You see (3, 3) on a worksheet and freeze. The two endpoints are equal. Is this even legal notation? If it is, what numbers does it contain — zero, or one, or is this a typo the teacher missed?
Here is the short answer, and it is the one to lock in first:
(3, 3) is valid, legal, unambiguous notation. It denotes the empty set. No real number belongs to it.
The notation is fine. The set is just empty. Those two facts do not contradict — an empty set is still a set, and writing down a notation that names the empty set is no more strange than writing \{x \in \mathbb{R} : x^2 = -1\} or \{x : x \ne x\}. The name is well-formed; the thing it names happens to have nothing in it.
Unpack the definition
Every open interval (a, b) is defined as
Plug in a = b = 3. You get
Read the condition aloud: "x is strictly greater than 3 and x is strictly less than 3." Can any real number satisfy both at once? No. If x > 3 then x \ne 3 and x is bigger than 3, so x cannot also be smaller than 3. The condition is self-contradictory for every real x.
A set defined by a contradiction contains nothing. So
That is the entire derivation. The notation parses; the set is empty because the defining condition is unsatisfiable.
The four cases at a = b = 3, side by side
The interesting thing is how sensitive this is to the brackets. Swap one round bracket for a square bracket and the answer can change dramatically. All four cases with a = b = 3:
| Notation | Defining condition | Elements | Size |
|---|---|---|---|
| [3, 3] | 3 \le x \le 3 | \{3\} | one element |
| (3, 3) | 3 < x < 3 | \varnothing | empty |
| [3, 3) | 3 \le x < 3 | \varnothing | empty |
| (3, 3] | 3 < x \le 3 | \varnothing | empty |
Only the fully closed interval [3, 3] contains anything — exactly the point 3 itself, because 3 \le 3 and 3 \le 3 both hold. The other three cases all require at least one strict inequality that the point 3 cannot satisfy against itself, so each is empty.
Why [3, 3] is a point and (3, 3) is empty
The difference between [3, 3] and (3, 3) is a single-character change, but the sets they name are as different as they can possibly be: one has one element, the other has zero. Why is the behaviour so sharp?
Think of the brackets as two independent questions about the endpoints:
- "Is the left endpoint a allowed in the set?" — [ says yes, ( says no.
- "Is the right endpoint b allowed in the set?" — ] says yes, ) says no.
When a < b the two endpoints are different points, and each question has its own answer. The "body" of the interval — everything strictly between — does not care about either bracket.
When a = b the two endpoints are the same point, and the two questions collapse into one: "Is the single point 3 allowed?" For 3 to survive, the answer must be yes on both sides simultaneously. That only happens when both brackets are square. Any round bracket kicks the only candidate out, leaving nothing.
When a = b there is no "body" — nothing strictly between 3 and 3 — so the endpoint is the only thing the interval could contain. Exclude it, and \varnothing is what remains.
A set can be named without being inhabited
A source of confusion is the feeling that writing (3, 3) is somehow asking a question with no answer. But the empty set is a genuine, well-defined object: it has a name (\varnothing), it is a subset of every set, and it can be produced by perfectly legal set-builder expressions.
The notation (3, 3) is one such producer. It is the set-builder expression \{x : 3 < x < 3\} under a compact name. The fact that no element satisfies the condition does not invalidate the name any more than "the set of cricketers who bowled left-handed and right-handed in the same delivery" — a perfectly clear description that happens to pick out no one.
If you want more on this distinction, see Can an Interval Be Empty — and When?.
When does (a, b) with a = b actually show up?
It seems artificial, but coincident endpoints arise naturally in two places:
Solving inequalities that turn out to have no solution. If you solve |x - 3| < 0, the translation gives -0 < x - 3 < 0, which is 3 < x < 3, i.e., (3, 3) = \varnothing. And the answer is correct — |x - 3| is always \ge 0, so |x - 3| < 0 has no solutions.
Limiting arguments. If you shrink an interval (3 - \varepsilon, 3 + \varepsilon) by letting \varepsilon \to 0, the open intervals shrink towards (3, 3) = \varnothing. Meanwhile the closed intervals [3 - \varepsilon, 3 + \varepsilon] shrink towards [3, 3] = \{3\}. The fact that one limit is empty and the other is a single point is a hint at why closed intervals are the ones that have nice limiting behaviour (they are compact; open ones are not).
The self-test
Write the set described by each notation.
- [7, 7] → \{7\}
- (7, 7) → \varnothing
- (7, 7] → \varnothing
- [7, 7) → \varnothing
- (5, 2) → \varnothing (when a > b the condition 5 < x < 2 is also unsatisfiable; this is sometimes left undefined, but the set-builder reading is empty)
If all five felt automatic, you own this distinction. If any felt like a guess, re-read the rule at the top and rerun the set-builder translation.
Back to the main article: Intervals and Inequalities Preview.