Universal and Existential Quantifiers

A game on the number line: the challenger names xx, and you answer with the bigger point y=x+1y = x + 1 — ∀x ∃y (y>x)\forall x \,\exists y \,(y > x). Swap the quantifier order and the same game collapses. The ε\varepsilon-δ\delta line is exactly this pattern, with δ=ε/2\delta = \varepsilon/2 answering f(t)=2tf(t) = 2t.

∀x ∃y (y>x)\forall x \,\exists y \,(y > x)
¬(∀x ∃y)=∃x ∀y\lnot(\forall x \,\exists y) = \exists x \,\forall y
∀ε>0 ∃δ>0 ∀t (∣t∣<δ→∣2t∣<ε)\forall \varepsilon > 0 \,\exists \delta > 0 \,\forall t \,(|t| < \delta \to |2t| < \varepsilon)

Examples & Exercises

Negate completely the ε\varepsilon-δ\delta definition of continuity of ff at aa — "for all ε>0\varepsilon > 0 there is δ>0\delta > 0 such that for all xx, ∣x−a∣<δ|x-a| < \delta implies ∣f(x)−f(a)∣<ε|f(x)-f(a)| < \varepsilon". State precisely what the negation says about ff.

¬(∀ε>0 ∃δ>0 ∀x (∣x−a∣<δ→∣f(x)−f(a)∣<ε))\lnot\big(\forall\varepsilon{>}0\ \exists\delta{>}0\ \forall x\,(|x-a|<\delta \to |f(x)-f(a)|<\varepsilon)\big)
∃ε>0 ∀δ>0 ∃x (∣x−a∣<δ  ∧  ∣f(x)−f(a)∣≥ε)\exists\varepsilon{>}0\ \forall\delta{>}0\ \exists x\,(|x-a|<\delta \;\land\; |f(x)-f(a)|\ge\varepsilon)
The ε\varepsilon-δ\delta statement (ε=2.5\varepsilon = 2.5, δ=ε/2=1.25\delta = \varepsilon/2 = 1.25), negation chain shown: final step ∃ε>0 ∀δ>0 ∃t (∣t∣<δ∧∣2t∣≥ε)\exists\varepsilon{>}0\ \forall\delta{>}0\ \exists t\,(|t| < \delta \land |2t| \ge \varepsilon).
Show proof
  1. Push the outer negation inward one step at a time, flipping each quantifier (¬∀=∃¬\lnot\forall = \exists\lnot, ¬∃=∀¬\lnot\exists = \forall\lnot). Step 1: ∃ε>0 ¬(∃δ>0 ∀x (∣x−a∣<δ→∣f(x)−f(a)∣<ε))\exists\varepsilon > 0\ \lnot\big(\exists\delta > 0\ \forall x\,(|x-a| < \delta \to |f(x)-f(a)| < \varepsilon)\big). Step 2: ∃ε>0 ∀δ>0 ¬(∀x (∣x−a∣<δ→∣f(x)−f(a)∣<ε))\exists\varepsilon > 0\ \forall\delta > 0\ \lnot\big(\forall x\,(|x-a| < \delta \to |f(x)-f(a)| < \varepsilon)\big).
  2. Step 3: ∃ε>0 ∀δ>0 ∃x ¬(∣x−a∣<δ→∣f(x)−f(a)∣<ε)\exists\varepsilon > 0\ \forall\delta > 0\ \exists x\ \lnot\big(|x-a| < \delta \to |f(x)-f(a)| < \varepsilon\big). Finally use the negation of an implication: ¬(P→Q)≡P∧¬Q\lnot(P \to Q) \equiv P \land \lnot Q (scene 1, problem 3). The final negation is ∃ε>0 ∀δ>0 ∃x (∣x−a∣<δ∧∣f(x)−f(a)∣≥ε)\exists\varepsilon > 0\ \forall\delta > 0\ \exists x\,\big(|x-a| < \delta \land |f(x)-f(a)| \ge \varepsilon\big).
  3. What this says about ff: there is a fixed gap ε>0\varepsilon > 0 such that, no matter how close to aa we look (however small δ\delta is taken), there is always a point xx with f(x)f(x) at least ε\varepsilon away from f(a)f(a) — i.e. ff stays "torn" at aa. This is exactly the definition of "ff is discontinuous at aa".

Prove ∀x∈R ∃y∈R (y>x)\forall x \in \mathbb{R}\ \exists y \in \mathbb{R}\ (y > x) (witness y=x+1y = x+1) and disprove ∃y∈R ∀x∈R (y>x)\exists y \in \mathbb{R}\ \forall x \in \mathbb{R}\ (y > x). Conclusion: quantifier order matters.

∀x ∃y (y>x) is true,∃y ∀x (y>x) is false\forall x\ \exists y\ (y > x)\ \text{is true},\qquad \exists y\ \forall x\ (y > x)\ \text{is false}
The statement ∀x ∃y (y>x)\forall x\,\exists y\,(y > x) (true), challenge x=2x = 2: witness y=3y = 3, and 3>23 > 2 holds.
Show proof
  1. Prove the first statement. Let an arbitrary x∈Rx \in \mathbb{R} be given. Set y=x+1y = x + 1; then y∈Ry \in \mathbb{R} and y−x=1>0y - x = 1 > 0, i.e. y>xy > x. Since xx was arbitrary, such a yy exists for every xx — ∀x ∃y (y>x)\forall x\ \exists y\ (y > x) is true. At the scene's challenge x=2x = 2 the witness is y=3y = 3 and 3>23 > 2 holds.
  2. Disprove the second statement. Whatever candidate y0∈Ry_0 \in \mathbb{R} is proposed, set x=y0+1x = y_0 + 1; then x∈Rx \in \mathbb{R} and y0>xy_0 > x is equivalent to y0>y0+1y_0 > y_0 + 1, i.e. 0>10 > 1, which is false. Hence no y0y_0 satisfies "y0>xy_0 > x for all xx" — ∃y ∀x (y>x)\exists y\ \forall x\ (y > x) is false. In the scene the candidate y=2y = 2 falls to x′=3x' = 3.
  3. Conclusion: quantifier order completely changes the meaning. In ∀x ∃y\forall x\ \exists y the choice of yy may depend on xx (y=x+1y = x+1), but in ∃y ∀x\exists y\ \forall x a single yy must work for every xx. This ordering is exactly why δ\delta in the ε\varepsilon-δ\delta definition is allowed to depend on ε\varepsilon.

Negate "ff is bounded on [0,1][0,1]" — i.e. ∃M ∀x∈[0,1] ∣f(x)∣≤M\exists M\ \forall x \in [0,1]\ |f(x)| \le M — to obtain ∀M ∃x ∣f(x)∣>M\forall M\ \exists x\ |f(x)| > M, then use that form to prove f(x)=1/xf(x) = 1/x is unbounded on (0,1](0,1].

¬(∃M ∀x∈[0,1] (∣f(x)∣≤M)) ≡ ∀M ∃x∈[0,1] (∣f(x)∣>M)\lnot\big(\exists M\ \forall x \in [0,1]\ (|f(x)| \le M)\big)\ \equiv\ \forall M\ \exists x \in [0,1]\ (|f(x)| > M)
x=1∣M∣+2∈(0,1],f(x)=∣M∣+2>Mx = \frac{1}{|M|+2} \in (0,1],\qquad f(x) = |M|+2 > M
The statement ∃y ∀x (y>x)\exists y\,\forall x\,(y > x) (false), negation chain shown: final step ∀y ∃x (y≤x)\forall y\,\exists x\,(y \le x) — the same quantifier pattern as the negation of boundedness.
Show proof
  1. Negate the definition of boundedness: ¬(∃M ∀x∈[0,1] (∣f(x)∣≤M))\lnot\big(\exists M\ \forall x \in [0,1]\ (|f(x)| \le M)\big). Pushing the negation inward gives ∀M ¬(∀x∈[0,1] (∣f(x)∣≤M))\forall M\ \lnot\big(\forall x \in [0,1]\ (|f(x)| \le M)\big), then ∀M ∃x∈[0,1] ¬(∣f(x)∣≤M)\forall M\ \exists x \in [0,1]\ \lnot(|f(x)| \le M), i.e. ∀M ∃x∈[0,1] (∣f(x)∣>M)\forall M\ \exists x \in [0,1]\ (|f(x)| > M). In words: "whatever bound MM is proposed, there is a point xx exceeding it" — that is the definition of unboundedness.
  2. Now apply it to f(x)=1/xf(x) = 1/x (x∈(0,1]x \in (0,1]). Let an arbitrary M∈RM \in \mathbb{R} be given. Set x=1∣M∣+2x = \frac{1}{|M|+2}; the denominator is positive so x>0x > 0, and ∣M∣+2≥2|M|+2 \ge 2 gives x≤12<1x \le \frac12 < 1 — hence x∈(0,1]x \in (0,1]. Then ∣f(x)∣=f(x)=∣M∣+2|f(x)| = f(x) = |M|+2, and since ∣M∣≥M|M| \ge M we have ∣M∣+2>M|M|+2 > M. So ∣f(x)∣>M|f(x)| > M holds at this xx.
  3. Since MM was arbitrary, for every MM there is an x∈(0,1]x \in (0,1] with ∣f(x)∣>M|f(x)| > M — i.e. ∀M ∃x (∣f(x)∣>M)\forall M\ \exists x\ (|f(x)| > M) holds, so ff is unbounded on (0,1](0,1]. The key is the quantifier order: because MM is given first (∀M\forall M), the witness xx may be chosen afterwards (∃x\exists x), which is why the MM-dependent choice x=1/(∣M∣+2)x = 1/(|M|+2) is legitimate.

1. Foundations of Proof