Vous êtes sur la page 1sur 24

f (g(y)) may not have a limit, even if f and g have limits.

Separately, each of these functions have limits at all x and y:


 
1 if x = 0, y sin 1 if y 6= 0,
y
f (x) = g(y) =
0 if x =


6 0, 0 if y = 0,




but lim f (g(y)) doesn’t exist:


y→0


*


'


$


!






1 if y = 0, ± 1 , ± 1 , . . .
π 2π
f (g(y)) =
0 other y 0s. %&

()

+,






"#




/0









Why doesn’t f (g(x)) have a limit? Consider the deltas and
epsilons:
lim f (x) = L would require that
x→c

For any 1 > 0, there is a δ1 > 0 (depending on 1, f and c)


such that 0 < | x − c | < δ1 would imply | f (x) − L | < 1.

If x were replaced by g(y), this would become

For any 1 > 0, there is a δ1 > 0 (depending on 1, f and c)


such that 0 < | g(y) − c | < δ1
would imply | f (g(y)) − L | < 1,
(In particular, this would need g(y) 6= c.)
On the other hand, lim g(y) = c would mean that
y→d

For our 2 = δ1 > 0 there is a δ2 > 0


(depending on δ1, g and d)
such that 0 < | y − d | < δ2
would imply only that | g(y) − c | < δ1 .
(This would not give g(y) 6= c.)

(Remember: The f -limit needs 0 < | g(y) − c | < δ1 ,


in particular, that g(y) 6= c.)

Could a change in the properties of f or of g


make these boxed inequalities match?
Yes, if f were continuous, or if g were strictly monotone.
A Limit, lim f (x) = L, in Terms of δ ’s and  ’s
x→c

For any  > 0, there is a δ > 0 (depending on , f and c)


such that 0 < | x − c | < δ would imply | f (x) − L | < .
↑ This requires that x 6= c.
Continuity, lim f (x) = f (c), in Terms of δ ’s and  ’s
x→c

For any  > 0, there is a δ > 0 (depending on , f and c)


such that | x − c | < δ would imply | f (x) − f (c) | < .
↑ This doesn’t.
This definition does not require 0 < | x − c |.
Even if we had 0 = | x − c |, we would then have x = c,
which would imply that f (x) = f (c)
and | f (x) − f (c) | = 0 < , anyway.
Suppose now that f were continuous at x = c :
lim f (x) = f (c) would now mean that:
x→c
For any 1 > 0, there would be a δ1 > 0
such that | x − c | < δ1 would imply | f (x) − f (c) | < 1.
If x were replaced by g(y), this would become:
For any 1 > 0, there would be a δ1 > 0 (depending on 1, f and c)
such that | g(y) − c | < δ1 implies: | f (g(y)) − f (c) | < 1,
while lim g(y) = c would still mean that:
y→d

For our δ1 > 0 there is a δ2 > 0


such that 0 < | y − d | < δ2 implies | g(y) − c | < δ1 .
The connecting, boxed inequalities would now match.
If f is continuous at x = c
and if lim g(y) equals this c,
y→d±
  
then lim f g(y) equals f lim g(y) equals f (c).
y→d± y→d±

Proof: Combine
f (c + o(1)) = f (c) + o(1), meaning lim f (x) = f (c),
x→c
with g(y) = c + o(1), meaning lim g(y) = c,
y→d
 
to obtain f g(y) = f c + o(1) = f (c) + o(1). 
Notes: This works whether or not g is continuous at y = d.
The ±’s above mean that the limits in y can be
all left-handed, all right-handed, or all two-sided.
If f is continuous at x = c
and if lim g(y) equals this c,
y→d±
  
then lim f g(y) equals f lim g(y) equals f (c).
y→d± y→d±

Examples:
√ p r √
for f (x) = x, lim g(y) = lim g(y) = c;
y→d y→d

for f (x) = ln x, lim ln g(y) = ln lim g(y) = ln c;


y→d+ y→d+

for f (x) = ex, lim eg(y) = elimy→d− g(y) = ec ;


y→d−
 2
for f (x) = x2, lim g(y)2 = lim g(y) = c2.

y→d± y→d±
If f is continuous at g(d): lim f (x) = f (g(d)),
x→g(d)

and if g is continuous at d: lim g(y) = g(d),


y→d

then f(g(y)) is also continuous at y = d.


and we have lim f (g(y)) = f (g(d)).
y→d

Proof: Combine
y = d + o(1), (since y → d)

g(y) = g d + o(1) = g(d) + o(1), (g is continuous)
 
and f g(d) + o(1) = f g(d) + o(1),
to obtain (since f is continuous)
    
f g(y) =f g d + o(1) =f g(d) + o(1) = f g(d) + o(1).
If f is continuous at g(d): lim f (x) = f (g(d)),
x→g(d)

and if g is continuous at d: lim g(y) = g(d),


y→d

then f (g(y)) is also continuous at y = d.


and we have lim f (g(y)) = f (g(d)).
y→d

Examples:
√ p r p
for f (x) = x, lim g(y) = lim g(y) = g(d);
y→d y→d
 2
for f (x) = x2, lim g(y) 2
= g(d)2.

= lim g(y)
y→d y→d
1 1 1 1
for f (x) = , lim = = ,
x y→d g(y) lim g(y) g(d)
y→d
if g(d) 6= 0.
Now, let’s return to assuming only that f has a right-hand limit
at x = c :
f (c+) = lim f (x) = L would require that
x→c+

For any 1 > 0, there is a δ1 > 0 (depending on 1, f and c)


such that 0 < x − c < δ1 would imply | f (x) − L | < 1.

If x were replaced by g(y), this would become

For any 1 > 0, there is a δ1 > 0 (depending on 1, f and c)


such that 0 < g(y) − c < δ1
would imply | f (g(y)) − L | < 1,
(In particular, this would need g(y) > c.)
On the other hand, lim g(y) = c would yield
y→d+

For our 2 = δ1 > 0 there is a δ2 > 0


(depending on δ1, g and d)
such that 0 < | y − d | < δ2
would imply only that | g(y) − c |< δ1 .
This alone would not be the same as 0 < g(y) − c < δ1 .

Now, if we also had g increasing for y near d+, with y > d,


g(y) would be greater than g(d+) = c
and we would have 0 < g(y) − c < δ1 ,
exactly as the right-limit for f requires:
0 < g(y) − c < δ1
A One-Sided Limit for Composite Functions:

If g(y) is increasing as y approaches d+,

if c = lim g(y) exists, and if lim f (x) also exists,


y→d+ x→c+

then lim f (g(y)) equals this lim f (x).


y→d+ x→c+

Example:
p p 
for g(y) = y − d, lim f y − d = lim f (x).
y→d+ x→0+

This theorem works


whenever the one-sided limits c = g(d+) and f (c+) exist,
whether or not f is continuous,
and whether or not f (x) is even defined at x = c.
Another One-Sided Limit for Composite Functions:

If g(y) is increasing as y approaches d±,

if c = lim g(y) exists, and if lim f (x) also exists,


y→d± x→c±

then lim f (g(y)) equals this lim f (x).


y→d± x→c±

And Still Another One-Sided Limit for Composite Functions:

If g(y) is decreasing as y approaches d±,

if c = lim g(y) exists, and if lim f (x) also exists,


y→d± x→c∓

then lim f (g(y)) equals this lim f (x).


y→d± x→c∓
If g(y) were increasing and, moreover, continuous
on an interval on either(±) side of y = d,
then the function g would have an inverse function h there,
such that h(g(y)) = y and g(h(x)) = x.
y
Assume, as before, that lim g(y) = c. h (x
) x=
y→d±
y= g(y
Then we also have lim h(x) = d. )
x→c±


(c, d)
x
We have seen that the existence of lim f (x)
x→c±
would imply the existence of lim f (g(y)),
y→d±
and their equality: lim f (g(y)) = lim f (x).
y→d± x→c±

Conversely, if lim f (g(y)) were known to exist, we could have


y→d±
lim f (x) = lim f (g(h(x))) = lim f (g(y)) similarly.
x→c± x→c± y→d±
Two Equivalences for One-Sided Limits of Composite Functions:

If g(y) is increasing and continuous as y approaches d±, with


c = lim g(y), then the limits lim f (x) and lim f (g(y))
y→d± x→c± y→d±

either both exist, and are equal, or both limits do not exist.

If g(y) is decreasing and continuous as y approaches d±, with


c = lim g(y), then the limits lim f (x) and lim f (g(y))
y→d± x→c∓ y→d±

either both exist, and are equal, or both limits do not exist.

Note that g does not have to be defined or continuous at y = d.


c = lim g(y) = g(d±) would not have to equal any g(d).
y→d±
An Equivalence for Two-Sided Limits of Composite Functions:

If g(y) is monotone and continuous for d − δ < y < d + δ.



Then the limits lim f (x) and lim f g(y)
x→g(d) y→d

either both exist, and are equal, or both limits do not exist.

x = g(y),
These equivalences justify limit substitutions:
y = g −1(x),
 
lim f g g −1(x)
 
lim f g(y) = = lim f (x),
y→d x→lim g(y) x→g(d)
y→d
 
lim f g(y) = lim −1 f g(y) = lim f (x).
y→d y→ lim g (x) x→g(d)
x→g(d)

x = g(y),
These equivalences justify limit substitutions:
y = g −1(x),
−1
 
lim f g(y) = lim f g g (x) = lim f (x).
y→d x→lim g(y) x→c
 y→d
x = g(y) = y 2,
Example: Using √ we have
y = g −1(x) = x,

2
√ 2
lim
√ cos y = lim 2 cos x = lim cos x = −1,
y→ π x→ lim
√ y
x→π
y→ π

√ p
lim2 sin x = lim √ sin y 2 = lim sin y = 0.
x→π y→ lim x y→π
x→π 2

Example: n = m + 1,
Using lim am+1 = lim an = lim an
m = n − 1, m→∞ n→ lim m+1 n→∞
m→∞
If g(y) is monotone and continuous for d − δ < y < d + δ.

Then the limits lim f (x) and lim f g(y)
x→g(d) y→d

either both exist, and are equal, or both limits do not exist.

Warning: If the function g were not monotone on both sides of


y = d, lim f (x) could imply lim f g(y), but not ⇐=.
x→g(d) y→d

Consider g(y) = |y| on both sides of y = 0, .


Existence of lim f (x) implies existence of lim f (|y|),
x→0 y→0
but existence of lim f (|y|) does not involve f (negative x).
y→0
However, lim f (x) = lim f (|y|),
x→0+ y→0+
since |y| is increasing for y > 0.

h = g(u),
These equivalences justify limit substitutions:
u = g −1(h),
−1
 
lim F g(u) = lim F g g (h) = lim F (h).
u→d h→lim g(u) h→c
u→d


h = g(u) = u − x,
Example: With x a constant and using
u = g −1(h) = x + h,

f (u) − f (x) f (x + h) − f (x)


we have lim = lim
u→x u−x h→lim (u−x)
u→x
(x + h) − x
⇑ ⇓
f (x + (u − x)) − f (x) f (x + h) − f (x)
lim = lim
u→lim (x+h)
h→0
u−x h→0 h
Substitution as an Alternative to Factoring

x = y + 3
If we let , we have
y = x − 3

x2 − 9 (y + 3)2 − 9
lim = lim
x→3 x2 −x−6 y→lim (x−3)
x→3
(y + 3)2 − (y + 3) − 6

y 2 + 6y + 9 − 9
= lim
y→0 y 2 + 6y + 9 − y − 3 − 6
y 2 + 6y
= lim
y→0 y 2 + 5y
y+6 6
= lim = .
y→0 y+5 5
xp − 1
lim = p, for any integer p
x→1 x−1

Proof:
If p > 0, we start with
(x − 1)(xp−1 +xp−2 + . . . +x +1)
=xp +xp−1 + . . . +x2 +x
−xp−1−xp−2− . . .−x −1
=xp −1.
The desired limit now becomes
xp − 1
lim = lim xp−1 + xp−2 + . . . + x + 1 (p terms)
x→1 x − 1 x→1

= 1 + 1 +...+ 1 + 1 (p ones)
= p.
For p = 0, the limit is much simpler:
xp − 1 x0 − 1 1−1
lim = lim = lim = lim 0 = 0 = p.
x→1 x − 1 x→1 x − 1 x→1 x − 1 x→1
 p
1 1

p −1 x = ,

x −1

u u
For p < 0, lim = lim , using
x→1 x − 1 u→1 1 1
−1 u = ,


u x

u−p − 1 u−p − 1
= lim u = lim (−u) = (−1)(−p) = p,
u→1 1−u u→1 u−1

by an application of the first part, since −p is positive. 


p
y −1
q p p
lim = , for any rational .
y→1 y−1 q q

Proof: If we use the previous result, we have


p p
y −1
q y −1
q
lim = lim q
y→1 y−1 y→1 y −1
q

xp − 1  y = xq ,
= lim , substituting
x→1 xq −1 x = y q1 ,
xp − 1
lim
x→1 x−1 p
= = . 
xq − 1 q
lim
x→1 x−1
d p p p
−1 p
x q = x q , unless (x = 0 and ≤ 1).
dx q q

Proof: For nonzero x, the difference quotient has the limit!


p
p uq
p p p p xq p − 1
(x + h) q − x q uq − xq xq
lim = lim = lim  ,
u

h→0 h u→x u − x u→x
u x −1
so, if we substitute v for , we have x
x
p p p
x q v −1
q p
−1 vq − 1 p
−1 p
= · lim u = xq lim = xq .
x v→lim u→x x
v−1 v→1 v − 1 q
p
When x equals zero, and when is greater than 1, we can verify
p p
q
(0 + h) q − 0 q p
−1 p p −1
that lim = lim h q = 0 = ·0 q . 
h→0 h h→0 q

Vous aimerez peut-être aussi