triangular_lattice_coverable_critical_spacings
The triangular lattice with spacing d is coverable by pairwise non-overlapping closed unit disks for d = 1 / Real.sqrt 13 and for d = 2 / Real.sqrt 21.
Statement
theorem triangular_lattice_coverable_critical_spacings : is_coverable (triangular_lattice (1 / Real.sqrt 13)) ∧ is_coverable (triangular_lattice (2 / Real.sqrt 21))- Formalizes
- arXiv:2611.13535 Theorem C -- Covering the Triangular Lattice with Non-Overlapping Unit Disks
- Verified
- 12 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 3
- is_coverableA subset S of the complex plane is coverable by non-overlapping closed unit disks if there exists a set C of disk centers such that ||c -…
- triangular_latticeThe triangular lattice with spacing d in the complex plane is the set of points of the form d * (a + b * eisensteinOmega) for integers a…
- eisensteinOmegaThe primitive cube root of unity ω = (-1 + √(-3))/2 in the complex numbers.
- Used by 2
- triangular_lattice_coverable_one_div_sqrt_thirteenThe triangular lattice with spacing d = 1 / Real.sqrt 13 is coverable by non-overlapping closed unit disks.
- triangular_lattice_coverable_two_div_sqrt_twenty_oneThe triangular lattice with spacing d = 2 / Real.sqrt 21 is coverable by non-overlapping closed unit disks.
Read back from the Lean
Both of the following statements hold: 1. The triangular lattice with spacing $1 / \sqrt{13}$ is coverable: there exists a subset $C \subseteq \mathbb{C}$ such that any two distinct points in $C$ have distance at least 2 from each other, and every point in the triangular lattice $\{ (1/\sqrt{13})(a + b\omega) \mid a, b \in \mathbb{Z} \}$ (where $\omega = (-1 + i\sqrt{3})/2$) is at distance at most 1 from some point in $C$. 2. The triangular lattice with spacing $2 / \sqrt{21}$ is coverable: there exists a subset $C' \subseteq \mathbb{C}$ such that any two distinct points in $C'$ have distance at least 2 from each other, and every point in the triangular lattice $\{ (2/\sqrt{21})(a + b\omega) \mid a, b \in \mathbb{Z} \}$ is at distance at most 1 from some point in $C'$.
Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.
The proof, explained
To prove that the triangular lattice with spacing d is coverable by pairwise non-overlapping unit disks, the argument constructs disk centers as an affine sublattice of the Eisenstein integers, C = { d(c₀ + β(x + yω)) | x, y ∈ ℤ }. Choosing β such that ‖β‖ = 2/d ensures that distinct centers are separated by at least d ‖β‖ = 2, so the unit disks do not overlap. Covering every lattice point d(a + bω) then reduces to finding integers x, y such that the squared distance to some center is at most n = 4/d². For the critical spacings d = 1 / √13 (with n = 52) and d = 2 / √21 (with n = 21), suitable integer witnesses (x, y) are obtained by modular arithmetic from explicit lookup tables verified by computational reflection (decide).
The Eisenstein generator ω satisfies ω² + ω + 1 = 0.
By definition, ω has real part -1/2 and imaginary part √3 / 2. By Complex.ext, two complex numbers are equal if their real and imaginary parts match. The real part of ω² + ω + 1 simplifies to (-1/2)² - (√3/2)² - 1/2 + 1 = 0 using (√3)² = 3 from Real.sq_sqrt, and the imaginary part simplifies to 2(-1/2)(√3/2) + √3/2 = 0.
uses
Real.sq_sqrt,Complex.extprivate lemma tlc_omega_sq : eisensteinOmega^2 + eisensteinOmega + 1 = 0 := by have h3 : (Real.sqrt 3)^2 = 3 := Real.sq_sqrt (by norm_num) have hre : eisensteinOmega.re = -1/2 := by simp [eisensteinOmega] have him : eisensteinOmega.im = Real.sqrt 3 / 2 := by simp [eisensteinOmega] apply Complex.ext · simp only [pow_two, Complex.mul_re, Complex.add_re, hre, him, Complex.zero_re, Complex.one_re] ring_nf nlinarith [h3] · simp only [pow_two, Complex.mul_im, Complex.add_im, hre, him, Complex.zero_im, Complex.one_im] ring_nf
The square of ω satisfies ω² = -ω - 1.
Subtracting ω + 1 from both sides of tlc_omega_sq via linear_combination yields the formula for ω².
uses
tlc_omega_sqprivate lemma tlc_omega_sq' : eisensteinOmega^2 = -eisensteinOmega - 1 := by linear_combination tlc_omega_sq
For any integers a and b, the squared complex modulus of a + bω is given by the quadratic form a² - ab + b².
The squared modulus is Re(z)² + Im(z)² by Complex.normSq_apply. Expanding a + bω = (a - b/2) + i(b√3/2) gives (a - b/2)² + 3b²/4 = a² - ab + b², which is verified using (√3)² = 3 from Real.sq_sqrt.
uses
Complex.normSq_apply,Real.sq_sqrtprivate lemma tlc_normSq_int (a b : ℤ) : Complex.normSq ((a:ℂ) + (b:ℂ)*eisensteinOmega) = ((a^2 - a*b + b^2 : ℤ) : ℝ) := by rw [Complex.normSq_apply] have hre : ((a:ℂ) + (b:ℂ)*eisensteinOmega).re = (a:ℝ) - (b:ℝ)/2 := by simp [eisensteinOmega, Complex.add_re, Complex.mul_re]; ring have him : ((a:ℂ) + (b:ℂ)*eisensteinOmega).im = (b:ℝ)*Real.sqrt 3 / 2 := by simp [eisensteinOmega, Complex.add_im, Complex.mul_im]; ring rw [hre, him] have h3 : (Real.sqrt 3)^2 = 3 := Real.sq_sqrt (by norm_num) push_cast nlinarith [h3]
Every nonzero Eisenstein integer a + bω ≠ 0 has Euclidean norm at least 1: ‖a + bω‖ ≥ 1.
Since a + bω ≠ 0, its squared modulus is strictly positive by Complex.normSq_pos. By tlc_normSq_int, this squared modulus is the integer a² - ab + b² > 0, so a² - ab + b² ≥ 1. Since ‖a + bω‖² = a² - ab + b² ≥ 1 by Complex.sq_norm, taking square roots yields ‖a + bω‖ ≥ 1 via norm_nonneg.
uses
Complex.normSq_pos,tlc_normSq_int,Complex.sq_norm,norm_nonnegprivate lemma tlc_one_le_norm (a b : ℤ) (h : ((a:ℂ) + (b:ℂ)*eisensteinOmega) ≠ 0) : 1 ≤ ‖(a:ℂ) + (b:ℂ)*eisensteinOmega‖ := by have hpos : 0 < Complex.normSq ((a:ℂ) + (b:ℂ)*eisensteinOmega) := by rwa [Complex.normSq_pos] rw [tlc_normSq_int] at hpos have hN : (1:ℤ) ≤ a^2 - a*b + b^2 := by have h0 : (0:ℤ) < a^2 - a*b + b^2 := by exact_mod_cast hpos nlinarith [h0] have hsq : (1:ℝ) ≤ ‖(a:ℂ) + (b:ℂ)*eisensteinOmega‖^2 := by rw [Complex.sq_norm, tlc_normSq_int] exact_mod_cast hN nlinarith [norm_nonneg ((a:ℂ) + (b:ℂ)*eisensteinOmega)]
Given parameters β, c₀ ∈ ℂ and n = 4/d² with ‖β‖² = n, the centers C = { d(c₀ + β(x + yω)) | x, y ∈ ℤ } are separated by pairwise distance at least 2.
From d² * n = 4 and d > 0, we have ‖β‖ = 2/d. Two distinct centers c, c' ∈ C differ by d * β * ((x - x') + (y - y')ω). Because (x - x') + (y - y')ω is a nonzero Eisenstein integer by sub_ne_zero.mpr, its norm is at least 1 by tlc_one_le_norm. Taking norms via norm_mul, Complex.norm_real, Real.norm_eq_abs, and abs_of_pos shows ‖c - c'‖ = d * (2/d) * ‖(x - x') + (y - y')ω‖ ≥ 2 * 1 = 2, so the unit disks around distinct centers do not overlap.
uses
tlc_one_le_norm,sub_ne_zero.mpr,norm_mul,Complex.norm_real,Real.norm_eq_abs,abs_of_posprivate lemma tlc_coverable_of_data (d : ℝ) (hd : 0 < d) (β c₀ : ℂ) (n : ℝ) (hβ : Complex.normSq β = n) (hdn : d^2 * n = 4) (hcov : ∀ a b : ℤ, ∃ x y : ℤ, ‖2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)‖^2 ≤ n) : is_coverable (triangular_lattice d) := by have hn : n = 4/d^2 := by field_simp at hdn ⊢ linarith have hβnorm : ‖β‖ = 2/d := by have hsq : ‖β‖^2 = (2/d)^2 := by rw [Complex.sq_norm, hβ, hn] field_simp ring nlinarith [norm_nonneg β, (by positivity : (0:ℝ) ≤ 2/d)] have hd2 : (0:ℝ) < d/2 := by positivity refine ⟨{z : ℂ | ∃ x y : ℤ, z = (d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega))}, ?_, ?_⟩ · rintro c c' ⟨x, y, rfl⟩ ⟨x', y', rfl⟩ hne have hq : ((x:ℂ) + (y:ℂ)*eisensteinOmega) ≠ ((x':ℂ) + (y':ℂ)*eisensteinOmega) := by intro h; exact hne (by rw [h]) have hdiff : ((x:ℂ) + (y:ℂ)*eisensteinOmega) - ((x':ℂ) + (y':ℂ)*eisensteinOmega) = ((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega := by push_cast; ring have hq' : ((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega ≠ 0 := by rw [← hdiff]; exact sub_ne_zero.mpr hq have hQ : 1 ≤ ‖((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega‖ := tlc_one_le_norm _ _ hq' have hnorm : ‖(d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) - (d:ℂ) * (c₀ + β*((x':ℂ) + (y':ℂ)*eisensteinOmega))‖ = d * (2/d) * ‖((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega‖ := by have he : (d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) - (d:ℂ) * (c₀ + β*((x':ℂ) + (y':ℂ)*eisensteinOmega)) = (d:ℂ) * β * (((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega) := by rw [← hdiff]; ring rw [he, norm_mul, norm_mul, Complex.norm_real, Real.norm_eq_abs, abs_of_pos hd, hβnorm] rw [hnorm, show d * (2/d) = 2 by field_simp] linarith
Every lattice point of the triangular lattice of spacing d is within distance 1 of some center in C, completing the proof that the lattice is coverable.
Any lattice point has the form z = d(a + bω). The hypothesis provides integers x, y such that ‖2(a + bω) - 2c₀ - 2β(x + yω)‖² ≤ n. Choosing the center c = d(c₀ + β(x + yω)) ∈ C, the vector z - c factors as (d/2) * (2(a + bω) - 2c₀ - 2β(x + yω)). Since the bracketed term has norm at most √n = 2/d, scaling by d/2 using mul_le_mul_of_nonneg_left yields ‖z - c‖ ≤ (d/2)(2/d) = 1.
uses
mul_le_mul_of_nonneg_left,norm_mul,Complex.norm_real,Real.norm_eq_abs,abs_of_pos· rintro z ⟨a, b, rfl⟩ obtain ⟨x, y, hxy⟩ := hcov a b refine ⟨(d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)), ⟨x, y, rfl⟩, ?_⟩ have hfac : (d:ℂ) * ((a:ℂ) + (b:ℂ)*eisensteinOmega) - (d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) = (d/2 : ℝ) * (2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) := by push_cast; ring rw [hfac, norm_mul, Complex.norm_real, Real.norm_eq_abs, abs_of_pos hd2] have hW : ‖2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)‖ ≤ 2/d := by have hW2 : ‖2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)‖^2 ≤ (2/d)^2 := by rw [show (2/d)^2 = n from by rw [hn] at *; field_simp; ring] exact hxy nlinarith [norm_nonneg (2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)), (by positivity : (0:ℝ) ≤ 2/d)] have := mul_le_mul_of_nonneg_left hW hd2.le rwa [show d/2 * (2/d) = 1 by field_simp] at this
The lookup table tlc_tbl13 contains, for all parity pairs (i, j) ∈ {0, 1}² and residues r ∈ {0, ..., 12}, a pair (u, v) with u ≡ i (mod 2), v ≡ j (mod 2), u - 4v ≡ r (mod 13), and u² - uv + v² ≤ 13. automation
The table explicitly enumerates candidate pairs for all 52 combinations. The universal quantification over the finite types Fin 2 and Fin 13 is checked by finite evaluation using the decide tactic.
private def tlc13row (i j : Fin 2) : Fin 13 → ℤ × ℤ := if (i : ℕ) = 0 ∧ (j : ℕ) = 0 then ![(0, 0), (-2, -4), (2, 0), (-2, 2), (-4, -2), (0, 2), (-2, -2), (2, 2), (0, -2), (4, 2), (2, -2), (-2, 0), (2, 4)] else if (i : ℕ) = 0 then ![(-4, -1), (0, 3), (-2, -1), (2, 3), (0, -1), (4, 3), (2, -1), (-2, 1), (-4, -3), (0, 1), (-2, -3), (2, 1), (0, -3)] else if (j : ℕ) = 0 then ![(-3, -4), (1, 0), (-1, -4), (3, 0), (-1, 2), (-3, -2), (1, 2), (-1, -2), (3, 2), (1, -2), (-3, 0), (1, 4), (-1, 0)] else ![(-1, 3), (-3, -1), (1, 3), (-1, -1), (3, 3), (1, -1), (-3, 1), (3, -1), (-1, 1), (-3, -3), (1, 1), (-1, -3), (3, 1)] private def tlc_tbl13 (i j : Fin 2) (r : Fin 13) : ℤ × ℤ := tlc13row i j r private lemma tlc_tbl13_spec (i j : Fin 2) (r : Fin 13) : let p := tlc_tbl13 i j r (p.1 % 2 = ((i:ℕ):ℤ) ∧ p.2 % 2 = ((j:ℕ):ℤ) ∧ (p.1 - 4*p.2) % 13 = ((r:ℕ):ℤ) ∧ p.1^2 - p.1*p.2 + p.2^2 ≤ 13) := by revert i j r decide
For any integers a and b, there exist integers x and y such that (a - 8x + 2y)² - (a - 8x + 2y)(b - 2x - 6y) + (b - 2x - 6y)² ≤ 13.
Set i = a mod 2, j = b mod 2, and r = (a - 4b) mod 13 using ZMod.val_intCast. The table entry (u, v) = tlc_tbl13 i j r satisfies the parities, residue mod 13, and norm bound ≤ 13 by tlc_tbl13_spec. Because b ≡ v (mod 2) and (a - u) - 4(b - v) is divisible by 26, setting K = (((a - u) - 4(b - v))/2)/13 gives an integer. Choosing x = (b - v)/2 + 3K and y = -K, linear integer arithmetic (omega) verifies that a - 8x + 2y = u and b - 2x - 6y = v, whose quadratic form is u² - uv + v² ≤ 13.
uses
tlc_tbl13_spec,ZMod.val_intCastprivate lemma tlc_cover13 (a b : ℤ) : ∃ x y : ℤ, (a - 8*x + 2*y)^2 - (a - 8*x + 2*y)*(b - 2*x - 6*y) + (b - 2*x - 6*y)^2 ≤ 13 := by set m : ZMod 13 := (a : ZMod 13) - 4*(b : ZMod 13) with hm set i : Fin 2 := ⟨(a : ZMod 2).val, ZMod.val_lt (a : ZMod 2)⟩ with hi set j : Fin 2 := ⟨(b : ZMod 2).val, ZMod.val_lt (b : ZMod 2)⟩ with hj set r : Fin 13 := ⟨m.val, ZMod.val_lt m⟩ with hr obtain ⟨h1, h2, h3, h4⟩ := tlc_tbl13_spec i j r set u : ℤ := (tlc_tbl13 i j r).1 with hu set v : ℤ := (tlc_tbl13 i j r).2 with hv rw [hu] at h1 h3 rw [hv] at h2 h3 simp only [hu, hv] at h4 have hi2 : ((i:ℕ):ℤ) = a % 2 := by simp only [hi] exact ZMod.val_intCast (n := 2) a have hj2 : ((j:ℕ):ℤ) = b % 2 := by simp only [hj] exact ZMod.val_intCast (n := 2) b have hm' : m = ((a - 4*b : ℤ) : ZMod 13) := by rw [hm]; push_cast; ring have hr2 : ((r:ℕ):ℤ) = (a - 4*b) % 13 := by simp only [hr] rw [hm'] exact ZMod.val_intCast (n := 13) (a - 4*b) have k1 : u % 2 = a % 2 := by rw [h1, hi2] have k2 : v % 2 = b % 2 := by rw [h2, hj2] have k3 : (u - 4*v) % 13 = (a - 4*b) % 13 := by rw [h3, hr2] set K : ℤ := (((a - u) - 4*(b - v))/2)/13 with hK refine ⟨(b - v)/2 + 3*K, -K, ?_⟩ have e1 : a - 8*((b - v)/2 + 3*K) + 2*(-K) = u := by omega have e2 : b - 2*((b - v)/2 + 3*K) - 6*(-K) = v := by omega rw [e1, e2] exact h4
The lookup table tlc_tbl21 contains, for every residue r ∈ {0, ..., 20}, a pair of odd integers (u, v) with u - 5v ≡ r (mod 21) and u² - uv + v² ≤ 21. automation
The 21 table entries are verified to be odd and to satisfy the modular equivalence and norm bound by finite computational reflection via decide.
private def tlc_tbl21 : Fin 21 → ℤ × ℤ := ![(-5, -1), (-3, -5), (-3, -1), (-1, -5), (-1, -1), (-1, 3), (1, -1), (1, 3), (3, -1), (3, 3), (-5, -3), (5, 3), (-3, -3), (-3, 1), (-1, -3), (-1, 1), (1, -3), (1, 1), (1, 5), (3, 1), (3, 5)] private lemma tlc_tbl21_spec (r : Fin 21) : let p := tlc_tbl21 r (p.1 % 2 = 1 ∧ p.2 % 2 = 1 ∧ (p.1 - 5*p.2) % 21 = ((r:ℕ):ℤ) ∧ p.1^2 - p.1*p.2 + p.2^2 ≤ 21) := by revert r decide
For any integers a and b, there exist integers x and y such that (2a - 3 - 10x + 2y)² - (2a - 3 - 10x + 2y)(2b + 1 - 2x - 8y) + (2b + 1 - 2x - 8y)² ≤ 21.
Let r = ((2a - 3) - 5(2b + 1)) mod 21 via ZMod.val_intCast. The table entry (u, v) = tlc_tbl21 r consists of odd integers satisfying u - 5v ≡ r (mod 21) and u² - uv + v² ≤ 21 by tlc_tbl21_spec. Since 2b + 1 and v are both odd and ((2a - 3) - u) - 5((2b + 1) - v) is divisible by 42, K = ((((2a - 3) - u) - 5((2b + 1) - v))/2)/21 is an integer. Choosing x = ((2b + 1) - v)/2 + 4K and y = -K, omega verifies (2a - 3) - 10x + 2y = u and (2b + 1) - 2x - 8y = v, giving value u² - uv + v² ≤ 21.
uses
tlc_tbl21_spec,ZMod.val_intCastprivate lemma tlc_cover21 (a b : ℤ) : ∃ x y : ℤ, (2*a - 3 - 10*x + 2*y)^2 - (2*a - 3 - 10*x + 2*y)*(2*b + 1 - 2*x - 8*y) + (2*b + 1 - 2*x - 8*y)^2 ≤ 21 := by set m : ZMod 21 := ((2*a - 3 : ℤ) : ZMod 21) - 5*((2*b + 1 : ℤ) : ZMod 21) with hm set r : Fin 21 := ⟨m.val, ZMod.val_lt m⟩ with hr obtain ⟨h1, h2, h3, h4⟩ := tlc_tbl21_spec r set u : ℤ := (tlc_tbl21 r).1 with hu set v : ℤ := (tlc_tbl21 r).2 with hv rw [hu] at h1 h3 rw [hv] at h2 h3 simp only [hu, hv] at h4 have hm' : m = (((2*a - 3) - 5*(2*b + 1) : ℤ) : ZMod 21) := by rw [hm]; push_cast; ring have h3' : (u - 5*v) % 21 = ((2*a - 3) - 5*(2*b + 1)) % 21 := by rw [h3] simp only [hr] rw [hm'] exact ZMod.val_intCast (n := 21) ((2*a - 3) - 5*(2*b + 1)) set K : ℤ := ((((2*a - 3) - u) - 5*((2*b + 1) - v))/2)/21 with hK refine ⟨((2*b + 1) - v)/2 + 4*K, -K, ?_⟩ have e1 : 2*a - 3 - 10*(((2*b + 1) - v)/2 + 4*K) + 2*(-K) = u := by omega have e2 : 2*b + 1 - 2*(((2*b + 1) - v)/2 + 4*K) - 8*(-K) = v := by omega rw [e1, e2] exact h4
The triangular lattice with spacing d = 1 / √13 is coverable by pairwise non-overlapping unit disks.
We apply tlc_coverable_of_data with d = 1 / √13, β = 8 + 2ω, c₀ = 0, and n = 52. The scaling relation d² * 52 = 4 holds by div_pow, one_pow, and Real.sq_sqrt, and normSq(β) = 8² - 8*2 + 2² = 52 by tlc_normSq_int. Given a, b ∈ ℤ, the integers x, y from tlc_cover13 simplify the expression 2(a + bω) - 2β(x + yω) using ω² = -ω - 1 (tlc_omega_sq') to (2a - 16x + 4y) + (2b - 4x - 12y)ω. By Complex.sq_norm and tlc_normSq_int, its squared norm equals 4 times the quadratic form from tlc_cover13, which is at most 4 * 13 = 52 = n.
uses
tlc_coverable_of_data,Real.sq_sqrt,div_pow,one_pow,tlc_normSq_int,tlc_cover13,tlc_omega_sq',Complex.sq_normprivate lemma tlc_coverable_one : is_coverable (triangular_lattice (1 / Real.sqrt 13)) := by have hd : (0:ℝ) < 1/Real.sqrt 13 := by positivity have hdn : (1/Real.sqrt 13)^2 * 52 = 4 := by rw [div_pow, one_pow, Real.sq_sqrt (by norm_num)]; norm_num refine tlc_coverable_of_data (1/Real.sqrt 13) hd (8 + 2*eisensteinOmega) 0 52 ?_ hdn ?_ · have h := tlc_normSq_int 8 2 norm_num at h exact h · intro a b obtain ⟨x, y, hxy⟩ := tlc_cover13 a b refine ⟨x, y, ?_⟩ have e : 2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*(0:ℂ) - 2*(8 + 2*eisensteinOmega)*((x:ℂ) + (y:ℂ)*eisensteinOmega) = ((2*a - 16*x + 4*y : ℤ) : ℂ) + ((2*b - 4*x - 12*y : ℤ) : ℂ)*eisensteinOmega := by push_cast ring_nf rw [tlc_omega_sq'] push_cast ring have hkey : ((2*a - 16*x + 4*y)^2 - (2*a - 16*x + 4*y)*(2*b - 4*x - 12*y) + (2*b - 4*x - 12*y)^2 : ℤ) = 4*((a - 8*x + 2*y)^2 - (a - 8*x + 2*y)*(b - 2*x - 6*y) + (b - 2*x - 6*y)^2) := by ring rw [e, Complex.sq_norm, tlc_normSq_int, hkey] have hZ : 4*((a - 8*x + 2*y)^2 - (a - 8*x + 2*y)*(b - 2*x - 6*y) + (b - 2*x - 6*y)^2) ≤ 52 := by linarith exact_mod_cast hZ
The triangular lattice with spacing d = 2 / √21 is coverable by pairwise non-overlapping unit disks.
We apply tlc_coverable_of_data with d = 2 / √21, β = 5 + ω, c₀ = (3 - ω)/2, and n = 21. The scaling relation d² * 21 = 4 holds by div_pow and Real.sq_sqrt, and normSq(β) = 5² - 5*1 + 1² = 21 by tlc_normSq_int. Given a, b ∈ ℤ, the integers x, y from tlc_cover21 simplify 2(a + bω) - 2c₀ - 2β(x + yω) using ω² = -ω - 1 (tlc_omega_sq') to (2a - 3 - 10x + 2y) + (2b + 1 - 2x - 8y)ω. By Complex.sq_norm and tlc_normSq_int, its squared norm is bounded by 21 via tlc_cover21.
uses
tlc_coverable_of_data,Real.sq_sqrt,div_pow,tlc_normSq_int,tlc_cover21,tlc_omega_sq',Complex.sq_normprivate lemma tlc_coverable_two : is_coverable (triangular_lattice (2 / Real.sqrt 21)) := by have hd : (0:ℝ) < 2/Real.sqrt 21 := by positivity have hdn : (2/Real.sqrt 21)^2 * 21 = 4 := by rw [div_pow, Real.sq_sqrt (by norm_num)]; norm_num refine tlc_coverable_of_data (2/Real.sqrt 21) hd (5 + eisensteinOmega) ((3 - eisensteinOmega)/2) 21 ?_ hdn ?_ · have h := tlc_normSq_int 5 1 norm_num at h exact h · intro a b obtain ⟨x, y, hxy⟩ := tlc_cover21 a b refine ⟨x, y, ?_⟩ have e : 2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*((3 - eisensteinOmega)/2) - 2*(5 + eisensteinOmega)*((x:ℂ) + (y:ℂ)*eisensteinOmega) = ((2*a - 3 - 10*x + 2*y : ℤ) : ℂ) + ((2*b + 1 - 2*x - 8*y : ℤ) : ℂ)*eisensteinOmega := by push_cast ring_nf rw [tlc_omega_sq'] push_cast ring rw [e, Complex.sq_norm, tlc_normSq_int] exact_mod_cast hxy
The triangular lattice is coverable by pairwise non-overlapping unit disks at both critical spacings d = 1 / √13 and d = 2 / √21.
Combining tlc_coverable_one and tlc_coverable_two as a conjunction proves the main theorem.
uses
tlc_coverable_one,tlc_coverable_two/-- Coverability of the triangular lattice at spacings 1 / √13 and 2 / √21 by non-overlapping unit disks. -/ theorem triangular_lattice_coverable_critical_spacings : is_coverable (triangular_lattice (1 / Real.sqrt 13)) ∧ is_coverable (triangular_lattice (2 / Real.sqrt 21)) := ⟨tlc_coverable_one, tlc_coverable_two⟩
Real.sq_sqrt- For any real x ≥ 0, (√x)² = x.
Complex.ext- Two complex numbers are equal if and only if their real and imaginary parts are equal.
tlc_omega_sq- The Eisenstein generator ω satisfies ω² + ω + 1 = 0.
tlc_omega_sq'- The square of ω satisfies ω² = -ω - 1.
Complex.normSq_apply- The squared modulus of a complex number z is Re(z)² + Im(z)².
Complex.normSq_pos- The squared modulus of a complex number is strictly positive if and only if the number is nonzero.
tlc_normSq_int- For integers a and b, the squared modulus of a + bω is a² - ab + b².
Complex.sq_norm- The squared norm ‖z‖² equals the squared modulus normSq(z).
norm_nonneg- The norm of any element in a seminormed group is non-negative.
tlc_one_le_norm- Any nonzero Eisenstein integer a + bω ≠ 0 has Euclidean norm at least 1.
sub_ne_zero.mpr- If x ≠ y, then x - y ≠ 0.
norm_mul- The norm of a product is the product of the norms: ‖a * b‖ = ‖a‖ * ‖b‖.
Complex.norm_real- The complex norm of a real number cast to ℂ equals its real norm.
Real.norm_eq_abs- The norm of a real number equals its absolute value.
abs_of_pos- The absolute value of a strictly positive number equals the number itself.
mul_le_mul_of_nonneg_left- Multiplying an inequality b ≤ c on the left by a non-negative element a ≥ 0 preserves the inequality: a * b ≤ a * c.
tlc_coverable_of_data- A triangular lattice of spacing d is coverable by non-overlapping unit disks if there exist parameters β, c₀ and n = 4/d² with ‖β‖² = n such that every lattice point can be approximated by a center within squared distance n.
tlc_tbl13_spec- Each entry in the 13-row lookup table satisfies the required parities modulo 2, residue modulo 13, and norm bound ≤ 13.
ZMod.val_intCast- The standard integer representative of (a : ℤ/nℤ) is the remainder a % n.
tlc_cover13- For all integers a, b, there exist integers x, y such that (a - 8x + 2y)² - (a - 8x + 2y)(b - 2x - 6y) + (b - 2x - 6y)² ≤ 13.
tlc_tbl21_spec- Each entry in the 21-element table consists of odd integers satisfying the residue condition modulo 21 and norm bound ≤ 21.
tlc_cover21- For all integers a, b, there exist integers x, y such that (2a - 3 - 10x + 2y)² - (2a - 3 - 10x + 2y)(2b + 1 - 2x - 8y) + (2b + 1 - 2x - 8y)² ≤ 21.
div_pow- The power of a quotient is the quotient of the powers: (a / b)ⁿ = aⁿ / bⁿ.
one_pow- One raised to any power is one: 1ⁿ = 1.
tlc_coverable_one- The triangular lattice of spacing d = 1 / √13 is coverable by pairwise non-overlapping unit disks.
tlc_coverable_two- The triangular lattice of spacing d = 2 / √21 is coverable by pairwise non-overlapping unit disks.
- The existence of integer covering witnesses (x, y) for every lattice point relies on explicit lookup tables (tlc_tbl13 with 52 entries and tlc_tbl21 with 21 entries) whose required properties are verified by computational reflection via `decide`.
- The set of disk centers is chosen as a dilated, rotated, and translated sublattice of the Eisenstein integers, with dilation factor ‖β‖ = 2/d ensuring that the minimum distance between distinct centers is exactly 2.
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.
open Complexprivate lemma tlc_omega_sq : eisensteinOmega^2 + eisensteinOmega + 1 = 0 := by have h3 : (Real.sqrt 3)^2 = 3 := Real.sq_sqrt (by norm_num) have hre : eisensteinOmega.re = -1/2 := by simp [eisensteinOmega] have him : eisensteinOmega.im = Real.sqrt 3 / 2 := by simp [eisensteinOmega] apply Complex.ext · simp only [pow_two, Complex.mul_re, Complex.add_re, hre, him, Complex.zero_re, Complex.one_re] ring_nf nlinarith [h3] · simp only [pow_two, Complex.mul_im, Complex.add_im, hre, him, Complex.zero_im, Complex.one_im] ring_nfprivate lemma tlc_omega_sq' : eisensteinOmega^2 = -eisensteinOmega - 1 := by linear_combination tlc_omega_sqprivate lemma tlc_normSq_int (a b : ℤ) : Complex.normSq ((a:ℂ) + (b:ℂ)*eisensteinOmega) = ((a^2 - a*b + b^2 : ℤ) : ℝ) := by rw [Complex.normSq_apply] have hre : ((a:ℂ) + (b:ℂ)*eisensteinOmega).re = (a:ℝ) - (b:ℝ)/2 := by simp [eisensteinOmega, Complex.add_re, Complex.mul_re]; ring have him : ((a:ℂ) + (b:ℂ)*eisensteinOmega).im = (b:ℝ)*Real.sqrt 3 / 2 := by simp [eisensteinOmega, Complex.add_im, Complex.mul_im]; ring rw [hre, him] have h3 : (Real.sqrt 3)^2 = 3 := Real.sq_sqrt (by norm_num) push_cast nlinarith [h3]private lemma tlc_one_le_norm (a b : ℤ) (h : ((a:ℂ) + (b:ℂ)*eisensteinOmega) ≠ 0) : 1 ≤ ‖(a:ℂ) + (b:ℂ)*eisensteinOmega‖ := by have hpos : 0 < Complex.normSq ((a:ℂ) + (b:ℂ)*eisensteinOmega) := by rwa [Complex.normSq_pos] rw [tlc_normSq_int] at hpos have hN : (1:ℤ) ≤ a^2 - a*b + b^2 := by have h0 : (0:ℤ) < a^2 - a*b + b^2 := by exact_mod_cast hpos nlinarith [h0] have hsq : (1:ℝ) ≤ ‖(a:ℂ) + (b:ℂ)*eisensteinOmega‖^2 := by rw [Complex.sq_norm, tlc_normSq_int] exact_mod_cast hN nlinarith [norm_nonneg ((a:ℂ) + (b:ℂ)*eisensteinOmega)]private lemma tlc_coverable_of_data (d : ℝ) (hd : 0 < d) (β c₀ : ℂ) (n : ℝ) (hβ : Complex.normSq β = n) (hdn : d^2 * n = 4) (hcov : ∀ a b : ℤ, ∃ x y : ℤ, ‖2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)‖^2 ≤ n) : is_coverable (triangular_lattice d) := by have hn : n = 4/d^2 := by field_simp at hdn ⊢ linarith have hβnorm : ‖β‖ = 2/d := by have hsq : ‖β‖^2 = (2/d)^2 := by rw [Complex.sq_norm, hβ, hn] field_simp ring nlinarith [norm_nonneg β, (by positivity : (0:ℝ) ≤ 2/d)] have hd2 : (0:ℝ) < d/2 := by positivity refine ⟨{z : ℂ | ∃ x y : ℤ, z = (d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega))}, ?_, ?_⟩ · rintro c c' ⟨x, y, rfl⟩ ⟨x', y', rfl⟩ hne have hq : ((x:ℂ) + (y:ℂ)*eisensteinOmega) ≠ ((x':ℂ) + (y':ℂ)*eisensteinOmega) := by intro h; exact hne (by rw [h]) have hdiff : ((x:ℂ) + (y:ℂ)*eisensteinOmega) - ((x':ℂ) + (y':ℂ)*eisensteinOmega) = ((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega := by push_cast; ring have hq' : ((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega ≠ 0 := by rw [← hdiff]; exact sub_ne_zero.mpr hq have hQ : 1 ≤ ‖((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega‖ := tlc_one_le_norm _ _ hq' have hnorm : ‖(d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) - (d:ℂ) * (c₀ + β*((x':ℂ) + (y':ℂ)*eisensteinOmega))‖ = d * (2/d) * ‖((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega‖ := by have he : (d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) - (d:ℂ) * (c₀ + β*((x':ℂ) + (y':ℂ)*eisensteinOmega)) = (d:ℂ) * β * (((x - x' : ℤ) : ℂ) + ((y - y' : ℤ) : ℂ)*eisensteinOmega) := by rw [← hdiff]; ring rw [he, norm_mul, norm_mul, Complex.norm_real, Real.norm_eq_abs, abs_of_pos hd, hβnorm] rw [hnorm, show d * (2/d) = 2 by field_simp] linarith · rintro z ⟨a, b, rfl⟩ obtain ⟨x, y, hxy⟩ := hcov a b refine ⟨(d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)), ⟨x, y, rfl⟩, ?_⟩ have hfac : (d:ℂ) * ((a:ℂ) + (b:ℂ)*eisensteinOmega) - (d:ℂ) * (c₀ + β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) = (d/2 : ℝ) * (2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)) := by push_cast; ring rw [hfac, norm_mul, Complex.norm_real, Real.norm_eq_abs, abs_of_pos hd2] have hW : ‖2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)‖ ≤ 2/d := by have hW2 : ‖2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)‖^2 ≤ (2/d)^2 := by rw [show (2/d)^2 = n from by rw [hn] at *; field_simp; ring] exact hxy nlinarith [norm_nonneg (2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*c₀ - 2*β*((x:ℂ) + (y:ℂ)*eisensteinOmega)), (by positivity : (0:ℝ) ≤ 2/d)] have := mul_le_mul_of_nonneg_left hW hd2.le rwa [show d/2 * (2/d) = 1 by field_simp] at thisprivate def tlc13row (i j : Fin 2) : Fin 13 → ℤ × ℤ := if (i : ℕ) = 0 ∧ (j : ℕ) = 0 then ![(0, 0), (-2, -4), (2, 0), (-2, 2), (-4, -2), (0, 2), (-2, -2), (2, 2), (0, -2), (4, 2), (2, -2), (-2, 0), (2, 4)] else if (i : ℕ) = 0 then ![(-4, -1), (0, 3), (-2, -1), (2, 3), (0, -1), (4, 3), (2, -1), (-2, 1), (-4, -3), (0, 1), (-2, -3), (2, 1), (0, -3)] else if (j : ℕ) = 0 then ![(-3, -4), (1, 0), (-1, -4), (3, 0), (-1, 2), (-3, -2), (1, 2), (-1, -2), (3, 2), (1, -2), (-3, 0), (1, 4), (-1, 0)] else ![(-1, 3), (-3, -1), (1, 3), (-1, -1), (3, 3), (1, -1), (-3, 1), (3, -1), (-1, 1), (-3, -3), (1, 1), (-1, -3), (3, 1)]private def tlc_tbl13 (i j : Fin 2) (r : Fin 13) : ℤ × ℤ := tlc13row i j rprivate lemma tlc_tbl13_spec (i j : Fin 2) (r : Fin 13) : let p := tlc_tbl13 i j r (p.1 % 2 = ((i:ℕ):ℤ) ∧ p.2 % 2 = ((j:ℕ):ℤ) ∧ (p.1 - 4*p.2) % 13 = ((r:ℕ):ℤ) ∧ p.1^2 - p.1*p.2 + p.2^2 ≤ 13) := by revert i j r decideprivate lemma tlc_cover13 (a b : ℤ) : ∃ x y : ℤ, (a - 8*x + 2*y)^2 - (a - 8*x + 2*y)*(b - 2*x - 6*y) + (b - 2*x - 6*y)^2 ≤ 13 := by set m : ZMod 13 := (a : ZMod 13) - 4*(b : ZMod 13) with hm set i : Fin 2 := ⟨(a : ZMod 2).val, ZMod.val_lt (a : ZMod 2)⟩ with hi set j : Fin 2 := ⟨(b : ZMod 2).val, ZMod.val_lt (b : ZMod 2)⟩ with hj set r : Fin 13 := ⟨m.val, ZMod.val_lt m⟩ with hr obtain ⟨h1, h2, h3, h4⟩ := tlc_tbl13_spec i j r set u : ℤ := (tlc_tbl13 i j r).1 with hu set v : ℤ := (tlc_tbl13 i j r).2 with hv rw [hu] at h1 h3 rw [hv] at h2 h3 simp only [hu, hv] at h4 have hi2 : ((i:ℕ):ℤ) = a % 2 := by simp only [hi] exact ZMod.val_intCast (n := 2) a have hj2 : ((j:ℕ):ℤ) = b % 2 := by simp only [hj] exact ZMod.val_intCast (n := 2) b have hm' : m = ((a - 4*b : ℤ) : ZMod 13) := by rw [hm]; push_cast; ring have hr2 : ((r:ℕ):ℤ) = (a - 4*b) % 13 := by simp only [hr] rw [hm'] exact ZMod.val_intCast (n := 13) (a - 4*b) have k1 : u % 2 = a % 2 := by rw [h1, hi2] have k2 : v % 2 = b % 2 := by rw [h2, hj2] have k3 : (u - 4*v) % 13 = (a - 4*b) % 13 := by rw [h3, hr2] set K : ℤ := (((a - u) - 4*(b - v))/2)/13 with hK refine ⟨(b - v)/2 + 3*K, -K, ?_⟩ have e1 : a - 8*((b - v)/2 + 3*K) + 2*(-K) = u := by omega have e2 : b - 2*((b - v)/2 + 3*K) - 6*(-K) = v := by omega rw [e1, e2] exact h4private def tlc_tbl21 : Fin 21 → ℤ × ℤ := ![(-5, -1), (-3, -5), (-3, -1), (-1, -5), (-1, -1), (-1, 3), (1, -1), (1, 3), (3, -1), (3, 3), (-5, -3), (5, 3), (-3, -3), (-3, 1), (-1, -3), (-1, 1), (1, -3), (1, 1), (1, 5), (3, 1), (3, 5)]private lemma tlc_tbl21_spec (r : Fin 21) : let p := tlc_tbl21 r (p.1 % 2 = 1 ∧ p.2 % 2 = 1 ∧ (p.1 - 5*p.2) % 21 = ((r:ℕ):ℤ) ∧ p.1^2 - p.1*p.2 + p.2^2 ≤ 21) := by revert r decideprivate lemma tlc_cover21 (a b : ℤ) : ∃ x y : ℤ, (2*a - 3 - 10*x + 2*y)^2 - (2*a - 3 - 10*x + 2*y)*(2*b + 1 - 2*x - 8*y) + (2*b + 1 - 2*x - 8*y)^2 ≤ 21 := by set m : ZMod 21 := ((2*a - 3 : ℤ) : ZMod 21) - 5*((2*b + 1 : ℤ) : ZMod 21) with hm set r : Fin 21 := ⟨m.val, ZMod.val_lt m⟩ with hr obtain ⟨h1, h2, h3, h4⟩ := tlc_tbl21_spec r set u : ℤ := (tlc_tbl21 r).1 with hu set v : ℤ := (tlc_tbl21 r).2 with hv rw [hu] at h1 h3 rw [hv] at h2 h3 simp only [hu, hv] at h4 have hm' : m = (((2*a - 3) - 5*(2*b + 1) : ℤ) : ZMod 21) := by rw [hm]; push_cast; ring have h3' : (u - 5*v) % 21 = ((2*a - 3) - 5*(2*b + 1)) % 21 := by rw [h3] simp only [hr] rw [hm'] exact ZMod.val_intCast (n := 21) ((2*a - 3) - 5*(2*b + 1)) set K : ℤ := ((((2*a - 3) - u) - 5*((2*b + 1) - v))/2)/21 with hK refine ⟨((2*b + 1) - v)/2 + 4*K, -K, ?_⟩ have e1 : 2*a - 3 - 10*(((2*b + 1) - v)/2 + 4*K) + 2*(-K) = u := by omega have e2 : 2*b + 1 - 2*(((2*b + 1) - v)/2 + 4*K) - 8*(-K) = v := by omega rw [e1, e2] exact h4private lemma tlc_coverable_one : is_coverable (triangular_lattice (1 / Real.sqrt 13)) := by have hd : (0:ℝ) < 1/Real.sqrt 13 := by positivity have hdn : (1/Real.sqrt 13)^2 * 52 = 4 := by rw [div_pow, one_pow, Real.sq_sqrt (by norm_num)]; norm_num refine tlc_coverable_of_data (1/Real.sqrt 13) hd (8 + 2*eisensteinOmega) 0 52 ?_ hdn ?_ · have h := tlc_normSq_int 8 2 norm_num at h exact h · intro a b obtain ⟨x, y, hxy⟩ := tlc_cover13 a b refine ⟨x, y, ?_⟩ have e : 2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*(0:ℂ) - 2*(8 + 2*eisensteinOmega)*((x:ℂ) + (y:ℂ)*eisensteinOmega) = ((2*a - 16*x + 4*y : ℤ) : ℂ) + ((2*b - 4*x - 12*y : ℤ) : ℂ)*eisensteinOmega := by push_cast ring_nf rw [tlc_omega_sq'] push_cast ring have hkey : ((2*a - 16*x + 4*y)^2 - (2*a - 16*x + 4*y)*(2*b - 4*x - 12*y) + (2*b - 4*x - 12*y)^2 : ℤ) = 4*((a - 8*x + 2*y)^2 - (a - 8*x + 2*y)*(b - 2*x - 6*y) + (b - 2*x - 6*y)^2) := by ring rw [e, Complex.sq_norm, tlc_normSq_int, hkey] have hZ : 4*((a - 8*x + 2*y)^2 - (a - 8*x + 2*y)*(b - 2*x - 6*y) + (b - 2*x - 6*y)^2) ≤ 52 := by linarith exact_mod_cast hZprivate lemma tlc_coverable_two : is_coverable (triangular_lattice (2 / Real.sqrt 21)) := by have hd : (0:ℝ) < 2/Real.sqrt 21 := by positivity have hdn : (2/Real.sqrt 21)^2 * 21 = 4 := by rw [div_pow, Real.sq_sqrt (by norm_num)]; norm_num refine tlc_coverable_of_data (2/Real.sqrt 21) hd (5 + eisensteinOmega) ((3 - eisensteinOmega)/2) 21 ?_ hdn ?_ · have h := tlc_normSq_int 5 1 norm_num at h exact h · intro a b obtain ⟨x, y, hxy⟩ := tlc_cover21 a b refine ⟨x, y, ?_⟩ have e : 2*((a:ℂ) + (b:ℂ)*eisensteinOmega) - 2*((3 - eisensteinOmega)/2) - 2*(5 + eisensteinOmega)*((x:ℂ) + (y:ℂ)*eisensteinOmega) = ((2*a - 3 - 10*x + 2*y : ℤ) : ℂ) + ((2*b + 1 - 2*x - 8*y : ℤ) : ℂ)*eisensteinOmega := by push_cast ring_nf rw [tlc_omega_sq'] push_cast ring rw [e, Complex.sq_norm, tlc_normSq_int] exact_mod_cast hxy/-- Coverability of the triangular lattice at spacings 1 / √13 and 2 / √21 by non-overlapping unit disks. -/theorem triangular_lattice_coverable_critical_spacings : is_coverable (triangular_lattice (1 / Real.sqrt 13)) ∧ is_coverable (triangular_lattice (2 / Real.sqrt 21)) := ⟨tlc_coverable_one, tlc_coverable_two⟩