# A two-term scalar detector and its remaining quadratic-encoding problem This checkpoint is based on notes of **Gabe Conant with Astra 6.2**. It proves a local scalar-detection lemma and an exact change of coordinates in the native pole equations. It does **not** give a smaller defining quartic. The established general counts remain **165 and 180**. The retained scalar and point conventions are those of `papers/quartic/shared_product_audit_20261119_v78.tex` and `papers/quartic/quartic193_20261117_v74.tex`. The only pole input used in the local implication below is the existing prohibition of negative odd valuations, at primes outside its stated exceptional set. This package does not establish that arithmetic input independently. ## 1. A universal local implication Let $K=\mathbb Q(\sqrt2)$, let $R>0$ be rational, and let ```math Z\in K,\qquad f\in K^\times,\qquad J=\frac{Z^2}{R^5}+\frac{f^2}{R^4},\qquad \tau=\frac{N_{K/\mathbb Q}(J)}{\operatorname{Tr}_{K/\mathbb Q}(J)}, \qquad b=R(R+9792)\tau^2. ``` At both real embeddings, $J$ is strictly positive. Thus its norm, trace, and $\tau$ are nonzero, and $b>0$. This includes $Z=0$. **Lemma.** Let $p\nmid 2\cdot9792$, let $\mu=v_p(R)>0$ be odd, and suppose that $b$ does not have negative odd $p$-adic valuation. There is a valuation slot $w$ of $K$ above $p$ for which ```math v_w(Z)>\frac{9\mu}{4},\qquad v_w(f)>\frac{7\mu}{4}. ``` Here valuations are normalized by $v_w(p)=1$; the two slots are counted with multiplicity at an inert prime, and $v_w(0)=+\infty$. **Proof.** Since $p\nmid9792$, ```math v_p(b)=\mu+2v_p(\tau) ``` is odd. The assumed obstruction makes it positive, so $v_p(\tau)>-\mu/2$. If $w_1,w_2$ are the two valuations of $J$, the trace inequality gives ```math v_p(\tau)=w_1+w_2-v_p(\operatorname{Tr}J) \le\max(w_1,w_2). ``` Choose a slot where $v_w(J)>-\mu/2$. The two finite summand valuations in $J$ are ```math 2v_w(Z)-5\mu\quad\hbox{and}\quad2v_w(f)-4\mu. ``` The first is odd and the second is even. They cannot tie; if $Z=0$, only the second is present. Therefore each present summand has valuation greater than $-\mu/2$. Rearranging proves both inequalities. No assumption on the parity of $v_w(f)$ was needed. $\square$ For the intended application, $f=f(H)=1/x(H)$ and $Z=2z(Q_2)/z(H)$. The lemma controls the quotient and the reference point at the **same** slot. Thus it supplies the local inequalities needed by the existing scalar formal-logarithm argument; it does not replace them by absolute depths of unrelated numerator points. ## 2. Real bounds and a precise local preparation condition If $R\ge1$, $|Z^{(i)}|\le M_Z$, and $|f^{(i)}|\le M_f$ at both real embeddings, put $A=M_Z^2+M_f^2$. Then ```math 05\mu/4$, which does not preserve the existing scalar height argument and its third powers of the contact radical. No complete input-preserving quadratic encoding below the current counts has been obtained here. The next concrete task is to remove this remaining fourth-power relation by a valid reassociation with the point or pole equations, or to find a quadratic point chart retaining the scaled quotient $Z/R$ without requiring its own $R^2$ register. Every downstream equation and every zero fiber would then have to be checked before claiming a smaller defining polynomial. ## Reproduction and scope of the certificate Run, from the repository root: ```sh python -B computation/variable_reductions_v81/general/check.py ``` The checker imports no construction producer. It verifies the rational coordinate identities in both directions, the rescaled norm/trace identities, and the exact parity and linear identities used in the universal valuation proof. It compares its result with the immutable `independent.json`; its default invocation writes no files. The certificate is not a finite-prime substitute for the proof, a global completeness certificate, or a new polynomial census.