The strict lower-half truncation
For \(d\in\mathbb N\) and \(p(q)=\sum_{n\ge0}a_nq^n\in\mathbb Z[q]\), define \[ \operatorname{low}_d(p)=\sum_{2n<d}a_nq^n. \] The inequality is strict. It retains the exponents below the midpoint \(d/2\).
Lean writes the finite index set as \[ 0\le n<\left\lfloor\frac{d+1}{2}\right\rfloor. \] For even \(d=2m\), the allowed exponents are \(0,\ldots,m-1\). For odd \(d=2m+1\), they are \(0,\ldots,m\). At \(d=0\), the sum is empty and the result is zero.
For \(p=4-3q+2q^2+7q^3\), \[ \operatorname{low}_3p=4-3q,\qquad \operatorname{low}_4p=4-3q,\qquad \operatorname{low}_5p=4-3q+2q^2. \] The operation discards coefficients according to the chosen bound \(d\), independently of their signs.
The low coefficients of a reciprocity equation
Suppose a polynomial \(P\) is supported strictly below \(d/2\), and \[ \operatorname{reflect}_dP=P+D. \] Every term of the reflected polynomial has degree strictly above \(d/2\). Therefore the coefficient of degree \(n\) with \(2n<d\) on the left is zero. The right coefficient is \(P_n+D_n\), so \(P_n=-D_n\).
This determines the candidate \[ P=-\operatorname{low}_dD. \] The anti-reciprocal solution theorem verifies that this candidate satisfies the whole equation when \(\operatorname{reflect}_dD=-D\).
For the polynomial example \(d=3\), \(D=q^3+2q^2-2q-1\), the candidate is \(P=1+2q\). Direct calculation gives \[ \operatorname{reflect}_3P=q^3+2q^2=P+D. \]
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.CanonicalP.lowPart: CanonicalKL.lean:23.