4.5: Monotone Function
( \newcommand{\kernel}{\mathrm{null}\,}\)
A function f:A→E∗, with A⊆E∗, is said to be nondecreasing on a set B⊆A iff
x≤y implies f(x)≤f(y) for x,y∈B.
It is said to be nonincreasing on B iff
x≤y implies f(x)≥f(y) for x,y∈B.
Notation: f↑ and f↓( on B), respectively.
In both cases, f is said to be monotone or monotonic on B. If f is also one to one on B (i.e., when restricted to B), we say that it is strictly monotone (increasing if f↑ and decreasing if f↓).
Clearly, f is nondecreasing iff the function −f=(−1)f is nonincreasing. Thus in proofs, we need consider only the case f↑. The case f↓ reduces to it by applying the result to −f.
If a function f:A→E∗(A⊆E∗) is monotone on A, it has a left and a right (possibly infinite) limit at each point p∈E∗.
In particular, if f↑ on an interval (a,b)≠∅, then
f(p−)=supa<x<pf(x) for p∈(a,b]
and
f(p+)=infp<x<bf(x) for p∈[a,b).
(In case f↓, interchange "sup" and "inf.")
- Proof
-
To fix ideas, assume f↑.
Let p∈E∗ and B={x∈A|x<p}. Put q=supf[B] (this sup always exists in E∗; see Chapter 2, §13). We shall show that q is a left limit of f at p (i.e., a left limit over B).
There are three possible cases:
(1) If q is finite, any globe Gq is an interval (c,d),c<q<d, in E1. As c<q=supf[B],c cannot be an upper bound of f[B] (why?, so c is exceeded by some f(x0),x0∈B. Thus
c<f(x0),x0<p.
Hence as f↑, we certainly have
c<f(x0)≤f(x) for all x>x0 (x∈B).
Moreover, as f(x)∈f[B], we have
f(x)≤supf[B]=q<d,
so c<f(x)<d; i.e., f(x)∈(c,d)=Gq.
We have thus shown that
(∀Gq)(∃x0<p)(∀x∈B|x0<x)f(x)∈Gq,
so q is a left limit at p.
(2) If q=+∞, the same proof works with Gq=(c,+∞]. Verify!
(3) If q=−∞, then
(∀x∈B)f(x)≤supf[B]=−∞,
i.e., f(x)≤−∞, so f(x)=−∞ (constant) on B. Hence q is also a left limit at p (§1, Example (a)).
In particular, if f↑ on A=(a,b) with a,b∈E∗ and a<b, then B= (a,p) for p∈(a,b]. Here p is a cluster point of the path B (Chapter 3, §14, Example (h)), so a unique left limit f(p−) exists. By what was shown above,
q=f(p−)=supf[B]=supa<x<pf(x), as claimed.
Thus all is proved for left limits.
The proof for right limits is quite similar; one only has to set
B={x∈A|x>p},q=inff[B].◻
Note 1. The second clause of Theorem 1 holds even if (a,b) is only a subset of A, for the limits in question are not affected by restricting f to (a,b). (Why?) The endpoints a and b may be finite or infinite.
Note 2. If Df=A=N (the naturals), then by definition, f:N→E∗ is a sequence with general term xm=f(m),m∈N (see §1, Note 2). Then setting p=+∞ in the proof of Theorem 1, we obtain Theorem 3 of Chapter 3, §15. (Verify!)
The exponential function F:E1→E1 to the base a>0 is given by
F(x)=ax.
It is monotone (Chapter 2, §§11-12, formula (1)), so F(0−) and F(0+) exist. By the sequential criterion (Theorem 1 of §2), we may use a suitable sequence to find F(0+), and we choose xm=1m→0+. Then
F(0+)=limm→∞F(1m)=limm→∞a1/m=1
(see Chapter 3, §15, Problem 20).
Similarly, taking xm=−1m→0−, we obtain F(0−)=1. Thus
F(0+)=F(0−)=limx→0F(x)=limx→0ax=1.
(See also Problem 12 of §2.)
Next, fix any p∈E1. Noting that
F(x)=ax=ap+x−p=apax−p,
we set y=x−p. (Why is this substitution admissible?) Then y→0 as x→p, so we get
limx→pF(x)=limap⋅limx→pax−p=aplimy→0ay=ap⋅1=ap=F(p).
As limx→pF(x)=F(p),F is continuous at each p∈E1. Thus all exponentials are continuous.
If a function f:A→E∗(A⊆E∗) is nondecreasing on a finite or infinite interval B=(a,b)⊆A and if p∈(a,b), then
f(a+)≤f(p−)≤f(p)≤f(p+)≤f(b−),
and for no x∈(a,b) do we have
f(p−)<f(x)<f(p) or f(p)<f(x)<f(p+);
similarly in case f↓ (with all inequalities reversed).
- Proof
-
By Theorem 1, f↑ on (a,p) implies
f(a+)=infa<x<pf(x) and f(p−)=supa<x<pf(x);
thus certainly f(a+)≤f(p−). As f↑, we also have f(p)≥f(x) for all x∈ (a,p); hence
f(p)≥supa<x<pf(x)=f(p−).
Thus
f(a+)≤f(p−)≤f(p);
similarly for the rest of (1).
Moreover, if a<x<p, then f(x)≤f(p−) since
f(p−)=supa<x<pf(x).
If, however, p≤x<b, then f(p)≤f(x) since f↑. Thus we never have f(p−)<f(x)<f(p). similarly, one excludes f(p)<f(x)<f(x)<f(p+). This completes the proof. ◻
Note 3. If f(p−),f(p+), and f(p) exist (all finite), then
|f(p)−f(p−)| and |f(p+)−f(p)|
are called, respectively, the left and right jumps of f at p; their sum is the (total) jump at p. If f is monotone, the jump equals |f(p+)−f(p−)|.
For a graphical example, consider Figure 14 in §1. Here f(p)=f(p−) (both finite ), so the left jump is 0. However, f(p+)>f(p), so the right jump is greater than 0. Since
f(p)=f(p−)=limx→p−f(x),
f is left continuous (but not right continuous) at p.
If f:A→E∗ is monotone on a finite or infinite interval (a,b) contained in A, then all its discontinuities in (a,b), if any, are "jumps," that
is, points p at which f(p−) and f(p+) exist, but f(p−)≠f(p) or f(p+)≠f(p).
- Proof
-
By Theorem 1, f(p−) and f(p+) exist at each p∈(a,b).
If, in addition, f(p−)=f(p+)=f(p), then
limx→pf(x)=f(p)
by Corollary 3 of §1, so f is continuous at p. Thus discontinuities occur only if f(p−)≠f(p) or f(p+)≠f(p).◻