amp_damping_shift_spectral_min_re
The smallest real part of an eigenvalue of $L_c = G + c\,\mathrm{id}$ is $c - 1$.
Statement
theorem amp_damping_shift_spectral_min_re (d : ℕ) (h : 2 ≤ d) (c : ℝ) :
spectral_min_re d (amp_damping_generator d h + (c : ℂ) • LinearMap.id) = c - 1- Formalizes
- arXiv:2610.37187 Proposition 9 (eq. 7.1), m(L_c) = c - 1
- Verified
- 08 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 4
- amp_damping_generatorThe $\mathbb{C}$-linear map $G : M_d(\mathbb{C}) \to M_d(\mathbb{C})$ (bundled as a linear endomorphism, for $d \ge 2$),…
- spectral_min_reThe minimal real part m(Φ) of an eigenvalue of a linear map Φ on d × d complex matrices is the infimum of the real parts of the elements…
- p_amp_dampingThe projection matrix P = |1⟩⟨1| onto the second standard basis state in dimension d ≥ 2.
- v_amp_dampingThe matrix unit V = |0⟩⟨1| in the standard basis of d × d complex matrices for d ≥ 2.
- Used by 1
- amp_damping_shift_sharp_spectral_boundFix $d \ge 2$, $V = |0\rangle\langle 1|$, $P = |1\rangle\langle 1|$, $G(X) = VXV^\dagger - \tfrac12(PX + XP)$ and…
Read back from the Lean
For every natural number $d$, every hypothesis $h : 2 \le d$, and every real number $c$, the minimum real part of the spectrum (over $\mathbb{C}$) of the linear map $(\Phi + c \cdot \mathrm{id})$ on the space of $d \times d$ complex matrices is equal to $c - 1$, where $\Phi = \text{amp\_damping\_generator}(d, h)$ is the amplitude damping generator.
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 find the smallest real part of an eigenvalue of the shifted Lindbladian $L_c = G + c\,\mathrm{id}$, the argument analyzes the entrywise action of $L_c$ on matrices. An explicit calculation shows that any eigenvalue $\mu$ of $L_c$ must satisfy $\mu \in \{c, c - 1/2, c - 1\}$, because an assumed eigenvector for any other scalar is forced entry by entry to vanish. Furthermore, the matrix difference $E_{1,1} - E_{0,0}$ is shown to be a genuine nonzero eigenvector with eigenvalue $c - 1$. Because all eigenvalues are real, bounded below by $c - 1$, and $c - 1$ is attained, the minimal real part of the spectrum is exactly $c - 1$.
The action of $L_c = G + c\,\mathrm{id}$ on a matrix $X$ is given entrywise by $(L_c(X))_{i,j} = \delta_{i,0}\delta_{j,0} X_{1,1} - \frac{1}{2}(\delta_{i,1} + \delta_{j,1}) X_{i,j} + c X_{i,j}$.
Unfolding the definition of the amplitude damping generator $G(X) = V X V^* - \frac{1}{2}(PX + XP)$ and the shift $c\,\mathrm{id}$, along with the definitions of $V = |0\rangle\langle 1|$ and $P = |1\rangle\langle 1|$, reduces matrix multiplication and conjugate transpose to sums over basis indicators. Simplifying the sums using indicator summation identities (`Finset.sum_ite_eq'`) and splitting into cases on whether $i = 1$ and whether $j = 1$ gives the entrywise expression.
uses
Matrix.mul_apply,Finset.sum_ite_eq',ite_andtheorem amp_damping_spmin_apply_entry (d : ℕ) (h : 2 ≤ d) (c : ℝ) (X : Matrix (Fin d) (Fin d) ℂ) (i j : Fin d) : ((amp_damping_generator d h + (c : ℂ) • LinearMap.id : Module.End ℂ (Matrix (Fin d) (Fin d) ℂ)) X) i j = (if i = ⟨0, by omega⟩ ∧ j = ⟨0, by omega⟩ then X ⟨1, by omega⟩ ⟨1, by omega⟩ else 0) - (1/2 : ℂ) * ((if i = ⟨1, by omega⟩ then X i j else 0) + (if j = ⟨1, by omega⟩ then X i j else 0)) + c * X i j := by simp only [amp_damping_generator, LinearMap.add_apply, LinearMap.sub_apply, LinearMap.comp_apply, LinearMap.smul_apply, LinearMap.mulLeft_apply, LinearMap.mulRight_apply, LinearMap.id_apply, Matrix.add_apply, Matrix.sub_apply, Matrix.smul_apply, smul_eq_mul] simp [v_amp_damping, p_amp_damping, Matrix.mul_apply, Matrix.star_apply, ite_and, Finset.sum_ite_eq'] simp only [apply_ite (starRingEnd ℂ), map_one, map_zero, mul_ite, mul_one, mul_zero, Finset.sum_ite_eq', Finset.mem_univ, if_true] by_cases hi : i = ⟨1, by omega⟩ <;> by_cases hj : j = ⟨1, by omega⟩ <;> simp [hi, hj]
To prove that every eigenvalue $\mu$ of $L_c$ is $c$, $c - 1/2$, or $c - 1$, it suffices to show that any eigenvector $X \ne 0$ for a different scalar $\mu$ must equal zero.
In finite dimensions, the spectrum coincides with the set of eigenvalues (`Module.End.hasEigenvalue_iff_mem_spectrum`). Choosing an eigenvector $X \ne 0$ with $L_c(X) = \mu X$ and assuming for contradiction that $\mu$ is neither $c$, $c - 1/2$, nor $c - 1$, the goal reduces to showing that $X = 0$.
uses
Module.End.hasEigenvalue_iff_mem_spectrumtheorem amp_damping_spmin_mem_spectrum_cases (d : ℕ) (h : 2 ≤ d) (c : ℝ) (μ : ℂ) (hμ : μ ∈ spectrum ℂ (amp_damping_generator d h + (c : ℂ) • LinearMap.id : Module.End ℂ (Matrix (Fin d) (Fin d) ℂ))) : μ = c ∨ μ = c - 1/2 ∨ μ = c - 1 := by rw [← Module.End.hasEigenvalue_iff_mem_spectrum] at hμ obtain ⟨X, hX⟩ := hμ.exists_hasEigenvector by_contra hne push Not at hne obtain ⟨h1, h2, h3⟩ := hne apply hX.2
The eigenvalue equation holds entrywise: $\mu X_{i,j} = \delta_{i,0}\delta_{j,0} X_{1,1} - \frac{1}{2}(\delta_{i,1} + \delta_{j,1}) X_{i,j} + c X_{i,j}$ for all indices $i, j$.
Rewriting the left-hand side of the entrywise formula with `amp_damping_spmin_apply_entry` and replacing $L_c(X)$ by $\mu X$ yields the entrywise relation.
uses
amp_damping_spmin_apply_entryhave eq : ∀ i j, μ * X i j = (if i = ⟨0, by omega⟩ ∧ j = ⟨0, by omega⟩ then X ⟨1, by omega⟩ ⟨1, by omega⟩ else 0) - (1/2 : ℂ) * ((if i = ⟨1, by omega⟩ then X i j else 0) + (if j = ⟨1, by omega⟩ then X i j else 0)) + c * X i j := by intro i j rw [← amp_damping_spmin_apply_entry d h c X i j, hX.apply_eq_smul] simp
The $(1, 1)$-entry of the eigenvector vanishes: $X_{1,1} = 0$.
Specializing the entrywise equation to $(i, j) = (1, 1)$ simplifies to $\mu X_{1,1} = - X_{1,1} + c X_{1,1}$, or equivalently $(\mu - (c - 1)) X_{1,1} = 0$. Because $\mu \ne c - 1$, the factor $X_{1,1}$ must be 0 (`mul_eq_zero.mp`).
uses
Fin.ext_iff,mul_eq_zero.mp,sub_eq_zero.mphave h01 : (⟨0, by omega⟩ : Fin d) ≠ ⟨1, by omega⟩ := by simp [Fin.ext_iff] have hx11 : X ⟨1, by omega⟩ ⟨1, by omega⟩ = 0 := by have e := eq ⟨1, by omega⟩ ⟨1, by omega⟩ simp only [h01.symm, false_and, if_false, if_true] at e have : (μ - (c - 1)) * X ⟨1, by omega⟩ ⟨1, by omega⟩ = 0 := by linear_combination e rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h3 · exact h'
Every entry of $X$ is zero, so $X = 0$, completing the characterization of the spectrum.
Examining the entrywise equation for all pairs $(i, j)$ by cases on whether each index equals $1$: if $i = 1$ and $j = 1$, $X_{1,1} = 0$; if $i = 1$ and $j \ne 1$, the equation gives $(\mu - (c - 1/2)) X_{1,j} = 0$, forcing $X_{1,j} = 0$ since $\mu \ne c - 1/2$; similarly, if $i \ne 1$ and $j = 1$, $X_{i,1} = 0$; finally, if both indices differ from $1$, the equation reduces to $(\mu - c) X_{i,j} = \delta_{i,0}\delta_{j,0} X_{1,1} = 0$, forcing $X_{i,j} = 0$ since $\mu \ne c$. Hence $X = 0$, contradicting $X \ne 0$.
uses
ite_self,mul_eq_zero.mp,sub_eq_zero.mpext i j have e := eq i j by_cases hi : i = ⟨1, by omega⟩ <;> by_cases hj : j = ⟨1, by omega⟩ · subst hi; subst hj; simpa using hx11 · subst hi simp only [h01.symm, false_and, hj, if_false, if_true] at e have : (μ - (c - 1/2)) * X ⟨1, by omega⟩ j = 0 := by linear_combination e rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h2 · simpa using h' · subst hj simp only [h01.symm, and_false, hi, if_false, if_true] at e have : (μ - (c - 1/2)) * X i ⟨1, by omega⟩ = 0 := by linear_combination e rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h2 · simpa using h' · simp only [hi, hj, if_false] at e have : (μ - c) * X i j = (if i = ⟨0, by omega⟩ ∧ j = ⟨0, by omega⟩ then X ⟨1, by omega⟩ ⟨1, by omega⟩ else 0) := by linear_combination e rw [hx11, ite_self] at this rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h1 · simpa using h'
$c - 1$ is an eigenvalue of $L_c$, and therefore lies in its spectrum.
Consider the candidate matrix $X = E_{1,1} - E_{0,0}$, where $E_{a,b}$ is the single-entry unit matrix. Evaluating entrywise via `amp_damping_spmin_apply_entry` across 9 index cases ($0$, $1$, or neither) shows that $L_c(X) = (c - 1) X$. Moreover, $X \ne 0$ because its $(1, 1)$-entry is $1 \ne 0$ (since $0 \ne 1$ in $\mathrm{Fin}(d)$ for $d \ge 2$). By `Module.End.hasEigenvalue_of_hasEigenvector`, $c - 1$ is an eigenvalue and thus in the spectrum.
uses
Module.End.hasEigenvalue_iff_mem_spectrum,Module.End.hasEigenvalue_of_hasEigenvector,Module.End.mem_eigenspace_iff,amp_damping_spmin_apply_entry,Matrix.single_apply,Ne.symm,congrFun,Fin.ext_ifftheorem amp_damping_spmin_mem_c_sub_one (d : ℕ) (h : 2 ≤ d) (c : ℝ) : ((c : ℂ) - 1) ∈ spectrum ℂ (amp_damping_generator d h + (c : ℂ) • LinearMap.id : Module.End ℂ (Matrix (Fin d) (Fin d) ℂ)) := by rw [← Module.End.hasEigenvalue_iff_mem_spectrum] apply Module.End.hasEigenvalue_of_hasEigenvector (x := Matrix.single ⟨1, by omega⟩ ⟨1, by omega⟩ 1 - Matrix.single ⟨0, by omega⟩ ⟨0, by omega⟩ 1) have h01 : (⟨0, by omega⟩ : Fin d) ≠ ⟨1, by omega⟩ := by simp [Fin.ext_iff] refine ⟨?_, ?_⟩ · rw [Module.End.mem_eigenspace_iff] ext i j rw [amp_damping_spmin_apply_entry] simp only [Matrix.sub_apply, Matrix.smul_apply, Matrix.single_apply, smul_eq_mul] have hc : ∀ x : Fin d, x = ⟨0, by omega⟩ ∨ x = ⟨1, by omega⟩ ∨ (x ≠ ⟨0, by omega⟩ ∧ x ≠ ⟨1, by omega⟩) := fun x => by tauto rcases hc i with rfl | rfl | ⟨hi0, hi1⟩ <;> rcases hc j with rfl | rfl | ⟨hj0, hj1⟩ <;> (try have hi0' := Ne.symm hi0) <;> (try have hi1' := Ne.symm hi1) <;> (try have hj0' := Ne.symm hj0) <;> (try have hj1' := Ne.symm hj1) <;> simp [h01, h01.symm, *] <;> ring · intro h0 have := congrFun (congrFun h0 ⟨1, by omega⟩) ⟨1, by omega⟩ simp [Matrix.single_apply, h01.symm] at this
The smallest real part of an eigenvalue of $L_c = G + c\,\mathrm{id}$ is $c - 1$.
By definition, `spectral_min_re` is the infimum of the real parts of eigenvalues. The previous lemmas established that $c - 1$ is an eigenvalue with real part $c - 1$ (`amp_damping_spmin_mem_c_sub_one`), and that every eigenvalue $\mu$ is either $c$, $c - 1/2$, or $c - 1$ (`amp_damping_spmin_mem_spectrum_cases`). Since each of these real parts is at least $c - 1$, $c - 1$ is the minimum (least element) of the set, so its infimum is $c - 1$ by `IsLeast.csInf_eq`.
uses
IsLeast.csInf_eq,amp_damping_spmin_mem_c_sub_one,amp_damping_spmin_mem_spectrum_casestheorem amp_damping_shift_spectral_min_re (d : ℕ) (h : 2 ≤ d) (c : ℝ) : spectral_min_re d (amp_damping_generator d h + (c : ℂ) • LinearMap.id) = c - 1 := by unfold spectral_min_re apply IsLeast.csInf_eq refine ⟨⟨(c : ℂ) - 1, amp_damping_spmin_mem_c_sub_one d h c, by simp⟩, ?_⟩ rintro x ⟨μ, hμ, rfl⟩ rcases amp_damping_spmin_mem_spectrum_cases d h c μ hμ with rfl | rfl | rfl <;> norm_num <;> linarith
Matrix.mul_apply- the $(i, k)$-entry of the matrix product $M N$ is given by the sum $\sum_j M_{i,j} N_{j,k}$
Finset.sum_ite_eq'- summing an indicator function over a finite set leaves the evaluated term if the target element belongs to the set, and 0 otherwise
ite_and- an if-then-else condition with a conjunction $P \wedge Q$ splits into nested conditions on $P$ and $Q$
Module.End.hasEigenvalue_iff_mem_spectrum- for a linear endomorphism of a finite-dimensional space over a field, a scalar is an eigenvalue if and only if it belongs to the spectrum
amp_damping_spmin_apply_entry- the $(i, j)$-entry of $(G + c\,\mathrm{id})(X)$ is $\delta_{i,0}\delta_{j,0} X_{1,1} - \frac{1}{2}(\delta_{i,1} + \delta_{j,1}) X_{i,j} + c X_{i,j}$
Fin.ext_iff- two elements of $\mathrm{Fin}(d)$ are equal if and only if their underlying natural numbers are equal
mul_eq_zero.mp- if a product of two numbers is zero, then at least one factor is zero
sub_eq_zero.mp- if $a - b = 0$, then $a = b$
ite_self- an if-then-else expression whose two branches are identical simplifies to that common value
Module.End.hasEigenvalue_of_hasEigenvector- an endomorphism has eigenvalue $\mu$ if it possesses an eigenvector for $\mu$
Module.End.mem_eigenspace_iff- a vector $x$ belongs to the eigenspace of $f$ for eigenvalue $\mu$ if and only if $f(x) = \mu x$
Matrix.single_apply- the $(i', j')$-entry of the single-entry matrix unit at $(i, j)$ with value $c$ is $c$ when $i = i'$ and $j = j'$, and 0 otherwise
Ne.symm- the inequality relation is symmetric: $a \ne b$ implies $b \ne a$
congrFun- equal functions take equal values at any argument
IsLeast.csInf_eq- if an element belongs to a set and is a lower bound for that set, it is the infimum of the set
amp_damping_spmin_mem_c_sub_one- $c - 1$ belongs to the spectrum of the shifted generator $G + c\,\mathrm{id}$
amp_damping_spmin_mem_spectrum_cases- any element $\mu$ in the spectrum of $G + c\,\mathrm{id}$ satisfies $\mu = c$, $\mu = c - 1/2$, or $\mu = c - 1$
- The entrywise formula in `amp_damping_spmin_apply_entry` and the eigenspace membership verification in `amp_damping_spmin_mem_c_sub_one` rely on exhaustive case splits on whether matrix indices equal 0, 1, or another index, with each branch closed by algebraic simplification (`simp` and `ring`).
- The dimension bound $2 \le d$ is repeatedly invoked via `omega` to justify that indices $0$ and $1$ are distinct valid elements of $\mathrm{Fin}(d)$.
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.
theorem amp_damping_spmin_apply_entry (d : ℕ) (h : 2 ≤ d) (c : ℝ) (X : Matrix (Fin d) (Fin d) ℂ) (i j : Fin d) : ((amp_damping_generator d h + (c : ℂ) • LinearMap.id : Module.End ℂ (Matrix (Fin d) (Fin d) ℂ)) X) i j = (if i = ⟨0, by omega⟩ ∧ j = ⟨0, by omega⟩ then X ⟨1, by omega⟩ ⟨1, by omega⟩ else 0) - (1/2 : ℂ) * ((if i = ⟨1, by omega⟩ then X i j else 0) + (if j = ⟨1, by omega⟩ then X i j else 0)) + c * X i j := by simp only [amp_damping_generator, LinearMap.add_apply, LinearMap.sub_apply, LinearMap.comp_apply, LinearMap.smul_apply, LinearMap.mulLeft_apply, LinearMap.mulRight_apply, LinearMap.id_apply, Matrix.add_apply, Matrix.sub_apply, Matrix.smul_apply, smul_eq_mul] simp [v_amp_damping, p_amp_damping, Matrix.mul_apply, Matrix.star_apply, ite_and, Finset.sum_ite_eq'] simp only [apply_ite (starRingEnd ℂ), map_one, map_zero, mul_ite, mul_one, mul_zero, Finset.sum_ite_eq', Finset.mem_univ, if_true] by_cases hi : i = ⟨1, by omega⟩ <;> by_cases hj : j = ⟨1, by omega⟩ <;> simp [hi, hj]theorem amp_damping_spmin_mem_spectrum_cases (d : ℕ) (h : 2 ≤ d) (c : ℝ) (μ : ℂ) (hμ : μ ∈ spectrum ℂ (amp_damping_generator d h + (c : ℂ) • LinearMap.id : Module.End ℂ (Matrix (Fin d) (Fin d) ℂ))) : μ = c ∨ μ = c - 1/2 ∨ μ = c - 1 := by rw [← Module.End.hasEigenvalue_iff_mem_spectrum] at hμ obtain ⟨X, hX⟩ := hμ.exists_hasEigenvector by_contra hne push Not at hne obtain ⟨h1, h2, h3⟩ := hne apply hX.2 have eq : ∀ i j, μ * X i j = (if i = ⟨0, by omega⟩ ∧ j = ⟨0, by omega⟩ then X ⟨1, by omega⟩ ⟨1, by omega⟩ else 0) - (1/2 : ℂ) * ((if i = ⟨1, by omega⟩ then X i j else 0) + (if j = ⟨1, by omega⟩ then X i j else 0)) + c * X i j := by intro i j rw [← amp_damping_spmin_apply_entry d h c X i j, hX.apply_eq_smul] simp have h01 : (⟨0, by omega⟩ : Fin d) ≠ ⟨1, by omega⟩ := by simp [Fin.ext_iff] have hx11 : X ⟨1, by omega⟩ ⟨1, by omega⟩ = 0 := by have e := eq ⟨1, by omega⟩ ⟨1, by omega⟩ simp only [h01.symm, false_and, if_false, if_true] at e have : (μ - (c - 1)) * X ⟨1, by omega⟩ ⟨1, by omega⟩ = 0 := by linear_combination e rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h3 · exact h' ext i j have e := eq i j by_cases hi : i = ⟨1, by omega⟩ <;> by_cases hj : j = ⟨1, by omega⟩ · subst hi; subst hj; simpa using hx11 · subst hi simp only [h01.symm, false_and, hj, if_false, if_true] at e have : (μ - (c - 1/2)) * X ⟨1, by omega⟩ j = 0 := by linear_combination e rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h2 · simpa using h' · subst hj simp only [h01.symm, and_false, hi, if_false, if_true] at e have : (μ - (c - 1/2)) * X i ⟨1, by omega⟩ = 0 := by linear_combination e rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h2 · simpa using h' · simp only [hi, hj, if_false] at e have : (μ - c) * X i j = (if i = ⟨0, by omega⟩ ∧ j = ⟨0, by omega⟩ then X ⟨1, by omega⟩ ⟨1, by omega⟩ else 0) := by linear_combination e rw [hx11, ite_self] at this rcases mul_eq_zero.mp this with h' | h' · exact absurd (sub_eq_zero.mp h') h1 · simpa using h'theorem amp_damping_spmin_mem_c_sub_one (d : ℕ) (h : 2 ≤ d) (c : ℝ) : ((c : ℂ) - 1) ∈ spectrum ℂ (amp_damping_generator d h + (c : ℂ) • LinearMap.id : Module.End ℂ (Matrix (Fin d) (Fin d) ℂ)) := by rw [← Module.End.hasEigenvalue_iff_mem_spectrum] apply Module.End.hasEigenvalue_of_hasEigenvector (x := Matrix.single ⟨1, by omega⟩ ⟨1, by omega⟩ 1 - Matrix.single ⟨0, by omega⟩ ⟨0, by omega⟩ 1) have h01 : (⟨0, by omega⟩ : Fin d) ≠ ⟨1, by omega⟩ := by simp [Fin.ext_iff] refine ⟨?_, ?_⟩ · rw [Module.End.mem_eigenspace_iff] ext i j rw [amp_damping_spmin_apply_entry] simp only [Matrix.sub_apply, Matrix.smul_apply, Matrix.single_apply, smul_eq_mul] have hc : ∀ x : Fin d, x = ⟨0, by omega⟩ ∨ x = ⟨1, by omega⟩ ∨ (x ≠ ⟨0, by omega⟩ ∧ x ≠ ⟨1, by omega⟩) := fun x => by tauto rcases hc i with rfl | rfl | ⟨hi0, hi1⟩ <;> rcases hc j with rfl | rfl | ⟨hj0, hj1⟩ <;> (try have hi0' := Ne.symm hi0) <;> (try have hi1' := Ne.symm hi1) <;> (try have hj0' := Ne.symm hj0) <;> (try have hj1' := Ne.symm hj1) <;> simp [h01, h01.symm, *] <;> ring · intro h0 have := congrFun (congrFun h0 ⟨1, by omega⟩) ⟨1, by omega⟩ simp [Matrix.single_apply, h01.symm] at thistheorem amp_damping_shift_spectral_min_re (d : ℕ) (h : 2 ≤ d) (c : ℝ) : spectral_min_re d (amp_damping_generator d h + (c : ℂ) • LinearMap.id) = c - 1 := by unfold spectral_min_re apply IsLeast.csInf_eq refine ⟨⟨(c : ℂ) - 1, amp_damping_spmin_mem_c_sub_one d h c, by simp⟩, ?_⟩ rintro x ⟨μ, hμ, rfl⟩ rcases amp_damping_spmin_mem_spectrum_cases d h c μ hμ with rfl | rfl | rfl <;> norm_num <;> linarith