Professional Documents
Culture Documents
In this part of the course, we will be concerned with how one can formalise the idea of summing an innite list of numbers a1 + a2 + a3 + . . . .
As you would expect, this will once again involve the notion of a limit.
We begin with a basic denition: Denition Let (an) n=1 be a sequence. For each n 1, let sn = a1 + a2 + . . . + a n =
n k=1
ak .
an ,
is, formally, the ordered pair ((an), (sn)). an is called the nth term of the series and sn is called the nth partial sum of the series.
Using this ordered pair notation to denote a series would be cumbersome and this is why we denote a series by the notation an. But
remember, that although this looks like an arithmetic expression, it is simply a notation which we use to denote an ordered pair of sequences.
Study of series is basically equivalent to the study of sequences. If we have an innite series of real numbers a1 + a2 + a3 + . . . , then we can study it via the sequence (sn), where sn = a1 + a2 + . . . + an .
Note that, as far as we are concerned, a series involves an innite list of numbers. We do not discuss nite series, since there are no convergence issues there.
Example
(1)n.
The nth term of this sequence is (1)n and the nth partial sum sn is 1 if n is odd and 0 if n is even.
Example
(1)2n.
The nth term of this sequence is 1 and the nth partial sum is n.
Note that the sequence (an) for the second series is actually a subsequence of that for the rst - but that the partial sums bear little resemblance.
Consider
1 2
n1
1 1 1 sn = 1 + + + + 2 2 2 1 (1/2)n = 1 (1 /2) n 1 = 2 1 . 2
n1
Convergence of series
Denition Let
an be a series. If the
sequence (sn) of partial sums converges to L (nite), then we say that the series converges to L, or has sum L. If (sn) diverges, then we say that an diverges.
If a series
an converges to L, we write
n=1
an = L.
In other words, convergence of a1 + a2 + a3 + . . . is dened as convergence of the sequence (sn) of partial sums. Conversely, given any sequence (sn) of numbers, we can regard it as the sequence of partial sums of a series dened as follows:
ar ,
r=1
ar .
In this usage, there is not yet any assumption that the series in question converges.
(b) If the series converges, we use to denote the number which is the sum of the series, i.e. the limit of partial sums.
1.
2.
Let us prove these. For simplicity, assume that a = 1. First prove that |b| < 1.
1 if bn1 converges to 1 b
1 . lim s = n n 1b
Now use algebra of limits or heredity properties to deduce that, for general value of a, for |b| < 1, a . lim s = n n 1b
On the other hand, if |b| > 1, then |bn| diverges as n , and so sn also diverges. If b = 1, then sn = 1 + 1 + . . . + 1 = n, so sn diverges. If b = 1, then s2n1 = 1 and s2n = 0 so sn diverges.
Theorem If
an converges, then an 0 as n .
Proof: If
If it did, then we would have to have br 0 as r . Hence |b|r 0 as r . But |b|r 1 for all r.
Converse is false: an tending to 0 does not necessarily mean that the series converges.
1 diverges, though n
an
For example,
Finding sucient conditions for a series to converge is the main aim in what follows, and its not easy. Life would be simple if it were the case that an converges if and only if an 0, but it isnt so.
BUT just because an tends to 0, it does not mean the series an converges!!!
Is that clear??!!
Series
1 ns 1/n
1 + + + > + +
1 1 1 +( + ) 2 3 4 1 1 1 1 ( + + + ) + ... + 5 6 7 8 1 1 1 + m1 + . . . + m) ( m1 2 +1 2 +2 2 1 1 1 1+ +( + ) 2 4 4 1 1 1 1 ( + + + ) + ... + 8 8 8 8 1 1 1 ( m + m + . . . + m) 2 2 2
1 1 1 1 = 1 + + + + ... + 2 2 2 2 m = 1+ . 2
So sn is unbounded.
For s < 1, note that, for each n, 1 1 . ns n Then 1 tn, sn = s r=1 r where tn =
n 1 r=1 r n
But we have already shown that tn is unbounded. Hence sn is unbounded also, and so must be divergent.
What about s > 1? Consider the sum of the rst 2m 1 terms. 1 + + + + 1 1 ( s + s) 2 3 1 1 1 1 ( s + s + s + s) + . . . + 4 5 6 7 1 1 1 + ... + m ) ( (m1)s + m1 s s (2 + 1) (2 1) 2 1 1 1 + (2 s ) + (4 s ) + . . . + 2 4 (2m1 2(m1)s).
Thus we have s2m1 1 + 2(s1) + 22(s1) + . . . + 2(m1)(s1) = 1 + b + b2 + . . . + b(m1), where b = 2(s1) satises |b| < 1.
Thus sn is increasing in n, and so the above implies that sn (1 b)1 for all n.
There are also exist some Algebra of Limits results which can be proved directly from the same results for sequences:
an and
bn converge,
a = L and n=1 n
b = M. n=1 n
Then, for any real number c, the series (an + bn) and can converge, and
ca = cL. n=1 n (a + b ) = L + M and n n=1 n
But . . . note that the same does not hold for products. For example, if n an = (1) / n, then, as we will shortly see, an converges. However, harmonic series. (a n a n )
1 , the n
This is not all that surprising: we cannot express the partial sums of the product series in terms of the partial sums of the original series in a simple way.
Then (CHECK!) un = = = =
n r=1 n r=1 n r=1 n r=1
= ?
Example
1 converges. r(r+1)
Proof:
1 1 . Then, as n , = Trick: r(r1 r r+1 +1)
Now
n 1 3 2 1 3 3 ) = +3 + ( r+2 r 2 n+1 r=1 r + 1 r=3 r n 1 2 2 2 n+1 n+2 r=3 r n 1 1 1 2 r=3 r 3 3 2 = + 2 n+1 n+1 2 1 1 n+2 2 2 2 3 = + . n+1 n+1 n+2 n
Note that, whether or not a series converges does not depend where we start summing it.
A series is nonnegative if all its terms are nonnegative, that is ar 0 for all r. Later we look at series which have some negative terms, but its easiest at the moment to stick to nonnegative series.
The reason why this is easy is that the sequence (sn) of partial sums is increasing. The aim now is to develop a range of tests for convergence.
Proposition. The sequence (sn) of partial sums converges if and only if it is bounded above. If it is not bounded above, then sn as n .
Proof. Since ar 0 for all r, (sn) is increasing. If it is bounded, then convergence follows from the fact that a bounded monotone sequence converges.
On the other hand, if (sn) does converge, then it must be bounded, as any convergent sequence is bounded. If (sn) is unbounded, then it follows from the fact that it is increasing that it must diverge to innity. ar (ar 0) either converges or diverges
So
to innity.
Example.
1 converges. r=0 r!
But 1 1 sn 1 + 1 + + 2 22 1 1 + . . . + n2 + 222 2 1 + 2 3.
1 + . . ..) (Think about 1 + 1 + 2 4
We dened a series
We said that such a series is convergent if the sequence sn of its partial sums converges to a nite limit L, i.e. limn sn = L, where sn = n r=1 ar . If ar converges to L, we write ar = L
We have seen examples of convergent and divergent series, and proved various results about divergent and convergent series.
We then said we would now for some time restrict attention to series with nonnegative terms, i.e. ar 0 for r = 1, 2, . . .. Our aim is to develop some tests for convergence of such series.
Comparison Test
Theorem [Comparison Test] Let (an), (bn) be nonnegative sequences such that an bn for all n 1. Then
1. If
bn converges, then
b . n=1 n
a n=1 n
2. If
an diverges, then
bn diverges.
Now, if (tn) converges, then it is bounded above, and so (sn) is bounded above. Since (sn) is increasing, it must converge. Since sn tn for all n, it is clear that limn sn limn tn.
ar diverges.
Then (sn) is unbounded above (if it were bounded then it would converge, as it is increasing).
This implies that (tn) is unbounded, since sn tn, and so (tn) cannot be convergent (it is also increasing).
It follows that
br diverges.
When using the Comparison Test, its important to use it in the right direction.
If you want to use CT to show converges, you need to nd 0 an bn for all n. If you wanted to use it to show with cn dn 0.
an
bn that you
cn dn
1 1 for each n, Proof: note that 0 n ns since ns n. Now apply Statement 2 in the
The Comparison Test can be weakened slightly as follows. (Here, what weve done is replace for all n with for all suciently large n.)
Theorem Let (an), (bn) be nonnegative sequences such that there is some constant C > 0 and some non-negative integer N such that an Cbn for all n N . Then
1. If
bn converges, then
an does also.
2. If
an diverges, then
bn diverges.
Proof: The trick is to remember that the convergence or divergence of a series does not depend on the values of nitely many of its terms.
(Although, if it does converge, then the value of its sum will be inuenced by the value of each and every individual term.)
With this in mind, dene an = aN +n1, and bn = bN +n1 for n = 1, 2, . . .. Then a1 = aN , a2 = aN +1, a3 = aN +2, and so on. And b1 = bN , b2 = bN +1, b3 = bN +2, and so on.
We have assumed that an Cbn for all n N . Therefore, by denition above, 0 an C bn for n = 1, 2, . . .. First let us prove Statement 1: if converges, then bn an
To do so, we will use Statement 1 in the (strong) Comparison Test. (C bn) an does.
an converges as long as
(C bn) converge if
bn
(C bn)
bn converges.
CHECK! (This follows easily from the denition of convergence of series, together with algebra of limits results for sequences.)
bn
br =
n+N 1
br
r=1
N 1 r=1
br = tn+N 1
N 1
br ,
r=1
where tn = n r=1 br . Then tn converges if and only if tn converges. (Though the respective limits will not in general be equal.)
bn (C bn)
converges, and so it follows from the above that bn converges, and hence that
converges.
ar =
r=1
ar
N 1 r=1
N 1
ar ,
r=1
This completes the proof of Statement 1 in the weak Comparison Test. Note that the weak test does not state that
a n=1 n b . n=1 n
The proof of Statement 2 is very similar, using Statement 2 in the strong Comparison Test. bn,
an and
Since
an diverges, also
an diverges.
(CHECK!)
bn diverges.
n2 + 1 Example Consider . The n th n5 + n + 1 term here behaves like 1/n3, because the dominant term on the numerator is n2 and the dominant term in the denominator is n5. But this needs to be made precise.
We can formally compare the series with 1/n3 by noting that n2 + n2 2 n2 + 1 = 3. 5 5 n +n+1 n n The series 2/n3 converges because 1/n3
does (this being a standard result from above, using Algebra of Limits results for sequences). Hence, by the CT, the given series converges also.
The following, more sophisticated, version of the Comparison Test, is sometimes more useful. We could call it the Limiting Comparison Test, but well just call it the Comparison Test (since it subsumes the previous versions).
Theorem [Comparison Test] Suppose (an), (bn) are positive and that an/bn L, where L = 0 (and L is nite) as n . Then an , bn either both converge or both diverge: that is, they have the same behaviour, with respect to convergence.
Proof: Suppose an 0, bn 0 for all n, and an/bn L as n where 0 < L < . By denition of limit, for every n N, an | L| < . bn Apply this with = L/2. > 0, there
Thus there exists an integer N such that, for all n N , an | L| < L/2. bn That is, for n N , we have an L 3L < L+ = bn 2 2 an L L > L = . bn 2 2
Suppose
the weak form of the Comparison Test (with C = 3L/2, 0 < C < ), also.
Suppose
the weak form of the Comparison Test (with C = 2/L, 0 < C < ), also.
n2 + 1 Example Consider again . n5 + n + 1 Using the limiting form of the Comparison Test, to compare the series with simply observe that, since n5 + n3 (n2 + 1)/(n5 + n + 1) = 5 3 1/n n +n+1 1 + n2 = 1 + n4 + n5 1 = 0, and since 1/n3 converges, then the given 1/n3, we
Example Consider
3 ( n 1) . 8 n +n+2
We have (n 1)3 n8 + n + 2 = = = n3(1 1/n)3 n8(1 + n7 + 2n8) n3(1 1/n)3 n4 1 + n7 + 2n8 (1 1/n)3 n 1 + n7 + 2n8
Thus we have 1 an = , 16n n 1 + n7 + 2n8 for all n 2. To see this, note that (1 1/n)3 1/8 for all n 2. Also, 1 + n7 + 2n8 4 = 2. (1 1/n)3
1 diverges, so n
Why do we bother stating all these dierent forms of Comparison Test though??
Well, because life will not always be so nice as to have an bn for all n, as in the strong version of the CT.
Also, there will be cases when the ratio an/bn does not converge, as in the limiting version of the CT, but we still may be able to use one of the other versions.
Ratio Test Theorem [Ratio Test] Let an be a nonnegative series such that an+1 (L = allowed). L = nlim an Then
1. L < 1
an converges.
2. L > 1 L = ).
Note that the Ratio Test says nothing if L = 1: in this case, the test is useless. Consider the series 1/n and 1/n2. In
both cases, an+1/an 1, yet the rst series is divergent and the second convergent.
Proof Let us prove (1). So suppose that L < 1. Evidently, we may choose an M such that L < M < 1. Take in the denition of limit to be
M L > 0.
Thus there exists N such that for n N an+1 | L| < M L. an Hence there exists N such that an+1 < L + (M L) = M. nN an In particular, aN +1 < M aN .
converges (since 0 < M < 1), we have by the Comparison Test that converges, and hence a N +n an converges.
Let us now prove (2). Assume then that the limit L > 1. As in part (1), there exists a positive integer N such that an+1/an > 1 for n N . In particular, aN > 0; otherwise, this is meaningless.
n7 . Letting 6n
br , where b > 0.
br converges.
br diveregs.
Warning
BEWARE: an+1/an may not converge at all, in which case the Ratio Test does not apply!!
L = ). Then,
1. L < 1
an converges.
2. L > 1
case L = ).
n7 . Here, 6n
n7/n (n1/n)7 = = . 6 6
Again, note that the Root Test says nothing about the case L = 1.
L < 1. Then, given L < M < 1, by the denition of limit with = M l > 0, there exists an
1/n integer N such that an M for all n N .
an 0. Hence no convergence.
Integral Test
The following test draws on the interpretation of an integral as a Riemann sum (see your calculus courses).
Theorem [Integral Test] Let g be a positive, decreasing, integrable (for example, continuous) function on [1, ).
n g (x) dx. Then the series Let G(n) = 1
g (n) converges if and only if the sequence (G(n)) converges. In other words,
g (x) dx exists. integral 1
g (n)
Take a positive integer r. Since g is decreasing, for any t [r, r + 1], g (r) g (t) g (r + 1). The function g is integrable so, integrating the above over [r, r + 1], we get
r+1 g (r)dt r r+1 r+1 g (t)dt r g (r + 1)dt r r+1 g ( t ) dt g ( r + 1) . r
or g (r )
g (r )
g (r) G(n + 1)
r=2
g (r )
As we have seen before, limn G(n) exists i (G(n)) is bounded above. Suppose that limn G(n) exists, so G(n) is bounded above. Then
n+1 r=2 g (r ) G(n + 1)
Since
n r=1 g (r ), that is
g (r) is convergent.
Conversely, if
Theorem [Integral Test] Let g be a positive, decreasing, integrable (for example, continuous) function on [a, ),
n g (x) dx. and suppose a 1. Let G(n) = a
if the sequence (G(n)) converges. In other words, g (n) converges if and only if the
g (x) dx exists. improper integral a
To see whether it converges, we can use the integral test. Let g (n) = 1/(n log n). Then 1 log n 1 n n dx = du, g ( x ) dx = log 2 u 2 2 x log x where we have made the substitution u = log x. So
n G(n) = [log u]log log 2 = log log n log log 2.
Alternating series
A series is alternating if its terms are alternately positive and negative. Such a series takes the form (1)n+1cn, where cn 0.
Theorem [Leibniz Alternating Series Test/LAST] Suppose that an = (1)n+1cn is an alternating series, where cn 0. Then, if (cn) is a decreasing sequence and limn cn = 0, the series converges. an
Corollary For any positive number s, (1)n+1 converges. ns Proof: Note that cn = 1/ns is decreasing and limn 1/ns = 0. Now apply LAST.
Proof of LAST. Consider s2n = (c1 c2) + (c3 c4) + . . . + (c2n1 c2n). The bracketing shows that (s2n) is an increasing subsequence of (sn), since each of the bracketed terms is nonnegative.
Also, we can write s2n = c1 (c2 c3) (c4 c5) . . . (c2n2 c2n1) c2n c1 .
But also we have s2n+1 = s2n + c2n+1. By assumption, c2n+1 0. Hence, by Algebra of Limits, s2n+1 s as n . The result follows.
Note: LAST says that if the sequence (cn) is decreasing and tends to 0, then the series (1)n+1cn converges. It says nothing at all if one of these two conditions fails to hold.
This does not mean that these two conditions are necessary for convergence of alternating series: it just means the Leibniz test doesnt work in those situations.
Absolute convergence
Denition Let
some of the terms may be negative). If |an| converges, we say that absolutely. If diverges, then conditionally.
an converges but
an is said to converge
Note that if
Theorem If a series
an is absolutely
Since
a+ n and
We have proved that the geometric series ar n converges absolutely if |r| < 1. It does not converge either absolutely or conditionally when |r| 1.
(1)n The fact (from Leibniz test) that n 1 converges and the earlier fact that n (1)n diverges, show that the series n converges conditionally.
Theorem [Comparison Test] Let (an), (bn) be sequences such that |an| |bn| for all n. Then
1. If
an
2. If
|bn| diverges.
Proof. Apply the Comparison Test to the non-negative series |an| and |bn|.
Theorem [Ratio Test] Let such that |an+1| L = nlim |an| Then
an be a series
(L = allowed).
1. L < 1
an converges absolutely
2. L > 1
an diverges.
Proof. To prove the rst part, apply the Ratio Test to the non-negative series |a n |.
To prove the second part, note the proof of the Ratio Test for nonnegative series shows that, if L > 1, then |an| 0. It follows that an 0, so the necessary condition of convergence of diverges. an fails, and so an
an be a series,
1. L < 1
an converges absolutely.
2. L > 1
case L = ).
Proof. To prove the rst part, apply the Root Test to the non-negative series |a n |.
To prove the second part, note the proof of the Root Test for nonnegative series shows that, if L > 1, then |an| 0. It follows that an 0, so the necessary condition of convergence of diverges. an fails, and so an
Power Series
For our purposes, a power series is a series of the form anxn, where x is a real xn/n!, used to
It turns out that, for any real number x, this series converges, and we may dene the exponential function by exp(x) =
xn n=0 n!
xn n=0 n!
. Its
easy to show that this converges absolutely for all x. We simply observe that xn+1/(n + 1)! |xn/n!| |x| = 0, n+1
for any x, and so, by the Ratio Test, absolute convergence follows.
Example Lets determine exactly those values of x for which the series convergent. xn/n is
The ratio test therefore tells us that the series converges absolutely if |x| < 1, and that it diverges if |x| > 1. But what if |x| = 1?
x = 1: here, ratio test is useless and we have to be more sophisticated. Well, |x| = 1 corresponds to two cases: x = 1 and x = 1. Treat each separately.
(1)n . n
So we have now determined exactly the values of x where the series converges: it converges for 1 x < 1 and diverges for all other values of x.
Theorem There is R such that the series anxn converges absolutely for all x (R, R), and diverges for all x with |X | > R. (This includes the case R = ).
In the case in which R is nite, what happens at R is not determined by this theorem, and has to be considered separately. The name radius of convergence is given to R. It is possible to have R = 0.
a n xn 0 converges then
Proof. We have that anxn 0 0 as n . Therefore, there exists N such that | a n xn 0 | 1 for n N . Let |x| < |x0|, and note that x n n n | a n x | = | a n x0 | | | x0 x n | x0 |
for all n N.
x |n is a convergent |x
0
anxn is