Definitions
(a) A set S of real numbers is called bounded from above if there is a real number k such that k ≥ s for all s in S.
The element k is called an upper bound of S.
(b) A set S of real numbers is called boundedif there is a real number m such that
|s| ≤ m for all s in S.
(c) b is called the least upper bound (or supremum) of S, denoted by b = lub S or sup S, if (i) b ≥ s for all s in S, i.e. b is an upper bound of S.
(ii) if k is another upper bound of S then k ≥ b.
Equivalently,
b = lub S ⇐⇒ ∀ > 0 ∃s ∈ S such that b − < s ≤ b =⇒ |b − s| < .
(d) A sequence {xn} ⊂ R is called a Cauchy sequence if for any > 0 there exists K ∈ N such that if n, m ≥ K then |xn− xm| < .
Facts: It is easy to show that the followings are true.
(i) If {xn} ⊂ R is a Cauchy sequence then {xn} is bounded, i.e. there exists L ∈ R such that |xn| ≤ L for all n ∈ N.
(ii) If {xn} isa convergent sequence then {xn} is a Cauchy sequence.
Remarks
(a) Completeness Axiom of R: Every nonempty, bounded from above set of real numbers has a least upper bound (or supremum) in the set of real numbers.
(b) Least Upper Bound Property: Let S be a nonempty, bounded from above subset of R and let b = lub S. By taking n = 1
n, for n = 1, 2, . . . , there exists xn∈ S such that b − 1
n = b − n < xn ≤ b for all n ∈ N.
This implies that
for each n ∈ N there exists xn∈ S such that lim
n→∞xn= lub S.
(c) Theorem R is complete iff every Cauchy sequence {xn} ⊂ R converges to a point in R.
Proof.
=⇒ Given > 0 since {xn} is Cauchy,
there exists K ∈ N such that if n, m ≥ K then |xn− xm| < 2.
Case 1: {xn | n ∈ N} = {r1, r2, . . . , rm} is a finite subset of R. Then there is an element appears infinitely many times in the sequence {xn}.
Let rj be an element that appears infinitely many times in the sequence {xn} and let {xnk | k ∈ N} be the subsequence of {xn} such that
xnk = rj for all k ∈ N =⇒|xnk− rj| = 0 for all k ∈ N. This implies that if n ≥ K,since nK ≥ K, we have
|xn− rj|= |xn− xnK + xnK − rj| ≤ |xn− xnK| + |xnK − rj|<
2+ 0 <.
Hence
n→∞lim xn= rj. Case 2: {xn | n ∈ N} is an infinite subset of R.
Since a Cauchy sequence is bounded, there exists L > 0 such that xn∈ [−L, L] for all n ∈ N.
Let I1 = [−L, L] and let xn1 = x1 ∈ I1.
Divide I1 into 2 equal length subintervals and let I2 be a closed subinterval such that I2∩ {xn | n > n1} = an infinite subset of R
and let
xn2 ∈ I2∩ {xn | n > n1}.
Continuing this process, we obtain a nested sequence of closed intervals {Ik} and a subse- quence {xnk} such that
xnk ∈ Ik∩ {xn | n > nk−1} = an infinite subset of R for all k ∈ N.
Since
I1 ⊃ I2 ⊃ · · · ⊃ Ik⊃ · · · ; lim
k→∞|Ik| = lim
k→∞
L 2k−2 = 0, the sequence {xnk} converges and since
∃! x ∈
∞
\
k=1
Ik6= ∅, lim
k→∞xnk = x.
With the given , since lim
k→∞xnk = x,
there exists K1 ≥ K such that |x − xnK1| < 2. This implies that if n ≥ K,since nK1 ≥ K1 ≥ K, we have
|xn− x|= |xn− xnK1 + xnK1 − x| ≤ |xn− xnK1| + |xnK1 − x|< 2 +
2 =.
Hence
n→∞lim xn = x.
⇐= Let S be a nonempty, bounded from above subset of R. Choose a sufficiently large M > 0 such that
M is an upper bound of S and [−M, M ] ∩ S 6= ∅.
Let I1 = [−M, M ]. Divide I1 into 2 equal length subintervals and let I2 be the closed subinterval such that
the right endpoint of I2 is an upper bound of S and I2∩ S 6= ∅.
Continuing this process, we obtain a sequence of closed intervals {In} such that the right endpoint of In is an upper bound of S, In∩ S 6= ∅ ∀ n ≥ 1.
and
I1 ⊃ I2 ⊃ · · · ⊃ In⊃ · · · ; lim
n→∞|In| = lim
n→∞
M 2n−2 = 0, For each n ∈ N, since In∩ S 6= ∅, let xn be a point in In∩ S.
Since lim
n→∞|In| = lim
n→∞
M
2n−2 = 0, {xn} is a Cauchy sequence and hence it converges to a point, say x, in R.
To show that x = lub S, we need to show that
(i) x ≥ s for all s in S, i.e. x is an upper bound of S.
(ii) if u is another upper bound of S then u ≥ x.
Proof of (i): Suppose that x is not an upper bound of S, i.e. suppose that there exists s ∈ S such that s > x. Let = s − x. Since lim
n→∞xn = x, xn ∈ In and |In| = M
2n−2, there exists K ∈ N such that
if n ≥ K then |In| = M 2n−2 <
2 and |xn− x| < 2. This implies that
xn−
2 ≤ y ≤ xn+
2 for all y ∈ In∩ S ⇐⇒ In ⊂ (xn−
2, xn+ 2), x −
2 < xn< x +
2 ⇐⇒ xn∈ (x −
2, x +
2) for all n ≥ K.
Combining these, we get
x − < y < x + = s for all y ∈ In∩ S ⇐⇒ In ⊂ (x − , x + ) = (x − , s)
xn+ 2 xn
xn− 2
x x + 2
x − 2 x + = s
x − y
which implies that the right endpoint of In is smaller than s ∈ S. This contradicts to the definition of In. Hence, x is an upper bound of S.
Proof of (ii): Let u be another upper bound of S. Suppose that u < x, then = x − u 2 > 0.
Since lim
n→∞xn= x, there exists K ∈ N such that if n ≥ K then |xn− x| < .
Thus
= x − u
2 =⇒u = x − 2,
x − < xn< x + ⇐⇒ xn∈ (x − , x + ) for all n ≥ K.
This implies that
u = x − 2 <x − <xn< x + for all xn∈ S ∩ In, n ≥ K=⇒ u < xn∈ S
x + x
xn
xn− 2 xn+ 2 x −
u = x − 2
which contradicts to the assumption that u being an upper bound of S.
Hence u ≥ x.
(d) Every bounded, monotonic sequence {xn} ⊂ R is convergent.
Proof: Without loss of generality, we may assume that {xn} is an increasing sequence and it is bounded from above. By the completeness of R,
x = lub {xn | n ∈ N} exists.
Thus
for each > 0 there exists K ∈ N such that x − < xK ≤ x.
Now {xn} is an increasing sequence, we also have
xK ≤ xn for each n ≥ K.
Hence
x − < xK ≤ xn≤ x =⇒ |xn− x| < for all n ≥ K.
This proves that
n→∞lim xn = x.
Definition The series
∞
X
k=1
ak is called convergent if lim
n→∞Sn = lim
n→∞
n
X
k=1
ak exists and the limit is called the sumof the series. If the sequence {Sn} of nth partial sum is divergent, then the series is calleddivergent.
Remarks
(a) Test for Divergence If lim
n→∞an does not exist or if lim
n→∞an 6= 0, then the series
∞
X
n=1
an is divergent.
Proof: Since
∞
X
n=1
an is convergent ⇐⇒ {Sn =
n
X
k=1
ak} is a Cauchy sequence
=⇒ lim
m, n→∞|Sm− Sn| = 0
=⇒ lim
n→∞|an| = lim
n→∞|Sn− Sn−1| = 0.
This proves that if lim
n→∞andoes not exist or if lim
n→∞an 6= 0, then the series
∞
X
n=1
anis divergent.
(b) The Integral Test Suppose f is a continuous, positive, decreasing function on [1, ∞) and let an = f (n). Then the series
∞
X
n=1
an := lim
k→∞
k
X
n=1
an is convergent if and only if the improper integral
Z ∞ 1
f (x) dx := lim
M →∞
Z M 1
f (x) dx is convergent, i.e.
Z ∞ 1
f (x) dx and
∞
X
n=1
an converge or diverge simultaneously.
Proof: For each M ∈ N, consider the partition {1, 2, , 3, . . . , n, n + 1, . . . , M } of [1, M], i.e. xn = n + 1 and ∆xn = 1 for each 1 ≤ n ≤ M − 1. Since f is a continuous, positive, decreasing function on [1, ∞),
M −1
X
n=1
f (n + 1) =
M −1
X
n=1
inf
x∈[n,n+1]f (x) ∆xn ≤ Z M
1
f (x) dx ≤
M −1
X
n=1
sup
x∈[n,n+1]
f (x) ∆xn =
M −1
X
n=1
f (n).
Tis implies that Z ∞
1
f (x) dx and
∞
X
n=1
an converge or diverge simultaneously.
(c) The Ratio Test Suppose |an| > 0 for all n ∈ N and assume that
n→∞lim
|an+1|
|an| = L ∈ [0, ∞].
Then the series
∞
X
n=1
an is
(absolute convergent if L < 1,
divergent if L > 1 or L = ∞.
However, the ratio test is inconclusive if L = 1.
Proof: Suppose that L 6= 1 then = |L − 1|
2 > 0. Since lim
n→∞
|an+1|
|an| = L ∈ (0, ∞), there exists K ∈ N such that if n ≥ K then
|an+1|
|an| − L
< .
Thus
L − < |an+1|
|an| < L + ∀ n ≥ K
=⇒ L − )|an| < |an+1| < L + |an| ∀ n ≥ K
=⇒ L − )2|an−1| < |an+1| < L + 2
|an−1| ∀ n ≥ K
· · · ·
=⇒ L − )n−K+1|aK| < |an+1| < L + n−K+1
|aK| ∀ n ≥ K
=⇒
|an+1| < L + n−K+1
|aK| ∀ n ≥ K if L < 1
L − n−K+1
|aK| < |an+1| ∀ n ≥ K if L > 1
⇐⇒
|an+1| < L + 1 − L 2
n−K+1
|aK| ∀ n ≥ K if L < 1
L − L − 1 2
n−K+1
|aK| < |an+1| ∀ n ≥ K if L > 1
⇐⇒
|an+1| < 1 + L 2
n−K+1
|aK| ∀ n ≥ K if L < 1 L + 1
2
n−K+1
|aK| < |an+1| ∀ n ≥ K if L > 1.
This imples that
∞
X
n=K
|an+1|
<
∞
X
n=K
1 + L 2
n−K+1
|aK| =
|aK| 1 + L 2
1 −1 + L
2
since 0 < 1 + L
2 < 1 if L < 1,
>
∞
X
n=K
L + 1 2
n−K+1
|aK| = ∞ since L + 1
2 > 1 if L > 1.
Hence the series
∞
X
n=1
an is
(absolute convergent if L < 1,
divergent if L > 1 or L = ∞.
(d) The Root Test Suppose that
n→∞lim |an|1/n = L ∈ [0, ∞].
Then the series
∞
X
n=1
an is
(absolute convergent if L < 1,
divergent if L > 1 or L = ∞.
However, the root test is inconclusive if L = 1.
Proof: Suppose that L 6= 1 then = |L − 1|
2 > 0. Since lim
n→∞|an|1/n = L ∈ (0, ∞), there exists K ∈ N such that if n ≥ K then
|an|1/n− L < .
Thus
L − < |an|1/n < L + ∀ n ≥ K
=⇒ L − )n < |an| < L + n
∀ n ≥ K
=⇒
|an| < L + n
∀ n ≥ K if L < 1
L − n
< |an+1| ∀ n ≥ K if L > 1
⇐⇒
|an| < L + 1 − L 2
n
∀ n ≥ K if L < 1
L −L − 1 2
n
< |an| ∀ n ≥ K if L > 1
⇐⇒
|an| < 1 + L 2
n
∀ n ≥ K if L < 1 L + 1
2
n
< |an| ∀ n ≥ K if L > 1.
This imples that
∞
X
n=K
|an|
<
∞
X
n=K
1 + L 2
n
=
1 + L 2
K
1 −1 + L 2
since 0 < 1 + L
2 < 1 if L < 1,
>
∞
X
n=K
L + 1 2
n
= ∞ since L + 1
2 > 1 if L > 1.
Hence the series
∞
X
n=1
an is
(absolute convergent if L < 1,
divergent if L > 1 or L = ∞.
(e) The Rearrangement Theorem Let
∞
X
k=1
ak be an absolutely convergent series in R. Then
any rearragement of
∞
X
k=1
ak converges absolutely to the same value.
Proof: Let s =
∞
X
k=1
ak, let
∞
X
j=1
bj be an rearrangement of
∞
X
k=1
ak, and let L =
∞
X
k=1
|ak|.
Since |ak| ≥ 0 for all k ∈ N, L is an upper bound for each partial sum of
∞
X
k=1
|ak|, i.e.
n
X
k=1
|ak| ≤ L ∀ n ∈ N.
This implies that
r
X
j=1
|bj| = |b1| + · · · + |br| ≤ L ∀ r ∈ N,
i.e. the sequence {
r
X
j=1
|bj| | r ∈ N} is bounded from above by L.
Also since |bj| ≥ 0 for all j ∈ N, the sequence {
r
X
j=1
|bj| | r ∈ N} is increasing in r.
Therefore, the series
∞
X
j=1
bj is absolutely convergent to some element t ∈ R.
Claim: s = t.
Proof of the Claim: For each n ∈ N, let sn =
n
X
k=1
ak be the nth partial sum of
∞
X
k=1
ak.
For each > 0, since
∞
X
k=1
ak is convergent absolutely to s,
there exists M ∈ N such that if m > n ≥ M, then |s − sn| < and
m
X
k=n+1
|ak| < .
Also since
∞
X
j=1
bj is absolutely convergent to t, there exists a partial sum tr =
r
X
j=1
bj such that
|t − tr| < and such that each a1, a2, · · · , an appears in tr.
Having done this, choose m > n so large that every bj appearing in tr also appears in sm. Therefore
|s − t| ≤ |s − sm| + |sm− tr| + |tr− t| < +
m
X
k=n+1
|ak| + < 3.
Since > 0 is arbitrary, this proves that s = t.
(f) Remarks on Rearrangement:
(i) Let ak ∈ R for all k ∈ N. If
∞
X
k=1
ak is conditionally convergent, i.e. it is convergent but not absolutely convergent, then the series of potive terms is divergent (to ∞) and the series of negative terms is divergent (to −∞).
(ii) If
∞
X
k=1
ak is conditionally convergent and if c is an arbitrary real number, then there
exists a rearragement
∞
X
j=1
bj of
∞
X
k=1
ak such that c =
∞
X
j=1
bj. Proof:
1. Construction of a rearrangement of
∞
X
k=1
ak:
Step 1: take positive terms until we obtain a partial sum greater than c,
Step 2: take negative terms from the remaining series until we obtain a partial sum of terms less than c,
Step 3: take positive terms from the remaining series until we obtain a partial sum of terms greater than c,
Step 4: take negative terms from the remaining series until we obtain a partial sum of terms less than c,
· · · ·
Step 2r − 1: take positive terms from the remaining series until we obtain a partial sum of terms greater than c,
Step 2r: take negative terms from the remaining series until we obtain a partial sum of terms less than c,
· · · etc.
2. Convergence of the rearrangement: For each > 0, since
∞
X
k=1
ak converges =⇒ ∃ K ∈ N such that if n ≥ K then |
∞
X
k=n+1
ak| < .
Fix n ≥ K, we may take a sufficiently large r such that each a1, a2, · · · , an appears in the partial sum of terms obtained from Step r. This implies that
r→∞lim
∞
X
j=r+1
bj = 0 ⇐⇒
∞
X
j=1
bj = lim
m→∞
m
X
j=1
bj exists.
Since any subsequece {
mk
X
j=1
bj | k ∈ N} converges to the same value, and since the partial sums are greater than c in the odd number steps and the partial sums are less than c in the even number steps, one conclude that the constructed rearrangement converges to c.
Power Series
(a) Definition Let R be the radius of convergence of the power seies
∞
X
n=0
anxn, let Sm =
m
X
n=0
anxn for each m = 0, 1, 2, . . . , and let E = [a, b] ⊂ (−R, R). Then Sm is said to be uniformly convergent on E = [a, b] if for each > 0, there exists K = K() ∈ N such that
if m ≥ K then |
∞
X
n=m+1
anxn| = |
∞
X
n=0
anxn −
m
X
n=0
anxn| < ∀ x ∈ E = [a, b].
(b) Cauchy Criterion Sm =
m
X
n=0
anxn converges uniformly to f (x) =
∞
X
n=0
anxn on E = [a, b] if and only if for each > 0, there exists K = K() ∈ N such that
if m, n ≥ K then |Sm(x) − Sn(x)| < ∀x ∈ E = [a, b].
(c) Let R be the radius of convergence of
∞
X
n=1
anxn and let K = [a, b] be a bounded, closed subset of the interval of convergence (−R, R). Then the power series converges uniformly on K.
(d) Continuity Theorem Let E = [a, b] ⊂ (−R, R) and let Sm =
m
X
n=0
anxn for each m =
0, 1, 2, . . . . If Sm converges uniformly to f (x) =
∞
X
n=0
anxn on E then f is continuous on E = [a, b].
Proof: For each > 0, since Sm converges uniformly to f (x) =
∞
X
n=0
anxn on E = [a, b], there exists K ∈ N such that
if m ≥ K then |f (x) − Sm(x)| = |
∞
X
n=0
anxn −
m
X
n=0
anxn| < ∀ x ∈ E = [a, b].
On the other hand, since SK is continuous on E = [a, b], SK is uniformly continuous on E = [a, b]. With the given > 0,there exists δK > 0 such that
if x, y ∈ E and |x − y| < δK then |SK(x) − SK(y)|| < .
This implies that if x, y ∈ E and |x − y| < δK then
|f (x) − f (y)| ≤ |f (x) − SK(x)| + |SK(x) − SK(y)| + |SK(y) − f (y)| < 3.
Since > 0 is arbitrary, this proves that f is (uniformly) continuous on E = [a, b].
(e) Differentiation Theorem: A power series can be differentiated term-by-term within the interval of convergence. In fact, if
f (x) =
∞
X
n=1
anxn, then f0(x) =
∞
X
n=1
nanxn−1.
Both series have the same radius of convergence.
Proof: Since lim
n→∞ n1/n
= 1, the sequence |nan|1/n is bounded if and only if the sequence
|an|1/n is bounded. Moreover, it is easily seen that
n→∞lim |nan|1/n = lim
n→∞ |an|1/n.
Therefore, the radius of convergence of the two series is the same, so Sm0 =
m
X
n=1
nanxn−1 is uniformly convergent on each bounded closed subset E = [a, b] ⊂ (−R, R).
Thus,for each > 0, there exists K = K() ∈ N such that
if m, n ≥ K then |Sm0 (x) − Sn0(x)| < ∀x ∈ E = [a, b], .
and, by the Mean Value Theorem, there exists z, depending on m, n ≥ K, such that {Sm(x) − Sn(x)} − {Sm(c) − Sn(c)} = (x − c){Sm0 (z) − Sn0(z)}
which implies that
Sm(x) − Sm(c)
x − c − Sn(x) − Sn(c) x − c
= |Sm0 (z) − Sn0(z)| < .
Taking the limit with respect to m, we have
f (x) − f (c)
x − c − Sn(x) − Sn(c) x − c
= lim
m→∞
Sm(x) − Sm(c)
x − c − Sn(x) − Sn(c) x − c
≤ .
Also, for each x 6= c ∈ [a, b], since SK =
K
X
n=1
anxn is differentiable on E = [a, b] ⊂ (−R, R), there exists δK() > 0 such that
if 0 < |x − c| < δK() then
SK(x) − SK(c)
x − c − SK0 (c)
< .
Therefore, it follows that if 0 < |x − c| < δK() and n ≥ K, then
f (x) − f (c)
x − c − SK0 (c)
=
f (x) − f (c)
x − c − Sn(x) − Sn(c)
x − c +Sn(x) − Sn(c)
x − c − SK(x) − SK(c) x − c + SK(x) − SK(c)
x − c − SK0 (c)
≤
f (x) − f (c)
x − c − Sn(x) − Sn(c) x − c
+
Sn(x) − Sn(c)
x − c − SK(x) − SK(c) x − c
+
SK(x) − SK(c)
x − c − SK0 (c)
< 3.
This shows that f0(c) exists and equals lim
K→∞SK0 (c) =
∞
X
n=1
nancn−1 for all c ∈ E = [a, b].
(f) Integration Theorem Let E = [a, b] ⊂ (−R, R) and let Sm =
m
X
n=0
anxn for each m =
0, 1, 2, . . . . If Sm converges uniformly to f (x) =
∞
X
n=0
anxn on E then
Z b a
f (x) dx =
∞
X
n=0
Z b a
anxn dx,
i.e.
Z b a
m→∞lim Sm(x) dx = lim
m→∞
Z b a
Sm(x) dx.
Proof: For each > 0, since Sm converges uniformly to f (x) =
∞
X
n=0
anxn on E = [a, b], there exists K ∈ N such that
if m ≥ K then |f (x) − Sm(x)| <
b − a ∀ x ∈ E = [a, b].
For any partition P = {a = x0 < x1 < · · · < xn= b} of [a, b], since
|
n
X
i=1
f (x∗i) − Sm(x∗i) ∆xi| ≤
n
X
i=1
|f (x∗i) − Sm(x∗i)| ∆xi < b − a
n
X
i=1
∆xi = ,
by letting max
1≤i≤n∆xi → 0, we get
| Z b
a
∞
X
n=0
anxn dx − Z b
a
Sm(x) dx| = | Z b
a
f (x) dx − Z b
a
Sm(x) dx| ≤ ∀ m ≥ K.
Since > 0 is arbitrary, this proves that Z b
a
∞
X
n=0
anxn dx = Z b
a
f (x) dx = lim
m→∞
Z b a
Sm(x) dx =
∞
X
n=0
Z b a
anxn dx.
Taylor’s Theorem
(a) Suppose that n ∈ N, that f and its derivatives f0, f00, . . . , f(n−1) are defined and continuous on [a, b], and that f(n) exists in (a, b). If α, β ∈ [a, b], then there exists a number γ between α and β such that
f (β) = f (α) + f0(α)
1! (β − α) + f00(α)
2! (β − α)2+ · · · + f(n−1)(α)
(n − 1)! (β − α)n−1+ f(n)(γ)
n! (β − α)n
=
n−1
X
k=0
f(k)(α)
k! (β − α)k+f(n)(γ)
n! (β − α)n
=
n−1
X
k=0
f(k)(α)
k! (β − α)k+Rn(β).
Proof: Let P be the real number defined by the relation (β − α)n
n! P = f (β) −
n−1
X
k=0
f(k)(α)(β − α)k
k! ,
and consider the function g : [a, b] → R defined by
g(x) = f (β) −
n−1
X
k=0
f(k)(x)(β − x)k
k! − P
n!(β − x)n ∀ x ∈ [a, b].
Then g is differentiable on [a, b] with
g0(x) = −
n−1
X
k=0
d dx
f(k)(x)(β − x)k
k! − d
dx P
(n − 1)!(β − x)n−1
= −
n−1
X
k=0
f(k+1)(x)(β − x)k
k! +
n−1
X
k=1
f(k)(x)(β − x)k−1
(k − 1)! + P
(n − 1)!(β − x)n−1
= −
n−1
X
k=0
f(k+1)(x)(β − x)k
k! +
n−2
X
k=0
f(k+1)(x)(β − x)k
k! + P
(n − 1)!(β − x)n−1
= −f(n)(x)(β − x)n−1
(n − 1)! + P
(n − 1)!(β − x)n−1
= P − f(n)(x)
(n − 1)! (β − x)n−1.
Clearly, g is continuous on [a, b] and is differentiable on (a, b). Also it is evident that g(β) = 0 and it follows from the definition of P that g(α) = 0. By the Rolle’s Theorem, there exists a point γ between α and β such that g0(γ) = 0.
Since g0(γ) = 0, then P = f(n)(γ) and this proves that
f (β) =
n−1
X
k=0
f(k)(α)
k! (β − α)k+ f(n)(γ)
n! (β − α)n=
n−1
X
k=0
f(k)(α)
k! (β − α)k+ Rn(β).
(b) Integral Form for the Remainder Suppose that f and its derivatives f0, f00, . . . , f(n) are continuous on [a, b] to R. Then
f (b) = f (a) + f0(a)
1! (b − a) + f00(a)
2! (b − a)2+ · · · + f(n−1)(a)
(n − 1)! (b − a)n−1+ Rn(b), where the remainder is given by
Rn(b) = 1 (n − 1)!
Z b a
(b − t)n−1f(n)(t) dt.
Proof: Integrate Rn by parts to obtain Rn(b) = 1
(n − 1)!
(b − t)n−1f(n−1)(t)|t=bt=a+ (n − 1) Z b
a
(b − t)n−2f(n−1)(t) dt
= −f(n−1)(a)
(n − 1)! (b − a)n−1+ 1 (n − 2)!
Z b a
(b − t)n−2f(n−1)(t) dt
= −f(n−1)(a)
(n − 1)! (b − a)n−1−f(n−2)(a)
(n − 2)!(b − a)n−2+ 1 (n − 3)!
Z b a
(b − t)n−3f(n−2)(t) dt
· · · ·
= −f(n−1)(a)
(n − 1)! (b − a)n−1−f(n−2)(a)
(n − 2)!(b − a)n−2− · · · − f0(a)
1! (b − a) + Z b
a
f0(t) dt
= −
n−1
X
k=0
f(k)(a)
k! (b − a)k+ f (b).
This proves the stated formula
f (b) =
n−1
X
k=0
f(k)(a)
k! (b − a)k+ Rn(b).
Examples:
(a) Let f (x) = sin x for x ∈ (−∞, ∞).
For each n = 0, 1, 2, . . . , since
f(4n)(x) = sin x, f(4n+1)(x) = cos x, f(4n+2)(x) = − sin x, f(4n+3)(x) = − cos x, thus, for all n and for all t,
|f(n)(t)| ≤ 1.
It follows that
|Rn(x)| ≤ |x|n
n! → 0 for all − ∞ < x < ∞ as n → ∞,
i.e. Maclaurin series (Taylor series at a = 0) of sin x converges for −∞ < x < ∞.
Also since
f(2n)(0) = 0 and f(2n+1)(0) = (−1)n for n = 0, 1, 2, . . . , we have
sin x =
∞
X
n=0
(−1)n
(2n + 1)!x2n+1 = x −x3 3! + x5
5! − · · · for all − ∞ < x < ∞.
(b) Let f (x) = ln(1 + x) for x ∈ (−1, ∞).
Since
f0(x) = 1
1 + x, f00(x) = − 1
(1 + x)2, f000(x) = 2
(1 + x)3, f(4)(x) = − 3!
(1 + x)4, · · · we have
f(n)(x) = (−1)n−1(n − 1)!
(1 + x)n for n ≥ 1, and
Rn(x) = 1 (n − 1)!
Z x 0
(x − t)n−1f(n)(t) dt
= 1
(n − 1)!
Z x 0
(x − t)n−1 (−1)n−1(n − 1)!
(1 + t)ndt
= (−1)n−1 Z x
0
(x − t)n−1 (1 + t)n dt.
For 0 ≤ x ≤ 1 and for 0 ≤ t ≤ x, we have
1 + t = |t − (−1)| ≥ 1 =⇒ 1
(1 + t)n ≤ 1 ∀ t ∈ [0, x]
and hence
|Rn(x)| = Z x
0
(x − t)n−1 (1 + t)n dt ≤
Z x 0
(x − t)n−1 dt = xn
n → 0 as n → ∞.
For −1 < x < 0 and for x ≤ t ≤ 0, we have
1 + t = |t − (−1)| ≥ |x − t| = t − x =⇒ |x − t|n−1
(1 + t)n−1 ≤ 1 ∀ t ∈ [x, 0]
and hence
|Rn(x)| =
Z x 0
(x − t)n−1 (1 + t)n dt
= Z 0
x
(t − x)n−1 (1 + t)n−1
1 1 + t dt.
By the Mean Value Theorem for Integrals [i.e. If f is continuous on [a, b], then there is at least one number c in (a, b) for which
Z b a
f (x)dx = f (c)(b − a)] and since (t − x)n−1 (1 + t)n−1
1 1 + t is continuous on [x, 0], there exists xn ∈ (x, 0) for each n ∈ N such that
Z 0 x
(t − x)n−1 (1 + t)n−1
1
1 + t dt = (xn− x)n−1 (1 + xn)n−1
1 1 + xn
(0 − x).
Since −x = |x| and 0 < 1 + x < 1 + xn, we conclude that
|Rn(x)| < xn− x 1 + xn
n−1 |x|
1 + x
. Since |x| < 1 and xn < 0, we have
xn< |x|xn, 0 < xn+ |x| < |x|xn+ |x| = |x| 1 + xn, and thus
xn+ |x|
1 + xn < |x| =⇒ xn− x 1 + xn < |x|.
It follows that
|Rn(x)| < |x|n−1 |x|
1 + x → 0 as n → ∞ since |x| < 1,
i.e.Maclaurin series (Taylor series at a = 0) of ln(1 + x) converges for −1 < x ≤ 1.
Hence
ln(1 + x) =
∞
X
n=1
(−1)n−1
n xn= x − x2 2 +x3
3 − · · · for − 1 < x ≤ 1.