Properties of the Riemann–Stieltjes Integral

Theorem (Linearity Properties)

Let a < c < d < b and A, B ∈ IR and f, g, α, β : [a, b] → IR.

(a) If f, g ∈ R(α) on [a, b], then Af + Bg ∈ R(α) on [a, b] and

Z b Z b Z b
 
Af + Bg dα = A f dα + B g dα
a a a

(b) If f ∈ R(α) ∩ R(β) on [a, b], then f ∈ R(Aα + Bβ) on [a, b] and
Z b Z b Z b
f d(Aα + Bβ) = A f dα + B f dβ
a a a

(c) If f ∈ R(α) on [a, c] and on [c, b], then f ∈ R(α) on [a, b] and
Z b Z c Z b
f dα = f dα + f dα
a a c

(d) If f ∈ R(α) on [a, b] then f ∈ R(α) on [c, d] ⊂ [a, b].

Proof: (a) Let ε > 0. Then, for any partition IP of [a, b] and choice T for IP,
 Z b Z b 
 
S(IP, T, Af + Bg, α) − A f dα − B g dα 
 
a a
 Z b Z b 
 
= A S(IP, T, f, α) + B S(IP, T, g, α) − A
 f dα − B g dα
a a
 Z b   Z b 
   
≤ |A| S(IP, T, f, α) −
 f dα + |B| S(IP, T, g, α) −
  g dα
a a
Assume that A and B are nonzero. (The cases that A and/or B are zero are similar, but easier.)
Since f ∈ R(α) on [a, b] there is a partition IPf,ε such that
 Z b 

S(IP, T, f, α) −
 ε
f dα < whenever IP ⊃ IPf,ε

a 2|A|
and since g ∈ R(α) on [a, b] there is a partition IPg,ε such that
 Z b 

S(IP, T, g, α) −
 ε
g dα < whenever IP ⊃ IPg,ε

a 2|B|
It now suffices to set IPε = IPf,ε ∪ IPg,ε and observe that
 Z b Z b 
 
S(IP, T, Af + Bg, α) − A f dα − B g dα < ε whenever IP ⊃ IPε

a a

(b) See Problem Set 1, #3.

(c) See Problem Set 1, #2.

(d) See Problem Set 2, #3.

c Joel Feldman. 2017. All rights reserved. January 20, 2017 Properties of the Riemann–Stieltjes Integral 1
Theorem (Integration by Parts)
Let a < b and f, α : [a, b] → IR. If f ∈ R(α) on [a, b], then α ∈ R(f ) on [a, b] and
Z b Z b
f (x) dα(x) + α(x) df (x) = f (b)α(b) − f (a)α(a)
a a

Remark.

(a) The integration by parts formula may also be written

Z b Z b Z b
f dα + α df = d(f α)
a a a

Rb
(b) We shall later see that if α has a continuous first derivative, then a f (x) dα(x) =
Rb
a
f (x)α′ (x) dx. So if both f and α have continuous first derivatives, we may write the integration
by parts formula as
Z b Z b
′
f (x)α (x) dx = f (b)α(b) − f (a)α(a) − α(x)f ′ (x) dx
a a

which is the integration by parts formula of first year calculus (though you probably used f (x) =
u(x) and α(x) = v(x).
Rb
Proof of integration by parts: Our goal is to show that a
α df exists and takes the value
Z b Z b
α df = f (b)α(b) − f (a)α(a) − f dα
a a

Rb
So let’s look at the difference between a Riemann sum for a α df and the right hand side. For any
 
partition IP = a = x0 , x1 , x2 , x3 , · · · , xn = b of [a, b] and choice T = t1 , t2 , t3 , · · · , tn for
IP,
n Z b o


S IP, T, α, f − f (b)α(b) − f (a)α(a) − f dα
a
n
X n
X n
X Z b
 
= α(ti ) f (xi ) − f (xi−1 − α(xi )f (xi ) + α(xi−1 )f (xi−1 ) + f dα
| {z } | {z } a
i=1 i=1 α(b)f (b)
i=1 α(a)f (a)
for i=n for i=1

(The 1 ≤ i ≤ n − 1 terms of the second sum cancel the

2 ≤ i ≤ n terms of the third sum.)
n
X n Z b
  X  
=− f (xi ) α(xi ) − α(ti ) − f (xj−1 ) α(tj ) − α(xj−1 ) + f dα
i=1 j=1 a

Note that

c Joel Feldman. 2017. All rights reserved. January 20, 2017 Properties of the Riemann–Stieltjes Integral 2
   Rb
◦ f (xi ) α(xi ) − α(ti ) looks like a term in a Riemann sum approximation to a f dα with subin-
terval [ti , xi ] and choice point xi ∈ [ti , xi ] and
  Rb
◦ f (xj−1 ) α(tj ) − α(xj−1 ) looks like a term in a Riemann sum approximation to a f dα with
subinterval [xj−1 , tj ] and choice point xj−1 ∈ [xj−1 , tj ].
Here is a figure which shows all of these subintervals.
j=1 i=1 j=2 i=2 i = n−1 j = n i = n
a = x0 t1 x1 t2 x2 ··· tn−1 xn−1 tn xn = b

So all of the subintervals fit together perfectly to form the partition1


IP ∪ T = x0 , t1 , x1 , t2 , x2 , t3 , · · · , xn−1 , tn , xn
and Z

n b o
S IP, T, α, f − f (b)α(b) − f (a)α(a) − f dα
a
Z b

′
= −S IP ∪ T, T , f, α + f dα
a
where the choice
j=1 i=1 j=2 i=2 j=3 i=3 i=n−1 j=n i=n
′
 z}|{ z}|{ z}|{ z}|{ z}|{ z}|{ z }| { z }| { z}|{
T = x0 , x1 , x1 , x2 , x2 , x3 , · · · , xn−1 , xn−1 , xn
Now let ε > 0. Since f ∈ R(α) for [a, b] there is a partition IPε such that
 Z b 


˜ T̃, f, α −
S IP, f dα < ε

a
˜ finer than IPε . If the partitition IP above is finer than IPε then the partition
for all partitions IP
IP ∪ T is also finer than IPε and we have
 Z b o  Z b 

n ′


S IP, T, α, f − f (b)α(b) − f (a)α(a) − f dα  = S IP ∪ T, T , f, α −
  f dα < ε

a a

Theorem (The Change of Variables x = g(y))

Let
◦ a < b and c < d,
◦ g : [c, d] → [a, b] be continuous, strictly monotonically increasing, and obey g(c) = a and
g(d) = b and
◦ f, α : [a, b] → IR.
Set



h(y) = f g(y) β(y) = α g(y)
If f ∈ R(α) on [a, b], then h ∈ R(β) on [c, d] and
Z d Z b
h(y) dβ(y) = f (x) dα(x)
c a

1 There is a subtlety hidden in these definitions. We are not allowed to use subintervals of width zero. So if, for
example, ti = xi , we merge the two subintervals [xi−1 , ti ], [ti , xi ] into the single subinterval [xi−1 , xi ] and use
f (xi−1 )[α(ti ) − α(xi−1 )] + f (xi )[α(xi ) − α(ti )] = f (xi−1 )[α(xi ) − α(xi−1 )].

c Joel Feldman. 2017. All rights reserved. January 20, 2017 Properties of the Riemann–Stieltjes Integral 3
 Rd Rb
Proof: Our goal is to prove that h(y) dβ(y) exists and equals a f (x) dα(x), so let’s consider
c
Rd Rb
the difference between a Riemann sum for c h(y) dβ(y) and a f (x) dα(x). For any partition
IP = {c = y0 , · · · , yn = d} of [c, d] and any choice T = {s1 , · · · , sn } for IP

 Z b   n Z b 
  X   
S(IP, T, h, β) − f dα = 
  h(si ) β(yi ) − β(yi−1 ) − f dα

a i=1 a
X 
 n
Z b



 
=  f g(si ) α g(yi ) − α g(yi−1 ) − f dα
i=1 a
 Z b 
 
= S(g(IP), g(T), f, α) − f dα
a

where  
g(IP) = g(y)  y ∈ IP
x0 x1 xn
 z }| { z }| { z }| {
= g(y0 ) = g(c) = a , g(y1 ) , · · · , g(yn ) = g(d) = b

is a partition of [a, b] because g is assumed to be strictly monotonic, so that

yi−1 < yi =⇒ xi−1 < xi

and is assumed to obey x0 = g(y0 ) = a and xn = g(yn ) = b and

t1 tn
 z }| { z }| {
g(T) = g(s1 ) , · · · , g(sn )

is a choice for g(IP) because g is assumed to be strictly monotonic so that

yi−1 ≤ si ≤ yi =⇒ xi−1 = g(yi−1 ) ≤ g(si ) = ti ≤ g(yi ) = xi

Now let ε > 0.

◦ We have assumed that f ∈ R(α) on [a, b]. So there is a partition IPf,ε of [a, b] such that
 Rb 
IPf ⊃ IPf,ε =⇒ S(IPf , Tf , f, α) − a f dα < ε for all choices Tf for IPf .
◦ The assumptions that we have made on g guarantee that the inverse function g −1 : [a, b] → [c, d]



exists and that g −1 IPf,ε is a partition of [c, d]. We choose IPε = g −1 IPf,ε .
Then
IP ⊃ IPε =⇒ g(IP) ⊃ g(IPε ) = IPf,ε
finer allowed
 Z b   than IPf,ε Tf
z }| { z}|{ Z b 
   
=⇒ S(IP, T, h, β) −
 f dα = S( g(IP) , g(T) , f, α) −
  f dα < ε
a a

as desired.

c Joel Feldman. 2017. All rights reserved. January 20, 2017 Properties of the Riemann–Stieltjes Integral 4
Theorem (Second Fundamental Theorem of Calculus) Let a < b and f : [a, b] → IR.
Assume that
◦ f is differentiable on [a, b] and
◦ f ′ ∈ R on [a, b]
Then Z b
f ′ (x) dx = f (b) − f (a)
a


Proof: Let IP = x0 , x1 , x2 , · · · , xn be any partition of [a, b]. Then, by the mean value theorem,
there exists, for each 1 ≤ i ≤ n a ti ∈ [xi−1 , xi ] with


f (xi ) − f (xi−1 ) = f ′ (ti ) xi − xi−1

So, setting T = t1 , t2 , · · · , tn , we have
n
X n

X  
S(IP, T, f ′ ) = f ′ (ti ) xi − xi−1 = f (xi ) − f (xi−1 ) = f (b) − f (a)
i=1 i=1

So now we just have to apply the definition of “f ′ ∈ R on [a, b]”.

Theorem (Basic Bounds)

Let a < b and f, g, α : [a, b] → IR. Assume that f, g ∈ R(α) on [a, b] and α is monotonically
increasing.

(a) If f (x) ≤ g(x) for all x ∈ [a, b], then

Z b Z b
f dα ≤ g dα
a a

(a) If |f (x)| ≤ g(x) for all x ∈ [a, b], then

Z b  Z b
 

 f dα ≤ g dα
a a

Proof: Let ε > 0. Since f ∈ R(α) on [a, b] there is a partition IPf,ε such that
 Z b 
 
S(IP, T, f, α) − f dα < ε

a

whenever IP ⊃ IPf,ε and T is a choice for IP. Since g ∈ R(α) on [a, b] there is a partition IPg,ε such
that  Z b 
 
S(IP, T, g, α) − g dα < ε

a

whenever IP ⊃ IPg,ε and T is a choice for IP. Set IPε = a = x0 , x1 , x2 , · · · , xn = b = IPf,ε ∪ IPg,ε

and let T = t1 , t2 , · · · , tn } be a choice for IPε .

c Joel Feldman. 2017. All rights reserved. January 20, 2017 Properties of the Riemann–Stieltjes Integral 5
(a) We have
Z b
f dα ≤ S(IPε , T, f, α) + ε
a
n
X  
= f (ti ) α(xi ) − α(xi−1 ) + ε
i=1
n
X  
≤ g(ti ) α(xi ) − α(xi−1 ) + ε
i=1
(since f (ti ) ≤ g(ti ) and α(xi ) − α(xi−1 ) ≥ 0)
= S(IPε , T, g, α) + ε
Z b
≤ g dα + 2ε
a
Rb Rb Rb Rb
As a
f dα ≤ a
g dα + 2ε is true for all ε > 0, we also have a
f dα ≤ a
g dα.

(b) We have

Z b 
   

 f dα ≤ S(IPε , T, f, α) + ε
a
n
X   
≤ f (ti ) α(xi ) − α(xi−1 ) + ε
i=1
Xn
 
≤ g(ti ) α(xi ) − α(xi−1 ) + ε
i=1
(since |f (ti )| ≤ g(ti ) and α(xi ) − α(xi−1 ) ≥ 0)
= S(IPε , T, g, α) + ε
Z b
≤ g dα + 2ε
a

R  R R  R
 b  b  b  b
As  a f dα ≤ a g dα + 2ε is true for all ε > 0, we also have  a f dα ≤ a g dα.

Theorem (First Fundamental Theorem of Calculus) Let a < b and f : [a, b] → IR. Assume
that
◦ f ∈ R on [a, b]
Set, for a ≤ x ≤ b,
Z x
F (x) = f (x) dx
a

Then
(a) F is continuous on [a, b] and
(b) if f is continuous at x0 ∈ [a, b], then F is differentiable at x0 and F ′ (x0 ) = f (x0 ).

c Joel Feldman. 2017. All rights reserved. January 20, 2017 Properties of the Riemann–Stieltjes Integral 6
Proof: (a) Since f ∈ R it is bounded. Suppose that |f (t)| ≤ M for all a ≤ t ≤ b. Then
Z y 
   
F (y) − F (x) =  f (t) dt ≤ M |y − x|

x

so F is uniformly continous.
(b) Let f be continuous at x0 ∈ [a, b]. Then, for all a ≤ x ≤ b
   R x 
 F (x) − F (x0 )   x0 f (t) dt − f (x0 )[x − x0 ] 
 − f (x0 ) =  
 x − x0  x − x0 
 Z x 
 1   
= f (t) − f (x0 ) dt
x − x0 x0
 
≤ sup f (t) − f (x0 )
t between
x0 and x

Since f is continuous at x0 , the right hand side converges to zero as x → x0 .

c Joel Feldman. 2017. All rights reserved. January 20, 2017 Properties of the Riemann–Stieltjes Integral 7