CHAPTER 24

Derivatives of Inverse Functions and Logarithms

W
£ §
e know that dx d
e x = e x . But what about derivatives of exponential
£ x§
function with bases other than e? In other words, what is dx d
a ?
£ § £ §
And what about dx d
ln(x) and dx d
loga (x) ? The main goal of this chapter is
to answer these questions and thus expand our list of derivative rules.
£ x§
Let’s start with dxd
a . Since ln(x) is the inverse of e x , we know a = eln(a) .
We can thus convert the power a x to a power of e:
≥ ¥x
ax = eln(a) = eln(a) x .

With this, we can get the derivative of a x with the chain rule:

d h xi d h ln(a) x i d h i
a = e = eln(a) x ln(a)x = eln(a) x ln(a) = a x ln(a).
dx |dx {z dx }
chain rule
So the derivative of a x is just a x times the constant ln(a). This is a new rule.

d h xi
Rule 16 a = ln(a)a x
dx

d h xi d h xi
For example, 10 = ln(10)10 x º 2.302 · 10 x . Also 2 = ln(2)2 x º
dx dx
d h xi
0.693 · 2 x . Notice how special the base e is: e = ln(e)e x = 1 · e x = e x .
dx
The base a = e is the only base for which the derivative of a x is 1 times a x .
£ §
Next we will get a rule for dx d
ln(x) . Our strategy will be to use the fact
that ln(x) is the inverse of e x , that is,

if f (x) = e x , then f °1 (x) = ln(x).

£ §
Our plan is to first develop a general rule for dxd
f °1 (x) and then use it to
£ §
get dx
d
ln(x) . (See Chapter 4 If you need to review inverses.)
278 Derivatives of Inverse Functions and Logarithms

d h °1 i
Thus our immediate question is: What is f (x) ?
dx
To answer this, think about the relation between f and its inverse f °1 :
° ¢
f f °1 (x) = x.

The two sides of this equation are equal functions, so if we differentiate

both sides the derivatives will be equal:

d h ° °1 ¢ i d h i
f f (x) = x
dx dx

The right-hand side of this equation is 1. The left-hand side is the derivative
of a composition, so we can apply the chain rule to it:
° ¢ d h °1 i
f 0 f °1 (x) f (x) = 1
dx
° ¢ £ °1 §
In applying the chain rule we multiplied f 0 f °1 (x) by dx
d
f (x) . We stopped
£ °1 §
there because we don’t know what dx d
f (x) is. But it’s exactly what we
° ¢
want to find! We can isolate it by dividing the above equation by f 0 f °1 (x) :

d h °1 i 1
f (x) = 0 ° °1 ¢
dx f f (x) .

This is our latest rule.

Rule 17 (The inverse rule) If f is a function having a derivative f 0 and an

inverse f °1 , then
d h °1 i 1
f (x) = 0 ° °1 ¢ .
dx f f (x)

p
To illustrate this rule, suppose f (x) = x3 , which has an inverse f °1 (x) = 3
x.
£ °1 §
Let’s find dx
d
f (x) . We know that f 0 (x) = 3x2 , so our new rule gives

d h °1 i 1 1 1
f (x) = 0 ° °1 ¢ = ° ¢2 = p 2
.
dx f f (x) 3 f °1 (x) 3 x
3

Granted, this is not all that impressive, since we can use the power rule to
get the same answer:

d h °1 i d hp i d h 1i 1 2 1 1
f (x) = 3
x = x 3 = x° 3 = = p 2
.
dx dx dx 3 2
3x 3 3 x
3
 279

But the inverse rule can be very useful. We’ll now use it to find the
derivative of ln(x). Say f (x) = e x , so f °1 (x) = ln(x). Then

d h i d h °1 i
ln(x) = f (x) because ln(x) = f °1 (x)
dx dx
1
= 0 ° °1 ¢ by inverse rule
f f (x)
1
= ° ¢ because f °1 (x) = ln(x)
f0 ln(x)
1
= becase f 0 (x) = e x
eln( x)
1
= because eln( x) = x.
x
d h i 1
Thus ln(x) = . Figure 24.1 (left) illustrates this remarkable fact.
dx x
It shows the function f (x) = ln(x) along with its derivative f 0 (x) = 1x . Notice
how if x is near 0, the tangent to ln(x) at x is very steep, and indeed the
derivative 1x is very large. But as x gets bigger, the tangent to ln(x) gets
closer to horizontal (slope 0) while the derivative 1
x approaches zero.

y= ln(x) y= ln | x|
1 1
y= y=
x x

Figure 24.1. Left: the graphs of f (x) = ln(x) (black) and f 0 (x) = 1x (blue) with
domain (0, 1). Right: the graphs of f (x) = ln | x| (black) and f 0 (x) = 1x (blue).

Notice however, that the domain of ln(x) is (0, 1) but the domain of 1x is
(°1, 0) [ (0, 1). So when we say that the derivative of ln(x) is 1x , we really
mean 1x with its domain restricted to (0, 1). Figure 24.1 (right) shows a
somewhat more complete scenario. It shows the function ln(| x|), which we
will abbreviate as ln | x|. This function has domain (°1, 0) [ (0, 1), and its
derivative is 1x with its usual domain. So our latest rule has two parts.
d h i 1 d h ØØ ØØi 1
Rule 18 ln(x) = and ln x = .
dx x dx x
280 Derivatives of Inverse Functions and Logarithms

Here it is understood that in the first formula the domain of ln(x) and 1x
is (0, 1). In the second formula the domain of ln | x| and 1x is all real numbers
except 0. Do not sweat the difference between the two versions of this rule –
they say almost the same thing, and the second implies the first. We will
mostly use the first version in parts 3 and 4 of this book, but the second
version becomes particularly useful in Part 5.
At the beginning of this chapter we said our main goals were to find
£ x§ d £ § £ §
formulas for dx d
a , dx ln(x) and dxd
loga (x) . We’ve done all but the last
one. For it we will use the change of base formula (Fact 5.1 in Chapter 5,
page 88) which states
ln(x)
loga (x) = .
ln(a)
Using this, the constant multiple rule and Rule 18, we get
∑ ∏
d h i d ln(x) 1 d h i 1 1
loga (x) = = · ln(x) = · .
dx dx ln(a) ln(a) dx ln(a) x

With our prior agreement about domains, we get another two-part formula.

d h i 1 d h Ø Øi 1
Rule 19 loga (x) = and loga Ø xØ = .
dx x ln(a) dx x ln(a)

d h i d h i d h i
Example 24.1 log3 (x) tan(x) = log3 (x) tan(x) + log3 (x) tan(x)
dx dx dx
1
(product rule) = tan(x) + log3 (x) sec2 (x) .
x ln(3)

d hp i
Example 24.2 Find 5 + x3 + ln(x) .
dx
This is the derivative of a function to a power, so we can use the generalized
power rule:

d hp i d h° ¢1/2 i
5 + x3 + ln(x) = 5 + x3 + ln(x)
dx dx
1° ¢1/2°1 d h i
= 5 + x3 + ln(x) 5 + x3 + ln(x)
2 dx
µ ∂
1° 3
¢°1/2 2 1
= 5 + x + ln(x) 3x +
2 x

3x2 + 1x
= p .
2 5 + x3 + ln(x)
 281

d h ° ¢3 i ° ¢2 d h i
Example 24.3 7 + x + ln(x) = 0 + 1 + 3 ln(x) ln(x)
dx dx
° ¢2
° ¢2 1 3 ln(x)
= 1 + 3 ln(x) = 1+ .
x x

d h i d h i 1
h i ln(x) · x + ln(x) · x · x + ln(x) · 1
d ln(x) dx dx x
Example 24.4 = =
dx x x2 x2
1 + ln(x)
(quotient rule) = .
x2

Example 24.5 Find the derivative of 10 x

2
+3 x+2
.
(
2 y = 10u
The composition y = 10 x +3 x+2
can be broken up as
u = x2 + 3x + 2.

dy d y du
The chain rule then gives = ·
dx du dx
= ln(10) 10u · (2x + 3 + 0)
2
+3 x+2
= ln(10) 10 x (2x + 3) .

In Example 24.5 we differentiated a function of form a g( x) . Let’s repeat

£ x§
our steps to get a chain rule generalization for the rule dx
d
a = ln(a) a x .

Example 24.6 Find the derivative of a g( x) .

(
y = au
The composition y = a g ( x)
can be broken up as
u = g(x).

dy d y du
The chain rule then gives = ·
dx du dx
= ln(a) a u · g0 (x)

= ln(a) a g( x) g0 (x) .

£ §
This examples shws dx d
a g( x) = ln(a)a g( x) g0 (x), a companion to the rule
£ §
d
dx e
g ( x)
= e g( x) g0 (x). We will summarize these and chain rule generalizations
of the other rules from this chapter on the bottom of the next page.
282 Derivatives of Inverse Functions and Logarithms

Ø Ø
Example 24.7 Find the derivative of y = ln Ø sin(x)Ø.
(
y = ln | u|
This is a composition, and the function can be broken up as
u = sin(x)
dy d y du 1 1 cos(x)
The chain rule gives = = cos(x) = cos(x) = .
dx du dx u sin(x) sin(x)

Example 24.7 illustrates

Ø Ø
a common pattern, which is to differentiate a
° ¢
function of from ln g(x) or ln g(x) . Let’s redo the example in this setting.
Ø Ø
Ø Ø
Example 24.8 Find the derivative of y = ln Ø g(x)Ø.
(
y = ln | u|
This is a composition, and the function can be broken up as
u = g(x)
dy d y du 1 1 0 g0 (x)
The chain rule gives = = g0 (x) = g (x) = .
dx du dx u g(x) g(x)
Example 24.8 has shown that

d h ØØ Øi 1 d h i g0 (x)
ln g(x)Ø = · g(x) = .
dx g(x) dx g(x)
£ §
This is the chain rule generalization of the rule dxd
ln | x| = 1x , and it is worth
£ ° ¢ § 0
g ( x)
remembering. It implies dx d
ln g(x) = g( x) , and we often us it this way.
° ¢
(Recall
Ø
ln g(x) is not defined when g(x) is negative, so the rule as stated
Ø
for ln g(x)Ø is more all-encompassing.)
Ø
Here is a summary of this chapter’s main rules, along side their chain
rule generalizations. Remember them and internalize them.

Differentiation rules for exponential and log functions

Rule Chain rule generalization
d h xi d h g ( x) i
e = ex e = e g( x) g0 (x)
dx dx
d h xi d h g ( x) i
a = ln(a) a x a = ln(a) a g( x) g0 (x)
dx dx
d h ØØ ØØ i 1 d h ØØ Øi g0 (x)
ln x = ln g(x)Ø =
dx x dx g(x)
d h Ø Ø i 1 d h Ø Ø i g0 (x)
loga Ø xØ = loga Ø g(x)Ø =
dx x ln(a) dx g(x) ln(a)

We prefer the base e, so you should expect to that the formulas for a x and
loga to play less of a role. (Though in computer science, log2 is significant!)
 283

Ø Ø
Example 24.9 Find the derivative of y = ln Ø4x5 + 6x3 + x + 3Ø.
We will do this in two different ways. First we will use the chain rule, and
£ Ø Ø§ g 0 ( x )
then we will use the formula dx ln g(x) = g( x) from the previous page.
d Ø Ø
(
y = ln | u|
Using the chain rule, we first break the function up as
u = 4x5 + 6x3 + x + 3

dy d y du
The chain rule gives =
dx du dx
1° ¢
= 20x4 + 18x2 + 1
u
1 ° 4 2
¢ 20x4 + 18x2 + 1
= 20x +18x +1 = .
4x5 +6x3 + x+3 4x5 + 6x3 + x + 3
Ø £ ا g 0 ( x )
Next, using the formula ln Ø g(x)Ø = g( x) , from the previous page, the
d
dx
d h ØØ 5 Øi 20x4 + 18x2 + 1
answer comes in one step: ln 4x + 6x3 + x + 3Ø = .
dx 4x5 + 6x3 + x + 3
So using the formula is quicker. But you should soon reach the point
where the above to approaches are equally automatic. Doing the chain rule
in your head is in essence using the formula.

Ø ≥p ¥Ø
Ø Ø
Example 24.10 Find the derivative of y = ln Øtan x + 3x Ø. 2
Ø Ø
This has the form of a compositionØ ln Ø g(x)
Ø
Ø, so we can use either the straight
£ §
chain rule or the formula d ln Ø g(x)Ø = 1 g0 (x) from the previous page.
dx g ( x)
Let’s try the formula.

d h ØØ ≥p ¥ Øi
Ø 1 d h ≥p ¥i
ln Øtan x2 +3x Ø = ≥p ¥ tan x2 +3x
dx tan x2 +3x dx
1 ≥p ¥ d hp i
= ≥p ¥ sec2 x2 +3x x2 + 3x
tan x2 +3x dx

1 ≥p ¥1° ¢ 1 °1 d h 2 i
= ≥p ¥ sec2 x2 +3x x2 +3x 2 x +3x
tan x2 +3x 2 dx

1 ≥p ¥1° ¢° 1
= ≥p ¥ sec2 x2 +3x x2 +3x 2 (2x + 3)
tan x2 +3x 2
≥p ¥
(2x + 3) sec2 x2 +3x
= ≥p ¥p
2 tan x +3x2 x2 +3x
 284 Derivatives of Inverse Functions and Logarithms

Exercises for Chapter 24

In exercises 1–20 differentiate the given function.
µ ∂
1 p 1
1. ln(x) + + x + 3 2. ln x +2
x x
ln(w) 1
3. 4.
w x2 + ln(x)
° ¢ ° ¢
5. ln sin3 (x) 6. ln tan(x)
p ° ° ¢¢
7. 5 + ln(ºµ ) + µ 3 8. ln sec x3
° ¢ ° ° ¢ ¢3
9. cos ln(x) 10. sec ln x
p 9 ° ¢
11. º2 + ln(5µ ) + µ 12. ln sec(x3 )
p ° ¢
13. ln(x2 + 1) 3x + 1 14. sec ln(x3 )
° ¢
15. ln xe x 16. ln (x) e x
° ¢ ° ° ¢¢
17. tan ln(x) + x 18. ln sin x3
µ ∂
1 x3 ln(x)
19. ln 1 + 20.
x x3 + 1
ln(3 + h) ° ln(3) 2z ° 8
21. Find lim 22. Find lim
h !0 h z !3 z ° 3

Exercise Solutions for Chapter 24

In exercises 1–20 differentiate the given function.
∑ ∏
d 1 p 1 1 1
1. ln(x) + + x + 3 = ° 2 + p
dx x x x 2 x
∑ ∏ 1
d ln(w) · w ° ln(w) · 1 1 ° ln(w)
3. = w =
dw w w2 w2
d h ° 3 ¢i 1 d h 3 i 1 d h i cos(x)
5. ln sin (x) = sin (x) = 3 sin 2
(x) sin(x) = 3
dx sin3 (x) dx sin3 (x) dx sin(x)

d h p i d h i º 3 1 3p
7. 5 + ln(ºµ ) + µ 3 = 5 + ln(ºµ ) + µ 3/2 = 0 + + µ 1/2 = + µ
dµ dµ ºµ 2 µ 2
° ¢
d h ° ¢i ° ¢1 sin ln(x)
9. cos ln(x) = ° sin ln(x) = °
dx x x
d h p 9 i d h i 5 9 7/2 1 9 p 7
11. º2 + ln(5µ ) + µ = º2 + ln(5µ ) + µ 9/2 = 0 + + µ = + µ
dx dx 5µ 2 µ 2
 285

d h p i d h ip d hp i
13. ln(x2 + 1) 3x + 1 = ln(x2 + 1) 3x + 1 + ln(x2 + 1) 3x + 1
dx dx dx
2x p 3
= 2 3x + 1 + ln(x2 + 1) p
x +1 2 3x + 1
d h ° x¢ i 1 d h xi 1 ° ¢ e x (1 + x) 1 + x
15. ln xe = x xe = x 1 · e x + xe x = =
dx xe dx xe xe x x
° ¢
d h ° ¢ i ° ¢1 sec2 ln(x)
17. 2
tan ln(x) + x = sec ln(x) + 1 = +1
dx x x
∑ µ ∂∏ ∑ ∏ µ ∂
d 1 1 d 1 1 °1 °1
19. ln 1 + = 1
1+ = 1 2
= 2
dx x 1 + x dx x 1+ x x x +x

ln(3 + h) ° ln(3)
21. Find lim . Solution: Let f (x) = ln(x).
h!0 h
ln(x + h) ° ln(x) 1 ln(3 + h) ° ln(3) 1
Then lim = f 0 (x) = . Thus lim = f 0 (3) = .
h!0 h x h !0 h 3