AFTD Auto-Formalizing Theoretical Domains

anticommuting_hermitian_expectation_sq_le_one

theoremverifiedformalization

If continuous linear operators P and Q on a complex Hilbert space are self-adjoint, have operator norm at most one, and anticommute (P * Q + Q * P = 0), then for any unit vector psi, the sum of the squares of the real expectation values satisfies (re <psi, P psi>)^2 + (re <psi, Q psi>)^2 <= 1.

Statement

theorem anticommuting_hermitian_expectation_sq_le_one
    {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℂ E] [CompleteSpace E]
    (P Q : E →L[ℂ] E) (hP : IsSelfAdjoint P) (hQ : IsSelfAdjoint Q)
    (hP_norm : ‖P‖ ≤ 1) (hQ_norm : ‖Q‖ ≤ 1)
    (h_anticomm : P * Q + Q * P = 0)
    (ψ : E) (hψ : ‖ψ‖ = 1) :
    (RCLike.re (inner ℂ ψ (P ψ))) ^ 2 + (RCLike.re (inner ℂ ψ (Q ψ))) ^ 2 ≤ 1
Formalizes
arXiv:2610.35838 Appendix E, Fact 1
Verified
01 Nov 2026
Axioms
Classical.choiceQuot.soundpropext

Read back from the Lean

Let $E$ be a complex Hilbert space (a complete normed additive commutative group with an inner product space structure over $\mathbb{C}$). Let $P, Q : E \toL[\mathbb{C}] E$ be continuous linear operators such that: 1. $P$ is self-adjoint (IsSelfAdjoint P), 2. $Q$ is self-adjoint (IsSelfAdjoint Q), 3. The operator norm of $P$ satisfies $\|P\| \le 1$, 4. The operator norm of $Q$ satisfies $\|Q\| \le 1$, and 5. $P$ and $Q$ anticommute, i.e., $P \circ Q + Q \circ P = 0$. Then for any unit vector $\psi \in E$ (satisfying $\|\psi\| = 1$), the sum of the squares of the real parts of the expectation values of $P$ and $Q$ on $\psi$ is at most $1$: $$(\operatorname{Re} \langle \psi, P\psi \rangle)^2 + (\operatorname{Re} \langle \psi, Q\psi \rangle)^2 \le 1.$$

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 $(\operatorname{Re}\langle\psi, P\psi\rangle)^2 + (\operatorname{Re}\langle\psi, Q\psi\rangle)^2 \le 1$, we set $a = \operatorname{Re}\langle\psi, P\psi\rangle$, $b = \operatorname{Re}\langle\psi, Q\psi\rangle$, and define the test vector $v = a P\psi + b Q\psi$. Using self-adjointness and the anticommutation relation $PQ + QP = 0$, the real cross term $\operatorname{Re}\langle P\psi, Q\psi\rangle$ vanishes, which implies $\|v\|^2 \le a^2 + b^2$ because $\|P\|, \|Q\| \le 1$ and $\|\psi\| = 1$. Meanwhile, computing $\operatorname{Re}\langle \psi, v\rangle = a^2 + b^2$ and applying Cauchy-Schwarz gives $a^2 + b^2 \le \|v\|$. Squaring this bound gives $(a^2 + b^2)^2 \le \|v\|^2 \le a^2 + b^2$, which immediately implies $a^2 + b^2 \le 1$.

  1. Define the real expectation values $a = \operatorname{Re}\langle \psi, P\psi\rangle$ and $b = \operatorname{Re}\langle \psi, Q\psi\rangle$, and form the test vector $v = a P\psi + b Q\psi$.

    We introduce real shorthands $a$ and $b$ for the real expectation values of $P$ and $Q$ in the quantum state $\psi$, and construct the linear combination $v = a P\psi + b Q\psi \in E$.

    lines 9–11

      set a := RCLike.re (inner ℂ ψ (P ψ))
      set b := RCLike.re (inner ℂ ψ (Q ψ))
      set v := (a : ℂ) • (P ψ) + (b : ℂ) • (Q ψ)
  2. The vectors $P\psi$ and $Q\psi$ are orthogonal with respect to the real part of the inner product: $\operatorname{Re}\langle P\psi, Q\psi\rangle = 0$.

    Using the self-adjointness of $P$ and $Q$, we have $\langle P\psi, Q\psi\rangle = \langle \psi, P Q\psi\rangle$ and $\langle Q\psi, P\psi\rangle = \langle \psi, Q P\psi\rangle$. By conjugate symmetry, $\langle P\psi, Q\psi\rangle + \overline{\langle P\psi, Q\psi\rangle} = \langle \psi, (PQ + QP)\psi\rangle$. Since $P$ and $Q$ anticommute ($PQ + QP = 0$), this sum is zero. Using the identity $z + \bar{z} = 2\operatorname{Re}(z)$, we get $2\operatorname{Re}\langle P\psi, Q\psi\rangle = 0$, which implies $\operatorname{Re}\langle P\psi, Q\psi\rangle = 0$.

    lines 12–33

    uses ContinuousLinearMap.adjoint_inner_right, ContinuousLinearMap.star_eq_adjoint, inner_conj_symm, inner_add_right, inner_zero_right, RCLike.add_conj, mul_eq_zero.mp, Complex.ofReal_eq_zero.mp

      have h_ortho : RCLike.re (inner ℂ (P ψ) (Q ψ)) = 0 := by
        have h1 : inner ℂ (P ψ) (Q ψ) = inner ℂ ψ (P (Q ψ)) := by
          have h := ContinuousLinearMap.adjoint_inner_right P ψ (Q ψ)
          rw [← ContinuousLinearMap.star_eq_adjoint, hP] at h
          exact h.symm
        have h2 : inner ℂ (Q ψ) (P ψ) = inner ℂ ψ (Q (P ψ)) := by
          have h := ContinuousLinearMap.adjoint_inner_right Q ψ (P ψ)
          rw [← ContinuousLinearMap.star_eq_adjoint, hQ] at h
          exact h.symm
        have h3 : starRingEnd ℂ (inner ℂ (P ψ) (Q ψ)) = inner ℂ (Q ψ) (P ψ) :=
          inner_conj_symm (Q ψ) (P ψ)
        have h4 : inner ℂ (P ψ) (Q ψ) + starRingEnd ℂ (inner ℂ (P ψ) (Q ψ)) = 0 := by
          rw [h3, h1, h2, ← inner_add_right]
          change inner ℂ ψ ((P * Q + Q * P) ψ) = 0
          rw [h_anticomm]
          exact inner_zero_right ψ
        have h5 := (RCLike.add_conj (inner ℂ (P ψ) (Q ψ))).symm.trans h4
        have h6 : (RCLike.re (inner ℂ (P ψ) (Q ψ)) : ℂ) = 0 := by
          cases mul_eq_zero.mp h5 with
          | inl h7 => norm_num at h7
          | inr h7 => exact h7
        exact Complex.ofReal_eq_zero.mp h6
  3. The squared norm of the test vector satisfies $\|v\|^2 \le a^2 + b^2$.

    Expanding $\|v\|^2 = \|a P\psi + b Q\psi\|^2$ gives $\|a P\psi\|^2 + 2ab\operatorname{Re}\langle P\psi, Q\psi\rangle + \|b Q\psi\|^2$. The cross term vanishes because $\operatorname{Re}\langle P\psi, Q\psi\rangle = 0$. Since $P$ and $Q$ have operator norm at most 1 and $\|\psi\| = 1$, we have $\|P\psi\| \le 1$ and $\|Q\psi\| \le 1$. Therefore $\|a P\psi\|^2 = a^2 \|P\psi\|^2 \le a^2$ and $\|b Q\psi\|^2 = b^2 \|Q\psi\|^2 \le b^2$, which yields $\|v\|^2 \le a^2 + b^2$.

    lines 35–61

    uses norm_add_sq, mul_zero, mul_le_mul_of_nonneg_right, zero_le_one, norm_smul, Complex.norm_real, Real.norm_eq_abs, mul_pow, sq_abs, norm_nonneg

      have h_norm_sq : ‖v‖ ^ 2 ≤ a ^ 2 + b ^ 2 := by
        rw [norm_add_sq (𝕜 := ℂ)]
        have h_cross : RCLike.re (inner ℂ ((a : ℂ) • (P ψ)) ((b : ℂ) • (Q ψ))) = 0 := by
          calc RCLike.re (inner ℂ ((a : ℂ) • (P ψ)) ((b : ℂ) • (Q ψ)))
            _ = a * b * RCLike.re (inner ℂ (P ψ) (Q ψ)) := by simp; ring
            _ = 0 := by rw [h_ortho, mul_zero]
        have hP_le : ‖P ψ‖ ≤ 1 := by
          have h := P.le_opNorm ψ
          have : ‖P‖ * ‖ψ‖ ≤ 1 * 1 := by rw [hψ]; exact mul_le_mul_of_nonneg_right hP_norm zero_le_one
          linarith
        have hQ_le : ‖Q ψ‖ ≤ 1 := by
          have h := Q.le_opNorm ψ
          have : ‖Q‖ * ‖ψ‖ ≤ 1 * 1 := by rw [hψ]; exact mul_le_mul_of_nonneg_right hQ_norm zero_le_one
          linarith
        have hP_sq : ‖(a : ℂ) • (P ψ)‖ ^ 2 ≤ a ^ 2 := by
          rw [norm_smul, Complex.norm_real, Real.norm_eq_abs, mul_pow, sq_abs]
          have : ‖P ψ‖ ^ 2 ≤ 1 := by
            have : 0 ≤ ‖P ψ‖ := norm_nonneg _
            nlinarith
          nlinarith
        have hQ_sq : ‖(b : ℂ) • (Q ψ)‖ ^ 2 ≤ b ^ 2 := by
          rw [norm_smul, Complex.norm_real, Real.norm_eq_abs, mul_pow, sq_abs]
          have : ‖Q ψ‖ ^ 2 ≤ 1 := by
            have : 0 ≤ ‖Q ψ‖ := norm_nonneg _
            nlinarith
          nlinarith
        linarith
  4. The real inner product of $\psi$ with $v$ equals $a^2 + b^2$: $\operatorname{Re}\langle \psi, v\rangle = a^2 + b^2$.

    Expanding $v = a P\psi + b Q\psi$ inside $\langle \psi, v\rangle$ by linearity in the second argument yields $\operatorname{Re}(a\langle\psi, P\psi\rangle + b\langle\psi, Q\psi\rangle)$. Pulling the real scalars $a$ and $b$ out of the real part and using the definitions of $a = \operatorname{Re}\langle\psi, P\psi\rangle$ and $b = \operatorname{Re}\langle\psi, Q\psi\rangle$, this evaluates to $a \cdot a + b \cdot b = a^2 + b^2$.

    lines 63–71

    uses inner_add_right, inner_smul_right, map_add, sq

      have h_inner : RCLike.re (inner ℂ ψ v) = a ^ 2 + b ^ 2 := by
        simp only [v, inner_add_right, inner_smul_right, map_add]
        have ha : RCLike.re ((a : ℂ) * inner ℂ ψ (P ψ)) = a ^ 2 := by
          have h : RCLike.re ((a : ℂ) * inner ℂ ψ (P ψ)) = a * RCLike.re (inner ℂ ψ (P ψ)) := by simp
          rw [h, sq]
        have hb : RCLike.re ((b : ℂ) * inner ℂ ψ (Q ψ)) = b ^ 2 := by
          have h : RCLike.re ((b : ℂ) * inner ℂ ψ (Q ψ)) = b * RCLike.re (inner ℂ ψ (Q ψ)) := by simp
          rw [h, sq]
        rw [ha, hb]
  5. By the Cauchy-Schwarz inequality, $a^2 + b^2 \le \|v\|$.

    We bound $\operatorname{Re}\langle \psi, v\rangle \le |\langle \psi, v\rangle|$. By Cauchy-Schwarz, $|\langle \psi, v\rangle| \le \|\psi\| \cdot \|v\|$. Since $\psi$ is a unit vector ($\|\psi\| = 1$), this gives $\operatorname{Re}\langle \psi, v\rangle \le \|v\|$, which translates to $a^2 + b^2 \le \|v\|$.

    lines 73–78

    uses RCLike.re_le_norm, norm_inner_le_norm, one_mul

      have h_cs : a ^ 2 + b ^ 2 ≤ ‖v‖ := by
        rw [← h_inner]
        calc RCLike.re (inner ℂ ψ v)
          _ ≤ ‖inner ℂ ψ v‖ := RCLike.re_le_norm _
          _ ≤ ‖ψ‖ * ‖v‖ := norm_inner_le_norm _ _
          _ = ‖v‖ := by rw [hψ, one_mul]
  6. Squaring the inequality and combining it with the norm bound yields $a^2 + b^2 \le 1$.

    Since $a^2 + b^2 \ge 0$, squaring $a^2 + b^2 \le \|v\|$ gives $(a^2 + b^2)^2 \le \|v\|^2$. Combining this with the upper bound $\|v\|^2 \le a^2 + b^2$ gives $(a^2 + b^2)^2 \le a^2 + b^2$. For any non-negative real number $x$, $x^2 \le x$ implies $x \le 1$; applying this to $x = a^2 + b^2$ concludes that $a^2 + b^2 \le 1$.

    lines 80–86

      have h_sq_le : (a ^ 2 + b ^ 2) ^ 2 ≤ a ^ 2 + b ^ 2 := by
        have hpos : 0 ≤ a ^ 2 + b ^ 2 := by positivity
        have h1 : (a ^ 2 + b ^ 2) ^ 2 ≤ ‖v‖ ^ 2 := by nlinarith
        linarith
    
      have h_final : a ^ 2 + b ^ 2 ≤ 1 := by nlinarith
      exact h_final
ContinuousLinearMap.adjoint_inner_right
For a continuous linear map $A$, the inner product $\langle x, A^* y\rangle$ equals $\langle A x, y\rangle$.
ContinuousLinearMap.star_eq_adjoint
The involution $A \mapsto \mathrm{star}(A)$ on continuous linear maps equals the Hilbert space adjoint operator $A \mapsto A^*$.
inner_conj_symm
The inner product has conjugate symmetry: $\overline{\langle y, x\rangle} = \langle x, y\rangle$.
inner_add_right
The inner product is additive in its right argument: $\langle x, y + z\rangle = \langle x, y\rangle + \langle x, z\rangle$.
inner_zero_right
The inner product of any vector with zero on the right is zero: $\langle x, 0\rangle = 0$.
RCLike.add_conj
A complex number plus its complex conjugate equals twice its real part: $z + \bar{z} = 2 \operatorname{Re}(z)$.
mul_eq_zero.mp
If a product equals zero, at least one factor equals zero.
Complex.ofReal_eq_zero.mp
If the complex embedding of a real number is zero, the real number is zero.
norm_add_sq
The expansion of the squared norm of a sum: $\|x + y\|^2 = \|x\|^2 + 2\operatorname{Re}\langle x, y\rangle + \|y\|^2$.
mul_zero
Multiplying any number by zero gives zero: $a \cdot 0 = 0$.
mul_le_mul_of_nonneg_right
If $b \le c$ and $0 \le a$, then $b \cdot a \le c \cdot a$.
zero_le_one
Zero is less than or equal to one: $0 \le 1$.
norm_smul
The norm of a scalar multiple is the product of norms: $\|r \cdot x\| = \|r\| \cdot \|x\|$.
Complex.norm_real
The complex modulus of a real number embedded into $\mathbb{C}$ equals its real absolute value.
Real.norm_eq_abs
The real norm of a real number is its absolute value: $\|r\| = |r|$.
mul_pow
Powers distribute over multiplication: $(ab)^n = a^n b^n$.
sq_abs
The square of an absolute value is the square of the number: $|a|^2 = a^2$.
norm_nonneg
The norm of any vector is non-negative: $0 \le \|a\|$.
inner_smul_right
The inner product is linear in its right argument under scalar multiplication: $\langle x, r \cdot y\rangle = r \langle x, y\rangle$.
map_add
An additive homomorphism preserves addition: $f(x + y) = f(x) + f(y)$.
sq
The square of an element is the element multiplied by itself: $a^2 = a \cdot a$.
RCLike.re_le_norm
The real part of a complex number is bounded by its modulus: $\operatorname{Re}(z) \le |z|$.
norm_inner_le_norm
The Cauchy-Schwarz inequality for the inner product: $|\langle x, y\rangle| \le \|x\| \cdot \|y\|$.
one_mul
Multiplying on the left by one is the identity: $1 \cdot a = a$.

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

/-- Fact 1: Anticommuting Hermitian operators with norm <= 1 satisfy <P>^2 + <Q>^2 <= 1 in any quantum state. -/theorem anticommuting_hermitian_expectation_sq_le_one    {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℂ E] [CompleteSpace E]    (P Q : E →L[ℂ] E) (hP : IsSelfAdjoint P) (hQ : IsSelfAdjoint Q)    (hP_norm : ‖P‖ ≤ 1) (hQ_norm : ‖Q‖ ≤ 1)    (h_anticomm : P * Q + Q * P = 0)    (ψ : E) (hψ : ‖ψ‖ = 1) :    (RCLike.re (inner ℂ ψ (P ψ))) ^ 2 + (RCLike.re (inner ℂ ψ (Q ψ))) ^ 2 ≤ 1 := by  set a := RCLike.re (inner ℂ ψ (P ψ))  set b := RCLike.re (inner ℂ ψ (Q ψ))  set v := (a : ℂ) • (P ψ) + (b : ℂ) • (Q ψ)  have h_ortho : RCLike.re (inner ℂ (P ψ) (Q ψ)) = 0 := by    have h1 : inner ℂ (P ψ) (Q ψ) = inner ℂ ψ (P (Q ψ)) := by      have h := ContinuousLinearMap.adjoint_inner_right P ψ (Q ψ)      rw [← ContinuousLinearMap.star_eq_adjoint, hP] at h      exact h.symm    have h2 : inner ℂ (Q ψ) (P ψ) = inner ℂ ψ (Q (P ψ)) := by      have h := ContinuousLinearMap.adjoint_inner_right Q ψ (P ψ)      rw [← ContinuousLinearMap.star_eq_adjoint, hQ] at h      exact h.symm    have h3 : starRingEnd ℂ (inner ℂ (P ψ) (Q ψ)) = inner ℂ (Q ψ) (P ψ) :=      inner_conj_symm (Q ψ) (P ψ)    have h4 : inner ℂ (P ψ) (Q ψ) + starRingEnd ℂ (inner ℂ (P ψ) (Q ψ)) = 0 := by      rw [h3, h1, h2, ← inner_add_right]      change inner ℂ ψ ((P * Q + Q * P) ψ) = 0      rw [h_anticomm]      exact inner_zero_right ψ    have h5 := (RCLike.add_conj (inner ℂ (P ψ) (Q ψ))).symm.trans h4    have h6 : (RCLike.re (inner ℂ (P ψ) (Q ψ)) : ℂ) = 0 := by      cases mul_eq_zero.mp h5 with      | inl h7 => norm_num at h7      | inr h7 => exact h7    exact Complex.ofReal_eq_zero.mp h6  have h_norm_sq : ‖v‖ ^ 2 ≤ a ^ 2 + b ^ 2 := by    rw [norm_add_sq (𝕜 := ℂ)]    have h_cross : RCLike.re (inner ℂ ((a : ℂ) • (P ψ)) ((b : ℂ) • (Q ψ))) = 0 := by      calc RCLike.re (inner ℂ ((a : ℂ) • (P ψ)) ((b : ℂ) • (Q ψ)))        _ = a * b * RCLike.re (inner ℂ (P ψ) (Q ψ)) := by simp; ring        _ = 0 := by rw [h_ortho, mul_zero]    have hP_le : ‖P ψ‖ ≤ 1 := by      have h := P.le_opNorm ψ      have : ‖P‖ * ‖ψ‖ ≤ 1 * 1 := by rw [hψ]; exact mul_le_mul_of_nonneg_right hP_norm zero_le_one      linarith    have hQ_le : ‖Q ψ‖ ≤ 1 := by      have h := Q.le_opNorm ψ      have : ‖Q‖ * ‖ψ‖ ≤ 1 * 1 := by rw [hψ]; exact mul_le_mul_of_nonneg_right hQ_norm zero_le_one      linarith    have hP_sq : ‖(a : ℂ) • (P ψ)‖ ^ 2 ≤ a ^ 2 := by      rw [norm_smul, Complex.norm_real, Real.norm_eq_abs, mul_pow, sq_abs]      have : ‖P ψ‖ ^ 2 ≤ 1 := by        have : 0 ≤ ‖P ψ‖ := norm_nonneg _        nlinarith      nlinarith    have hQ_sq : ‖(b : ℂ) • (Q ψ)‖ ^ 2 ≤ b ^ 2 := by      rw [norm_smul, Complex.norm_real, Real.norm_eq_abs, mul_pow, sq_abs]      have : ‖Q ψ‖ ^ 2 ≤ 1 := by        have : 0 ≤ ‖Q ψ‖ := norm_nonneg _        nlinarith      nlinarith    linarith  have h_inner : RCLike.re (inner ℂ ψ v) = a ^ 2 + b ^ 2 := by    simp only [v, inner_add_right, inner_smul_right, map_add]    have ha : RCLike.re ((a : ℂ) * inner ℂ ψ (P ψ)) = a ^ 2 := by      have h : RCLike.re ((a : ℂ) * inner ℂ ψ (P ψ)) = a * RCLike.re (inner ℂ ψ (P ψ)) := by simp      rw [h, sq]    have hb : RCLike.re ((b : ℂ) * inner ℂ ψ (Q ψ)) = b ^ 2 := by      have h : RCLike.re ((b : ℂ) * inner ℂ ψ (Q ψ)) = b * RCLike.re (inner ℂ ψ (Q ψ)) := by simp      rw [h, sq]    rw [ha, hb]  have h_cs : a ^ 2 + b ^ 2 ≤ ‖v‖ := by    rw [← h_inner]    calc RCLike.re (inner ℂ ψ v)      _ ≤ ‖inner ℂ ψ v‖ := RCLike.re_le_norm _      _ ≤ ‖ψ‖ * ‖v‖ := norm_inner_le_norm _ _      _ = ‖v‖ := by rw [hψ, one_mul]  have h_sq_le : (a ^ 2 + b ^ 2) ^ 2 ≤ a ^ 2 + b ^ 2 := by    have hpos : 0 ≤ a ^ 2 + b ^ 2 := by positivity    have h1 : (a ^ 2 + b ^ 2) ^ 2 ≤ ‖v‖ ^ 2 := by nlinarith    linarith  have h_final : a ^ 2 + b ^ 2 ≤ 1 := by nlinarith  exact h_final
See it in the graphDiscuss this resultChallenge it