n 1 ν(36ν2 1) = 3 + 3Σ3n+1 Σn Sn
where Sn = 1 + 1 2 + + 1 n, Σn = 1 + 1 3 + + 1 2n 1. Hence prove that the sum of the series when continued to infinity is
3 + 3 2 log3 + 2log2.
(Math. Trip. 1905.)

22. Show that

1 1 n(4n2 1) = 2log2 1, 1 1 n(9n2 1) = 3 2(log3 1).

23. Prove that the sums of the four series

1 1 4n2 1, 1(1)n1 4n2 1 , 1 1 (2n + 1)2 1, 1 (1)n1 (2n + 1)2 1
are 1 2, 1 4π 1 2, 1 4, 1 2 log2 1 4 respectively.

24. Prove that n!(a/n)n tends to 0 or to  according as a < e or a > e.

[If un = n!(a/n)n then un+1/un = a{1 + (1/n)}n a/e. It can be shown that the function tends to  when a = e: for a proof, which is rather beyond the scope of the theorems of this chapter, see Bromwich’s Infinite Series, pp. 461 et seq.]

25. Find the limit as x of

a0 + a1x + + arxr b0 + b1x + + brxr λ0+λ1x,
distinguishing the different cases which may arise.
(Math. Trip. 1886.)

26. Prove that

log 1 + x n(x > 0)
diverges to . [Compare with (x/n).] Deduce that if x is positive then
(1 + x)(2 + x)(n + x)/n!
as n . [The logarithm of the function is 1n log 1 + x ν.]

27. Prove that if x > 1 then

1 (x + 1)2 = 1 (x + 1)(x + 2) + 1! (x + 1)(x + 2)(x + 3) + 2! (x + 1)(x + 2)(x + 3)(x + 4) + .

(Math. Trip. 1908.)

[The difference between 1/(x + 1)2 and the sum of the first n terms of the series is

1 (x + 1)2 n! (x + 2)(x + 3)(x + n + 1).]

28. No equation of the type

Aeαx + Beβx + = 0,
where A, B, … are polynomials and α, β, … different real numbers, can hold for all values of x. [If α is the algebraically greatest of α, β, …, then the term Aeαx outweighs all the rest as x .]

29. Show that the sequence

a1 = e,a2 = ee2 ,a3 = eee3 ,
tends to infinity more rapidly than any member of the exponential scale.

[Let e1(x) = ex, e2(x) = ee1(x), and so on. Then, if ek(x) is any member of the exponential scale, an > ek(n) when n > k.]

30. Prove that

d dx{φ(x)}ψ(x) = d dx{φ(x)}α + d dx{βψ(x)}
where α is to be put equal to ψ(x) and β to φ(x) after differentiation. Establish a similar rule for the differentiation of φ(x)[{ψ(x)}χ(x)] .

31. Prove that if Dxnex2 = ex2 φn(x) then (i) φn(x) is a polynomial of degree n, (ii) φn+1 = 2xφn + φ, and (iii) all the roots of φn = 0 are real and distinct, and separated by those of φn1 = 0. [To prove (iii) assume the truth of the result for κ = 1, 2, …, n, and consider the signs of φn+1 for the n values of x for which φn = 0 and for large (positive or negative) values of x.]

32. The general solution of f(xy) = f(x)f(y), where f is a differentiable function, is xa, where a is a constant: and that of

f(x + y) + f(x y) = 2f(x)f(y)
is coshax or cosax, according as f(0) is positive or negative. [In proving the second result assume that f has derivatives of the first three orders. Then
2f(x) + y2{f(x) + ε y}= 2f(x)[f(0) + yf(0) + 1 2y2{f(0) + ε}],
where εy and εtend to zero with y. It follows that f(0) = 1, f(0) = 0, f(x) = f(0)f(x), so that a = f(0) or a = f(0).]

33. How do the functions xsin(1/x), xsin 2(1/x) , xcosec(1/x) behave as x +0?

34. Trace the curves y = tanxetan x, y = sinxlogtan 1 2x.

35. The equation ex = ax + b has one real root if a < 0 or a = 0, b > 0. If a > 0 then it has two real roots or none, according as aloga > b a or aloga < b a.

36. Show by graphical considerations that the equation ex = ax2 + 2bx + c has one, two, or three real roots if a > 0, none, one, or two if a < 0; and show how to distinguish between the different cases.

37. Trace the curve y = 1 xlog ex 1 x , showing that the point (0, 1 2) is a centre of symmetry, and that as x increases through all real values, y steadily increases from 0 to 1. Deduce that the equation

1 xlog ex 1 x = α
has no real root unless 0 < α < 1, and then one, whose sign is the same as that of α 1 2. [In the first place
y 1 2 = 1 x log ex1 x loge1 2x = 1 x log sinh 1 2x 1 2x
is clearly an odd function of x. Also
dy dx = 1 x2 1 2xcoth 1 2x 1 log sinh 1 2x 1 2x .
The function inside the large bracket tends to zero as x 0; and its derivative is
1 x 1 1 2x sinh 1 2x2 ,
which has the sign of x. Hence dy/dx > 0 for all values of x.]

38. Trace the curve y = e1/xx2 + 2x, and show that the equation

e1/xx2 + 2x = α
has no real roots if α is negative, one negative root if
0 < α < a = e1/22 + 22,
and two positive roots and one negative if α > a.

39. Show that the equation fn(x) = 1 + x + x2 2! + + xn n! = 0 has one real root if n is odd and none if n is even.

[Assume this proved for n = 1, 2, … 2k. Then f2k+1(x) = 0 has at least one real root, since its degree is odd, and it cannot have more since, if it had, f2k+1(x) or f2k(x) would have to vanish once at least. Hence f2k+1(x) = 0 has just one root, and so f2k+2(x) = 0 cannot have more than two. If it has two, say α and β, then f2k+2(x) or f2k+1(x) must vanish once at least between α and β, say at γ. And

f2k+2(γ) = f2k+1(γ) + γ2k+2 (2k + 2)! > 0.
But f2k+2(x) is also positive when x is large (positively or negatively), and a glance at a figure will show that these results are contradictory. Hence f2k+2(x) = 0 has no real roots.]

40. Prove that if a and b are positive and nearly equal then

log a b = 1 2(a b) 1 a + 1 b,
approximately, the error being about 1 6{(a b)/a}3. [Use the logarithmic series. This formula is interesting historically as having been employed by Napier for the numerical calculation of logarithms.]

41. Prove by multiplication of series that if 1 < x < 1 then

1 2{log(1 + x)}2 = 1 2x2 1 3(1 + 1 2)x3 + 1 4(1 + 1 2 + 1 3)x4 , 1 2(arctanx)2 = 1 2x2 1 4(1 + 1 3)x4 + 1 6(1 + 1 3 + 1 5)x6 .

42. Prove that

(1 + αx)1/x = eα{1 1 2a2x + 1 24(8 + 3a)a3x2(1 + εx)},
where εx 0 with x.

43. The first n + 2 terms in the expansion of log 1 + x + x2 2! + + xn n! in powers of x are

x xn+1 n! 1 n + 1 x 1!(n + 2) + x2 2!(n + 3) + (1)n xn n!(2n + 1).
(Math. Trip. 1899.)

44. Show that the expansion of

exp x x2 2 xn n
in powers of x begins with the terms
1 x + xn+1 n + 1 s=1n xn+s+1 (n + s)(n + s + 1).
(Math. Trip. 1909.)

45. Show that if 1 < x < 1 then

1 3x + 1 4 3 622x2 + 1 4 7 3 6 932x3 + = x(x + 3) 9(1 x)7/3, 1 3x + 1 4 3 623x2 + 1 4 7 3 6 933x3 + = x(x2 + 18x + 9) 27(1 x)10/3 .

[Use the method of Ex. XCII. 6. The results are more easily obtained by differentiation; but the problem of the differentiation of an infinite series is beyond our range.]

46. Prove that

0 dx (x + a)(x + b) = 1 a blog a b, 0 dx (x + a)(x + b)2 = 1 (a b)2b a b blog a b, 0 xdx (x + a)(x + b)2 = 1 (a b)2 alog a b a + b, 0 dx (x + a)(x2 + b2) = 1 (a2 + b2)b 1 2πa blog a b , 0 xdx (x + a)(x2 + b2) = 1 a2 + b2 1 2πb + alog a b ,

provided that a and b are positive. Deduce, and verify independently, that each of the functions

a 1 loga,aloga a + 1,1 2πa loga,1 2π + aloga
is positive for all positive values of a.

47. Prove that if αβγ are all positive, and β2 > αγ, then

0 dx αx2 + 2βx + γ = 1 β2 αγlog β + β2 αγ αγ ;
while if α is positive and αγ > β2 the value of the integral is
1 αγ β2 arctan αγ β2 β , that value of the inverse tangent being chosen which lies between 0 and π. Are there any other really different cases in which the integral is convergent?

48. Prove that if a > 1 then

1 dx (x + a)x2 1 =0 dt cosht + a = 21 du u2 + 2au + 1;
and deduce that the value of the integral is
2 1 a2 arctan1 a 1 + a if 1 < a < 1, and
1 a2 1log a + 1 + a 1 a + 1 a 1 = 2 a2 1argtanha 1 a + 1 if a > 1. Discuss the case in which a = 1.

49. Transform the integral 0 dx (x + a)x2 + 1, where a > 0, in the same ways, showing that its value is

1 a2 + 1log a + 1 + a2 + 1 a + 1 a2 + 1 = 2 a2 + 1argtanh a2 + 1 a + 1 .

50. Prove that

01 arctanxdx = 1 4π 1 2 log2.

51. If 0 < α < 1, 0 < β < 1, then

11 dx (1 2αx + α2)(1 2βx + β2) = 1 αβlog 1 + αβ 1 αβ.

52. Prove that if a > b > 0 then

dθ acoshθ + bsinhθ = π a2 b2.

53. Prove that

01 logx 1 + x2dx = 1 logx 1 + x2dx,0 logx 1 + x2dx = 0,
and deduce that if a > 0 then
0 logx a2 + x2dx = π 2aloga.

[Use the substitutions x = 1/t and x = au.]

54. Prove that

0log 1 + a2 x2 dx = πa
if a > 0. [Integrate by parts.]