rational_quadratic_coloring_from_integer_scales
If for every positive integer q there is an integer coloring avoiding false length-3 and true length-17 quadratics with leading coefficient q and arbitrary integer lower coefficients, then a rational coloring avoids those two types with leading coefficient one and arbitrary rational lower coefficients.
Statement
theorem rational_quadratic_coloring_from_integer_scales (h : ∀ q : ℕ, 0 < q → ∃ f : ℤ → Bool, (∀ a b : ℤ, ¬ ∀ k : Fin 3, f ((q : ℤ) * (k.val : ℤ) ^ 2 + a * k.val + b) = false) ∧ (∀ a b : ℤ, ¬ ∀ k : Fin 17, f ((q : ℤ) * (k.val : ℤ) ^ 2 + a * k.val + b) = true)) : ∃ f : ℚ → Bool, (∀ a b : ℚ, ¬ ∀ k : Fin 3, f ((k.val : ℚ) ^ 2 + a * k.val + b) = false) ∧ (∀ a b : ℚ, ¬ ∀ k : Fin 17, f ((k.val : ℚ) ^ 2 + a * k.val + b) = true)- Source
- Compactness reduction for the rational quadratic approach to OP-1482.
- Verified
- 13 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext
The proof, explained
To obtain a rational coloring avoiding false length-3 and true length-17 quadratic progressions with leading coefficient one, the problem is formulated as finding a point in the intersection of closed constraint sets in the compact product space ℚ → Bool. For any finite collection of rational constraints, clearing denominators by setting q to the product of all coefficient denominators allows us to scale a hypothesis integer coloring with leading coefficient q into a rational coloring that simultaneously satisfies every constraint in the collection. By compactness and the finite intersection property, the total intersection of all constraint sets is nonempty, yielding a global rational coloring.
For q ∈ ℕ and a ∈ ℚ with a.den ∣ q, the integer product (q / a.den) * a.num equals q * a in ℚ.
Since the denominator of a divides q, we have q = a.den * (q / a.den). Multiplying (q / a.den) by a.num, and rewriting a.num as a * a.den, ring arithmetic rearranges the expression to ((q / a.den) * a.den) * a = q * a.
uses
Nat.cast_mul,Nat.mul_div_cancel',Int.cast_natCast,Rat.mul_den_eq_num,mul_commprivate lemma rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd (q : ℕ) (a : ℚ) (h : a.den ∣ q) : (((q / a.den : ℕ) : ℤ) * a.num : ℚ) = (q : ℚ) * a := by have hq : (q : ℚ) = (a.den : ℚ) * ((q / a.den : ℕ) : ℚ) := by rw [← Nat.cast_mul, Nat.mul_div_cancel' h] calc (((q / a.den : ℕ) : ℤ) * a.num : ℚ) = ((q / a.den : ℕ) : ℚ) * (a.num : ℚ) := by simp only [Int.cast_natCast] _ = ((q / a.den : ℕ) : ℚ) * (a * (a.den : ℚ)) := by rw [← Rat.mul_den_eq_num a] _ = (((q / a.den : ℕ) : ℚ) * (a.den : ℚ)) * a := by ring _ = (q : ℚ) * a := by rw [mul_comm ((q / a.den : ℕ) : ℚ) (a.den : ℚ), ← hq]
If integers A and B represent q * a and q * b in ℚ, then the numerator of q * (k² + a k + b) is the integer q k² + A k + B.
Let J be the integer q k² + A k + B. Expanding q * (k² + a k + b) and substituting A = q * a and B = q * b shows that q * (k² + a k + b) equals J as rational numbers. Taking numerators on both sides and using the fact that the numerator of an integer cast is itself gives the identity.
uses
congrArg,Rat.num_intCastprivate lemma rational_quadratic_coloring_from_integer_scales_rat_num_mul_int_quadratic (q : ℕ) (a b : ℚ) (A B : ℤ) (hA : (A : ℚ) = (q : ℚ) * a) (hB : (B : ℚ) = (q : ℚ) * b) (n : ℕ) (k : Fin n) : ((q : ℚ) * ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b)).num = (q : ℤ) * (k.val : ℤ) ^ 2 + A * (k.val : ℤ) + B := by let J : ℤ := (q : ℤ) * (k.val : ℤ) ^ 2 + A * (k.val : ℤ) + B have hq : (q : ℚ) * ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b) = (J : ℚ) := by dsimp [J] push_cast rw [hA, hB] ring have hnum := congrArg Rat.num hq simpa [Rat.num_intCast] using hnum
An integer coloring avoiding monochromatic quadratics with leading coefficient q scales to a rational coloring avoiding monochromatic quadratics whose coefficients have denominators dividing q.
Define the integer coefficients A = (q / a.den) * a.num and B = (q / b.den) * b.num, which equal q * a and q * b respectively. If the scaled coloring F(x) = f((q * x).num) were constant on the rational quadratic progression k² + a k + b for k < n, then by the previous numerator identity, f would be constant on the integer progression q k² + A k + B for k < n, contradicting the hypothesis on f.
uses
Int.cast_mul,rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd,rational_quadratic_coloring_from_integer_scales_rat_num_mul_int_quadraticprivate lemma rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constraint (q : ℕ) (f : ℤ → Bool) {n : ℕ} (target : Bool) (hf : ∀ A B : ℤ, ¬ ∀ k : Fin n, f ((q : ℤ) * (k.val : ℤ) ^ 2 + A * (k.val : ℤ) + B) = target) (a b : ℚ) (ha : a.den ∣ q) (hb : b.den ∣ q) (F : ℚ → Bool) (hF : ∀ x, F x = f ((q : ℚ) * x).num) : ¬ ∀ k : Fin n, F ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b) = target := by let A : ℤ := ((q / a.den : ℕ) : ℤ) * a.num let B : ℤ := ((q / b.den : ℕ) : ℤ) * b.num have hA : (A : ℚ) = (q : ℚ) * a := by dsimp [A] simp only [Int.cast_mul] exact rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd q a ha have hB : (B : ℚ) = (q : ℚ) * b := by dsimp [B] simp only [Int.cast_mul] exact rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd q b hb intro h apply hf A B intro k have hp := rational_quadratic_coloring_from_integer_scales_rat_num_mul_int_quadratic q a b A B hA hB n k have hk := h k rw [hF ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b), hp] at hk exact hk
Set up the index set for all quadratic constraints and define the corresponding sets of colorings.
We define the index type Index as the disjoint sum (ℚ × ℚ) ⊕ (ℚ × ℚ), where the left summand indexes pairs (a, b) for avoiding false length-3 progressions, and the right summand indexes pairs (a, b) for avoiding true length-17 progressions. For each index, T(i) is defined as the set of rational colorings F : ℚ → Bool that avoid the forbidden monochromatic progression.
theorem rational_quadratic_coloring_from_integer_scales (h : ∀ q : ℕ, 0 < q → ∃ f : ℤ → Bool, (∀ a b : ℤ, ¬ ∀ k : Fin 3, f ((q : ℤ) * (k.val : ℤ) ^ 2 + a * k.val + b) = false) ∧ (∀ a b : ℤ, ¬ ∀ k : Fin 17, f ((q : ℤ) * (k.val : ℤ) ^ 2 + a * k.val + b) = true)) : ∃ f : ℚ → Bool, (∀ a b : ℚ, ¬ ∀ k : Fin 3, f ((k.val : ℚ) ^ 2 + a * k.val + b) = false) ∧ (∀ a b : ℚ, ¬ ∀ k : Fin 17, f ((k.val : ℚ) ^ 2 + a * k.val + b) = true) := by let Index : Type := (ℚ × ℚ) ⊕ (ℚ × ℚ) let aa : Index → ℚ := fun i => match i with | Sum.inl p => p.1 | Sum.inr p => p.1 let bb : Index → ℚ := fun i => match i with | Sum.inl p => p.2 | Sum.inr p => p.2 let T : Index → Set (ℚ → Bool) := fun i => match i with | Sum.inl p => {F : ℚ → Bool | ¬ ∀ k : Fin 3, F ((k.val : ℚ) ^ 2 + p.1 * (k.val : ℚ) + p.2) = false} | Sum.inr p => {F : ℚ → Bool | ¬ ∀ k : Fin 17, F ((k.val : ℚ) ^ 2 + p.1 * (k.val : ℚ) + p.2) = true}
Each constraint set T(i) is closed in the product topology on ℚ → Bool.
A coloring F belongs to T(i) when not all evaluations on the progression equal the target boolean value. This is equivalent to saying that at least one evaluation equals the opposite value, so T(i) is a finite union over k of cylinder sets {F | F(x(k)) = opposite}. In the product topology, each evaluation map is continuous and singletons in Bool are closed, so each cylinder set is closed; hence their finite union T(i) is closed.
uses
Set.mem_iUnion,Bool.eq_false_of_not_eq_true,isClosed_iUnion_of_finite,IsClosed.preimage,continuous_apply,isClosed_singleton,Bool.eq_true_of_not_eq_falsehave hTclosed : ∀ i : Index, IsClosed (T i) := by intro i rcases i with ⟨a, b⟩ | ⟨a, b⟩ · dsimp [T] let x : Fin 3 → ℚ := fun k => (k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b have hset : {F : ℚ → Bool | ¬ ∀ k : Fin 3, F (x k) = false} = ⋃ k : Fin 3, {F : ℚ → Bool | F (x k) = true} := by ext F simp only [Set.mem_iUnion] constructor · intro hh by_contra hne apply hh intro k exact Bool.eq_false_of_not_eq_true (fun hk => hne ⟨k, hk⟩) · rintro ⟨k, hk⟩ hall simp [hall] at hk rw [hset] apply isClosed_iUnion_of_finite intro k exact IsClosed.preimage (continuous_apply (x k)) isClosed_singleton · dsimp [T] let x : Fin 17 → ℚ := fun k => (k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b have hset : {F : ℚ → Bool | ¬ ∀ k : Fin 17, F (x k) = true} = ⋃ k : Fin 17, {F : ℚ → Bool | F (x k) = false} := by ext F simp only [Set.mem_iUnion] constructor · intro hh by_contra hne apply hh intro k exact Bool.eq_true_of_not_eq_false (fun hk => hne ⟨k, hk⟩) · rintro ⟨k, hk⟩ hall simp [hall] at hk rw [hset] apply isClosed_iUnion_of_finite intro k exact IsClosed.preimage (continuous_apply (x k)) isClosed_singleton
The family of constraint sets {T(i)} satisfies the finite intersection property.
Given a finite set of indices s, let q be the product of all denominators a_i.den * b_i.den for i ∈ s. Since all denominators are positive, q is a positive integer. By hypothesis, there is an integer coloring f avoiding false length-3 and true length-17 quadratics with leading coefficient q. For every index i ∈ s, the denominators of a_i and b_i divide q, so the scaled coloring F(x) = f((q * x).num) satisfies the constraint T(i) by the reduction lemma. Thus F belongs to the intersection ⋂_{i ∈ s} T(i), proving it is nonempty.
uses
Finset.prod_pos,Nat.mul_pos,Rat.den_pos,Set.mem_iInter,Finset.dvd_prod_of_mem,dvd_trans,Nat.dvd_mul_right,Nat.dvd_mul_left,rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constrainthave hFIP : ∀ s : Finset Index, (⋂ i ∈ s, T i).Nonempty := by intro s let d : Index → ℕ := fun i => (aa i).den * (bb i).den let q : ℕ := ∏ i ∈ s, d i have hqpos : 0 < q := by dsimp [q] apply Finset.prod_pos intro i hi exact Nat.mul_pos (aa i).den_pos (bb i).den_pos rcases h q hqpos with ⟨f, hf3, hf17⟩ let F : ℚ → Bool := fun x => f ((q : ℚ) * x).num have hFdef : ∀ x : ℚ, F x = f ((q : ℚ) * x).num := fun x => rfl refine ⟨F, ?_⟩ rw [Set.mem_iInter] intro i rw [Set.mem_iInter] intro hi have hdvd : d i ∣ q := by dsimp [q] exact Finset.dvd_prod_of_mem d hi have hda : (aa i).den ∣ q := dvd_trans (Nat.dvd_mul_right (aa i).den (bb i).den) hdvd have hdb : (bb i).den ∣ q := dvd_trans (Nat.dvd_mul_left (bb i).den (aa i).den) hdvd rcases i with ⟨a, b⟩ | ⟨a, b⟩ · dsimp [T] exact rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constraint q f (n := 3) (target := false) hf3 (aa (Sum.inl (a, b))) (bb (Sum.inl (a, b))) hda hdb F hFdef · dsimp [T] exact rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constraint q f (n := 17) (target := true) hf17 (aa (Sum.inr (a, b))) (bb (Sum.inr (a, b))) hda hdb F hFdef
There exists a rational coloring avoiding all forbidden length-3 and length-17 quadratic progressions.
By Tychonoff's theorem, the space ℚ → Bool is compact. Since the family of closed sets T(i) has the finite intersection property, their total intersection ⋂_{i : Index} T(i) is nonempty. Any coloring F in this intersection satisfies F ∈ T(inl(a, b)) and F ∈ T(inr(a, b)) for all a, b ∈ ℚ, which gives the required non-monochromatic conditions for all rational coefficients.
uses
CompactSpace.iInter_nonempty,Set.mem_iInterhave hinter : (⋂ i : Index, T i).Nonempty := CompactSpace.iInter_nonempty hTclosed hFIP rcases hinter with ⟨F, hF⟩ have hFall : ∀ i : Index, F ∈ T i := by simpa only [Set.mem_iInter] using hF refine ⟨F, ?_, ?_⟩ · intro a b have hh := hFall (Sum.inl (a, b)) simpa [T] using hh · intro a b have hh := hFall (Sum.inr (a, b)) simpa [T] using hh
Nat.cast_mul- The natural number cast preserves multiplication: ↑(m * n) = ↑m * ↑n.
Nat.mul_div_cancel'- If n divides m, then n * (m / n) = m.
Int.cast_natCast- The integer cast of a natural number cast into a ring equals the direct natural number cast: ↑(↑n : ℤ) = ↑n.
Rat.mul_den_eq_num- For any rational q, multiplying by its denominator yields its numerator: q * ↑q.den = ↑q.num.
mul_comm- Multiplication is commutative: a * b = b * a.
congrArg- Equal inputs to a function yield equal outputs: a₁ = a₂ implies f a₁ = f a₂.
Rat.num_intCast- The numerator of an integer cast to the rationals is the integer itself: (↑a : ℚ).num = a.
Int.cast_mul- The integer cast preserves multiplication: ↑(m * n) = ↑m * ↑n.
rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd- If the denominator of a rational a divides q, then the integer (q / a.den) * a.num equals q * a as a rational number.
rational_quadratic_coloring_from_integer_scales_rat_num_mul_int_quadratic- If integers A and B represent q * a and q * b, then the numerator of q * (k² + a k + b) equals q k² + A k + B.
Set.mem_iUnion- An element belongs to a union of sets if and only if it belongs to at least one set in the family.
Bool.eq_false_of_not_eq_true- A boolean value that is not true must be false.
isClosed_iUnion_of_finite- A finite union of closed sets is closed.
IsClosed.preimage- The preimage of a closed set under a continuous function is closed.
continuous_apply- In a product space, the coordinate projection (evaluation at a point) is continuous.
isClosed_singleton- Singletons are closed in a T₁ topological space.
Bool.eq_true_of_not_eq_false- A boolean value that is not false must be true.
Finset.prod_pos- A finite product of strictly positive elements is strictly positive.
Nat.mul_pos- The product of two strictly positive natural numbers is strictly positive.
Rat.den_pos- The denominator of every rational number is strictly positive.
Set.mem_iInter- An element belongs to an intersection of sets if and only if it belongs to every set in the family.
Finset.dvd_prod_of_mem- An element of a finite set divides the product of all elements in the set.
dvd_trans- Divisibility is transitive: if a divides b and b divides c, then a divides c.
Nat.dvd_mul_right- A natural number a divides a * b.
Nat.dvd_mul_left- A natural number a divides b * a.
rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constraint- If an integer coloring avoids monochromatic progressions with leading coefficient q, its numerator scaling gives a rational coloring avoiding monochromatic progressions whose coefficients have denominators dividing q.
CompactSpace.iInter_nonempty- In a compact space, any family of closed sets with the finite intersection property has a nonempty intersection.
Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
private lemma rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd (q : ℕ) (a : ℚ) (h : a.den ∣ q) : (((q / a.den : ℕ) : ℤ) * a.num : ℚ) = (q : ℚ) * a := by have hq : (q : ℚ) = (a.den : ℚ) * ((q / a.den : ℕ) : ℚ) := by rw [← Nat.cast_mul, Nat.mul_div_cancel' h] calc (((q / a.den : ℕ) : ℤ) * a.num : ℚ) = ((q / a.den : ℕ) : ℚ) * (a.num : ℚ) := by simp only [Int.cast_natCast] _ = ((q / a.den : ℕ) : ℚ) * (a * (a.den : ℚ)) := by rw [← Rat.mul_den_eq_num a] _ = (((q / a.den : ℕ) : ℚ) * (a.den : ℚ)) * a := by ring _ = (q : ℚ) * a := by rw [mul_comm ((q / a.den : ℕ) : ℚ) (a.den : ℚ), ← hq]private lemma rational_quadratic_coloring_from_integer_scales_rat_num_mul_int_quadratic (q : ℕ) (a b : ℚ) (A B : ℤ) (hA : (A : ℚ) = (q : ℚ) * a) (hB : (B : ℚ) = (q : ℚ) * b) (n : ℕ) (k : Fin n) : ((q : ℚ) * ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b)).num = (q : ℤ) * (k.val : ℤ) ^ 2 + A * (k.val : ℤ) + B := by let J : ℤ := (q : ℤ) * (k.val : ℤ) ^ 2 + A * (k.val : ℤ) + B have hq : (q : ℚ) * ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b) = (J : ℚ) := by dsimp [J] push_cast rw [hA, hB] ring have hnum := congrArg Rat.num hq simpa [Rat.num_intCast] using hnumprivate lemma rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constraint (q : ℕ) (f : ℤ → Bool) {n : ℕ} (target : Bool) (hf : ∀ A B : ℤ, ¬ ∀ k : Fin n, f ((q : ℤ) * (k.val : ℤ) ^ 2 + A * (k.val : ℤ) + B) = target) (a b : ℚ) (ha : a.den ∣ q) (hb : b.den ∣ q) (F : ℚ → Bool) (hF : ∀ x, F x = f ((q : ℚ) * x).num) : ¬ ∀ k : Fin n, F ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b) = target := by let A : ℤ := ((q / a.den : ℕ) : ℤ) * a.num let B : ℤ := ((q / b.den : ℕ) : ℤ) * b.num have hA : (A : ℚ) = (q : ℚ) * a := by dsimp [A] simp only [Int.cast_mul] exact rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd q a ha have hB : (B : ℚ) = (q : ℚ) * b := by dsimp [B] simp only [Int.cast_mul] exact rational_quadratic_coloring_from_integer_scales_rat_num_mul_eq_of_den_dvd q b hb intro h apply hf A B intro k have hp := rational_quadratic_coloring_from_integer_scales_rat_num_mul_int_quadratic q a b A B hA hB n k have hk := h k rw [hF ((k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b), hp] at hk exact hktheorem rational_quadratic_coloring_from_integer_scales (h : ∀ q : ℕ, 0 < q → ∃ f : ℤ → Bool, (∀ a b : ℤ, ¬ ∀ k : Fin 3, f ((q : ℤ) * (k.val : ℤ) ^ 2 + a * k.val + b) = false) ∧ (∀ a b : ℤ, ¬ ∀ k : Fin 17, f ((q : ℤ) * (k.val : ℤ) ^ 2 + a * k.val + b) = true)) : ∃ f : ℚ → Bool, (∀ a b : ℚ, ¬ ∀ k : Fin 3, f ((k.val : ℚ) ^ 2 + a * k.val + b) = false) ∧ (∀ a b : ℚ, ¬ ∀ k : Fin 17, f ((k.val : ℚ) ^ 2 + a * k.val + b) = true) := by let Index : Type := (ℚ × ℚ) ⊕ (ℚ × ℚ) let aa : Index → ℚ := fun i => match i with | Sum.inl p => p.1 | Sum.inr p => p.1 let bb : Index → ℚ := fun i => match i with | Sum.inl p => p.2 | Sum.inr p => p.2 let T : Index → Set (ℚ → Bool) := fun i => match i with | Sum.inl p => {F : ℚ → Bool | ¬ ∀ k : Fin 3, F ((k.val : ℚ) ^ 2 + p.1 * (k.val : ℚ) + p.2) = false} | Sum.inr p => {F : ℚ → Bool | ¬ ∀ k : Fin 17, F ((k.val : ℚ) ^ 2 + p.1 * (k.val : ℚ) + p.2) = true} have hTclosed : ∀ i : Index, IsClosed (T i) := by intro i rcases i with ⟨a, b⟩ | ⟨a, b⟩ · dsimp [T] let x : Fin 3 → ℚ := fun k => (k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b have hset : {F : ℚ → Bool | ¬ ∀ k : Fin 3, F (x k) = false} = ⋃ k : Fin 3, {F : ℚ → Bool | F (x k) = true} := by ext F simp only [Set.mem_iUnion] constructor · intro hh by_contra hne apply hh intro k exact Bool.eq_false_of_not_eq_true (fun hk => hne ⟨k, hk⟩) · rintro ⟨k, hk⟩ hall simp [hall] at hk rw [hset] apply isClosed_iUnion_of_finite intro k exact IsClosed.preimage (continuous_apply (x k)) isClosed_singleton · dsimp [T] let x : Fin 17 → ℚ := fun k => (k.val : ℚ) ^ 2 + a * (k.val : ℚ) + b have hset : {F : ℚ → Bool | ¬ ∀ k : Fin 17, F (x k) = true} = ⋃ k : Fin 17, {F : ℚ → Bool | F (x k) = false} := by ext F simp only [Set.mem_iUnion] constructor · intro hh by_contra hne apply hh intro k exact Bool.eq_true_of_not_eq_false (fun hk => hne ⟨k, hk⟩) · rintro ⟨k, hk⟩ hall simp [hall] at hk rw [hset] apply isClosed_iUnion_of_finite intro k exact IsClosed.preimage (continuous_apply (x k)) isClosed_singleton have hFIP : ∀ s : Finset Index, (⋂ i ∈ s, T i).Nonempty := by intro s let d : Index → ℕ := fun i => (aa i).den * (bb i).den let q : ℕ := ∏ i ∈ s, d i have hqpos : 0 < q := by dsimp [q] apply Finset.prod_pos intro i hi exact Nat.mul_pos (aa i).den_pos (bb i).den_pos rcases h q hqpos with ⟨f, hf3, hf17⟩ let F : ℚ → Bool := fun x => f ((q : ℚ) * x).num have hFdef : ∀ x : ℚ, F x = f ((q : ℚ) * x).num := fun x => rfl refine ⟨F, ?_⟩ rw [Set.mem_iInter] intro i rw [Set.mem_iInter] intro hi have hdvd : d i ∣ q := by dsimp [q] exact Finset.dvd_prod_of_mem d hi have hda : (aa i).den ∣ q := dvd_trans (Nat.dvd_mul_right (aa i).den (bb i).den) hdvd have hdb : (bb i).den ∣ q := dvd_trans (Nat.dvd_mul_left (bb i).den (aa i).den) hdvd rcases i with ⟨a, b⟩ | ⟨a, b⟩ · dsimp [T] exact rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constraint q f (n := 3) (target := false) hf3 (aa (Sum.inl (a, b))) (bb (Sum.inl (a, b))) hda hdb F hFdef · dsimp [T] exact rational_quadratic_coloring_from_integer_scales_quadratic_coloring_constraint q f (n := 17) (target := true) hf17 (aa (Sum.inr (a, b))) (bb (Sum.inr (a, b))) hda hdb F hFdef have hinter : (⋂ i : Index, T i).Nonempty := CompactSpace.iInter_nonempty hTclosed hFIP rcases hinter with ⟨F, hF⟩ have hFall : ∀ i : Index, F ∈ T i := by simpa only [Set.mem_iInter] using hF refine ⟨F, ?_, ?_⟩ · intro a b have hh := hFall (Sum.inl (a, b)) simpa [T] using hh · intro a b have hh := hFall (Sum.inr (a, b)) simpa [T] using hh