Indian Institute of Technology Kanpur

Department of Mathematics and Statistics

Analysis I (MTH 301)
Exercise Sheet 4
1. Limit point
Throughout this section, we assume that pX, dq is a metric space and A Ň X.

1.1. Show that the following are equivalent for a P X:

(L.1) a is a limit point of A.
(L.2) For every r ą 0, Bpa; rq X pAztauq is infinite.
(L.3) There exists a sequence tan u8
n“1 in Aztau such that an ÝÝÝÑ a.
nÑ8
(L.4) There exists a sequence tan u8 8
n“1 in Aztau such that tdpan , aqun“1 decreases to 0 strictly.

Remark. In view of (L.2), it is clear that finite sets cannot have limit points.

1.2.˚ Show that, the following are equivalent:

(C.1) All limit points of A are contained in A.
(C.2) @x P XzA, Dr ą 0 such that Bpx; rq X A “ H.
A is said to be closed (in X) if it satisfies (C.1). Some basic properties of closed sets are mentioned
in 1.5.
def
1.3. Show that, for any x P X and r ą 0, Brx; rs “ ty P X : dpx, yq ď ru is a closed subset of
def
X. What about Srx; rs “ ty P X : dpx, yq “ ru?

1.4. Show that the following subsets of R are closed (in R):
(a) ra, bs, where a ă b P R.
(b) p´8, αs, where α P R.
(c) rα, `8q, where α P R.
(d) Any finite subset.

1.5.˚ (a) H and X are closed. č

(b) Show that, if tAi uiPI is an arbitrary family of closed subsets of X, then Ai is closed.
iPI
(c) If n P N and A1 , . . . , An are closed subsets of X then A1 Y ¨ ¨ ¨ Y An is closed.
(d) Show that the conclusion of 1.5.c for countably infinite family of closed sets.

1.6.˚ We say that A is perfect if it is closed, and every point of A is a limit point of A. Show
that the Cantor’s middle-third set is perfect.
Hint. 1.16 of Exercise Sheet 2 may be useful.

1.7.˚ Show that, if A Ň R is perfect then A is uncountable.

Hint. Assume contrary and let A “ tx1 , x2 , . . . u. Choose a closed and bounded interval
ra1 , b1 s such that pa1 , b1 q X A ‰ H and x1 R ra1 , b1 s. Then choose ra2 , b2 s Ň ra1 , b1 s such
that pa2 , b2 q X A ‰ H and x2 R ra2 , b2 s. Continue this process to obtain a point in A which
different from all x1n s.
2 Exercise Sheet 4 (MTH 301)

1.7. provides another argument for the uncountablity of K.

1.8.˚ Assume that A is countable. Show that the set of all limit points of A has cardinality at
most |R|.
Hint. Use the equivalence given by 1.1..

a P A is said to be an isolated point of A if it is not a limit point of A.

1.9.˚ Show that if X is infinite then it admits an infinite subset all of whose points are isolated
points.
2. Limit of a function
In 2.1. and 2.2., we let X, Y be metric spaces, A Ň X and f : A ÝÑ Y . Assume that a P X is
a limit point of A.

2.1. Show that f can have at most one limit at a.

2.2.˚ Show that, if f satisfies Cauchy criteria at a then lim f pxq exists, provided Y is complete.
xÑa

2.3. Let pX, dq be a complete metric space, A Ň X and ρ be the metric on A induced by d.
Show that pA, ρq is complete if and only if A is a closed in X.
2.4. shows that long term behaviours like convergence/Cauchy of a sequence may completely
change with the choice of the metric.
def
2.4. (a) For n P N, consider fn pxq “ xn , @x P r0, 1s. Show that tfn u8 n“1 does not converge
with respect to the metric induced by the } ¨ }8 norm, but it does converge if } ¨ }8 is
replaced by } ¨ }1 .
(b) Let txn u8
n“1 be asequence of integers. Fix a prime p. Find a necessary and sufficient
condition for xn ÝÝÝÑ 0 with respet to the metric dp (defined in Exercise Sheet 3).
nÑ8 ˇ ˇ
ˇ1 1ˇ
(c) Let ρ be the metric on p0, 1s defined as ρpx, yq “ ˇ x ´ y ˇ. Show that txn u8
n“1 Ň p0, 1s
is Cauchy with respect to ρ if and only if it is convergent with respect to the usual
metric and lim xn ‰ 0. Conclude that p0, 1s is complete with respect to ρ.
nÑ8

2.5. Let pX, dq be a metric space, txn u8n“1 be a sequence in X and ℓ P X. Assume that
8
every subsequence of txn un“1 admits a further subsequence that converges to ℓ. Show that
xn ÝÝÝÑ ℓ.
nÑ8

2.6. Let pX, dq be a metric space and txn u8 8

n“1 be a seqeunce in X. Show that, if txn un“1 is
Cauchy then dpxn , xn`1 q ÝÝÝÑ 0, but the converse is not true. Does the converse hold if d
nÑ8
is an ultrametric?

2.7. Let pYi , di q be a metric space for i “ 1, . . . , n. For any 1 ď p ď `8, define
$ 1
&pd1 px1 , y1 qp ` ¨ ¨ ¨ ` dn pxn , yn qp q p if p ă `8
def
dp px, yq “
% sup di pxi , yi q if p “ `8,
1ďiďn

where x “ px1 , . . . , xn q and y “ py1 , . . . , yn q P Y1 ˆ ¨ ¨ ¨ ˆ Yn . Suppose that pX, dq is a metric

space, A Ň X and a is a limit point of X. For f : A ÝÑ Y1 ˆ ¨ ¨ ¨ ˆ Yn , denote by fi the
Exercise Sheet 4 (MTH 301) 3

i-th coordinate of f , for all i “ 1, . . . , n. Show that f has a limit at a if and only if each fi
has a limit at a, and furthermore in that case, one has
´ ¯
lim f pxq “ lim f1 pxq, . . . , lim fn pxq .
xÑa xÑa xÑa

2.8. and 2.9. deals with the convergence of sequence of matrices. We let d P N and K “ R or C.
Assume that Md pKq is equipped with the sup norm, i.e.,
def
}A} “ sup |aij |,
1ďi,jďn

where A “ paij qi,j .

2.8. Let tAn u8 8

n“1 and tBn un“1 be sequences in Md pKq. Show the following:
(a) If An ÝÝÝÑ A and Bn ÝÝÝÑ B, then An ` Bn ÝÝÝÑ A ` B and An Bn ÝÝÝÑ AB.
nÑ8 nÑ8 nÑ8 nÑ8
(b) If An ÝÝÝÑ A, then P An Q ÝÝÝÑ P AQ, for all P, Q P Md pKq.
nÑ8 nÑ8
(c) If An ÝÝÝÑ A, then det An ÝÝÝÑ det A.
nÑ8 nÑ8

2.9.˚ Show that, for every A P Md pKq, there exists a sequence tAn u8
n“1 of invertible matrices
such that An ÝÝÝÑ A.
nÑ8

Hint. Row and column reductions may be useful.

2.10.˚ For any n P N, let fn : r´1, 1s ÝÑ R be defined by

$
&0
’ if ´ 1 ď x ď 0
def
fn pxq “ nx if ´ 0 ď x ď n1
’
%1 if n1 ď x ď 1.
(a) Show that tfn u8
n“1 is Cauchy in pCr´1, 1s, } ¨ }1 q.
(b) Show that tfn u8
n“1 is not convegrent in pCr´1, 1s, } ¨ }1 q.
ż0
Hint. If fn ÝÝÝÑ f in pCr´1, 1s, } ¨ }1 q then first show that |f | “ 0. Next show
nÑ8 ´1
ż1
that |f ´ 1| “ 0, for all N P N.
1
N

3. Upper and lower limits of a function

Let X be a metric spaces, A Ň X and a P X is a limit point of A. For f : A ÝÑ R, define
˜ ¸
def
lim sup f pxq “ inf sup f pxq ,
xÑa δą0 xPBpa;δqXpAztauq

where the supremum and infimum has been taken in R.

˜ ¸ ¨ ˛

3.1. Show that lim sup f pxq “ lim sup f pxq “ lim ˝ sup f pxq‚.
xÑa δÑ0` xPBpa;δqXpAztauq nÑ`8 1
xPB pa; n qXpAztauq

Remark. The above generalizes the concept of upper limit of a real sequence.
4 Exercise Sheet 4 (MTH 301)

3.2.˚ (a) Show that lim sup f pxq “ `8 if and only if f is unbounded above in any Bpa; δq.
xÑa
(b) Show that lim sup f pxq “ ´8 if and only if lim f pxq “ ´8.
xÑa xÑa

3.3.˚ Assume now that lim sup f pxq exists in R. Denote lim sup f pxq by ℓ. Show the following:
xÑa xÑa
(LS.1) p@ε ą 0qpDδ ą 0 s.t. @x P Aztau X Bpa; δq, f pxq ă ℓ ` εq.
(LS.2) p@ε ą 0qp@δ ą 0 Dx P Aztau X Bpa; δq s.t. f pxq ą ℓ ´ εq.

3.4.˚ Suppose that ℓ P R is such that both (LS.1) and (LS.2) hold for every ε ą 0. Then show
that ℓ “ lim sup f pxq.
xÑa

3.5.˚ From 3.3. and 3.4. conclude that, if lim sup f pxq exists in R, then it must be the unique
xÑa
number for which one has both (LS.1) and (LS.2).

Remark. 3.5. provides an equivalent characterization of lim sup f pxq, provided it exists in R.
xÑa
Finally, to include the cases lim sup f pxq “ ˘8, we adopt the following conventions:
xÑa

(i) For ℓ “ `8, we consider (LS.1) null and void, and interpret (LS.2) as follows:
p@M P Rqp@δ ą 0 Dx P Aztau X Bpa; δq s.t. f pxq ą M q.
(ii) For ℓ “ ´8, (LS.2) is to be considered null and void, and (LS.1) is to be interpreted as
follows:
p@M P RqpDδ ą 0 s.t. @x P Aztau X Bpa; δq, f pxq ă M q.
Now combining 3.2.-3.5., one obtains that, for ℓ P R, ℓ “ lim sup f pxq ðñ (LS.1) and (LS.2) hold
xÑa
together.
3.6. Define and prove the analogues of 3.1.-3.5. for lim inf f pxq.
xÑa

3.7.˚ Let L be the set of all ℓ P R such that lim f pxn q “ ℓ, for some sequence txn u8
n“1 in Aztau
nÑ`8
with lim xn “ a. Show that lim sup f pxq “ max L and lim inf f pxq “ min L.
nÑ`8 xÑa xÑa

3.8.˚ (a) Show that lim inf f pxq ď lim sup f pxq.
xÑa xÑa
(b) Show that lim inf f pxq “ lim sup f pxq if and only if lim f pxq exists, and in that case
xÑa xÑa xÑa
one has,
lim inf f pxq “ lim sup f pxq “ lim f pxq.
xÑa xÑa xÑa

3.9.˚ Using 3.8.b, show that lim f pxq exists if f satisfies the Cauchy criterion at a.
xÑa

The rest of this section deals with some basic properties of lim sup and lim inf. It is customary to
make the following conventions first:
def def x x
(i) For x P R, x ` 8 “ ` 8, x ´ 8 “ ´ 8 and `8
“ ´8
“ 0.
def def
(ii) For x ą 0, x ¨ p`8q “ ` 8 and x ¨ p´8q “ ´ 8.
def def
(iii) For x ă 0, x ¨ p`8q “ ´ 8 and x ¨ p´8q “ ` 8.
We continue with the X, A and a as above. Let f, g : A ÝÑ R.
Exercise Sheet 4 (MTH 301) 5

3.10. Show that, excluding the indeterminate forms `8 ´ 8 and ´8 ` 8, the following inequal-
ities hold:
lim inf f pxq ` lim inf gpxq ď lim inf pf pxq ` gpxqq
xÑa xÑa xÑa
ď lim inf f pxq ` lim sup gpxq
xÑa xÑa
ď lim suppf pxq ` gpxqq ď lim sup f pxq ` lim sup gpxq. (3.1)
xÑa xÑa xÑa

3.11. Assume that, f pxq, gpxq ě 0, for all x P A. Formulate and prove the analogues of (3.1) for
the product f g (exclude the indeterminate forms 0 ¨ p`8q and p`8q ¨ 0).

3.12. In each of the cases appearing in 3.10. and 3.11., provide an example in which “ă” occurs.

Remark. It is important to realize that some simple properties that hold for limit of functions do
not hold for upper and lower limits.

3.13. (a) Show that, if lim f pxq exists finitely, then equality occurs in the first and last inequality
xÑa
of (3.1).
(b) What will be the analogues of 3.13.a for product of two functions?

3.14. (a) Show that lim inf p´f pxqq “ ´ lim sup f pxq and lim supp´f pxqq “ ´ lim inf f pxq.
xÑa xÑa xÑa xÑa
(b) Assume that there exists r ą 0 such that f ą 0 on Bpa; rq X pAztauq. Then show that
1 1 1 1
lim sup “ and lim inf “ .
xÑa f pxq lim inf f pxq xÑa f pxq lim sup f pxq
xÑa xÑa

1
Here 0
is interpreted as `8.

3.15. (a) Let f be as in 3.14.b. Assume that

ˆ ˙ˆ ˙
1
lim sup f pxq lim sup “ 1.
xÑa xÑa f pxq
Show that lim f pxq exists finitely.
xÑa
(b) If for any g : A ÝÑ R, either lim suppf pxq ` gpxqq “ lim sup f pxq ` lim sup gpxq or
xÑa xÑa xÑa
lim inf pf pxq ` gpxqq “ lim inf f pxq ` lim inf gpxq, then show that lim f pxq exists.
xÑa xÑa xÑa xÑa
(c) What will be the analogue of 3.15.b for product?

3.16. Show that, for any sequence tan u8

n“1 of positive reals,
an`1 1 1 an`1
lim inf ď lim inf ann ď lim sup ann ď lim sup .
nÑ`8 an nÑ`8 nÑ`8 nÑ`8 an

def x1 ` ¨ ¨ ¨ ` xn
3.17. Let txn u8
n“ be a real sequence. Define yn “ , for all n P N. Show that
n
lim inf xn ď lim inf yn ď lim sup yn ď lim inf xn .
nÑ`8 nÑ`8 nÑ`8 nÑ`8

Give an example for which all of the above limits are finite and inequalities are strict.
6 Exercise Sheet 4 (MTH 301)

Recall that, a subsequenctial limit of a sequence txn u8

n“1 in a metric space X is a point ℓ P X
8
such that xnk ÝÝÝÝÑ ℓ, for some subsequence txnk uk“1 . In many applications, 3.18. yields the
kÑ`8
subsequential limits of the sequence.
k
ď
3.18.˚ Suppose that A1 , . . . , Ak are pairwise disjoint infinite subsets of N such that N “ Ai .
i“1
For each i “ 1, . . . , k, denote by Li the set of all subsequential limit of the subsequence
ďk
given by Ai . Show that Li contains precisely all subsequential limits of txn u8
n“1 .
i“1

Hint. First try to prove the above for k “ 2.

Some typical applications of 3.18. are mentioned in

3.19. Find all subssequential limits for each of the following sequences:
8
!´ πnαu
(a) tsin n“1 , where α P Q.
nπ ¯n )8
(b) cos .
3 )8 n“1
! nπ
(c) n sin .
" 2 3„ n“1 2
ȷ*8
2n 2n
(d) ´ .
7 7 n“1
(e) tnα ´ rnαsu8 n“1 , where α P Q.

Remark. However, in many occasions unfortunately there is no such natural partition as help.
For instance tcos nu8 8
n“1 or trn un“1 , where r1 , r2 , . . . is a listing of all rational numbers in p0, 1q.
Can you find all the subseqeuntial limits of these two sequences?

3.20. Find lim sup xn and lim inf xn where

nÑ`8 nÑ`8
def
(a) xn “ p´1qn n,
def n
(b) xn “ np´1q n ,
1
def ` n ˘
(c) xn “ 1 ` 2p´1q n n ,
def ` ˘n
(d) xn “ 2 cos 2nπ 3
,
def log n ´ p1 ` cos nπqn
(e) xn “ ,
log 2n
def
(f) xn “ cos nα, where α P R.

3.21.˚ Assume that Rd is endowed with the euclidean metric, where d P N. Suppose that txn u8 n“1 is
a sequencwe of unit vectors, i.e., }xn } “ 1, for all n P N. Show that, if lim inf }xn `x} ě }x}
nÑ`8
d d
holds for all x P R , then xxn , xy ÝÝÝÑ 0, for all x P R .
nÑ8

Remark. Can you generalize it to any inner product space?