This section develops continuity and the intermediate value theorem. It assumes the least-upper-bound property of the real numbers and the – definition of continuity.
A value cannot be skipped on an interval
Definition 3.5.1
(Interval)
A set is an interval if it contains every point between any two of its members: and imply .
The first result says that a continuous real function on a closed interval cannot jump over a value. The idea of the proof is to collect the points whose function value is still at most , then study the rightmost possible boundary of that collection. The collection need not itself have a largest element; completeness supplies a supremum instead.
Theorem 3.5.2
(Intermediate value theorem)
Let be continuous and suppose lies between and . Then there is with . If lies strictly between the endpoint values, then can be taken in .
Proof
(The supremum of the sublevel set)
Assume first . Set
The set is nonempty because , and is bounded above by . To justify the value at , for each positive integer choose with ; this uses the defining property of a supremum. Then , so continuity gives .
If , continuity at provides such that whenever and . Since , we know ; choose with . This lies in , contradicting that is an upper bound. Hence . The endpoint inequalities show . If the endpoint values occur in the opposite order, apply the argument to and ; if equals either endpoint value, that endpoint itself is a solution.
Proof X-ray
Where completeness and continuity enter
The argument has three load-bearing points. First, exists only because the real numbers have the least-upper-bound property; nothing else produces a candidate point. Second, continuity is used twice and in two different directions. Along the sequence taken from inside , it passes the inequality to the limit and gives . At itself, it propagates a strict inequality to a whole neighborhood and so pushes points of to the right of . Third, the hypothesis guarantees there is room to the right of in which to place .
The proof closes by trichotomy: was shown directly, and leads to a contradiction, so .
To see the first use of continuity break, let on and on , with . Then and , but : the inequality did not pass to the limit because is not continuous at .
The role of completeness deserves its own test: the statement fails on a number system that has gaps, even for a very tame function.
Counterexample mode
Completeness cannot be dropped
Claim: a continuous function on an interval of rational numbers takes every value between two of its values. This is false. Let on . It is continuous, with . A rational zero would be in lowest terms with . Then is even, say , so and is even as well, contradicting lowest terms. Hence never takes the value on . The real zero lies between and , where , but it is not a point of the domain. The proof above needed to be a point of the domain, which the least-upper-bound property provides in but not in .
Notice what has been proved: at least one solution exists. The theorem says neither that the solution is unique nor how quickly an approximation converges.
Common mistake
(Checking only the endpoints of a function that is not continuous on the whole interval)
Let on . Then , yet has no zero. The theorem needs continuity at every point of , not merely a sign change between the endpoint values.
Common mistake
(A zero need not be a sign change)
The function has the zero on , but , so no sign change at the endpoints detects it. The intermediate value theorem gives a sufficient condition for a zero, never a necessary one.
Continuous images of intervals
Proposition 3.5.3
(The continuous image of an interval is an interval)
If is an interval and is continuous, then is an interval.
Proof
(Between two values of the image)
Take with . Choose with and . The entire closed interval between and lies in , so the intermediate value theorem shows that every belongs to .
Continuity alone does not make the image closed or bounded: maps to , which is not closed, while maps to the unbounded interval . For a continuous function on a closed bounded interval, the extreme value theorem supplies a minimum and maximum ; combining that theorem with intermediate values gives . Both the interval structure and compactness have a job in that conclusion.
Quick check
(Is the image of on the open interval an open interval?)
Find the largest value and decide whether it is attained.
Solution · Answer
No. For we have , and is attained at . For the equation has the root . So the image is , an interval that is neither open nor closed.
Worked examples
Worked example 3.5.4
(A root and an explicit bracket)
Show that has exactly one real solution and locate it in . Let . It is continuous, , and . The intermediate value theorem gives a root .
For ,
The quadratic factor is positive because . Thus is strictly increasing, so two distinct roots are impossible. Existence came from continuity and the sign change; uniqueness came from strict increase. Neither step should be silently substituted for the other.
To continue bisection, the midpoint is and , giving the smaller bracket . At each step, continuity and opposite endpoint signs guarantee that the retained half still contains a root. A bracket after halvings has length times the original length; its midpoint is within half that length of the root.
Worked example 3.5.5
(A fixed point of the cosine function)
Show that has exactly one solution in . Put , which is continuous. Then and , because . The intermediate value theorem gives a zero in .
Uniqueness: on the cosine is strictly decreasing and is strictly decreasing, so is strictly decreasing as a sum of two strictly decreasing functions. Hence cannot vanish twice. As a numerical check, and , which places the root in .
Worked example 3.5.6
(An image that is an interval but not closed)
Let on . By the previous result the image is an interval. Since for , the image lies in . Every is attained at , so . The minimum is attained, but the supremum is not, because is not in the domain. This is why the closed-and-bounded hypothesis is needed for .
Strictly monotone functions on intervals have continuous inverses
Suppose is strictly increasing on an interval . Strict increase gives injectivity, and the definition of gives surjectivity onto ; hence exists. The inverse is strictly increasing.
Theorem 3.5.7
(Continuity of the inverse of a strictly monotone function)
Let be an interval and strictly monotone. Then is a bijection of onto , and is continuous relative to . Continuity of is not assumed.
Proof
(Gaps on both sides of the point)
Let be strictly increasing. Fix and . If contains a point below , choose with : start with any in and use a point between and sufficiently close to . Similarly, if contains a point above , choose with . For each chosen point, strict increase makes the corresponding gap or positive. Take to be the minimum of the gaps that exist; if neither exists, and continuity is immediate.
For with , a chosen must satisfy , since would imply . If there is no such , every is at least . The corresponding upper argument gives if exists, and otherwise . Thus . This is exactly continuity of at relative to . If is strictly decreasing, apply the increasing case to ; since , the inverse is again continuous relative to .
Proof X-ray
What the interval hypothesis buys
The proof never differentiates or uses continuity of . Strict monotonicity is used twice: to make the gaps and positive, so that is a genuine positive number, and to turn the inequality back into . The interval property of is used in the one non-obvious step, the choice of and arbitrarily close to : because contains every point between and , it has points arbitrarily near on that side. Without that, the nearest points of below or above could be far away, and no could squeeze close to .
The argument closes because every with within of is squeezed between and , hence within of .
The hypothesis on the domain is not decoration, as the following example with a disconnected domain shows.
Counterexample mode
Strictly increasing and continuous, but the inverse is not
Claim: if is continuous and strictly increasing on its domain, its inverse is continuous. This fails when the domain is not an interval. Let , and put on and on . This is continuous relative to and strictly increasing, with image . Its inverse satisfies for , but . Thus as , so the inverse is discontinuous at . The hole in is invisible in the image, which is why the interval assumption on the domain is needed.
Continuity of has a different role from the one in the theorem: it ensures that is an interval. For example, for and for is strictly increasing on , yet its image is not an interval. Its inverse is nevertheless continuous relative to that image.
Practice with solutions
Exercise 3.5.8
(Let be continuous. Prove that for some .)
CoreProof
Hint(1)
Look at the sign of at the two endpoints.
Solution(3)
- 1
Reduce to a zero of one function
Define . It is continuous on , and exactly when .
- 2
Compare the endpoint signs
Key step
and .
Because: Since maps into .
- 3
Apply the intermediate value theorem
∎If either endpoint value is zero, that endpoint works. Otherwise and the theorem gives with . This conclusion does not assert uniqueness: makes every point a solution.
Exercise 3.5.9
(Let be continuous on , with . Show that some satisfies .)
CoreProof
Solution(3)
- 1
Form the difference function
is continuous on .
- 2
Relate the endpoint values
Key step
.
Because: The hypothesis gives .
- 3
Conclude
∎If , take ; otherwise the endpoint values have opposite signs, so the intermediate value theorem gives with , that is, .
Quick check
(A continuous has and . Must have exactly one solution? Must its range be ?)
Test with a function that oscillates.
Solution · Answer
There is at least one zero, but neither stronger statement follows. Take . Its endpoint values are and , and , , , , . Intermediate values give zeros in each of , and , while the range contains and as well. The endpoint data do not identify the extrema.